Split (3)

train · 4.35k rows

article stringlengths 1.98k 169k	summary stringlengths 1.01k 4.15k	section_headings listlengths 2 38	keywords listlengths 0 12	year stringclasses 11 values	title stringlengths 30 189
In this study , we investigated how microtubule motors organize microtubules in Drosophila neurons . We showed that , during the initial stages of axon outgrowth , microtubules display mixed polarity and minus-end-out microtubules push the tip of the axon , consistent with kinesin-1 driving outgrowth by sliding antiparallel microtubules . At later stages , the microtubule orientation in the axon switches from mixed to uniform polarity with plus-end-out . Dynein knockdown prevents this rearrangement and results in microtubules of mixed orientation in axons and accumulation of microtubule minus-ends at axon tips . Microtubule reorganization requires recruitment of dynein to the actin cortex , as actin depolymerization phenocopies dynein depletion , and direct recruitment of dynein to the membrane bypasses the actin requirement . Our results show that cortical dynein slides ‘minus-end-out’ microtubules from the axon , generating uniform microtubule arrays . We speculate that differences in microtubule orientation between axons and dendrites could be dictated by differential activity of cortical dynein . Neuronal development involves the dramatic morphologic reorganization of spherical morphologically undifferentiated cells to highly polarized mature neurons . Developing neurons grow long microtubule-based axons and dendrites . Over the last decades , the origin of the mechanical forces that drive process formation in neurons has been the subject of numerous studies ( Suter and Miller , 2011; Kapitein and Hoogenraad , 2015 ) . Recently , our lab demonstrated that kinesin-1 , a major microtubule motor , slides microtubules against each other in many cell types ( Jolly et al . , 2010; Barlan et al . , 2013 ) , including neurons , and this sliding plays an important role in neuronal polarization . We have shown that microtubule-microtubule sliding is required for initial axon formation and regeneration after injury ( Lu et al . , 2013b; 2015 ) . This process is likely required for neurodevelopment in many organisms besides Drosophila , as kinesin-driven microtubule sliding has been implicated in generating the minus-end-out microtubule pattern observed in dendrites of Caenorhabditis elegans neurons ( Yan et al . , 2013 ) . The main well-established function of kinesin-1 ( also known as conventional kinesin ) is the transport of cargoes along microtubules in the cytoplasm . Each kinesin-1 molecule is a heterotetramer that consists of two heavy chains ( KHC ) and two light chains ( Kuznetsov et al . , 1988 ) . Each KHC polypeptide contains two microtubule-binding domains: one ATP-dependent site in the motor domain and a second ATP-independent site at the C-terminus ( Hackney and Stock , 2000; Seeger and Rice , 2010; Yan et al . , 2013 ) . Kinesin-1 is thought to slide microtubules against each other with these two heavy chain domains; one microtubule is used as a track , while the other is transported as a cargo; kinesin light chains are not required for sliding ( Jolly et al . , 2010; Yan et al . , 2013 ) . Axons contain microtubule arrays of uniform orientation with plus-ends facing the axon tip ( Baas et al . , 1988; Stone et al . , 2008 ) . However , kinesin-1 is a plus-end motor , and therefore can only slide microtubules with their minus-ends leading and plus-ends trailing ( Figure 1A ) , which is inconsistent with the final orientation of microtubules in mature axons . To address this apparent contradiction , we asked two questions: First , are microtubules indeed pushed with their minus-ends out at the initial stages of axon outgrowth , as would be expected if they are pushed by kinesin-1 ? Second , if this is the case , how are microtubules with the ‘wrong’ orientation replaced by microtubules with normal ( plus-end-out ) orientation in mature axons ? To address these questions , we imaged and tracked markers of microtubule plus-ends and minus-ends in cultured Drosophila neurons and S2 cells at different stages of process growth . Our results showed that , at the initial stages of neurite formation , microtubules have mixed polarity with minus-ends being pushed against the plasma membrane; later , cytoplasmic dynein , attached to the actin cortex , removes minus-end-out microtubules to the cell body , creating microtubule arrays with uniform plus-end-out orientation . We speculate that regulation of dynein’s microtubule sorting activity could explain the differences in microtubule orientation between axons and dendrites . 10 . 7554/eLife . 10140 . 003Figure 1 . Microtubule minus-ends push the plasma membrane during the initial stages of neurite outgrowth . ( A ) Model of microtubule-microtubule sliding driven by kinesin-1 . Kinesin-1 slides antiparallel microtubules apart with their minus-ends leading ( left panel ) . When kinesin-1 binds to parallel microtubules ( right panel ) , forces applied by the two motors to the two microtubules are counteracted resulting in no net movement; instead , kinesin-1 crosslinks these microtubules . Large green arrows indicate direction of microtubule sliding; small orange arrows indicate the direction of kinesin-1 movement relative to microtubules . ( B ) A representative S2 cell expressing GFP-CAMSAP3 and mCherry-tubulin . Note that CAMSAP3 molecules accumulate at microtubule ends . Two different regions of the cell body ( labeled 1 and 2 ) were magnified in the insets ( see Video 2 ) . Scale bar , 5 µm . ( C and D ) Minus-ends of microtubules localize at the tips of growing processes during the initial stages of process formation in S2 cells . GFP-CAMSAP3 expressing S2 cells were plated on coverslips and imaged 5 min after plating . The plasma membrane was stained with a Deep Red membrane dye ( red ) . ( C ) Last frame of a time-lapse video . Images at different time points of the growing process in the white box are shown at higher magnification . Green arrows indicate positions of the most distal CAMSAP3 dot; magenta arrows show the position of the tip of the process ( see Video 4 ) . Scale bars are 10 µm and 3 µm for main and inset panels , respectively . ( D ) A graph showing the position of the process tip and the microtubule minus-ends shown in the inset of ( C ) as a function of time . ( E–F ) Microtubule plus-ends do not colocalize with the tip of growing processes in S2 cells . ( E ) Representative kymographs of growing processes from cells expressing GFP-CAMSAP3 ( left panel ) or EB1-GFP ( right panel ) . The plasma membrane was stained with a Deep Red membrane dye . Note that CAMSAP3 consistently localizes at the tips of the processes during outgrowth events , however EB1 comets do not colocalize with the tip of the growing processes ( horizontal scale bar , 10 µm; vertical scale bar , 25 s ) . ( F ) Graph depicting the fraction of time that CAMSAP3 or EB1 colocalize with the tips of the processes during the growing events . Error bars indicate s . d . ( CAMSAP3 , n=55 growing processes; EB1 , n=51 growing processes ) . Data collected from four independent experiments . ***p<0 . 0001 . ( G–I ) Localization of microtubule minus-ends at the tips of the processes during the initial stages of neurite formation in cultured neurons . ( G ) A still image of 4 hr-cultured neurons expressing elav>GFP-CAMSAP3 . The plasma membrane was labeled with Deep Red dye . Note that CAMSAP3 mostly localized to the tips of neurites . Scale bar , 5 µm . ( H ) Diagram showing the position of the neurite tip and the microtubule minus-ends of the axon showed in ( I ) as a function of time . ( I ) Still images from a time-lapse of a 4 hr-cultured neuron plated as described in ( G ) . Yellow dashed lines are guides to visualize the neurite growth . Green arrows indicate positions of CAMSAP3; magenta arrows show position of the tip of the process ( see Video 5 ) . Scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 00310 . 7554/eLife . 10140 . 004Figure 1—figure supplement 1 . CAMSAP3 labels minus-ends of microtubules in Drosophila S2 cells . ( A ) Microtubules in S2 cells expressing GFP-CAMSAP3 were fragmented by 1 hr treatment with 25 µM vinblastine; cells were fixed and stained with primary antitubulin antibody DM1α and TRITC-labeled secondary antibody . Note that GFP-CAMSAP3 decorates only one end of each microtubule fragment . Scale bar , 5 µm . ( B ) A still image of a S2 cell coexpressing EB1-GFP and mCherry-CAMSAP3 ( see Video 3 ) . The yellow box represents the area magnified in ( C ) . Scale bar , 10 µm . ( C ) Still frames from a time-lapse sequence of the cell depicted in ( B ) captured at different time points . The white bracket and the white arrowhead show localization of mCherry-CAMSAP3 or EB1-GFP , respectively , on a single microtubule . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 004 We previously demonstrated that kinesin-1 slides microtubules against each other , and this sliding generates the forces that drives outgrowth at the initial stages of neurite outgrowth ( Lu et al . , 2013b ) and axon regeneration ( Lu et al . , 2015 ) . Because kinesin-1 is a plus-end microtubule motor , it can only slide microtubules with their minus-ends leading and plus-ends trailing ( Figure 1A ) . If this model is correct , it suggests that kinesin-1 must extend neurites by pushing microtubule minus-ends against the plasma membrane during the initial stages of neurite formation . Furthermore , because the model predicts that two microtubules have to be in antiparallel orientation to slide against each other , sliding by kinesin-1 will result in the simultaneous transport of two microtubules in opposite directions ( see Figure 1A and the legend for the explanation ) . Bidirectional microtubule movement can indeed be observed in growing axons of cultured Drosophila neurons using tubulin tagged with a photoconvertible probe ( Video 1 ) . 10 . 7554/eLife . 10140 . 005Video 1 . Microtubules slide in both directions in Drosophila-cultured neurons . Time-lapse video of photoconverted microtubules in Drosophila-cultured neurons expressing tdEOS-αtubulin . A small area of the nascent axon was photoconverted by 405 nm light . Note that microtubules slide in both directions . Scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 005 To initially test this hypothesis , we first took advantage of Drosophila S2 tissue culture cells . S2 cells provide a good model system to explore the mechanism of process formation because they canbe induced to form cellular processes when the integrity of the actin filament network is impaired by treatment with either Cytochalasin D or Latrunculin B ( LatB ) ( Kim et al . , 2007; Lu et al . , 2013a ) . In addition , this system enables us to efficiently study the mechanisms of process formation by knocking down candidate proteins with double-stranded RNA ( dsRNA ) ( Rogers and Rogers , 2008 ) . To study microtubule minus-end distribution in live cells , we ectopically expressed a fluorescently tagged microtubule minus-end binding protein called calmodulin-regulated spectrin-associated protein ( CAMSAP ) , also known as Patronin or Nezha . CAMSAP proteins bind to microtubule minus-ends and stabilize them against depolymerization , making them the perfect candidate to label microtubule minus-ends ( Akhmanova and Hoogenraad , 2015 ) . We initially performed experiments with GFP-tagged Patronin , the single Drosophila member of CAMSAP family ( Wang et al . , 2013 ) , but its expression level in S2 cells was very low and GFP signal was not robustly found on microtubules ( data no shown ) . On the other hand , its mammalian ortholog CAMSAP3 tagged with GFP expressed at consistently higher levels and reliably decorated microtubule ends ( Figure 1B ) . First , we wanted to test whether GFP-CAMSAP3 decorates only one end of microtubules in Drosophila cells . Because the microtubule network is normally too dense to identify both ends of microtubules , we induced the formation of short microtubules in cells by partial depolymerization with 25 µM Vinblastine for 1 hr . Examination of these short microtubule fragments demonstrated that only one end of each microtubule contained a GFP-CAMSAP3 patch ( Figure 1—figure supplement 1A ) . In untreated S2 cells , spontaneous growth and shrinkage events associated with the dynamic instability of microtubule plus-ends were not seen in microtubule ends decorated by GFP-CAMSAP3 , suggesting that GFP-CAMSAP3 labels microtubule minus-ends in Drosophila cells ( Video 2 ) . Furthermore , EB1-GFP and mCherry-CAMSAP3 never colocalized when expressed in the same cell , further confirming the minus-end localization of CAMSAP3 ( Figure 1—figure supplement 1B and 1C; Video 3 ) . These results , together with published data ( Tanaka et al . , 2012; Hendershott and Vale , 2014; Jiang et al . , 2014; Akhmanova and Hoogenraad , 2015 ) , demonstrated that mammalian GFP-CAMSAP3 reliably marks microtubule minus-ends in Drosophila . 10 . 7554/eLife . 10140 . 006Video 2 . CAMSAP3 decorates microtubule-ends in Drosophila cells . Related to Figure 1B . A time-lapse video of S2 cells expressing GFP-CAMSAP3 and mCherry-tubulin . Scale bar 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 00610 . 7554/eLife . 10140 . 007Video 3 . Microtubule plus-ends and minus-ends binding proteins do not colocalize in Drosophila S2 cells . Related to Figure 1—figure supplement 1B , C . A time-lapse video of a S2 cell coexpressing EB1-GFP and mCherry-CAMSAP3 . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 007 To study localization of microtubule minus-ends in growing processes , we induced the formation of processes in S2 cells expressing GFP-CAMSAP3 and started collecting images 5 min after plating the cells . At this time point , nascent processes were actively growing . We simultaneously tracked GFP-CAMSAP3 and the plasma membrane using a membrane dye ( CellMask Deep Red ) . We found that a significant fraction of growing processes contained GFP-CAMSAP3 dots at their tips , and that these processes only elongated when microtubule minus-ends were present at their tips ( Figure 1C , D; Video 4 ) . Interestingly , we often observed retraction of the GFP-CAMSAP3 marker from the process tip; these events always coincided with a pause in process outgrowth ( Figure 1C , inset ) . Quantitative analysis demonstrated that while CAMSAP3 almost always colocalized with the tips of the growing processes , the plus-ends marker EB1 could only be found in the tips of the growing processes approximately 30% of the time ( Figure 1E , F ) , suggesting that at this stage microtubule dynamics does not play a major role in process outgrowth . 10 . 7554/eLife . 10140 . 008Video 4 . Initial process outgrowth in Drosophila S2 cells is driven by microtubule minus-ends . Related to Figure 1C . A time-lapse video of Drosophila S2 cells expressing GFP-CAMSAP3 plated for 5 min . Deep red dye was used to stain the membrane . Scale bars 10 µm and 5 µm , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 008 To investigate the localization of microtubule minus-ends in Drosophila neurons , we created a transgenic fly that expresses GFP-CAMSAP3 under the UAS promoter . Neurons were harvested and cultured from the brains of larvae expressing GFP-CAMSAP3 driven by the pan-neuronal promoter elav-Gal4 ( Egger et al . , 2013; Lu et al . , 2015 ) ( see Material and methods ) . We visualized microtubule minus-ends in growing neurites at the initial stages of growth ( 4 hr after plating ) , when neurons started to develop processes ( length= 9 . 96 µm , s . d . ± 4 . 5 µm , n=50 axons ) . We found that , like in S2 cells , the growing neurites contained GFP-CAMSAP3 dots at the tips of neurites , and localization of the dots to the tips precisely correlated with neurite outgrowth ( Figure 1G–I; Video 5 ) . All together , these data show that at least a fraction of microtubules in growing neurites have the ‘wrong’ orientation ( minus-end-out ) . Localization of microtubule minus-end ( s ) at the neurite tip correlates with neurite outgrowth , consistent with kinesin-1 pushing the minus-ends of microtubules against the plasma membrane driving the initial outgrowth . 10 . 7554/eLife . 10140 . 009Video 5 . Microtubule minus-ends push the plasma membrane in growing neurites of young cultured Drosophila neurons . Related to Figure 1I . Time-lapse video of a Drosophila neuron expressing elav>GFP-CAMSAP3 cultured for 4 hr . Deep red dye was used to stain the membrane . Scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 009 If kinesin-1 slides antiparallel microtubules at the initial stages of axon formation , the growing neurites should initially contain microtubules with mixed orientation ( Figure 1A and the legend for the explanation ) . To test this prediction , we imaged and tracked the direction of the plus-end microtubule marker , EB1-GFP , in the axons of cultured neurons at different time points after plating . Quantitative analysis of EB1-GFP comets using tracking software ( see Material and methods ) demonstrated that , shortly after plating , growing neurites contained EB1 comets moving in both anterograde and retrograde directions ( Figure 2A , B; Video 6 ) . The microtubule orientation in axons remained mixed during the 1st day in culture . At 36 hr , the fraction of retrograde EB1 comets started to decline , and at 48 hr , developed axons were mostly filled with plus-end-out microtubules ( Figure 2A , B; Video 6 ) . These results demonstrated that axons initially contain microtubule arrays with mixed orientation . 10 . 7554/eLife . 10140 . 010Figure 2 . Dynein specifies microtubule orientation in axons . ( A and B ) Axonal microtubules gradually acquire uniform orientation during development . ( A ) Representative still images of EB1-GFP expressing neurons cultured for 4 hr , 21 hr , or 48 hr . Kymographs of EB1 comets are shown below corresponding images . Magenta and green arrows indicate direction of EB1 comet movement ( plus-end-in and plus-end-out , respectively ) . Dashed yellow lines define the area of the axon used for plotting EB1-GFP kymographs ( see Video 6 ) . Scale bars , 10 µm . ( B ) Fraction of EB1-GFP comets directed toward the tip of the neurites ( plus-end-out ) or the cell body ( plus-end-in ) . See Material and methods for an explanation of EB1 comet quantification . Error bars indicate s . d . ( 4 hr , n=35 axons with 761 comets; 21 hr , n=33 axons with 408 comets; 36 hr , n=33 axons with 526 comets; 48 hr , n=25 axons with 299 comets ) . p=0 . 034 , **p<0 . 0001 , n . s . = not significant . Data collected from three independent experiments . ( C and D ) Dynein knockdown causes mixed orientation of microtubules in axons . ( C ) Representative still images of control ( elav-gal4 ) and dynein knockdown ( two different DHC shRNAs driven by elav-Gal4 ) 48 hr-cultured neurons expressing EB1-GFP . Magenta and green arrows indicate directions of EB1 comet movement ( plus-end-in and plus-end-out , respectively ) ( see Video 7 ) . Scale bars , 10 µm . ( D ) Fraction of EB1-GFP comets directed toward the tip of the neurites ( plus-end-out ) or the cell body ( plus-end-in ) . Error bars indicate s . d . ( Control , n=28 axons with 269 comets; DHC RNAi#1 , n=45 axons with 928 comets; DHC RNAi#2 , n=30 axons with 454 comets ) . p<0 . 0001 . Data collected from three independent experiments . ( E ) Dynein knockdown in neurons results in accumulation of microtubule minus-ends at the tips of neurites . Images of 48 hr-cultured neurons expressing elav>GFP-CAMSAP3 ( left panel ) or elav>GFP-CAMSAP3 + DHC RNAi ( right panel ) . Bottom panels are magnifications of the yellow-boxed areas . Scale bars , 10 µm . ( F–G ) Dynein inactivation induces antiparallel microtubule arrays in S2 cell processes . ( F ) Representative images of untreated ( control ) or DHC dsRNA-treated S2 cells expressing EB1-GFP . Magenta and green arrows indicate plus-end-in or plus-end-out direction of EB1 comet movement , respectively ( see Video 8 ) . Scale bars , 10 µm . ( G ) Graphs depict the direction of EB1-GFP comets in the processes of control S2 cells , and cells after knockdown of DHC , p150Glued , Lis1 , or NudE . Error bars indicate s . d . ( Control , n=55 cells with 1747 comets; DHC RNAi , n=50 cells with 1929 comets; p150 RNAi , n=26 cells with 2282 comets; Lis1 RNAi , n=33 cells with 3359 comets; NudE RNAi , n=24 cells with 2518 comets ) . p=0 . 001–0 . 0001 , *p<0 . 0001 . Data collected from three independent experiments . ( H–I ) Dynein inactivation in S2 cells results in accumulation of microtubule minus-ends in the process tips . ( H ) Representative images of untreated ( control ) or DHC dsRNA-treated S2 cells expressing GFP-CAMSAP3 . In control S2 cells , CAMSAP3 particles display a scattered distribution with few minus-ends at process tips . In dynein RNAi S2 cells , CAMSAP3 particles robustly accumulate at the tips of the processes . Scales bars , 10 µm . ( I ) Graphs show the fraction of S2 cells displaying the phenotypes depicted in ( H ) . ( Control , n=117 cells; DHC RNAi , n=84 cells; p150Glued=90 cells; Lis1 RNA1 , n=83 cells; NudE RNAi=79 cells ) . DHC , dynein heavy chain . *p<0 . 0001 . Data collected from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 01010 . 7554/eLife . 10140 . 011Figure 2—figure supplement 1 . Knockdown efficiency of DHC and dynein cofactors in Drosophila S2 cells . ( A ) Western blot analysis of S2 cell extracts obtained from control ( untreated ) or DHC dsRNA-treated cells using anti-DHC and anti-KHC ( loading control ) antibodies . ( B ) Temporal color code hyper-stacks representing the displacement of lysosomes in S2 cells . Lysosomes were stained by addition of Lysotracker to the medium . Lysosome transport of untreated ( control ) or dsRNA-treated S2 cells was recorded for 1 min and analyzed by temporal code . White color represents static lysosomes . DHC , dynein heavy chain . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 01110 . 7554/eLife . 10140 . 012Figure 2—figure supplement 2 . Distribution of GFP-CAMSAP3 in DHC RNAi S2 processes . ( A ) Distribution of GFP-CAMSAP3 in DHC depleted S2 process . CAMSAP3 molecules accumulate mainly at the end of the tip of the processes ( upper panel ) . Adjustment of contrast in the boxed area reveals multiple GFP-CAMSAP3 speckles in the intermediate segment of the process ( bottom panel ) . Scale bars , 10 µM . ( B ) Schematic representation of microtubules organization in DHC depleted S2 process . In the absence of dynein , S2 cells develop processes containing microtubule arrays of mixed polarity . The tips of the processes contain massive accumulation of minus-ends , other minus-ends are scattered in the shafts of the processes . DHC , dynein heavy chain . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 01210 . 7554/eLife . 10140 . 013Video 6 . Axonal microtubule organization switches from mixed to uniform orientation during neuron development . Related to Figure 2A . Time-lapse videos of Drosophila-cultured neurons expressing ubi-EB1-GFP cultured for 4 hr , 21 hr and 48 hr . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 013 Our data suggest that some sorting mechanism is responsible for the elimination of microtubules with the ‘wrong’ ( plus-ends toward the soma ) orientation from maturing axons . We hypothesized that this sorting factor is cytoplasmic dynein because it has been reported that mutations in dynein light intermediate chain or dynein-cofactors ( NudE ) are required for uniform microtubule orientation in axons of Drosophila dendritic arborization ( da ) neurons ( Zheng et al . , 2008; Arthur et al . , 2015 ) . To characterize the role of dynein , we expressed two different shRNAs targeting dynein heavy chain ( DHC ) using elav-Gal4 to specifically knock down dynein in neurons . elav>DHC RNAi animals developed to the third instar larval stage; however , their locomotion was severely impaired and most died before reaching the pupae stage ( data not shown ) . In addition , none of the surviving elav>DHC RNAi pupae eclosed into adults . We cultured neurons obtained from brains of third instar elav>DHC RNAi larvae . Our lab has previously shown that mitochondrial movement was substantially diminished in elav>DHC RNAi neurons , indicating that the activity of dynein is impaired in those neurons ( Lu et al . , 2015 ) . To track the microtubule orientation after dynein depletion , we genetically combined transgenes encoding DHC RNAi with EB1-GFP or GFP-CAMSAP3 . We first quantified the direction of the EB1 comets in neurons grown for 48 hr; at this time point , microtubules in the axons of control neurons are mostly oriented with plus-end-out ( Figure 2B ) . Analysis of EB1 comets in elav>DHC RNAi 48 hr-neurons revealed that axons contained microtubule arrays of mixed orientation ( Figure 2C , D; Video 7 ) , suggesting that dynein is necessary to remove minus-end-out microtubules from axons . Interestingly , while microtubule minus-ends in control neurons , as visualized by GFP-CAMSAP3 , were scattered throughout the length of the axons , inactivation of dynein resulted in dramatic accumulation of the minus-end markers at neurite tips ( Figure 2E; control , left panels; DHC-RNAi , right panels ) . 10 . 7554/eLife . 10140 . 014Video 7 . Dynein knockdown causes axons to contain antiparallel microtubules . Related to Figure 2C . Time-lapse videos of a control and two elav>DHC shRNA Drosophila cultured neurons expressing EB1-GFP . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 014 To further investigate the role of dynein in organizing microtubule arrays in cell processes , we again examined Drosophila S2 cells . The treatment of S2 cells with DHC dsRNA efficiently knocked down DHC ( Figure 2—figure supplement 1A , B ) . Recapitulating the neuronal phenotype , dynein depletion in S2 cells resulted in processes with EB1-GFP comets traveling in both directions ( Figure 2F , G; Video 8 ) . We also examined the distribution of the GFP-CAMSAP3 in S2 processes . In control cells , GFP-CAMSAP3 has a broad , scattered dot distribution ( Figure 2H , left panel ) . Dynein depletion induced a striking en masse accumulation of CAMSAP3 at the tips of processes ( Figure 2H , I ) . Overexposed images of those processes revealed that minus-ends could be still found in other areas of the processes ( Figure 2—figure supplement 2 ) . Identical phenotypes ( mixed polarity of GFP-EB1 comets and accumulation of CAMSAP3 at the tips of the processes ) were observed after knocking down several dynein cofactors ( p150Glued , Lis1 or NudE; Figure 2G–I; Figure 2—figure supplement 1B for knockdown efficiency ) . The en masse accumulation of microtubule minus-ends at neurite tips and S2 processes in dynein RNAi cells is consistent with the idea while kinesin-driven sliding stays intact , the sorting mechanism is inactivated by dynein depletion . 10 . 7554/eLife . 10140 . 015Video 8 . Dynein RNAi causes mixed microtubule orientation in processes of Drosophila S2 cells . Related to Figure 2F . Time-lapse videos of a control ( untreated ) and DHC RNAi Drosophila S2 cells expressing EB1-GFP . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 015 In many organisms , the organization and orientation of the mitotic spindle requires attachment of cytoplasmic dynein to the cellular cortex ( Laan et al . , 2012; Kiyomitsu and Cheeseman , 2013 ) . We hypothesized that organization of microtubules in axons , like their organization in the mitotic spindle , requires anchoring of dynein to the cortical network of actin filaments . To test this idea , we depolymerized actin filaments in cultured Drosophila neurons by treatment with LatB and quantified microtubule orientation in axons . We found that increasing concentrations of LatB gradually reduced the amount of F-actin , as judged by the intensity of rhodamine-phalloidin staining ( data not shown ) . In parallel with F-actin depolymerization , axons displayed an increased fraction of microtubules with plus-ends facing the cell body ( Figure 3A ) . At a high concentration of LatB ( 10 µM ) , when virtually no F-actin can be detected , about half of all EB1 comets moved toward the cell body , demonstrating that depolymerization of actin induced random microtubule orientation in neurons ( Figure 3A ) . Treatment with high LatB concentration also induced accumulation of microtubule minus-ends at the tips of the axons in elav>GFP-CAMSAP3 neurons ( Figure 3B ) . 10 . 7554/eLife . 10140 . 016Figure 3 . Dynein recruitment to the cortex is required for microtubule sorting . ( A ) Actin depolymerization in cultured neurons results in formation of axons with antiparallel microtubules . Graph depicts the fraction of axonal EB1 comets moving in each direction in 48 hr-cultured neurons treated with different concentrations of LatB . Error bars indicate s . d . ( 0 µM LatB , n=34 axons with 905 comets; 1 µM LatB , n=20 axons with 852 comets; 2 µM LatB , n=37 axons with 102 comets; 10 µM LatB , n=15 axons with 547 comets ) . p=0 . 0144 , **p<0 . 0001 . Data collected from three independent experiments . ( B ) LatB treatment induces accumulation of minus-ends in neurite tips . A 48 hr-cultured neuron expressing elav>GFP-CAMSAP3 . Panels on the right are magnified areas of the neurite tip before or 14 hr after LatB washout . Images are overexposed to show the outline of the neurite . Note that LatB washout induces retrograde movement of CAMSAP3 decorated microtubules . Scale bars are 10 µm and 5 µm , respectively . ( C–E ) Actin depolymerization in S2 cells results in the formation of processes with antiparallel microtubule orientation and an accumulation of microtubule minus-ends at the tips of processes ( compare with Figure 2F–H ) . ( C ) Graph depicting the direction of EB1-GFP comets in S2 processes . Error bars indicate s . d . ( Control , n=20 cells with 1833 comets; 10 µM LatB , n=22 cells with 2408 comets ) . p<0 . 0001 . Data collected from three independent experiments . ( D ) Graph depicting the distribution of GFP-CAMSAP3 in the processes of S2 cells treated with 0 . 5 µM ( control ) or 10 µM LatB ( Control , n=82 cells; 10 µM LatB , n=75 cells ) . **p<0 . 0001 . Data collected from three independent experiments . ( E ) Confocal image of a S2 cell expressing GFP-CAMSAP3 plated in 10 µM LatB . Note that GFP-CAMSAP3 accumulates at process tips . Scale bar , 10 µm . ( F and G ) Recruitment of dynein to cortical actin activates the sorting activity of dynein . ( F ) Representative still images from time-lapses of S2 cells expressing GFP-CAMSAP3 plated for 4 hr in the presence of 10 µM of LatB . In control cells , the microtubule minus-ends remain at the tips of the processes ( left panels ) . LatB washout resulted in a clearing of microtubules minus-ends from processes ( middle panels ) . Dynein knockdown impairs the microtubule sorting activity after LatB washout and microtubule minus-ends remained clustered at the tips ( see Video 9 ) . Scale bars , 10 µm . ( G ) Schematic representation of the LatB washout assays ( microtubule minus-ends are represented in magenta ) . In the presence of LatB , dynein is decoupled from the plasma membrane , preventing its sorting activity ( top panel ) . After washout , dynein is recruited to the plasma membrane by cortical actin , resulting in robust microtubule sorting and transport of minus-end-out microtubules toward the cell body ( bottom panel ) . ( H–I ) Direct recruitment of dynein to the membrane bypasses the F-actin requirement for microtubule sorting . Endogenous dynein can be recruited to the plasma membrane in S2 cells coexpressing FRB-GFP-BicD and GAP43-FKBP . In the presence of 10 µM LatB , cortical actin is depolymerized and therefore dynein remains soluble in the cytoplasm . Addition of 1 µM rapalog induces the direct recruitment of the dynein-BicD complex to the plasma membrane ( see Material and methods ) . The sorting activity of dynein was tracked by imaging mCherry-CAMSAP3 . ( H ) Still images from time-lapses of a S2 cell before ( left panel ) and after addition of rapalog ( middle and right panels ) . Magenta and green arrowheads represent the position of the membrane and the CAMSAP3 , respectively . Note that the CAMSAP3 signal moves toward the cell body when rapalog is added while there is not a substantial retraction of the processes ( see Video 10 ) . If the same experiment is performed in DHC RNAi cells , addition of rapalog does not induce retrograde transport of the minus-end-out microtubules from the tips of processes ( see Video 10 ) . Scale bars , 10 µm . ( I ) Schematic representation of the BicD-dynein recruitment assays ( microtubule minus-ends are represented in magenta ) . In the presence of LatB , dynein is soluble in the cytoplasm . Addition of rapalog directly recruits dynein to the membrane in the presence of BicD recruitment proteins , activating dynein’s sorting activity and retrograde transport of microtubules to the cell body . ( J ) Fraction of processes that displayed retrograde movement of minus-end-out microtubules as imaged by CAMSAP3 signal . ( Control , n=122 processes; DHC RNAi , n=99 processes ) . DHC , dynein heavy chain . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 01610 . 7554/eLife . 10140 . 017Figure 3—figure supplement 1 . Validation of the rapalog recruitment assays in S2 cells . ( A ) S2 cells expressing FRB-GFP and PEX3-RFP-FKBP before ( upper panels ) and after ( bottom panels ) addition of rapalog . Note that in the presence of rapalog , GFP signal concentrates to peroxisomes . Scale bars , 10 µm . ( B ) TIRF images of S2 cells expressing FRB-GFP-BicD and GAP43-FKBP before and after addition of rapalog . Both images were acquired using the same laser power and exposure time . ( C ) Quantification of the GFP TIRF signal in cells expressing FRB-GFP-BicD with or without GAP43-FKBP after addition of rapalog . Note that the increase of the GFP signal at the membrane only happens in the presence of GAP43-FKBP , demonstrating that BicD was successfully recruited to the plasma membrane . Error bars indicate s . d . ( FRB-GFP-BicD , n=12 cells; FRB-GFP-BicD + GAP43-FKBP , n=14 cells ) . TIRF , total internal reflection fluorescence . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 017 These data show that actin depolymerization phenocopied dynein depletion in axons , suggesting that cortical dynein sorts microtubules . If this hypothesis is correct , LatB washout should restore dynein recruitment to the cortex , and the microtubule sorting activity of cortical dynein should remove minus-ends-out microtubules from the tip of the axon . To test this prediction , we washed out LatB and tracked the localization of GFP-CAMSAP3 . We observed that the LatB washout induced removal of GFP-CAMSAP3 from axon tips without affecting the length of the axon ( Figure 3B ) . It should be mentioned that LatB washout did not result in the normal scattered distribution of minus-ends along the length of the axon observed in control neurons ( Figure 2E , left panels ) . Instead , they remained clustered in the shaft , most likely because by the time of drug washout these minus-end-out microtubules were already crosslinked into a bundle . To observe the sorting activity of cortical dynein in real time , we again used S2 cells . As in the case of neurons , the treatment of S2 cells with high concentration of LatB induced formation of processes containing microtubules of mixed polarity ( Figure 3C ) and massive accumulation of minus-ends in their tips ( Figure 3D–F; Video 9 ) . Strikingly , LatB washout resulted in robust movement of GFP-CAMSAP3 labeled minus-ends from the tips of processes toward the cell body ( Figure 3F , middle panel; Figure 3G; Video 9 ) . However , if the same experiment was performed after dynein knockdown , retrograde transport of microtubule minus-ends was not observed and the minus-ends caps stay at the tips of processes ( Figure 3F , right panel; Video 9 ) , directly demonstrating the role of cortical dynein in removal of minus-end-out microtubules from processes to the cell body . 10 . 7554/eLife . 10140 . 018Video 9 . Cortical dynein sorts cytoplasmic microtubules . Related to Figure 3F . Time-lapse images of S2 cells expressing GFP-CAMSAP3 cultured with 10 µM LatB . The drug was then washed out in control and DHC RNAi cells . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 018 To further confirm that dynein and actin filaments are components of the same microtubule-sorting pathway , we decided to bypass the requirement for actin filaments by directly recruiting dynein to the cell membrane . For these experiments , we used a dynein recruitment tool developed by the Akhmanova and Hoogenraad labs ( Hoogenraad et al . , 2003; Kapitein et al . , 2010 ) , which contains the dynein activator Bicaudal D ( BicD ) fused to a FRB domain ( FRB-GFP-BicD ) . The FRB-BicD-dynein complex can then be recruited to any region of interest by coexpression of a targeting protein coupled to a FKBP domain . This FKBP domain chemically dimerizes with the FRB domain in the presence of rapalog ( a cell-permeable small molecule analog of rapamycin ) ( Clackson et al . , 1998 ) . To confirm that the FRB-FKBP dimerization system for dynein recruitment works in Drosophila S2 cells , we first coexpressed a GFP construct fused to FRB ( FRB-GFP ) and the peroxisome membrane-targeting signal peptide coupled to the red fluorescent protein ( PEX3-RFP-FKBP ) . In the absence of rapalog , the GFP signal appeared soluble in the cytoplasm . The addition of rapalog recruited FRB-GFP to peroxisomes ( Figure 3—figure supplement 1A ) . We next used this system to directly recruit dynein to the plasma membrane . For membrane targeting , we fused the transmembrane domain of GAP-43 ( Heim and Griesbeck , 2004 ) with FKBP and coexpressed this construct with FRB-GFP-BicD in S2 cells . To visualize the recruitment of BicD to the membrane in these cells , we used total internal reflection fluorescence ( TIRF ) microscopy to image the GFP signal before and after addition of rapalog . Quantification showed that addition of rapalog significantly increased the intensity of the GFP fluorescence due to recruitment of cytoplasmic BicD to the plasma membrane ( Figure 3—figure supplement 1B ) . We next tested if direct recruitment of dynein to the membrane using rapalog could drive microtubule sorting in the absence of cortical actin . To test this hypothesis , mCherry-CAMSAP3 was used to track the localization of microtubule minus-ends . In the absence of rapalog , CAMSAP3 accumulated in the processes of S2 cells treated with 10 µM LatB . Time-lapse imaging before addition of the drug revealed that these CAMSAP3 clusters , marking positions of minus-ends , were either static or pushing against the plasma membrane ( Video 10 ) . However , when the BicD-dynein complex was recruited to the membrane by addition of rapalog , CAMSAP3-labeled microtubule minus-ends robustly moved away from the tip , toward the cell body ( Figure 3H , top panels; Figure 3I; Video 10 ) . The same assay performed in DHC knockdown cells did not show this removal of microtubule minus-ends ( Figure 3H , bottom panels; Video 10 ) . Taken together , these data confirmed that dynein is responsible for microtubule sorting , and this activity requires the attachment of the motor to the cortex mediated by actin filaments . 10 . 7554/eLife . 10140 . 019Video 10 . Direct recruitment of cytoplasmic dynein to the plasma membrane activates its microtubule sorting activity in the absence of F-actin . Related to Figure 3H . S2 cells expressing GAP43-FKBP , FBP-GFP-BicD and mCherry-CAMSAP3 ( control and DHC RNAi ) were cultured in the presence of 10 µM LatB for 2 hr . To recruit BicD to the membrane 1 µM rapalog was added to the medium . Time-lapse videos were taken before and after addition of rapalog . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 019 We previously demonstrated that initial neurite formation in Drosophila neurons requires microtubule-microtubule sliding driven by kinesin-1 . Knockdown of kinesin-1 in primary neurons impairs the motility of interphase microtubules and causes neurons to fail to develop axons ( Lu et al . , 2013b ) . This was surprising because kinesin-1 is a plus-end microtubule motor and thus can only slide microtubules with minus-ends leading and plus-ends trailing . Furthermore , symmetry considerations dictate that kinesin slides antiparallel microtubules against each other; if microtubules are oriented parallel to each other kinesin would bundle rather than slide them ( Figure 1A ) . Both of these considerations appear to contradict established literature about microtubule polarity in axons . In this work , we performed two experiments that further support the idea that kinesin-1’s sliding activity drives the initial stages of neurite formation ( Lu et al . , 2013b ) . First , nascent growing neurites contain microtubule arrays with mixed polarity . As mentioned above , this antiparallel orientation is required for microtubule sliding by kinesin-1 ( Figure 1A ) . Second , microtubule minus-ends are pushed against the plasma membrane at the tips of processes , generating the force for neurite outgrowth . Recent experimental data revealed that ‘mitotic’ kinesins , such as kinesin-5 ( Nadar et al . , 2008; Nadar et al . , 2012 ) and kinesin-6 ( Lin et al . , 2012; del Castillo et al . , 2015 ) , play an important role in the regulation of axon outgrowth . The well established function of these mitotic motors is to reorganize and stabilize the mitotic spindle to facilitate chromosome segregation . These motors accumulate in the spindle midzone where antiparallel microtubules coming from opposite poles overlap . Our work shows that developing Drosophila axons , like the spindle midzone , are filled with antiparallel microtubule arrays . This microtubule configuration suggests that the same mechanisms controlling the mitotic spindle could be used to regulate axonal outgrowth ( Baas , 1999 ) . Interestingly , ‘mitotic kinesins’ control neurite outgrowth both in mammalian and Drosophila systems , which supports the idea that the molecular players that drive axon initiation and outgrowth in Drosophila may be conserved across species . Our data show that minus-end-out microtubules are consistently observed at the early stages of neurite formation , while at later stages the vast majority of microtubules in axons have their plus-ends out , as have been reported in multiple published papers ( Baas et al . , 1988; Stepanova et al . , 2003; Stone et al . , 2008 ) . Therefore , neurons must activate an additional sorting mechanism that eliminates these ‘wrong’ polarity microtubules from axons . Previous studies reported that dynein and dynein-associated proteins are required for correct microtubule orientation in axons of Drosophila da sensory neurons ( Zheng et al . , 2008; Arthur et al . , 2015 ) . Our results using cultured neurons are in full agreement with these studies , as dynein knockdown caused mixed microtubule polarity in axons . Interestingly , several studies have linked dynein-mediated microtubule reorganization to actin , both in interphase ( Mazel et al . , 2014 ) and in mitosis ( Kotak et al . , 2012; Kiyomitsu and Cheeseman , 2013 ) . In this study , we showed that depolymerization of actin filaments ( including cortical actin ) using high concentrations of LatB causes the same defects in microtubule orientation as dynein knockdown . Recovery of cortical actin after drug washout results in efficient dynein-driven removal of minus-end-out microtubules from the tips of processes . Furthermore , the requirement for actin can be efficiently bypassed by direct recruitment of the dynein machinery directly to the plasma membrane . Therefore , we propose that microtubule sorting depends on the activity of cytoplasmic dynein attached to the cortical actin filament meshwork . However , as we did not get complete recovery of the wild type distribution of microtubules in our LatB washout or dynein-membrane recruitment experiments , it is likely that additional proteins that link actin and microtubule pathways in the neuron are involved in microtubule organization; one good candidate is the well characterized actin-microtubule crosslinking protein Short stop ( Lee and Kolodziej , 2002 ) . In addition to the sorting activity described in this work , other groups have observed that dynein activity is required for axon elongation ( Ahmad and Baas , 1995; Ahmad et al . , 1998; Grabham et al . , 2007 ) . A more recent study by Miller et al . has reported that dynein generates forces that push the cytoskeletal meshwork forward during axonal elongation in cultured chick sensory neurons ( Roossien et al . , 2014 ) . In agreement with their data , we predict that cortical dynein can promote axon outgrowth after the initial stages of neurite outgrowth , when microtubules are sorted into a plus-end-out orientation . Because the direction of dynein-powered forces is dictated by the intrinsic orientation of microtubules , microtubules with minus-end-out are redirected toward the cell body , whereas plus-end-out microtubules are transported toward the tip of the axon . Therefore , dynein-driven microtubule transport can not only remove microtubules with the wrong orientation ( minus-end-out ) but also push microtubules of the right orientation ( plus-end-out ) toward the axon tip . In addition to the role of dynein in developing neurons , dynein may play an important role in axonal microtubule maintenance . Microtubule polymers , like other protein structures , require subunit turnover to maintain their integrity . It has been proposed that several microtubule severing proteins are able to fragment microtubules into short pieces ( McNally and Vale , 1993 ) . In this scenario , short microtubule seeds are generated which may be reoriented by diffusion forces in either direction ( plus-end-in or plus-end-out ) with equal probability . Indeed , active bidirectional transport of short microtubule fragments in the axons has been described in cultured rat hippocampal neurons ( Liu et al . , 2010 ) . The authors proposed that the transport of these short microtubule fragments is driven by dynein . Our prediction is that microtubule fragments transported in the retrograde direction should be oriented with plus-ends toward the cell body while anterogradely transported fragments should have the opposite orientation . It has been shown in many different model systems that kinesin-1 and dynein are the two major motors involved in reorganizing cytoplasmic microtubules ( Fink and Steinberg , 2006; Straube et al . , 2006; Jolly et al . , 2010; Mazel et al . , 2014 ) . Both motors play an important role in axon formation . However , the contribution of each motor likely differs at different developing stages ( see model described in Figure 4 ) . In spherically shaped undifferentiated cells , kinesin-1 motors slide antiparallel microtubules perpendicularly to the plasma membrane with their minus-end-leading ( Figure 4A ) . This sliding generates the force that breaks the symmetry of the cell and induces a deformation of the plasma membrane to initiate neurite outgrowth . During this stage , the role of cortical dynein is probably limited by steric restrictions . Cortical dynein can only slide microtubules tangential to the plasma membrane , and therefore dynein-mediated movement cannot deform the membrane and initiate outgrowth . At the next stage , nascent neurites contain antiparallel microtubule arrays ( Figure 4B ) . This configuration allows kinesin-1 to continue to drive antiparallel microtubule sliding . Once the processes start to form , the new geometry will allow cortical dynein to contribute to both microtubule organization and neurite outgrowth . Cortical dynein in the neurite cortex can only slide microtubules parallel to the axis of the process . This contribution of cortical dynein likely increases as the processes grow longer ( engaging more cortical dynein molecules ) and thinner ( allowing dynein to reach a larger fraction of microtubules in the processes ) . 10 . 7554/eLife . 10140 . 020Figure 4 . Model of microtubule sliding and axon formation . ( A ) Kinesin-1 induced sliding of antiparallel microtubules initiates formation of processes . Note that at this stage cortical dynein can only slide microtubules parallel to the plasma membrane , suggesting that dynein is not involved in the initiation of processes . ( B ) Short neurites contain antiparallel microtubule arrays . Under this configuration , kinesin-1 still slides microtubules with their minus-ends out , and cortical dynein can start removing minus-end-out microtubules to the cell body . ( C ) Due to continuous dynein-powered sorting activity , the growing axon is mostly filled with uniformly oriented plus-end-out microtubules . At this stage , dynein can continue removing minus-end-out microtubules , and can contribute to axon elongation by pushing plus-end-out microtubules toward the tip ( see Roossien et al . , 2014 ) . Note that at this stage , kinesin-1 motors only bundle parallel microtubules and no longer contribute to microtubule sliding or outgrowth . The triangles on the right represent the contribution of the motors at each stage . DOI: http://dx . doi . org/10 . 7554/eLife . 10140 . 020 There are two possible microtubule orientations in the neurite that can interact with cytoplasmic dynein . Microtubules with minus-end-out ( microtubule of the ‘wrong’ polarity ) will be moved by dynein toward the cell body , therefore eliminating them from the processes . At the same time , dynein can interact with and push plus-end-out microtubules moving them toward axon tips , and thus contributing to the forces that drive axonal growth ( Figure 4B ) . As a result of this continuous sorting activity of dynein , developed axons contain uniform microtubule arrays with their plus-end distal ( Figure 4C ) . Under this new microtubule configuration , kinesin-1 contributes little to microtubule sliding ( as antiparallel microtubule arrays are needed for kinesin to engage in sliding ) , but instead favors its bundling activity ( Figure 1A , right panel ) . In addition to its contribution to neurite outgrowth , cortical dynein may have an important role in axon maintenance , removing nascent microtubules of wrong orientation that may appear in the axon either due to new microtubule polymerization or due to the severing of pre-existing microtubules . Because axons and dendrites are common morphologic features observed in neurons from ancestral metazoans to mammals , we postulate that the same molecular mechanism presented here for axon formation in Drosophila neurons is conserved in other organisms . Of course , this ‘microtubule-centric’ model does not describe all the mechanisms involved in axon formation , and clearly leaves out the critical question of axon guidance , but it provides clear roles for multiple microtubule motors involved in axon formation and microtubule organization . Microtubule orientation differs between axons and dendrites . While axons are filled with uniformly oriented microtubules with plus-end-out , dendrites contain either microtubules of mixed orientation ( mammalian neurons ) ( Baas et al . , 1988 ) or a majority of microtubules minus-end-out ( Drosophila and C . elegans neurons ) ( Stone et al . , 2008; Yan et al . , 2013 ) . We speculate that neurons selectively employ cortical dynein to dictate which of the nascent neurites will become the future axon . It is likely that the microtubule sliding activity or efficiency of dynein recruitment may be downregulated in dendrites . This hypothesis is in agreement with a recent study in Drosophila da sensory neurons showing that disruption of NudE , a dynein cofactor , impairs microtubule orientation in axons , without affecting the orientation of dendritic microtubules ( Arthur et al . , 2015 ) . Furthermore , Yan et al . directly demonstrated that kinesin-1 ( unc-116 ) is required for minus-end-out orientation of microtubules in dendrites ( Yan et al . , 2013 ) . Those observations support the idea that the activity of cortical dynein is downregulated in dendrites , thus preserving the initial minus-end-out orientation of microtubules created by kinesin . Future experiments are required to unravel how dynein interacts with cortical actin and how dynein’s microtubule-sorting activity is regulated in neurons . To visualize microtubule minus-ends in the Drosophila S2 cell system , a cDNA encoding mouse CAMSAP3 ( Jiang et al . , 2014 ) was cloned into the pMT-GFP and pMT-mCherry backbones by NotI-AgeI restriction enzyme sites to generate GFP-CAMSAP3 and mCherry-CAMSAP3 . GFP-CAMSAP3 was also cloned into UASp backbone by KpnI-XbaI to create a transgenic Drosophila fly line containing UASp-GFP-CAMSAP3 by standard P element-mediated transformation . A plasmid encoding EB1-GFP under endogenous EB1 promoter ( pMT-EB1:EB1-GFP ) was used to visualize microtubule plus-ends in S2 cells . For recruitment experiments , all constructs were cloned into pAC . V2014 , a modified version of pAC5 . 1 containing the following multiple cloning site ( KpnI-NheI-BmtI-HindIII-AscI-EcoRI-NotI-XbalI-EcoRV-XhoI ) . FRB-GFP and PEX3-mRFP-FKBP fragments from pβActin-GFP-FRB and pβactin-PEX3-mRFP-FKBP plasmids ( Kapitein et al . , 2010 ) were subcloned in the HindIII-NotI and HindIII-EcoRI sites to generate pAC . V2014-FRB-GFP and pAC . V2014-PEX3-mRFP-FKBP . Drosophila BicD cDNA ( LD17129 from DGRC ) was amplified by PCR and ligated in the AscI-NotI sites of pAC . V2014-FRB-GFP to create pAC . V2014-FRB-GFP-BicD . FKBP fragment was inserted in pAC . V2014 using the AscI-EcoRI sites to create pAC . 2014-FKBP . To recruit the FKBP domain to the membrane , the DNA sequence encoding the transmembrane domain GAP-43 ( MLCCMRRTKQVEKNDEDQKI ) was ligated in the KpnI-AscI sites of pAC2014-FKBP to create pAC2014-GAP43-FKBP . Fly stocks and crosses were cultured on standard cornmeal food based on Bloomington Stock Center’s recipe at room temperature . The following fly stock lines were used in this study: UASp-tdEOS2-αtub84B ( 2nd and 3rd chromosome insertions ) ( Lu et al . , 2013b ) , ubi-EB1-GFP ( 3rd chromosome insertion ) ( Shimada et al . , 2006 ) ; UASt-EB1-GFP ( 3rd chromosome insertion ) ( Rolls et al . , 2007 ) ; elav-Gal4 ( 3rd chromosome insertion , Bloomington stock #8760 ) ( Luo et al . , 1994 ) ; UASp-GFP-CAMSAP3 ( 2nd chromosome insertion , created in this study ) ; DHC64C-RNAi TRiP lines , ( Valium 20 , Bloomington stock #36698 , 3rd chromosome attP2 insertion , targeting DHC64C CDS 1302–1322; Valium 22 , Bloomington stock #36583 , 2nd chromosome attP40 insertion , targeting DHC64C CDS 10044–10064 ) . Stocks of yw; DHC64C-TRiP RNAi-Valium22; ubi-EB1-GFP , yw; UASt-EB1-GFP; DHC64C-TRiP RNAi-Valium20 , and yw; UASp-GFP-CAMSAP3; DHC64C-TRiP RNAi-Valium20 were generated using standard balancing procedures , and crossed with elav-Gal4 to examine EB1-GFP or GFP-CAMSAP3 in DHC64C knockdown . Primary neurons were obtained from brains of 3rd instar larva as previously described ( Lu et al . , 2015 ) . Neurons were plated onto Concanavalin A-coated coverslips in supplemented Schneider’s medium ( 20% fetal bovine serum , 5 μg/ml insulin , 100 μg/ml penicillin , 100 μg/ml streptomycin , and 10 μg/ml tetracycline ) . For actin depolymerization assays , neurons were plated in Xpress medium for 2 hr before addition of LatB . Drosophila S2 cells were cultured as previously described ( Barlan et al . , 2013 ) . To induce the formation of processes in S2 cultures , cells were plated in the presence of 0 . 5 µM LatB . For knockdown assays in S2 cells , cultures at 1 . 5 x 106 cells/mL were treated twice with 20 µg of dsRNA ( day 1 and day 3 ) and cell analysis was performed on day 5 . Double-stranded RNA was transcribed in vitro with T7 polymerase , and purified using LiCl extraction . Primers used to create T7 templates from fly genomic DNA were as follows . T7 promoter sequences ( TAATACGACTCACTATAGGG ) were added to the 5′ end of each primer ) . DHC , forward , AAACTCAACAGAATTAACGCCC; reverse , TTGGTACTTGTCACACCACT ( Jolly et al . , 2010 ) ; p150Glued , forward GAGTTTGAGGAGACGATGGACCACC , reverse GTTGCACGATGGGGTTTCCTTTGCAG; Lis1 , forward GGTTGAATTACGCGATCATGAGCATACTGTGGA , reverse GGAGGTGCAGAAATGCTGATGCGCGTATAG; NudE , forward , GCTCAAGTTGGAATCGCATGGCATCGATATGTC , reverse , CTCTCGTCTCATCCATTAATCGCTGTAGTTTTTCCTGC . To induce the formation of short microtubule fragments , Drosophila S2 cells stably expressing GFP-CAMSAP3 were treated with 25 µM Vinblastine for 1 hr . Soluble fraction of tubulin was removed using BRB80 buffer ( 80 mM PIPES buffer ( pH 6 . 8 ) , 1 mM EGTA , 1 mM DTT , and 1 mM MgCl2 ) supplemented with 1% Triton X-100 . Extracted cells were fixed and microtubules were immunostained with mouse anti-α-tubulin ( DM1α ) . To image EB1 comets , CAMSAP3 localization and lysosome transport in Drosophila S2 cells and primary neurons , a Nikon Eclipse U2000 inverted microscope equipped with a Yokogawa CSU10 spinning disk head , Perfect Focus system ( Nikon ) and a 100 X 1 . 45 NA lens was used . Images were acquired using Evolve EMCCD ( Photometrics ) and controlled by Nikon Elements 4 . 00 . 07 software . For EB1 and CAMSAP3 time-lapses , images were collected every 2 s for 1 min ( EB1 comets ) and every 1 min or 5 min for 16 min or 60 min , respectively ( CAMSAP3 ) . To image the plasma membrane , CellMask Deep Red dye ( 1:10 , 000 ) was added in both cultured neurons and S2 cells 5 min before imaging . For To visualize sliding , a small fraction of microtubules in cultured neurons expressing tdEOS-αTub84B was photoconverted . Photoconversion was performed by confining the illumination of a heliophore laser ( 405 nm ) in the epifluorescence pathway using a diaphragm . Images were collected once per 30 s for 5 min . To image the recruitment of BicD to the plasma membrane and the distribution of CAMSAP3 in mature neurons , TIRF images were collected using a Nikon Eclipse U2000 inverted microscope equipped with a Plan-Apo TIRF 100×/1 . 45 NA objective and a Hamamatsu CMOS Orca Flash 4 . 0 camera ( Hamamatsu Photonics , Hamamatsu , Japan ) , controlled by MetaMorph 7 . 7 . 7 . 0 software ( Molecular Devices , Downingtown , PA ) . Statistical significance for CAMSAP3 populations was determined using two-tailed Fischer’s test with a confidence interval of 95% . The orientations of EB1 comets both in cultured neurons and S2 cells were quantified using MATLAB and ImageJ . EB1 comets were tracked using the plus-end tracking algorithm in u-track 2 . 0 , developed by Gaudenz Danuser’s group ( Jaqaman et al . , 2008; Matov et al . , 2010 ) . The x-y positions of EB1 comets were extracted and loaded into a custom ImageJ plugin . This semi-automated plugin defined the center of cells and determined the angle of each comet’s trajectory as compared with the cell center . For curved and serpentine axons , trajectories were subdivided into linear segments before ImageJ analysis to ensure that comet orientation was correctly identified . Comet trajectories with an angle >290° or <70° compared with the cell center were defined as plus-end-out . Comet trajectories with an angle >110 and <250 degrees were defined as plus-end-in . Only comets contained in processes were included in the analysis . To create kymographs presented in Figure 1E and Figure 2A the Reslice plugin developed in FIJI was used . Statistical significance for EB1 comets was determined using the non-parametric Mann-Whitney test with a confidence interval of 95% . This test analysis compares the distributions of two unmatched groups . For Western blot analysis of S2 cell extracts , the following primary antibodies were used: anti-DHC monoclonal antibody 2C11-2 ( Sharp et al . , 2000 ) and rabbit polyclonal antibody against KHC head domain provided by A . Minin ( Institute of Protein Research , Russian Academy of Sciences , Moscow , Russia ) . The formation of the FRB-FKBP complex in all recruitment experiments was induced by addition of 1 µM rapalog ( final concentration ) ( A/C Heterodimerizer , Clontech ) to the culture medium . To test the efficiency of recruitment experiments , Drosophila S2 cells were transiently cotransfected either with plasmids encoding FRB-GFP and PEX3-mRFP-FKBP or FRB-GFP-BicD and GAP43-FKBP . Cells were imaged before and 30 min after addition of rapalog to the medium . To directly recruit endogenous dynein to the membrane , S2 cells were cotransfected with plasmids codifying GAP43-FKBP , FRB-GFP-BicD and mCherry-CAMSAP3 in the following ratio ( 3:1:1 ) . Cells were plated with 10 µM LatB for 2 hr to allow the accumulation of microtubule minus-ends at process tips . Distribution of mCherry-CAMSAP3 was tracked by time-lapse imaging before and after addition of the drug ( to the final concentration of 1 µM ) .	Motor proteins can move along filaments called microtubules to transport proteins and other materials to different parts of the cell . Microtubules are “polar” filaments , meaning that they have two distinct ends that have different chemical properties . Motor proteins can only move along these filaments in one direction , for example , the kinesin motor proteins generally move toward the so-called “plus-end” , while dynein motors move in the opposite direction . A typical nerve cell ( or neuron ) is composed of a cell body , a long projection called an axon and many small branch-like structures called dendrites . Within the axon , the microtubules are arranged so that their plus-ends point outwards , but the microtubules in dendrites are arranged differently so that many minus-ends point outwards instead . This polarity is important for the neuron in deciding which proteins should be transported to axons , and which should go to the dendrites . However , it is not clear how these different microtubule arrangements are established . Here , del Castillo et al . used microscopy to study microtubules in the axons of fruit fly neurons . The experiments show that in the very early stages of neuron development , the axons contained microtubules of mixed polarity . However , by the later stages , the microtubules had become uniform with all the plus-ends directed outwards . Further experiments show that dynein is responsible for this organization as it pushes the minus-end-out microtubules out of the axons . Dynein uses a scaffold made of a protein called actin to attach to the inner surface of the cell and move the minus-end microtubules to the cell body of the neuron . Thus , del Castillo et al . ’s findings reveal that these dynein motors are responsible for the polarity of microtubules in mature axons . The next challenge is to understand how dynein is attached to the actin scaffold and why it rearranges microtubules in axons , but not in dendrites .	[ "Abstract", "Introduction", "Results", "Discussion", "Material", "and", "methods" ]	[ "cell", "biology" ]	2015	Interplay between kinesin-1 and cortical dynein during axonal outgrowth and microtubule organization in Drosophila neurons
PTEN controls three-dimensional ( 3D ) glandular morphogenesis by coupling juxtamembrane signaling to mitotic spindle machinery . While molecular mechanisms remain unclear , PTEN interacts through its C2 membrane-binding domain with the scaffold protein β-Arrestin1 . Because β-Arrestin1 binds and suppresses the Cdc42 GTPase-activating protein ARHGAP21 , we hypothesize that PTEN controls Cdc42 -dependent morphogenic processes through a β-Arrestin1-ARHGAP21 complex . Here , we show that PTEN knockdown ( KD ) impairs β-Arrestin1 membrane localization , β-Arrestin1-ARHGAP21 interactions , Cdc42 activation , mitotic spindle orientation and 3D glandular morphogenesis . Effects of PTEN deficiency were phenocopied by β-Arrestin1 KD or inhibition of β-Arrestin1-ARHGAP21 interactions . Conversely , silencing of ARHGAP21 enhanced Cdc42 activation and rescued aberrant morphogenic processes of PTEN-deficient cultures . Expression of the PTEN C2 domain mimicked effects of full-length PTEN but a membrane-binding defective mutant of the C2 domain abrogated these properties . Our results show that PTEN controls multicellular assembly through a membrane-associated regulatory protein complex composed of β-Arrestin1 , ARHGAP21 and Cdc42 . PTEN ( phosphatase and tensin homolog ) is the second most commonly mutated tumor suppressor gene in human cancer ( Cantley and Neel , 1999 ) and has a central role in multicellular morphogenesis ( Martin-Belmonte et al . , 2007; Jagan et al . , 2013a; Deevi et al . , 2016 ) . While PTEN antagonizes the phosphoinositol 3-kinase ( PI3K ) /AKT pathway via its N-terminal phosphatase domain ( Cantley and Neel , 1999 ) , three-dimensional ( 3D ) multicellular assembly was unaffected by forced variation of PI3K activity in colorectal organotypic model systems ( Jagan et al . , 2013a; Magudia et al . , 2012 ) . The PTEN domain structure includes an N-terminal phosphatase domain , a C2 domain , a C-terminal tail and a PDZ-binding domain . The C2 domain binds to membrane phospholipids by inserting a hydrophobic ( CBR3 ) loop into the membrane bilayer and thereby provides a scaffold for juxtamembrane signaling ( Lee et al . , 1999 ) . Furthermore , the PTEN C2 domain regulates polarized migration ( Raftopoulou and Hall , 2004 ) , multicellular morphology ( Leslie et al . , 2007; Jagan et al . , 2013b ) and has an important but poorly understood tumor suppressor function ( Caserta et al . , 2015 ) . Within complex systems , protein scaffolding enhances signaling efficiency by assembly of spatially distinct subcellular complexes for different cellular tasks ( Weng et al . , 1999; Pertz , 2010 ) . The PTEN C2 domain binds the plasma membrane and interacts with the scaffold protein β-Arrestin1 ( Lima-Fernandes et al . , 2011 ) that in turn binds and suppresses ARHGAP21 ( Anthony et al . , 2011 ) , a member of a highly conserved class of RhoGAPs ( Bos et al . , 2007; Anderson et al . , 2008 ) . ARHGAP21 regulates the small GTPases , Cdc42 ( Dubois et al . , 2005 ) and RhoA ( Anthony et al . , 2011 ) . These GTPases have overlapping , complementary functions required for mitotic spindle orientation and consequent control of the cell division axis , cytokinetic furrow positioning , daughter cell size and tissue morphogenesis ( Morin and Bellaïche , 2011 ) . Both Cdc42 and RhoA drive actin nucleation and cortical stiffening ( Ma et al . , 1998; Eisenmann et al . , 2007 ) required for spindle orientation ( Johnston et al . , 2013 ) . Furthermore , Cdc42 crosstalk with protein kinase c zeta [PRKCZ] ( Noda et al . , 2001; Durgan et al . , 2011 ) localizes force generators within the cell cortex that act via astral microtubules to orientate the spindle ( Hao et al . , 2010 ) . ARHGAP21 has high GAP activity for Cdc42 ( Dubois et al . , 2005 ) and its Pac-1 homologue regulates multicellular patterning in C . elegans by spatial regulation of Cdc42 ( Anderson et al . , 2008; Klompstra et al . , 2015 ) . Here , we investigate PTEN spatiotemporal coordination of mammalian glandular morphogenesis through conserved juxtamembrane β-Arrestin1-ARHGAP21 interactions , using 3D colorectal cancer ( CRC ) model systems . To substantiate physiological relevance of these processes , we also investigate their role in morphogenesis of 3D multicellular organoids isolated from normal colon . β-Arrestin1 scaffolds juxtamembrane signaling networks ( Kovacs et al . , 2009 ) , binds ARHGAP21 ( Anthony et al . , 2011 ) and governs PTEN catalytic and noncatalytic functions ( Lima-Fernandes et al . , 2011 ) . To ascertain whether PTEN regulates membrane-associated β-Arrestin1 and ARHGAP21 , we conducted expression and simple fractionation studies in PTEN-expressing [Caco-2 and HCT116] and -deficient [Caco-2 ShPTEN ( ShPTEN ) and PTEN -/- HCT116 ( PTEN -/- ) ] cells . We found near-significant or significant differences of total lysate β-Arrestin1 and ARHGAP21 between PTEN-expressing and -deficient cells [Caco-2 vs ShPTEN ( Figure 1A , B ) and HCT116 vs PTEN -/- cells ( Figure 1—figure supplements 1 and 2 ) ] . To infer subcellular localization of β-Arrestin1 and ARHGAP21 , we performed membrane fractionation studies and normalized each protein’s densitometry value against its total lysate level , to investigate relative proportions of β-Arrestin1 and ARHGAP21 associated with membrane . We found greater β-Arrestin1 but lower ARHGAP21 levels in Caco-2 than in ShPTEN membrane fractions ( Figure 1C , D ) . As β-Arrestins are known to localize to activated lysophosphatidic acid receptors [LPARs] ( Urs et al . , 2005; Li et al . , 2009 ) that are expressed in Caco-2 and HCT116 cell membranes ( Yun et al . , 2005 ) , we investigated effects of PTEN on lysophosphatidic acid ( LPA ) -induced membrane recruitment of β-Arrestin1 . We found greater LPA-mediated membrane enrichment of β-Arrestin1 in Caco-2 and HCT116 cells than in PTEN-deficient ShPTEN or PTEN -/- HCT116 ( PTEN -/- ) subclones ( Figure 1E , F; Figure 1—figure supplements 3 and 4 ) . We next used confocal microscopy to determine PTEN effects on β-Arrestin1 subcellular distribution in whole cells . We expressed the β-Arrestin1-mCherry fusion protein and mCherry only controls in PTEN-expressing and -deficient cells . We assessed colocalization with Alexa 488-labeled wheat germ agglutinin ( WGA ) , a widely used fluorescent probe for cell and Golgi complex membranes ( Crossman et al . , 2015 ) by confocal microscopy . β-Arrestin1-mCherry was predominantly cytosolic in vehicle only ( VO ) -treated cells , in accord with cytosolic accumulation of unlabelled β-Arrestin1 in fractionation studies . On treatment with LPA , the β-Arrestin1-mCherry fusion protein colocalized with WGA at the plasma membrane in PTEN-expressing Caco-2 and HCT116 control cells ( Figure 1G; Figure 1—figure supplement 5 ) . Line scanning analysis revealed overlap of β-Arrestin1-mCherry and Alexa 488 fluorescence signals in plasma membrane peaks in PTEN-expressing Caco-2 and HCT116 cells after LPA treatment ( Figure 1G; Figure 1—figure supplement 5 ) . While LPA had limited effects in ShPTEN cells that have residual low level PTEN ( Figure 1G ) , this treatment had no effects on β-Arrestin1-mCherry subcellular distribution in PTEN-null ( PTEN -/- ) cells ( Figure 1—figure supplement 5 ) . mCherry only did not localize at the plasma membrane ( data shown for control PTEN-expressing HCT116 and PTEN -/- cells only; Figure 1—figure supplement 6 ) . To exclude a non-specific effect of PTEN on ligand-mediated protein translocation to the cell membrane , we investigated 1 , 25 ( OH ) 2D3-mediated membrane localization of E-Cadherin ( Pálmer et al . , 2001 ) in Caco-2 and ShPTEN cells . We found that 1 , 25 ( OH ) 2D3 treatment induced equivalent E-Cadherin translocation to the plasma membrane in PTEN-expressing and -deficient cells , compared to control VO treatment ( data not shown ) . Collectively , these findings show that PTEN functions within a regulatory scaffolding network that couples β-Arrestin1 to ARHGAP21 at the plasma membrane . Within signaling scaffolds , β-Arrestin1 regulates monomeric GTPases ( Barnes et al . , 2005 ) and orchestrates cytoskeletal rearrangements ( Ge et al . , 2003 ) . We investigated β-Arrestin1-ARHGAP21 coregulation of Cdc42 , mitotic spindle orientation and morphogenesis in 3D organotypic model systems . SiRNA knockdown ( KD ) of β-Arrestin1 in control PTEN-expressing Caco-2 cells suppressed Cdc42 activation as assessed by Cdc42-GTP levels in cell lysates on Western blots ( Figure 2A , B ) . In contrast , siRNA KD of ARHGAP21 enhanced Cdc42 activation in PTEN-deficient cells ( Figure 2C , D ) . During normal organotypic 3D glandular morphogenesis , mitotic spindle planes are orientated at approximately right angles to gland centres ( GCs ) by Cdc42-dependent mechanisms . Conversely , ShPTEN cells show deficiencies of these processes ( Jagan et al . , 2013a; Deevi et al . , 2016; Jagan et al . , 2013b ) . In 3D Caco-2 cultures , SiRNA β-Arrestin1 KD suppressed Cdc42-GTP ( Figure 2E , F ) , induced mitotic spindle misorientation and abnormal multilumen formation ( Figure 2E , Figure 2—figure supplements 1 and 2 ) . Conversely , ARHGAP21 KD enhanced Cdc42-GTP ( Figure 2G , H ) , restored mitotic spindle orientation and promoted single lumen formation in 3D ShPTEN cultures ( Figure 2G , Figure 2—figure supplements 3 and 4 ) . Because of the previously reported relationship between ARHGAP21 and RhoA ( Lima-Fernandes et al . , 2011 ) , we assessed relationships between β-Arrestin1 , ARHGAP21 and RhoA . We found that activation of RhoA related directly to β-Arrestin1 and inversely to ARHGAP21 expression . β-Arrestin1 and RhoA-GTP were suppressed , while ARHGAP21 expression was enhanced by PTEN knockdown ( data for RhoA-GTP not shown ) . Taken together , these data indicate that PTEN regulates β-Arrestin1-ARHGAP21 interactions to control GTPase signaling , mitotic spindle orientation and 3D multicellular morphology . β-Arrestin1 has previously been shown to bind the PTEN C2 domain directly and modulate PTEN function ( Lima-Fernandes et al . , 2011 ) . To investigate PTEN regulation of β-Arrestin1-ARHGAP21 interactions , we conducted co-immunoprecipitation ( CoIP ) studies and normalized β-Arrestin1-associated ARHGAP21 against total ARHGAP21 densitometry values in cell lysates . Here , we show greater β-Arrestin1-associated ARHGAP21 levels in PTEN-expressing Caco-2 or HCT116 cells versus ShPTEN or PTEN-/- cells or IgG negative controls ( Figure 3A , B ) . To investigate involvement of PTEN catalytic and noncatalytic domains in these processes , we conducted transient expression studies of GFP-labeled full-length wild type ( wt ) PTEN or mutants ( Figure 3C ) , in PTEN-deficient cells . Mutants included full-length PTEN with a mutation at the CBR3 membrane-binding loop within the C2 domain ( PTEN-MCBR3 ) , full-length phosphatase-dead ( PTEN C124S-based ) constructs with mutations in key C-terminal phosphorylation sites , namely PTEN C124S-T383A ( CS-T383A ) and PTEN C124S-A4 ( CS-A4 with S380A , T382A , T383A and S385A mutations combined ) . CS-T383A has been proposed to contain an unmasked C2 domain ( Raftopoulou et al . , 2004 ) that effectively binds β-Arrestin1 while CS-A4 lacks β-Arrestin1 binding capacity ( Lima-Fernandes et al . , 2011 ) . We also used the isolated PTEN C2 domain ( C2 ) and a membrane-binding mutant of the C2 domain ( C2-MCBR3 ) . We found that expression of C2 enhanced β-Arrestin1-associated ARHGAP21 levels in CoIPs conducted in ShPTEN ( Figure 3D , E , ) and PTEN -/- cells ( Figure 3—figure supplements 1 and 2 ) . Conversely , C2-MCBR3 had no significant effect on β-Arrestin1-associated ARHGAP21 levels in CoIPs ( Figure 3D , E , Figure 3—figure supplements 1 and 2 ) . β-Arrestin1-associated ARHGAP21 levels were normalized against total ARHGAP21 expression in cell lysates . Collectively , these findings indicate that the membrane-binding function of PTEN is important for scaffolding ARHGAP21 and β-Arrestin1 . We then used an intramolecular bioluminescence resonance energy transfer ( BRET ) -based PTEN biosensor ( Lima-Fernandes et al . , 2014; Misticone et al . , 2016 ) to test if the full-length C124S C-terminal phosphorylation mutants ( Figure 3C ) that have different β-Arrestin1 binding capacities ( Lima-Fernandes et al . , 2011 ) , display different conformations . The biosensor contains PTEN sandwiched between the energy donor Renilla luciferase ( Rluc ) and the energy acceptor YFP . Changes in the BRET signal depend on the relative distance and orientation of the donor and acceptor proteins within the fusion and therefore provide readout for conformational change of PTEN in live cells ( Figure 3—figure supplement 3 ) . Wild-type ( wt ) PTEN , CS-T383A and CS-A4 mutants in the Rluc-PTEN-YFP construct produced different BRET signals ( Figure 3—figure supplement 3 ) . These findings show that the phosphatase-dead mutants do indeed adopt different conformations , which is consistent with differences in β-Arrestin1-binding capacity . We further investigated protein-protein interactions in vivo using sensitive proximity ligation assays [PLA] ( Söderberg et al . , 2006 ) . We expressed PTEN phosphatase-dead mutants or C2 domain constructs in PTEN -/- cells . We found prominent PLA signals for PTEN-β-Arrestin1 interactions in PTEN -/- cells expressing either PTEN CS-T383A or the intact C2 domain . Conversely , PTEN-/- cells expressing PTEN CS-A4 or C2-MCBR3 mutants showed markedly reduced levels of these interaction signals . PTEN-β-Arrestin1 interaction PLA signals in HCT116 and GFP-only transfected PTEN -/- cells were used as positive and negative controls , respectively ( Figure 3F; Figure 3—figure supplement 4 ) . These findings indicate that PTEN-β-Arrestin1 interactions can occur independently of PTEN phosphatase activity . Next , we investigated effects of PTEN on β-Arrestin1-ARHGAP21 interactions . Transfection of PTEN -/- cells with GFP-labeled-wt PTEN or -C2 domain enhanced β-Arrestin1-ARHGAP21 interactions compared to cells transfected with PTEN-MCBR3 or C2-MCBR3 . β-Arrestin1-ARHGAP21 interaction signals in HCT116 or GFP-only transfected PTEN -/- cells were used as positive and negative controls , respectively . ( Figure 3G; Figure 3—figure supplement 5 ) . Collectively , these data implicate the PTEN C2 domain in phosphatase-independent binding of β-Arrestin1 and in promoting β-Arrestin1-ARHGAP21 interactions . To explore effects of the isolated PTEN C2 domain on membrane recruitment of β-Arrestin1 or ARHGAP21 , we conducted expression , fractionation and CoIP assays in PTEN-deficient cells . Expression of the C2 domain enhanced total β-Arrestin1 and suppressed that of ARHGAP21 in PTEN-deficient cell lysates while C2-MCBR3 had no significant effects ( Figure 4A–C; Figure 4—figure supplements 1–3 ) . C2 expression also enriched β-Arrestin1 and suppressed ARHGAP21 in membrane fractions of PTEN-deficient colorectal cell lines ( Figure 4D–F; Figure 4—figure supplement 4 ) . Membrane fraction values were normalized against total expression of each protein in lysate . Furthermore , expression of C2 but not C2-MCBR3 also increased β-Arrestin1-associated ARHGAP21 levels in PTEN-deficient cell membrane fractions . β-Arrestin1-associated ARHGAP21 levels were normalized against total ARHGAP21 in the membrane fraction ( Figure 4G–J ) . In PTEN -/- cells , expression of C2 and full-length PTEN had greater effects on β-Arrestin1-associated ARHGAP21 levels than PTEN-MCBR3 , C2-MCBR3 or control ( Figure 4I , J ) . PTEN-MCBR3 had small but significant effects on β-Arrestin1-associated ARHGAP21 levels in excess of control ( Figure 4I , J ) . Taken together , these data indicate that PTEN enhances β-Arrestin1 membrane recruitment and β-Arrestin-ARHGAP21 interactions , predominantly through its membrane-binding C2 domain . To investigate PTEN coordination of morphogenic processes through its C2 domain , we conducted 3D organotypic culture studies . We found greater expression and membrane localization of β-Arrestin1 in 3D control PTEN-expressing Caco-2 cultures compared to ShPTEN cultures ( Figure 5A , Figure 5—figure supplement 1 ) , in agreement with our biochemical analysis . Conversely , ARHGAP21 immunoreactivity was lower in control Caco-2 compared to ShPTEN 3D cultures ( Figure 5B; Figure 5—figure supplement 2 ) . We have shown previously that the abnormal ShPTEN 3D phenotype can be rescued by expression of the PTEN C2 domain ( Jagan et al . , 2013b ) . Here , we show that the GFP-tagged PTEN C2 domain enhances β-Arrestin1 membrane enrichment ( Figure 5C , D ) , rescues mitotic spindle orientation ( Figure 5E , F ) as well as apical membrane alignment and single lumen morphology ( Figure 5E; Figure 5—figure supplements 3 and 4 ) in 3D ShPTEN cultures . These effects were not observed in ShPTEN cultures expressing control GFP or C2-MCBR3-GFP ( Figure 5C–F; Figure 5—figure supplements 3 and 4 ) . To investigate any potential for PTEN ShRNA off-target effects , we investigated effects of full-length ShRNA-resistant PTEN ( ShR PTEN ) on the integrated ShPTEN 3D morphology phenotype . We show that expression of ShR PTEN rescued defective morphogenesis of 3D ShPTEN cultures ( Figure 5—figure supplements 5 and 6 ) . Collectively , these studies show that the membrane-bound PTEN C2 domain coordinates multicellular gland assembly by β-Arrestin1 membrane recruitment , mitotic spindle orientation , apical membrane alignment and lumen formation . Precise spatiotemporal coordination of Cdc42 activity is central to multicellular morphogenesis ( Meitinger et al . , 2013 ) . To investigate PTEN non-catalytic regulation of Cdc42 via β-Arrestin1 and ARHGAP21 , we conducted transfection and peptide inhibitor studies . Expression of the catalytically inactive PTEN CS-T383A construct that binds β-Arrestin1 but not the PTEN CS-A4 binding-defective mutant , enhanced Cdc42-GTP levels in PTEN -/- cells ( Figure 6A , B ) . While Cdc42 can be inhibited by ARHGAP21 ( Dubois and Chavrier , 2005 ) , competitive β-Arrestin1 binding to the GAP domain can release the active GTPase from ARHGAP21 inhibition ( Anthony et al . , 2011 ) . To investigate the specific role of β-Arrestin1-ARHGAP21 interactions on Cdc42-dependent 3D morphogenesis , we used a cell-permeant 24-mer peptide analogue of the ARHGAP21 GAP domain that was designed to disrupt the β-Arrestin1-ARHGAP21 interaction ( Figure 6C ) ( Anthony et al . , 2011 ) . Here , we show that treatment with this β-Arrestin1-ARHGAP21 peptide binding inhibitor ( pep24 ) but not a scrambled control peptide attenuated the association between β-Arrestin1 and ARHGAP21 , resulting in lower levels of β-Arrestin1-associated ARHGAP21 ( Figure 6D , E; Figure 6—figure supplements 1 and 2 ) . Treatment by pep24 also suppressed Cdc42 activation in Caco-2 ( Figure 6F , G ) and HCT116 cells ( Figure 6—figure supplements 3 and 4 ) . Furthermore , pep24 treatment induced dysmorphogenesis of 3D Caco-2 cultures characterized by mitotic spindle misorientation ( Figure 6H , Figure 6—figure supplement 5 ) , apical membrane misalignment , aberrant epithelial configuration and loss of single central lumen formation ( Figure 6I , Figure 6—figure supplement 6 ) . Taken together , these data show that PTEN controls 3D morphogenesis by non-catalytic C2 domain scaffolding of β-Arrestin1-ARHGAP21 interactions and release of Cdc42 from ARHGAP21 inhibition . To avoid any compromise of experimental interpretation by intrinsic Caco-2 cancer cell mutations , we investigated our key observations from cell culture experiments in organoids formed from primary normal murine colon cells ( Clevers , 2016 ) . In this study , we cultured colonic crypt progenitor epithelium in Matrigel supplemented with growth factors as previously described ( Sato et al . , 2011 ) . By these methods , we generated colorectal organoids with appropriate mitotic spindle orientation , apical membrane alignment , luminogenesis and epithelial organization , in 3D cultures ( Figure 7A–C ) . To investigate the role of β-Arrestin1-ARHGAP21 interactions on colorectal homeostasis , we assessed effects of pep24 vs control peptide treatment on 3D organoid morphogenesis . Here , we show that pep24 treatment perturbed mitotic spindle orientation , disrupted 3D glandular morphology and lumen formation in normal colorectal organoids ( Figure 7A–C ) . Conversely , control peptide treatment had no discernible effect on 3D glandular morphology ( Figure 7A–C ) . Collectively , these findings highlight a significant role for β-Arrestin1-ARHGAP21 interaction in multicellular morphogenesis of normal colorectal epithelium . Scaffolding proteins have unique properties for assembling target molecules into cooperative networks within subcellular compartments ( Rock et al . , 2013 ) to control diverse biological functions ( Oh and Schnitzer , 2001; Eroglu et al . , 2003; Irazoqui et al . , 2003; Smith and Scott , 2013 ) . β-Arrestin1 acts as a molecular scaffold for G-protein-coupled receptors [GPCRs] ( Luttrell et al . , 1999 ) , the largest family of signaling receptors . Key GPCRs enhance β-Arrestin1 recruitment to the plasma membrane ( Urs et al . , 2005; Li et al . , 2009; Décaillot et al . , 2011 ) , activate PTEN ( Song et al . , 2009; Sanchez et al . , 2005 ) and promote PTEN-β-Arrestin1 interactions ( Lima-Fernandes et al . , 2011 ) . β-Arrestin1 also suppresses ARHGAP21 ( Anthony et al . , 2011 ) that is independently recruited to the plasma membrane by ADP-ribosylation factor 1 [ARF-1] ( Kumari and Mayor , 2008 ) . In this study , we investigated PTEN coregulation of β-Arrestin1 and ARHGAP21 . We found greater β-Arrestin1 levels in lysates of PTEN-expressing HCT116 cells than in the isogenic PTEN-null ( PTEN -/- ) subclone and near-significant differences in corresponding PTEN-expressing and -deficient Caco-2 cells . While the precise mechanisms of this effect remain unclear , PTEN mutation or deficiency and changes in β-Arrestin1 expression levels characterize various human cancers ( Cantley and Neel , 1999; Enslen et al . , 2014 ) . As well as protein abundance , stoichiometry and post-translational targeting machinery modulate the assembly of spatially restricted scaffolding complexes ( Boisvert et al . , 2012 ) . LPAR is a lysophosphatidic acid ( LPA ) activated GPCR that couples heterotrimeric G proteins , to control membrane recruitment of β-Arrestin1 ( Urs et al . , 2005; Li et al . , 2009 ) , GTPase activity ( Ueda et al . , 2001 ) and cell polarization processes ( Nagasaki and Gundersen , 1996 ) . To investigate these phenomena , we assessed spontaneous and LPA-mediated cell membrane localization of β-Arrestin1 and ARHGAP21 in subcellular fractions of PTEN-expressing and -deficient cells . We found proportionately greater differences of constitutive and LPA-mediated β-Arrestin1 membrane localization in PTEN-expressing Caco-2 vs PTEN-deficient ShPTEN cells , indicative of PTEN involvement in β-Arrestin1 membrane recruitment . Similarly , LPA promoted greater β-Arrestin1 membrane localization in PTEN-expressing HCT116 cells than in PTEN -/- cells . In both cell types , we found reciprocal differences of ARHGAP21 membrane localization , consistent with β-Arrestin1-mediated suppression ( Anthony et al . , 2011 ) . To investigate PTEN effects on β-Arrestin1 plasma membrane recruitment in whole cells , we used confocal microscopy to track the spatial distribution of transfected β-Arrestin1-mCherry against plasma membrane localization of Alexa 488-labeled WGA . Line scans of fluorescence intensities across image focal planes and high-resolution confocal z-stack reconstructions ( Furia et al . , 2014 ) revealed that PTEN enhanced LPA-mediated β-Arrestin1-mCherry colocalization with Alexa 488 -labelled WGA at the plasma membrane . LPA treatment had limited effects on β-Arrestin1-mCherry plasma membrane recruitment in ShPTEN cells that have residual low level PTEN but was ineffective in PTEN -/- cells . We cannot attribute these findings to nonspecific fluorophore diffusion , since mCherry distribution was unaffected by PTEN status or LPA treatment . Furthermore , we can exclude PTEN nonspecific effects on cell membrane protein localization because 1 , 25 ( OH ) 2D3-induced plasma membrane recruitment of E-Cadherin ( Pálmer et al . , 2001 ) was equivalent in PTEN-expressing and -deficient cells . Collectively , these findings show that PTEN is an essential coregulator of plasma membrane recruitment of β-Arrestin1 and of β-Arrestin1:ARHGAP21 functional interactions . β-Arrestin1 and ARHGAP21 coregulate GTPases ( Anthony et al . , 2011 ) that function as a signaling hub for diverse cytokinetic processes ( Glover et al . , 2008; Jaffe et al . , 2008 ) . We investigated PTEN regulation of GTPase activity through β-Arrestin1-ARHGAP21 scaffolding and conducted perturbation experiments in organotypic 3D culture models that are ideal for precise , image-based assays of multiscale epithelial homeostasis . We found that siRNA knockdown ( KD ) of β-Arrestin1 suppressed Cdc42 activation in PTEN-expressing cells and 3D cultures while siRNA ARHGAP21 KD had reciprocal effects by increasing Cdc42 activity in PTEN-deficient cultures . Furthermore , these perturbation experiments affected sequential layers of homeostatic controls . Suppression of β-Arrestin1 induced mitotic spindle misorientation , abnormal epithelial configuration , defective apical membrane positioning and formation of multiple lumens during assembly of PTEN-expressing 3D Caco-2 glandular structures . Conversely , ARHGAP21 KD increased Cdc42 activation , restored spindle orientation and rescued aberrant morphogenesis of isogenic PTEN-deficient 3D ShPTEN cultures . Because of the previously reported relationship between ARHGAP21 and RhoA ( Lima-Fernandes et al . , 2011 ) , we assessed relationships between PTEN , β-Arrestin1 , ARHGAP21 and RhoA . PTEN KD suppressed β-Arrestin1 , enhanced ARHGAP21 and suppressed RhoA activation . Collectively , these findings indicate that PTEN orchestrates the cell division plane and apical membrane dynamics during multicellular morphogenesis by coordination of β-Arrestin1-ARHGAP21 functional interactions and GTPase activity . In accord with the above findings , we found that PTEN enhanced β-Arrestin1-ARHGAP21 interactions . PTEN interacts with β-Arrestin1 ( Lima-Fernandes et al . , 2011 ) , localizes to the plasma membrane ( Lee et al . , 1999 ) and modulates multicellular morphogenesis ( Jagan et al . , 2013b ) via its C2 domain . Conversely , PTEN C2 domain mutations perturb multicellular patterning during development and neoplastic progression ( Caserta et al . , 2015 ) by unclear mechanisms . C2 domain molecular interactions are masked by a closed PTEN intramolecular conformation ( Rahdar et al . , 2009 ) . However , studies of the isolated C2 domain or an unmasked C2 mediated by alanine substitution at T383 [T383A] ( Raftopoulou et al . , 2004 ) within a PTEN C124S mutant construct , enables study of PTEN-phosphatase independent C2 domain molecular interactions . A C124S A4 mutant containing alanine substitutions at Ser380 , Thr382 , Thr383 and Ser385 suppresses β-Arrestin1 binding ( Lima-Fernandes et al . , 2011 ) . We used these constructs together with appropriate tools for detection of protein binding or conformational change to investigate scaffolding functions . We found in CoIP studies that the isolated PTEN C2 domain enhanced β-Arrestin1-associated ARHGAP21 expression in excess of that induced by the C2-MCBR3 membrane-binding mutant . Bioluminescence energy transfer ( BRET ) analysis of Rluc-PTEN-YFP biosensor constructs containing PTEN CS-T383A or PTEN-CS-A4 ( 35 ) revealed different conformational dynamics consistent with differential protein binding . Proximity ligation assay ( PLA ) is a sensitive method for visualization of signals generated by protein-protein interactions ( Söderberg et al . , 2006 ) . We found strong interactions between unmasked or isolated PTEN C2 domains and β-Arrestin1 , in excess of that observed with the C2-MCBR3 or C124S A4 mutants . Furthermore , expression of wt PTEN or the C2 domain in PTEN -/- HCT116 cells enhanced β-Arrestin1-ARHGAP21 interaction signals , in excess of PTEN-MCBR3 ( full-length PTEN mutated at the CBR3-membrane-binding loop ) or C2-MCBR3 . Interestingly , the PTEN-MCBR3 mutant had some limited scaffolding function , indicated by increased β-Arrestin1-ARHGAP21 binding in excess of control or C2-MCBR3 in PLA assays . Notwithstanding this finding , our data indicate that PTEN binds β-Arrestin1 and promotes β-Arrestin1-ARHGAP21 interactions predominantly through its intact C2 domain . C2 domains are phospholipid and protein binding modules involved in membrane recruitment and localization of signaling molecules ( Corbalán-García and Gómez-Fernández , 2010 ) . Here , we show that PTEN C2 enhances β-Arrestin1 expression . While precise mechanisms remain unclear , PTEN C2 binds thioredoxin-1 ( Meuillet et al . , 2004 ) that regulates β-Arrestin1 in a context-dependent manner ( Jia et al . , 2014 ) . In this study , PTEN C2 also promoted membrane enrichment of β-Arrestin1 in excess of total lysate concentrations and enhanced β-Arrestin1-ARHGAP21 interactions . Interestingly , wt PTEN promoted greater β-Arrestin1-associated ARHGAP21 expression in membrane fractions than the isolated C2 domain . Within its N-terminal domain , full-length PTEN contains a conserved polybasic phosphatidylinositol[4 , 5] biphosphate ( PtdIns [4 , 5P2] ) [PIP2]-binding site that participates in membrane-targeting ( Walker et al . , 2004 ) and could complement PTEN C2 domain-mediated interactions between scaffold complexes and membrane lipids . Similarly , in CoIP assays PTEN-MCBR3 promoted a small but significant increase of β-Arrestin1-associated ARHGAP21 expression in excess of C2-MCBR3 . Taken together , our findings indicate that PTEN promotes β-Arrestin1-ARHGAP21 interactions predominantly through its C2 domain , although the PTEN N-terminal domain has a weak additional effect . PTEN , β-Arrestins and GTPase-activating proteins modulate the activity of Rho GTPases CdcC42 and RhoA ( Martin-Belmonte et al . , 2007; Anthony et al . , 2011; Bos et al . , 2007; Anderson et al . , 2008 ) . Furthermore , the PTEN C2 domain has morphogenic properties ( Leslie et al . , 2007; Jagan et al . , 2013b; Caserta et al . , 2015 ) . We investigated PTEN orchestration of multicellular gland assembly through its C2 domain , in organotypic culture studies . We found greater β-Arrestin1 and lower ARHGAP21 immunoreactivity in control Caco-2 vs ShPTEN 3D cultures in accord with our findings in cell monolayers . Transfection of ShPTEN cells with a C2 domain construct promoted β-Arrestin1 membrane localization , rescued mitotic spindle orientation , single central lumen formation and 3D multicellular morphogenesis . Conversely , expression of the membrane-binding mutant C2-MCBR3 domain did not rescue the morphology phenotype . ShPTEN cells stably express PTEN ShRNA that targets the phosphatase domain ( Jagan et al . , 2013a ) . To assess PTEN ShRNA specificity and test for potential off-target effects , we investigated effects of full-length ShRNA-resistant PTEN ( ShR PTEN ) on the integrated ShPTEN 3D morphology phenotype . Expression of ShR PTEN rescued 3D ShPTEN morphogenesis and thus confirmed shRNA functional specificity . Collectively , the above data show that PTEN has important noncatalytic morphogenic functions mediated through its C2 domain and β-Arrestin1 membrane targeting . To investigate Cdc42 activation by the unmasked PTEN C2 domain , we conducted expression studies in PTEN -/- cells . Transfection of PTEN CS-T383A robustly enhanced Cdc42-GTP , while the β-Arrestin1-binding defective CS-A4 mutant ( Lima-Fernandes et al . , 2011 ) had no effect . Subsequent to these Cdc42 activation studies , we investigated the specific role of β-Arrestin1-ARHGAP21 interactions on Cdc42-dependent multicellular morphogenesis . We used a cell-permeant peptide analogue of the β-Arrestin1 docking site within the ARHGAP21 GAP domain ( pep24 ) to disrupt β-Arrestin-ARHGAP21 interactions ( Anthony et al . , 2011 ) . Pep24 treatment suppressed β-Arrestin1-associated ARHGAP21 expression , inhibited Cdc42 activation , induced spindle misalignment and aberrant morphogenesis of 3D Caco-2 cultures . These morphogenic effects phenocopied those of Cdc42 knockdown ( Jagan et al . , 2013a; Jaffe et al . , 2008 ) . Collectively , our findings indicate that PTEN C2 coordinates β-Arrestin1-ARHGAP21 and Cdc42-dependent multicellular morphogenesis in a 3D colorectal cancer model system . PTEN is frequently downregulated in human colorectal ( Naguib et al . , 2011 ) and other cancers , even in the absence of genetic loss or mutation ( Salmena et al . , 2008 ) . Regulatory cues may have a central role in tumorigenesis ( Song et al . , 2012 ) . Activated GPCRs modulate β-Arrestin1 conformation ( Shukla et al . , 2008 ) , membrane recruitment ( Urs et al . , 2005; Li et al . , 2009; Décaillot et al . , 2011 ) and PTEN-β-Arrestin1 interactions ( Lima-Fernandes et al . , 2011 ) . By these processes , GPCRs may influence β-Arrestin1-dependent GTPase activation , cytoskeletal dynamics and neoplastic multicellular patterning . Studies in cancer cell models provide useful mechanistic data ( Hanahan and Weinberg , 2011 ) but intrinsic mutations may compromise physiological relevance . Studies of 3D multicellular organoids isolated from normal tissues have provided basic insights into normal tissue morphogenesis ( Clevers , 2016 ) . We have previously generated intestinal crypt organoids for study of multicellular assembly , patterning and lineage commitment ( Campbell et al . , 1993; Tait et al . , 1994; Slorach et al . , 1999 ) . Suppression of β-Arrestin1-ARHGAP21 binding in organoid systems by pep24 treatment perturbed spindle orientation and apical membrane alignment to induce a multilumen phenotype , surrounded by disorganized epithelium . Collectively , these findings demonstrate the importance of β-Arrestin1-ARHGAP21 interactions in control of normal colorectal multicellular architecture . Within the PTEN C2 domain , the CBR3 loop can localize cytoplasmic PTEN to early endosomes arranged along the microtubule cytoskeleton , by binding endosomal PIP3 ( Naguib et al . , 2015 ) . Restriction of PTEN to a punctate vesicular distribution along microtubules may enable dephosphorylation of PIP3 signals generated by plasma membrane receptor tyrosine kinases and parcelled in endosomes ( Naguib et al . , 2015 ) . However , it is difficult to envisage that wide endosomal distribution of scaffolding interactions along the microtubule cytoskeleton could regulate the compartmentalized focus of GTPase activity ( Pertz , 2010 ) required for control of spindle dynamics and multicellular morphogenesis ( Durgan et al . , 2011 ) . Hence , dephosphorylation of PIP3 on endosomes and scaffolding of β-Arrestin1-ARHGAP21 may represent spatiotemporally distinct PTEN tumor suppressor functions . This study shows that β-Arrestin1-ARHGAP21 interactions represent an essential component of the PTEN morphogenic pathway and sheds light on conserved developmental mechanisms . In C . elegans , PTEN/DAF18 conducts nutrient-sensing through its phosphatase domain ( Ogg and Ruvkun , 1998 ) in a negative feedback loop with the insulin/IGF axis ( Narbonne et al . , 2015 ) and casein kinase II [CKII] ( Liu Tj et al . , 2001 ) . CKII phosphorylates PTEN to induce the closed conformation ( Rahdar et al . , 2009; Torres and Pulido , 2001 ) that suppresses plasma membrane binding ( Rahdar et al . , 2009 ) . We show that the PTEN membrane-binding C2 domain is essential for multicellular morphogenesis . Hence , our findings may provide a rationale for PTEN multifaceted control of embryonic development by nutrient-sensing ( Ogg and Ruvkun , 1998 ) and regulation of morphogenic growth ( Rouault et al . , 1999 ) according to the available nutrient energy balance ( Hietakangas and Cohen , 2009 ) . Our study also has oncological relevance , since disruption of PTEN C2 domain-mediated β-Arrestin1-ARHGAP21 interactions drive evolution of morphology phenotypes in 3D cultures that are evocative of colorectal cancer ( Deevi et al . , 2016; Jaffe et al . , 2008 ) . Dissection of these phenomena may yield novel targets for therapy aimed at suppression of aggressive cancer morphology phenotypes that predict early metastasis . All laboratory chemicals were purchased from Sigma-Aldrich , Dorset , England unless otherwise stated . RNAiMAX and X-tremeGENE transfection reagents were purchased from Thermofisher , Dublin , Ireland and Roche , Basel , Switzerland , respectively . Antibodies used in this study were anti-β-Actin ( A5316; Sigma Aldrich , Dorset , England [RRID:AB_476743] ) ; anti-β-Arrestin1 ( ab32099; Abcam , Cambridge , UK [RRID:AB_722896] ) ; anti-ARHGAP21 ( 55139-1-AP; Proteintech Manchester , UK [RRID:AB_10794449] ) ; anti-E-Cadherin ( 562526; BD Biosciences , Oxford , UK [RRID:AB_11153868] ) ; anti-GAPDH ( ab8245; Abcam , Cambridge , UK [RRID:AB_2107448] ) ; anti-GFP ( ab8245; Abcam , Cambridge , UK [RRID:AB_298911 ] ) ; anti-HSP90 ( sc-7947; Santa Cruz , Dallas , Texas , USA [RRID:AB_2121235] ) ; anti-Pericentrin ( PCN:ab4448; Abcam , Cambridge , UK [RRID:AB_304461] ) ; anti-Protein Kinase C ζ [PRKCZ] ( ab51157; Abcam , Cambridge , UK [RRID:AB_882057] ) ; anti-PTEN ( ab32199; Abcam , Cambridge , UK [RRID:AB_777535] ) ; anti-α-Tubulin ( Ab7291; Abcam , Cambridge , UK [RRID:AB_2241126] ) ; anti-Cdc42 ( ab41429; Abcam , Cambridge , UK [RRID_726768] ) and anti-Cdc42-GTP ( 26905; New East Biosciences , PA , USA , [RRID:AB_1961759] ) . These primary antibodies were used where appropriate in conjunction with Li-Cor IRDye 680 ( anti-rabbit ) [RRID:AB_621841] and IRDye 800 ( anti-mouse ) [RRID:AB_10793856] secondary antibodies , for use with the Li-Cor Infra-Red imaging systems ( Li-Cor Biosciences , Lincoln , Nebraska , USA ) in Western blots or with Alexa Fluor 568 ( anti-rabbit ) [RRID:AB_143011] and Alexa Fluor 488 ( anti-mouse ) [RRID:AB_141626;Molecular probes , Invitrogen , CA , USA] and/or anti-mouse CY5 ( Jackson Immunoresearch , Newmarket , Suffolk , UK[RRID:AB_[RRID:AB_2340152] ) for fluorescence or confocal microscopy . We obtained Alexa 488-labeled wheat germ agglutinin ( WGA ) from ThermoScientific Dublin ( Product No W11261 ) . DNA was imaged with DAPI ( Vector Scientific , Belfast , NI ) while FITC-labeled phalloidin ( p5282; Sigma-Aldrich , Dorset , England ) was used to image apical actin in organoid cultures . For PLA , studies we used mouse anti-β-Arrestin1 from ThermoScientific , Paisley , UK with Duolink in situ fluorescence kits , ( Sigma-Aldrich , Dorset , England ) according to manufacturer’s instructions . SiRNA oligonucleotides targeted against β-Arrestin1 ( Qiagen Flexitube; 1027417 ) or ARHGAP21 ( Dharmacon SmartPool; L-009382-01-0005 ) or nontargeting ( NT ) scrambled controls were purchased from Fisher Scientific , Dublin , Ireland . The cell permeant β-Arrestin1-ARHGAP21 binding disruptor peptide [pep24 - based on amino acids 1331 to 1355 within the ARHGAP21 GAP domain ( Anthony et al . , 2011 ) ] and scrambled control peptide were purchased from EZ Biolabs , Carmel , IN 46032 USA . Pep24 and control peptides were prepared in dimethyl sulfoxide ( DMSO ) , according to manufacturer’s instructions . For the pep24 experiments , cells were incubated in 2D or 3D cultures as outlined below for 48 hr , then treated with either 10 μM pep24 or 10 μM control peptide . Incubations were continued for 24 hr for assays of protein binding or Western blots in cell monolayers . In 3D morphogenesis assays , test and control peptides were added to the media in the above concentrations , changed at 48 hr intervals and effects on morphogenesis assessed at 4 days of culture . Stable PTEN-deficient Caco-2 ShPTEN ( ShPTEN ) cells were generated by transfection of parental Caco-2 cells ( obtained from the American Type Culture Collection , Manassas , VA [RRID:CVCL_0025] ) with replication-defective retroviral vectors encoding PTEN short hairpin RNA ( ShRNA ) , using the Phoenix retroviral expression system ( Orbigen , San Diego , CA USA ) , as previously described ( Jagan et al . , 2013a; Deevi et al . , 2011 ) . PTEN +/+ HCT116 [RRID:CVCL_0291] and PTEN -/- HCT116 ( here known as HCT116 and PTEN -/- ) colorectal epithelial cells were a gift from Dr Tod Waldman , Georgetown ( Lee et al . , 2004 ) and were cultured in McCoys 5A media supplemented with 10% FCS ( fetal calf serum ) , 1 mM L-glutamine and 1 mM sodium pyruvate . Caco-2 and ShPTEN cells were cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented 10% FCS , 1 mM non-essential Amino Acids and 1 mM L-Glutamine at 37°C in 5% CO2 . In 3D cultures , Caco-2 , ShPTEN cells and subclones transfected with SiRNAs , PTEN C2 domain or empty vector ( EV ) control constructs , were cultured embedded in a Matrigel matrix ( BD Biosciences , Oxford , UK ) , as previously described ( Jagan et al . , 2013a; Jagan et al . , 2013b ) . Caco-2 and HCT116 cells were characterized in terms of PTEN expression , AKT signaling , GTPase activation ( Jagan et al . , 2013a ) and Caco-2 morphogenic growth ( Jagan et al . , 2013a ) . Furthermore , short tandem repeat ( STR ) profiling ( Capes-Davis et al . , 2013 ) conducted by LGC Standards , Middlesex , UK confirmed authenticity by 100% and 94% matches , respectively , between study parental Caco-2 and HCT116 cells and original American Type Culture Collection ( ATCC ) derivatives . We carried out mammalian SiRNA and DNA transfections using RNAiMAX and X-tremeGENE transfection reagents respectively , according to manufacturer’s protocols . Cells were plated at 2 × 105 cells/35 mm dish for 24 hr , then transfected with 10 µM siRNA or 500 ng DNA/2 × 105 cells for all respective siRNA oligonucleotides or DNA constructs . Cells were incubated with RNA/RNAiMAX or DNA/X-tremeGENE lipofectamine complexes for 48 hr , before lysis and probing . In 2 ShPTEN cells , the stably expressed PTEN ShRNA targets a 58 base pair region within the PTEN phosphatase coding region and C2 domain constructs are unaffected ( Boehm et al . , 2005 ) . In membrane localization studies , PTEN-expressing and -deficient Caco-2 and HCT116 clones were transfected with β-Arrestin1-mCherry against mCherry only controls . In expression , co-immunoprecipitation and morphogenesis studies , ShPTEN cells were transiently transfected with empty vector ( EV ) only , the isolated PTEN C2 domain ( C2 ) or a C2 domain construct mutated at the CBR3 membrane-binding loop [C2-MCBR3] ( Lee et al . , 1999 ) in pEGFP expression vectors . PTEN-C124S-A4 and PTEN-C124S-T383A were generated by introduction of four alanine substitutions at Ser380 , Thr382 , Thr383 and Ser385 and by alanine substitution at Thr383 only , respectively ( Raftopoulou et al . , 2004 ) into lipid and protein phosphatase dead PTEN C124S ( Maier et al . , 1999 ) . These mutants suppress or enhance β-Arrestin1 binding , respectively ( Lima-Fernandes et al . , 2011 ) . PTEN-MCBR3 , the isolated C2 domain and C2-MCBR3 were gifts from Dr N Leslie , Dundee and were generated by replacement of 263-K-M-L-K-K-D-K-269 in the C2 domain CBR3 membrane targeting loop with the 263-A-A-G-A-A-D-A-269 sequence ( Lee et al . , 1999 ) , as previously described ( Jagan et al . , 2013b ) . Sequence specificities of C2-MCBR3 and PTEN-MCBR3 mutants were confirmed by sequencing studies . We conducted these experiments using a subcellular fractionation kit ( Thermo Fisher Scientific , Dublin , Ireland ) according to manufacturer’s protocol . Briefly , cells were trypsinized and lysed in cytoplasmic extraction buffer for 10 min at 4°C , then centrifuged at 500 g for 5 min . The supernatant was collected as the cytoplasmic fraction while the pellet was resuspended in membrane extraction buffer , vortexed for 5 s and mixed gently for 10 min at 4°C . The mix was centrifuged at 3000 g for 5 min and the supernatant was collected as the membrane fraction . In separate experiments , we conducted protein extraction , Western blotting and co-immunoprecipitation ( Co-IP ) assays in isolated cell membrane and cytosolic fractions . Equivalent amounts of membrane fraction and cytosol were loaded in immunoblots and Co-Ips . As previously described ( Jagan et al . , 2013a; Jagan et al . , 2013b ) , proteins were resolved using gel electrophoresis , followed by blotting onto nitrocellulose membranes . Membranes were probed using antibodies as indicated in the text . Experiments were repeated in triplicate . Cells were lysed on ice in buffer containing 100 mM Tris-HCl , pH 7 . 5 , 1% Triton X-100 , 5 mM EDTA , 5 mM EGTA , 50 mM NaCl , 5 mM NaF , 1 mM Na3VO4 and protease inhibitor . Cell lysates were centrifuged ( for 10 mins at 15 , 000 g ) and protein concentrations were measured by the BCA method . 1000 µg of protein was precleared overnight with control IgG and 15 μl of Protein A/G Sepharose beads ( Santa Cruz , Dallas , Texas , USA ) . The protein was then immunoprecipitated with the appropriate antibody-beads conjugate and incubated on a rotating wheel for 2 hr . The beads were collected by centrifugation and washed five times in wash buffer ( 50 mM Hepes , pH 7 . 4 , 1% Triton X-100 , 0 . 1% , SDS , 150 mM NaCl , 1 mM Na3VO4 ) . The beads were subsequently resuspended in 40 µl Laemmli sample buffer and processed for gel electrophoresis . Experiments were conducted as previously described ( Deevi et al . , 2016; Jagan et al . , 2013b; ) . Briefly , cells were grown on 90 mm dishes then lysed in buffer comprising 50 mM Tris-HCl ( pH 7 . 5 ) , 1% Triton X-100 , 100 mM NaCl , 10 mM MgCl2 , 5% glycerol , 1 mM Na3VO4 and protease inhibitor cocktail ( Roche ) and centrifuged at 12 , 500 g for 10 min . We assayed the GTP-bound form of RhoA by adding GST-Rhotekin fusion protein coupled with gluthathione sepharose 4B beads ( Sigma-Aldrich , Dorset , England ) to 1 mg of cell lysate . Beads were collected after 1 hr by centrifugation , washed x3 and resuspended in Laemmli buffer with 1 mM DTT . RhoA -GTP levels were then assayed by western blotting . Experiments were repeated in triplicate . We assessed protein-protein proximities using the Duolink II red kit ( Sigma-Aldrich , Dorset , UK ) according to the manufacturer’s instructions . Briefly , we transfected PTEN-/- cells with GFP tagged-EV or -PTEN constructs and cultured the cells in Millipore eight well chambers . After 24 hr , we fixed the cells with 4% paraformaldehyde ( PFA ) at room temperature for 20 min . We then permeabilized the cells with 0 . 05% TritonX100 in PBS for 10 min . Cells were blocked with immunofluorescence ( IF ) buffer ( Duolink , Sigma-Aldrich , Dorset ) , England for 2 hr according to manufacturer’s instructions and incubated with primary antibody overnight at 4°C . Cells were washed twice with buffer A , and incubated with PLA probe , ligase and polymerase according to the manufacturer’s protocol . Finally , cells were washed with buffer B and slides were mounted with a cover slip using Duolink in situ mounting medium with DAPI . BRET investigations were performed as described previously ( Lima-Fernandes et al . , 2014 ) . Briefly , HEK cells were transfected with the indicated plasmids 24 hr after seeding . At 24 hr post transfection , cells were detached , resuspended in full media , and distributed into poly-l-orthinine coated white 96-well optiplates ( Perkin Elmer ) . The following day , cells were washed with PBS and then overlayed with HBSS . Coelenterazine h was added to a final concentration of 5 mM and incubated for 3 min at 25°C . BRET readings were collected using a Multimode Reader Mithras2 LB 943 ( Berthold Technologies ) . Substrate and light emissions were detected at 480 nm ( Rluc ) and 540 nm ( YFP ) for 1 s . The BRET signal was calculated by ratio of the light emitted by YFP and the light emitted by Rluc ( YFP/Rluc ) . The ratio values were corrected by substracting background BRET signals detected when Rluc-PTEN was expressed . mBRET values were calculated by multiplying these ratios by 1000 . ∆mBRET values are shown to demonstrate the shift in BRET signal compared to wt signal , which is set to zero , or between the two mutants ( C124S-T383A and C124S-A4 ) that were tested . Caco-2 and Caco-2 ShPTEN cells were cultured and embedded in Matrigel matrix ( BD Biosciences , Oxford , UK ) , then imaged by confocal microscopy as we previously described ( Jagan et al . , 2013a; Jagan et al . , 2013b ) . Briefly , 6 × 104 trypsinized cells were mixed with Hepes buffer ( 20 mM , pH 7 . 4 ) and Matrigel ( 40% ) in a final volume of 100 μl , placed in each well of eight-well multichambers ( BD Falcon , Fisher Scientific , Dublin , Ireland ) , allowed to solidify for 30 min at 37°C and subsequently overlayed with 400 μl of media/well . We imaged the 3D cultures at progressive stages of morphogenesis as previously described ( Jagan et al . , 2013a; Jagan et al . , 2013b ) . We used C57B/6 wild-type mice ( 1–6 weeks old ) for experiments and conducted all animal procedures in accordance with local and national regulations . We isolated organoids as previously described ( Tait et al . , 1994; Slorach et al . , 1999 ) . Briefly , murine colons were opened longitudinally , cut into 0 . 5 cm fragments , washed 7–10 times in 1x HBSS ( low calcium , low magnesium ( Gibco-BRL ) , 2% D-glucose , 0 . 035% NaHCO3 ) to remove all luminal contents . The fragments were then finely chopped with a scalpel and digested in HBSS solution containing collagenase and dispase I neutral proteases ( Sigma-Aldrich , Dorset , UK ) at 1 mg/ml for 20 min at room temperature on a shaking platform . Digestion was stopped by the addition of 30 ml DMEM/F12 culture medium ( Life Technologies , Renfrew UK ) supplemented with 5% FCS containing penicillin and streptomycin . Large fragments and muscle sheets were allowed to settle to the bottom of the flask . We removed the supernatant containing the organoids and centrifuged it for 3 min at 250 rpm , to pellet the organoids . We removed the supernatant and gently resuspended the organoid pellet in 20 ml of the DMEM/F12 solution . We repeated the centrifugation step 5–6 times until the pellet contained a homogeneously sized organoid preparation . Organoids thus prepared were resuspended in a 2x volume of Matrigel ( growth factor reduced , phenol red free; BD Biosciences , Oxford UK ) supplemented with 50 ng/ml murine EGF , murine Noggin 100 ng/ml ( PeproTech , NJ , USA ) and 1 μg/ml human R-Spondin , as indicated for organoid culture ( Sato et al . , 2011 ) . Eight well multichambers were coated with a thin layer of undiluted Matrigel and allowed to solidify . Organoid preparations in Matrigel ( 100 μl suspension ) were placed into each well , then overlaid with 250 μL/well culture medium ( Dulbecco's modified Eagle medium/F12 ) supplemented with penicillin/streptomycin , 10 mmol/L HEPES , Glutamax supplements 1× N2 , 1 × B27 [Invitrogen] , 1 mmol/L N-acetylcysteine [Sigma] ) , 50 ng/ml murine EGF , Noggin 100 ng/ml and 1 μg/ml human R-Spondin ( Sato et al . , 2011 ) . We cultured the organoids for 4 days with peptide treatments as defined . Membrane and cytosolic localization of β-Arrestin1-mCherry or mCherry only were imaged against Alexa 488-labelled WGA , a widely used fluorescent marker that binds to cell membranes ( Crossman et al . , 2015 ) in Caco-2 , ShPTEN , HCT116 and PTEN −/− cells , with or without LPA treatment . We used a Leica SP8 confocal microscope and Leica LAS-X software for line scanning of fluorescent images . Caco-2 ShPTEN ( ShPTEN ) glands and organoid cultures were incubated in 4% PFA for 20 min and processed for immunofluorescence as previously described ( Jagan et al . , 2013a; Deevi et al . , 2016; Jagan et al . , 2013b ) . Briefly , 3D cultures were fixed in PFA for 20 min at room temperature and permeabilized for 10 min in 0 . 5% Triton X-100 in PBS . The 3D cultures were rinsed with PBS/glycine buffer for 15 min to reduce autofluorescence and blocked by incubation in IF Buffer ( PBS with 0 . 1% bovine serum albumin , 0 . 2% Triton X-100 , 0 . 05% Tween-20 ) +10% goat serum , for 1–1 . 5 hr at room temperature . Primary antibodies were diluted in blocking buffer and incubated overnight at 4°C . The 3D cultures were incubated with secondary antibodies and/or FITC-labeled phalloidin for 1 hr . DNA was stained using Vectashield mounting medium containing DAPI ( Vector Scientific , Belfast , NI ) . Sequential scan images were taken the midsection of glands/organoids at room temperature using a Leica SP5 confocal microscope [RRID:SCR_012314] on a HCX PL APO lambda blue 63 × 1 . 40 oil immersion objective at 1x or 2x zoom . Images were collected and scale bars added using LAS AF confocal software ( Leica ) [RRID:SCR_013673] . We assessed effects of transfection or treatment on signal intensity , spindle orientation , lumen formation and/or epithelial configuration in in 3D glands or organoids at 4 days of culture . Because imaging for apical protein kinase C zeta ( PRKCZ ) was unsuccessful in organoids , the apical domain was imaged using FITC-labelled phalloidin as a marker of apical actin . In cultured cells , centrosomes ( Csms ) were identified using anti-pericentrin ( PCN ) and microtubules by anti-α-Tubulin antibodies , respectively , and we identified chromosomal DNA by DAPI staining . We defined bipolar mitotic spindle architecture by convergence of microtubules towards each of 2 spindle poles , as we previously described ( Deevi et al . , 2016 ) . Caco-2 and Caco-2 shPTEN glands were cultured in Matrigel for 4 days , fixed with 4% PFA and stained with anti α-Tubulin , PRKCZ and PCN primary antibodies . Gland midsections were imaged by confocal microscopy to identify cells containing well-formed mitotic spindles , during metaphase or anaphase . Lines connecting each spindle extremity were drawn using ImageJ and the line center was considered as the spindle midpoint . Angles between spindle planes and lines connecting spindle midpoints to gland centres were measured , as outlined previously ( Deevi et al . , 2016; Jaffe et al . , 2008 ) . We used a similar approach or imaging of organoid morphogenesis and identified apical domains and spindles using FITC-labeled phalloidin and anti-α-Tubulin antibodies , respectively . Confocal microscopy images were processed using Leica Fw4000 Imaging software and cropped using Adobe Ilustrator [RRID:SCR_014198] . Confocal images were processed , merged and mean area quantified using LAS AF Leica imaging software , as previously described ( Jagan et al . , 2013a; Jagan et al . , 2013b ) . We assessed lumen formation , spindle orientation and signal intensities using 50 , 15 or 10 × 3D Caco-2 or ShPTEN glandular cultures ( glands ) in triplicate for each experimental condition , respectively . We selected glands with mitotic figures for spindle orientation assays . Organoids were fewer in number and we assessed lumen formation and spindle orientation in 10 organoids per experimental condition , in triplicate . Multicellular structures with single central lumens were expressed as a percentage and spindle orientation angles relative to gland centres were calculated using ImageJ . Descriptive statistics were expressed as the mean ± sem . Statistical analyses were by one or two-way ANOVA with the Tukey post hoc test or Student’s paired t test using Graphpad Prism software ( v5; Graphpad CA 92037 USA [RRID:SCR_002798] ) . Scatterplots and bar charts were used for display of quantitative numerical or categorical data .	The protein PTEN helps to organize cells in the body to form complex structures . In particular , it collects signals from a cells’ surroundings and changes where cells divide so new cells are produced in the right places . The control of cell division by PTEN is also thought to help limit the progression and spread of cancer . PTEN can interact with another protein called β-Arrestin1 , which behaves as a so-called scaffolding protein – in other words , one that helps groups of proteins to interact with each other . β-Arrestin1 has been found to control cell division via a series of other proteins , including ARHGAP21 and Cdc42 . The relationship between PTEN and these other proteins in dividing cells is still not fully understood . Javadi , Deevi et al . studied PTEN in human cells grown in the laboratory to show that a part of PTEN known as the C2 domain allows it to help organize cells by moving β-Arrestin1 to the outer edge of the cell – the cell membrane . This relocation allows β-Arrestin1 to interact with ARHGAP21 and Cdc42 , and control cell division . Active Cdc42 changes the orientation of cell division , allowing cells to organize into single layers of regular cells and similar tightly controlled structures . Further experiments revealed that these proteins are important to form tubes inside the glands of the gut . The C2 region of PTEN also helps to detect signals carried by fat molecules in the cell membrane , so these results provide a direct link between signaling and cell organization via PTEN . The work of Javadi , Deevi et al . provides new understanding of how PTEN links nutrient availability to cell organization during development and may also lead to new insights into the role of PTEN in limiting the growth of tumors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2017	PTEN controls glandular morphogenesis through a juxtamembrane β-Arrestin1/ARHGAP21 scaffolding complex
SpoIIIE is a membrane-anchored DNA translocase that localizes to the septal midpoint to mediate chromosome translocation and membrane fission during Bacillus subtilis sporulation . Here we use cell-specific protein degradation and quantitative photoactivated localization microscopy in strains with a thick sporulation septum to investigate the architecture and function of the SpoIIIE DNA translocation complex in vivo . We were able to visualize SpoIIIE complexes with approximately equal numbers of molecules in the mother cell and the forespore . Cell-specific protein degradation showed that only the mother cell complex is required to translocate DNA into the forespore , whereas degradation in either cell reverses membrane fission . Our data suggest that SpoIIIE assembles a coaxially paired channel for each chromosome arm comprised of one hexamer in each cell to maintain membrane fission during DNA translocation . We show that SpoIIIE can operate , in principle , as a bi-directional motor that exports DNA . The transport of DNA across cellular membranes is an essential part of bacterial processes such as transformation and conjugation ( Errington et al . , 2001; Burton and Dubnau , 2010 ) . A paradigmatic example is the segregation of chromosomes that are trapped in the septum during cell division , which requires specialized DNA translocases of the SpoIIIE/FtsK/HerA protein superfamily . The members of this superfamily use the energy of ATP to translocate DNA and peptides through membrane pores ( Bath et al . , 2000; Iyer et al . , 2004; Tato et al . , 2005; Burton and Dubnau , 2010 ) . SpoIIIE and FtsK contain an N-terminal domain that anchors the protein to the septal membrane ( Wu and Errington , 1997; Wang and Lutkenhaus , 1998; Yu et al . , 1998 ) , a poorly conserved linker domain , and a cytoplasmic motor domain with ATPase activity that is responsible for DNA translocation . The motdata-left-gapor domain consists of three subdomains: α , β , and γ ( Massey et al . , 2006 ) . α and β form the core ATPase domain and are responsible for chromosome translocation , while the γ subdomain regulates translocation directionality ( Pease et al . , 2005; Ptacin et al . , 2008 ) . During Bacillus subtilis sporulation , an asymmetrically-positioned septum creates two daughter cells of different size: the bigger mother cell and the smaller forespore . SpoIIIE is made before polar septation ( Foulger and Errington , 1989 ) and localizes to the leading edge of the constricting septum ( Fleming et al . , 2010; Fiche et al . , 2013 ) ( Figure 1A ) . As the sporulation septum closes around the chromosome , SpoIIIE forms a stable focus at the septal midpoint ( Wu and Errington , 1997; Fleming et al . , 2010 ) , where it mediates two key events . First , it keeps the mother cell and forespore septal membranes separated in the presence of a septum-trapped chromosome , playing an important role in septal membrane fission ( Liu et al . , 2006; Fleming et al . , 2010 ) ( Figure 1 ) . Second , it translocates the chromosome remaining in the mother cell ( about 2/3 of its total length ) to the forespore ( Wu and Errington , 1994; Bath et al . , 2000 ) . This vectorial DNA translocation is dictated by the interaction of the γ domain with SpoIIIE recognition sequences ( SRS ) that are distributed in a skewed manner along the B . subtilis chromosome from the origin of replication towards the terminus ( Figure 1 ) ( Pease et al . , 2005; Ptacin et al . , 2008 ) . It has been proposed that SpoIIIE exports DNA ( Sharp and Pogliano , 2002; Ptacin et al . , 2008 ) and that the interaction between the γ subdomain of SpoIIIE and the SRS favors either the selective assembly of SpoIIIE in the mother cell or , equivalently , the inactivation or disassembly of motor domains in the forespore ( Sharp and Pogliano , 2002; Becker and Pogliano , 2007; Ptacin et al . , 2008; Fiche et al . , 2013 ) . 10 . 7554/eLife . 06474 . 003Figure 1 . Chromosome translocation during B . subtilis sporulation . ( A ) The sporulation septum traps the oriC-proximal region of the forespore chromosome in the forespore ( F ) , the rest in the mother cell ( MC ) . SpoIIIE ( green ) localizes at the leading edge of the constricting septum , and assembles a translocation complex at the septal midpoint . The SpoIIIE complex maintains separation of the daughter cell membranes in the presence of trapped DNA . The direction of translocation ( white arrow ) is determined by the orientation-specific interaction between the SpoIIIE γ domain and the skewed chromosomal recognition sequences known as SRS ( black arrowheads on the chromosome , indicating the direction that SpoIIIE motor domains move on the DNA ) . Engulfment commences during DNA translocation , producing a curved septum and movement of the mother cell membrane around the forespore . ( B ) The aqueous channel model for SpoIIIE , showing two chromosome arms . Green represents the SpoIIIE channels formed by motor domains , grey the transmembrane domains , and red the membrane . ( C ) The paired channel model for SpoIIIE , in which each chromosome arm passes through a proteinaceous channel with subunits in both cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 003 Structural studies of the motor domains of B . subtilis SpoIIIE ( Cattoni et al . , 2013; 2014 ) and of Escherichia coli and Pseudomonas aeruginosa FtsK ( Massey et al . , 2006; Löwe et al . , 2008 ) reveal that each assembles a hexamer with a central channel large enough to accommodate one dsDNA molecule . Cell biological studies have shown that both arms of the circular chromosome are translocated simultaneously ( Burton et al . , 2007 ) , suggesting that the translocation complex is made of at least two SpoIIIE hexamers , one for each chromosome arm . However , it remains unclear how SpoIIIE is organized at the septum . One model is that DNA is translocated through an aqueous membrane pore ( Figure 1B ) ( Wu and Errington , 1997; Fiche et al . , 2013 ) . According to this model , the transmembrane domains of SpoIIIE simply localize the complex to the septum , and it has been proposed that tension generated during DNA translocation generates an asymmetric complex with hexameric motor domains oriented towards the mother cell cytoplasm ( Fiche et al . , 2013 ) . A second model proposes that SpoIIIE transmembrane domains in opposite septal membranes pair to form DNA-conducting channel that traverses both septal membranes , with the motor domains in the respective cytosols ( Figure 1C ) ( Liu et al . , 2006; Burton et al . , 2007; Fleming et al . , 2010 ) . This model postulates that assembly of the paired channel mediates septal membrane fission and that the assembled channel maintains separation of daughter cell membranes during DNA translocation ( Liu et al . , 2006; Fleming et al . , 2010 ) . Here , we investigate the organization of the SpoIIIE DNA translocation complex in living cells . We developed a cell-specific protein degradation system that selectively removes SpoIIIE from each cell after polar septation and used this system with GFP tagging and quantitative PALM ( qPALM ) to investigate SpoIIIE architecture and function . We show that SpoIIIE forms a complex with approximately equal numbers of molecules in the mother cell and the forespore , enough to assemble at least two hexamers in each cell . Both the forespore and mother cell SpoIIIE subcomplexes are required to maintain septal membrane fission during DNA translocation . However , only the mother cell SpoIIIE is essential for chromosome translocation into the forespore . In the absence of the mother cell protein , forespore SpoIIIE translocates the chromosome out of the forespore , indicating that SpoIIIE exports DNA . Together our results are consistent with the paired channel model in which SpoIIIE assembles a DNA-conducting channel that spans the mother cell and forespore septal membranes . Moreover , we show that SpoIIIE can operate , in principle , as a bi-directional motor that exports DNA from a given cell compartment . To investigate the organization and the cell-specific function of SpoIIIE we developed a system that allows degradation of specific proteins selectively in the mother cell or forespore during sporulation . Our system is based on a method developed by Griffith and Grossman ( 2008 ) in which a heterologous SsrA tag from E . coli ( SsrA* ) is fused to the C-terminus of target proteins . The SsrA* tag is recognized by B . subtilis ClpXP protease only in the presence of the cognate SspB adaptor protein from E . coli ( SspBEc ) . To achieve cell-specific degradation of SsrA-tagged proteins during sporulation , we expressed sspBEc in a cell-specific manner by using promoters ( PspoIIQ and PspoIID ) dependent on the sigma factors σF and σE , which become active only in the forespore or in the mother cell , respectively , immediately after polar septation ( Clarke et al . , 1986; Rong et al . , 1986; Londoño-Vallejo et al . , 1997; Sharp and Pogliano , 2002 ) ( Figure 2A ) . 10 . 7554/eLife . 06474 . 004Figure 2 . Cell-specific degradation of proteins during sporulation . ( A ) Cell-specific protein degradation system ( see text ) . Red indicates the cell membranes , green the target protein . F = forespore , MC = mother cell . ( B ) Fluorescence microscopy of GyrA-GFP-SsrA ( green ) during sporulation , without degradation ( strain JLG917 ) , forespore degradation ( F , JLG919 ) and mother cell degradation ( MC , JLG1281 ) . Membranes are stained with FM4-64 ( red ) . Scale bar , 1 μm . ( C and D ) Quantification of the loss of GyrA-GFP-SsrA* fluorescence after degradation in the ( C ) mother cell and ( D ) forespore . No degradation controls express GyrA-GFP-SsrA* but not sspBEc ( blue circles and diamonds ) . The ratio of the mean GFP intensity in the ( C ) mother cell/forespore ( red squares ) and ( D ) forespore/mother cell ( red triangles ) were calculated for sporangia with flat , slightly curved and engulfing septa 2 . 5 hr after the initiation of sporulation ( t2 . 5 ) . 25–66 sporangia were analyzed for each cell type . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 00410 . 7554/eLife . 06474 . 005Figure 2—figure supplement 1 . Description of the different stages of engulfment analyzed in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 00510 . 7554/eLife . 06474 . 006Figure 2—figure supplement 2 . Cell-specific degradation of σA-GFP-SsrA* . ( A ) Cell-specific degradation of a σA-GFP-SsrA* fusion protein during sporulation . When sspBEc is not expressed ( strain JLG247 ) , GFP signal is detected in both forespore and mother cell . Expressison of sspBEc in the forepore ( strain JLG261 ) leads to GFP disappearance in the forespore , while expression of sspBEc in the mother cell ( strain JLG259 ) leads to GFP disappearance in the mother cell . Sporangia were harvested 2 . 5 h after resuspension ( t2 . 5 ) and images of σA-GFP-SsrA* ( green ) and membrane ( red ) were taken . Membrane was stained with FM4-64 ( red ) . Scale bar is 1 μm . ( B and C ) Quantification of the loss of σA-GFP-SsrA* fluorescence after degradation in the ( B ) mother cell and ( C ) forespore . No degradation controls express σA-GFP-SsrA* but not sspBEc ( blue circles and diamonds ) . The ratio of the mean GFP intensity in the ( B ) mother cell/forespore ( red squares ) and ( C ) forespore/mother cell ( red triangles ) cell were calculated for sporangia with flat , slightly curved and engulfing at t2 . 5 . 20–46 sporangia were analyzed for each cell type . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 006 We first analyzed the cell-specific degradation of DNA gyrase subunit A ( GyrA ) by constructing a GyrA-GFP-SsrA* fusion protein . The GFP signal from this protein was initially detected in both the forespore and the mother cell ( Figure 2B ) , but disappeared in the SspBEc–containing cell ( Figure 2B ) . To estimate the degradation kinetics , we correlated loss of the GFP signal with the stages of engulfment , a phagocytosis like process that occurs during DNA translocation . Immediately after polar septation , the septum is first flat , then curved , and finally the mother cell membrane engulfs the forespore ( Figure 1A , Figure 2—figure supplement 1 ) . Quantification of GFP fluorescence intensity showed that the protein was lost at similar rates in the cell expressing sspBEc ( forespore or mother cell ) ( Figure 2C , D ) . In both cases , the fluorescence decreased close to zero when sporangia entered engulfment , indicating that most GyrA-GFP-SsrA* was degraded at this sporulation stage ( Figure 2C , D ) . Similar results were obtained after cell-specific degradation of a σA-GFP-SsrA* fusion protein ( Figure 2—figure supplement 2 ) . Thus , cell-specific expression of sspBEc provides a rapid and efficient way to selectively degrade SsrA-tagged proteins in the mother cell or in the forespore . SpoIIIE translocates DNA across a septum comprised of two membranes , yet it remains unclear if the DNA translocation complex contains SpoIIIE monomers in both the mother cell and the forespore or just one cell . To address this question we used the cell-specific degradation system described above . If SpoIIIE assembles in one cell , then triggering degradation in that cell would cause the focus to disappear , while triggering degradation in the other cell would have no effect . However , if the translocation complex is present in both cells , then degradation in only one cell would leave SpoIIIE monomers in the other cell , and the focus would be expected to persist until SpoIIIE was degraded in both cells simultaneously . The SpoIIIE-GFP-SsrA fusion protein , as expected , formed a bright focus at the septal midpoint and persisted around the forespore during engulfment ( Figure 3A ) . When SpoIIIE was degraded in either cell , more than 80% of the sporangia still contained a SpoIIIE focus . However , when SpoIIIE was degraded in both cells simultaneously just 25% of sporangia retained a focus , and these were faint and present only in sporangia with flat and curved septa ( Figure 3A ) , which are at early stages of sporulation . 10 . 7554/eLife . 06474 . 007Figure 3 . Cell-specific SpoIIIE degradation . ( A ) Visualization of SpoIIIE-GFP-SsrA* ( green ) and FM4-64 stained membranes ( red ) by deconvolution fluorescence microscopy at t2 . 5 . SpoIIIE was not degraded ( None , JLG451 ) or degraded in the forespore ( F , JLG452 ) ; mother cell , ( MC , JLG453 ) or both ( F and MC , JLG454 ) . Percent sporangia with detectable SpoIIIE foci and the number ( n ) of scored sporangia are indicated for each strain . Strains containing SpoIIIE-GFP-SsrA* without SspBEc or expressing SspBEc with untagged-SpoIIIE supported wild-type sporulation , chromosome translocation and membrane fission ( Supplemental data ) . Scale bar , 1 μm . ( B ) Fluorescence intensity of SpoIIIE-GFP foci without degradation ( green ) or after degradation in the forespore ( red ) or mother cell ( blue ) in sporangia with flat , slightly curved and engulfing septa . The mean focus intensity of sporangia with flat septa in the non-degradation strain was normalized to 100 . Each dot represents the intensity of a single focus; black dotted lines represent the mean of each data set . Between 35 and 95 foci were analyzed for each data set . ( C ) Model for SpoIIIE organization at the septal midpoint based on cell-specific degradation . The translocation complex contains SpoIIIE molecules on both sides of the septum so cell-specific degradation will only remove a fraction of the molecules . Degradation in both cells will remove all molecules . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 007 Quantification of focus intensity after degradation in either cell showed that the average focus intensity was reduced by ∼50% in sporangia with curved and engulfing septa ( Figure 3B ) . After degradation in the forespore , cells with flat septa showed a detectable reduction of GFP intensity ( Figure 3B ) , suggesting that SpoIIIE degradation starts slightly faster in the forespore , likely reflecting the earlier activation of σF vs σE . These results indicate that cell-specific degradation of SpoIIIE commences shortly after polar septation and that the SpoIIIE translocation complex is comprised of monomers with C-terminal motor domains ( and the SsrA* tags ) in the cytosol of each cell ( Figure 3C ) . They also suggest that approximately half of the SpoIIIE molecules in the septal focus are in the mother cell and half in the forespore . If SpoIIIE assembles a channel with subunits located in each cell , then super-resolution microscopy might show a SpoIIIE cluster in each cell . However , examination of PALM images revealed mainly single foci ( Figure 4A ) , consistent with previous reports ( Fleming et al . , 2010; Fiche et al . , 2013 ) . Most likely this observation reflects the fact that the distance between the forespore and mother cell membranes at the septum is just ∼20 nm ( Tocheva et al . , 2013 ) , which is close to the limit of resolution via PALM ( ∼25 nm ) . Consistent with this idea , occasionally we observed sporangia with what appear to be two clusters of molecules aligned across the septum ( Figure 4A , right column ) . To improve the ability to resolve these two subcomplexes , we used a B . subtilis mutant that retains thick septal peptidoglycan because it lacks the mother cell transcription factor ( σE ) that controls septal thinning , the first step in engulfment ( Figure 4B ) . In this ΔσE strain , the distance between the mother cell and forespore septal membranes is ∼40 nm , more than twice the distance of wild type ( Illing and Errington , 1991 ) , potentially allowing resolution of the SpoIIIE subcomplexes in each cell by PALM . We therefore introduced SpoIIIE-tdEOS into the ΔσE strain . As expected ( Pogliano et al . , 1999 ) , this strain did not initiate engulfment or block the second potential division septum , producing a high number of disporic sporangia ( Figure 4C , Figure 4—figure supplement 1 ) without impairing DNA translocation ( Figure 4—figure supplement 2 ) . 10 . 7554/eLife . 06474 . 008Figure 4 . Direct visualization of two SpoIIIE clusters at the septal midpoint . ( A ) PALM image of single focus and a rare dual focus of SpoIIIE-Dendra2 in wild type B . subtilis ( TCF25 ) , with FM5-95 stained membranes ( white ) . The bottom PALM images are zoomed in from the yellow boxes . Bar is 500 nm and 50 nm for overlaid and PALM images , respectively . Membrane is diffraction-limited image . ( B ) Schematic diagram of septal thinning . ( I ) After septation , septal peptidoglycan is degraded by a complex containing the SpoIID , SpoIIM and SpoIIP proteins ( pacman ) and the second potential division site is blocked . ( II ) Elimination of SpoIIDMP in the ΔspoIIDMP ( spoIID , spoIIM , spoIIP ) strain inhibits septal thinning without impairing DNA segregation . ( III ) Elimination of σE in the ΔσE ( spoIIGB ) strain inhibits septal thinning and produces disporic cells without impairing DNA segregation . ( C ) PALM images of SpoIIIE-tdEos in ΔσE strain ( JS03 ) . Classification of PALM images as dual foci or single foci were defined according to the parameters of our cluster analysis . Scale bar is 500 nm and 50 nm for overlaid and PALM images , respectively . Membrane is diffraction-limited image . ( D ) PALM images of SpoIIIE-tdEos in ΔSpoIIDMP strain ( JLG571 ) . The diffraction-limited images of the membranes ( white ) and the DNA ( green ) were obtained by staining with FM5-95 and DAPI , respectively . The relative forespore DAPI intensity was ∼25% in both sporangia . SpoIIIE dual foci are shown in the left panels . In-gel fluorescence and western blots of the different SpoIIIE fusion proteins used here can be found in Figure 4—figure supplement 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 00810 . 7554/eLife . 06474 . 009Figure 4—figure supplement 1 . Classification of cells in vegetative , monosporic and disporic cells at t1 . 75 . Mutation in σE produces a much higher percentage of disporic cells compared to the σE+ strain PY79 . More than 700 cells were classified per strain . The strains used were KP161 , TCF25 and JS03 . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 00910 . 7554/eLife . 06474 . 010Figure 4—figure supplement 2 . SpoIIIE segregates the chromosomes in the ΔσE strain . ( A ) Mutation in σE does not impair DNA translocation . Forespores show brighter DNA signal ( single arrowhead ) at t2 . 5 compared to t1 . 75 of sporulation time showing chromosome translocation from the mother cell into the forespore . ΔσE cells expressing SpoIIIE fused to tdEos ( JS03 ) were stained with DAPI and FM5-95 to visualize DNA ( Green ) and membrane ( red ) , respectively . DAPI signal was false-colored green for a better contrast . ( B ) Quantification of DNA translocation at different sporulation times . The amount of DNA in the forespore increases as sporulation progresses ( from ∼60–100% , inset ) . Our PALM experiments using ΔσE strains ( Figures 4 , 5 ) were done at t1 . 75 to maximize the percentage of cells are actively translocating DNA . Error bar represents standard error of the mean from 20 cells for each of the time points . ( C ) Cells were sporulated at 37°C and samples were taken at indicated times and stained with DAPI and FM5-95 to visualize DNA and membrane ( red ) , respectively . The amount of DAPI ( labeled-DNA ) in the forespore increased from t1 . 5 to t2 in ΔσE sporangia expressing wild type SpoIIIE ( SpoIIIEWT ) fused to tdEos ( JYS03 ) , indicating that the forespore received the chromosomal complement from the mother cell . In contrast , ΔσE cells expressing SpoIIIE ATPase mutant ( SpoIIIEATP− ) fused to tdEos ( JYS04 ) did not translocate DNA . SpoIIIEWT fused to Dendra2 ( JYS00 ) also translocated DNA in ΔσE strain . Altogether , these results indicate that SpoIIIE translocates chromosome during sporulation in ΔσE strain and the fusion protein does not impair the ability of this translocation . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01010 . 7554/eLife . 06474 . 011Figure 4—figure supplement 3 . Examples of single foci in the ΔσE strain when SpoIIIEWT is fused to tdEos . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01110 . 7554/eLife . 06474 . 012Figure 4—figure supplement 4 . Distribution of single and double foci in monosporic and disporic sporangia . Classification of monosporic and disporic sporangia in relation to the presence of either single focus or dual foci at the sporulation septum of cells expressing either SpoIIIE fused to tdEos ( JS03 ) or Dendra2 ( JS00 ) in the ΔσE strain . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01210 . 7554/eLife . 06474 . 013Figure 4—figure supplement 5 . SpoIIIE ATPase mutant ( SpoIIIEATP− ) fused to tdEos in ΔσE strain also organizes into dual foci . To demonstrate that SpoIIIEWT dual foci ( Figure 4 ) are observed when the chromosome is trapped in the septum and discard that are only at septa where chromosome transport is complete we imaged SpoIIIEATP− fused to tdEos by PALM . Single mutation G467S in the conserved region of the SpoIIIE ATPase motor domain impairs DNA translocation without affecting SpoIIIE foci assembly , and the forespore chromosome remains trapped in the septum ( Sharp and Pogliano , 1999; Fleming et al . , 2010 ) . qPALM shows that SpoIIIEATP− also assembles into dual foci ( PALM: left panels and , first and second right panels ) and single foci ( third right panel ) similar to the SpoIIIEWT ( Figure 4 ) . As expected sporangia expressing SpoIIIEATP− did not translocate DNA ( Figure 4—figure supplement 2C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01310 . 7554/eLife . 06474 . 014Figure 4—figure supplement 6 . SpoIIIE organizes into dual foci at the septum of sporangia where chromosome transport is incomplete . To demonstrate that dual foci represent SpoIIIE complexes that are actively translocating DNA , we imaged SpoIIIEWT in a different Bacillus strain , spoIIDMP triple mutant strain , which mantains a thick sporulation septum due to the inhibition of septal thinning ( Pogliano et al . , 1999; Abanes De Mello et al . , 2002; Chastanet and Losick , 2007; Gutierrez et al . , 2010 ) . In contrast to the ΔσE mutant , the completion of the second polar septum is inhibited in the spoIIDMP triple mutant , leading to the formation of fewer disporic sporangia . This feature allows the estimation of the degree of chromosome translocation using DNA-specific dyes and measuring the dye-intensity in the forespore vs the total intensity ( forespore plus mother cell ) , a standard procedure in the field ( Becker and Pogliano , 2007; Ptacin et al . , 2008 ) . The relative forespore DAPI intensity of sporangia showing dual foci in the spoIIDMP mutant ranged between 0 . 1 and 0 . 32 , suggesting that they were actively translocating DNA . These observations suggest that dual foci represent the organization of the SpoIIIE complex during the process of DNA translocation . Criteria for complete and incomplete DNA translocation is described in the chromosome translocation section of the results and in Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01410 . 7554/eLife . 06474 . 015Figure 4—figure supplement 7 . In-gel fluorescence and western blot of the SpoIIIE fusion proteins used in this study . Our single SpoIIIE molecule quantification relies on the fluorescent protein fused to SpoIIIE . Therefore , it is crucial that all SpoIIIE molecules are fused to the fluorescent protein and also all the fluorescent proteins fused to SpoIIIE . ( A ) Fluorescent proteins ( GFP , Dendra2 and tdEos ) fused to SpoIIIE migrate as a single band in 7% semi-denaturing in-gel fluorescence PAGE ( Drew et al . , 2006 ) . As expected the tandem dimer Eos ( tdEos ) fusion proteins migrate slightly higher than the GFP and the Dendra2 fusion proteins . ( B ) SpoIIIE fused to fluorescent proteins migrates as a higher molecular weight compared to the native SpoIIIE ( lanes 2 and 3 ) , indicating that all the SpoIIIE molecules are fused to the fluorescent proteins . The molecular weight of the fusion proteins corresponded to the in-gel fluorescence . The faint band at ∼130 kDa is unspecific . ( C ) Total protein visualized by Coomassie-Blue staining . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 015 Visualization of SpoIIIE-tdEOS via PALM in this strain showed that 39% of septa contained a medial SpoIIIE assembly comprised of two separate subcomplexes of molecules ( hereafter called ‘dual foci’ , Figure 4C , D ) . Most dual foci were aligned across the division septum and a few were slightly tilted ( Figure 4C ) . The average distance between the centers of each focus was 55 nm ( Figure 5A , B , Figure 5—figure supplement 1 ) . Electron cryotomography shows that before septal thinning the cytoplasmic faces of the septum are separated by ∼40 nm ( Tocheva et al . , 2013 ) , suggesting that the centers of each resolved subcomplexes of SpoIIIE lie in separate cells . Dual foci were also observed in 26% of sporangia when septal thinning was blocked by the absence of the SpoIID , SpoIIM and SpoIIP proteins that degrade septal peptidoglycan ( Figure 4B , D ) ( Pogliano et al . , 1999; Abanes De Mello et al . , 2002; Chastanet and Losick , 2007; Gutierrez et al . , 2010 ) . Thus using two strains with thick septa allowed visualization of SpoIIIE foci comprised of molecules in both the mother cell and forespore ( Figure 4 ) . 10 . 7554/eLife . 06474 . 016Figure 5 . quantitative PALM ( qPALM ) of SpoIIIE complexes . ( A ) The distances ( D ) between the centers of each cluster from SpoIIIE-tdEos dual foci in ΔσE strain ( JYS03 ) . The average distance between clusters , indicated by the main peak value of the kernel density estimator ( blue line ) , is 55 nm . ( B ) Schematic diagram of SpoIIIE ( blue ) at the division septum in ΔσE strain . To estimate the thickness of the peptidoglycan ( PG ) we subtracted twice the width of the lipid bilayer ( 3 nm; Lewis and Engelman , 1983 ) and twice the height of SpoIIIE ( 6 nm ) , based on the FtsK crystal structure ( Massey et al . , 2006 ) from 55 nm , to give 37 nm . FP , fluorescent protein . ( C ) Dimensions of single SpoIIIE-Dendra2 foci and resolved clusters in dual foci . Focus widths in foci parallel ( x ) and perpendicular ( y ) to the septum were calculated as described in supplementary methods . For single foci the parallel- and perpendicular-width ( in nm ) are 86 ± 10 and 94 ± 12 respectively , for dual foci , the forespore proximal cluster is 64 ± 11 and 57 ± 8 , the mother cell proximal cluster 77 ± 12 and 52 ± 7 . Error bars , standard deviation from Nsingle = 221 , Ndual foci = 51 . ( D ) Distribution of the number of SpoIIIE-Dendra2 molecules determined by qPALM at the septum in the mother cell , in the forespore and in both . Data ( blue bars ) are represented by Kernel Density Estimator ( red solid lines ) . Mean and standard deviations from Nfoci = 51 are shown . More details about qPALM and the algorithm employed for quantification can be found in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01610 . 7554/eLife . 06474 . 017Figure 5—figure supplement 1 . Distance distribution between the centers of each of the clusters from the SpoIIIE-Dendra2 dual foci in ΔσE strain . The main peak value of the kernel density estimator ( blue line ) is 68 nm , which is slightly higher than the measurement with SpoIIIE-tdEos strain ( 55 nm , Figure 5A ) , most likely because Dendra2 is dimmer and thus its localization is less accurate than mEos2 . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 01710 . 7554/eLife . 06474 . 018Figure 5—figure supplement 2 . Fermi-photoactivation . ( Top ) Approximate analytic solution of the optimal τc agrees well with the exact solution obtained from full stochastic simulations . The experimental time tF and the number of molecules Nmol were varied , whereas the kinetic rates were fixed to the rates of Dendra2 obtained from in vitro single-molecule experiments ( Lee et al . , 2012 ) . ( Bottom ) Photoactivation time of Dendra2 fused to SpoIIIE using Fermi-photoactivation scheme corresponding to the Figure 5D . Details about the quantification of the number of molecules using PALM can be found in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 018 We reasoned that if the 61% of septa that showed single foci in the ΔσE strain ( Figure 4C , Figure 4—figure supplement 3 ) were unresolved dual foci , they should be elongated across the septum compared to each focus in septa with dual-foci . We therefore determined the width of the SpoIIIE clusters in single foci and in each resolved subcomplexes from dual foci both perpendicular and parallel to the septum . On average , the width perpendicular to the septum of the single foci ( 94 nm ) was ∼1 . 7× larger than that of each individually resolved cluster in dual foci ( 54 nm ) , while their width parallel to the septum was similar ( ∼80 nm , Figure 5C ) , suggesting that single foci are unresolved dual foci . Consistent with this hypothesis , PALM images of SpoIIIE-Dendra2 showed a larger ( 83% ) number of septa with a single focus than SpoIIIE-tdEOS ( Figure 4—figure supplement 4 ) , as expected for a dimmer fluorescent protein with reduced localization accuracy ( Lee et al . , 2012 ) . We performed two experiments to determine if the dual foci represented the architecture of the SpoIIIE complex during or after DNA translocation . First , we examined PALM images of the translocation-deficient SpoIIIE ATPase mutant ( G467S; hereafter SpoIIIEATP− ) in ΔσE background . This strain showed dual foci in 23% of septa ( Figure 4—figure supplement 5 ) , suggesting that dual foci represent the structure of SpoIIIE when it assembles around the trapped chromosome , not after DNA translocation . Second , we used PALM to localize wild type SpoIIIE-tdEOS in ΔSpoIIDMP strain in sporangia stained with both the membrane stain FM5-95 and the DNA stain DAPI , so we could discriminate between cells in which DNA translocation was ongoing and those in which it was complete . Quantification of the forespore DNA relative to the total ( mother cell plus forespore ) , demonstrated that 81% of sporangia with dual foci had not completed DNA translocation ( Figure 4D , Figure 4—figure supplement 6 ) . These data suggest that dual foci represent the SpoIIIE DNA translocation complex . We next used qPALM to determine the relative numbers of SpoIIIE molecules in each cell in ΔσE sporangia with resolved dual foci , applying an algorithm recently developed in our laboratory ( Lee et al . , 2012; see ‘Materials and methods’ for details ) . We detected , on average , 34 ± 17 molecules in the entire dual foci , consistent with the number of molecules of wild type SpoIIIE and SpoIIIEATP− in strains with thin septa ( Figure 5D , Table 1 ) . We detected 19 ± 11 and 15 ± 8 molecules in the mother cell- and forespore-proximal subcomplexes from dual foci , respectively . Thus each subcomplex contains approximately half the number of SpoIIIE molecules detected in the entire dual focus , consistent with our quantification of the fluorescence intensity after cell-specific degradation of SpoIIIE-GFP ( Figure 3B , Table 1 ) . Together these results suggest that the SpoIIIE translocation complex consist of two complexes containing equivalent numbers of SpoIIIE molecules , one in the mother cell and one in the forespore . 10 . 7554/eLife . 06474 . 019Table 1 . Summary of SpoIIIE quantification obtained by qPALM and diffraction-limited images in the wild type strain , cell-specific degradation system and the thick septum strain ( ΔσE ) DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 019SpoIIIE-dendra2 quantification by qPALM , absolute numbersSpoIIIE-GFP intensity , %Wild typeΔσE ( dual foci ) Cell-specific degradationStrainSpoIIIEWTSpoIIIEATP−MC + FMCFNoneFMCMean ( SD ) 31 ( 18 ) 31 ( 14 ) 34 ( 17 ) 19 ( 11 ) 15 ( 8 ) 100 ( 29 ) 51 ( 25 ) 52 ( 31 ) N911285151517266123 The last step of cell division is a membrane fission event that divides the septal membrane to render two physically separated daughter cells . During sporulation , septal membrane fission occurs in the presence of a trapped chromosome ( Burton et al . , 2007; Fleming et al . , 2010 ) and depends on the assembly of a stable SpoIIIE translocation complex ( Fleming et al . , 2010 ) . It has been proposed that SpoIIIE mediates membrane fission by assembling a paired channel spanning both septal membranes ( Liu et al . , 2006; Fleming et al . , 2010 ) , with both halves of the channel required to mediate and maintain septal membrane fission . To test this hypothesis , we used cell-specific SpoIIIE degradation together with fluorescence recovery after photobleaching ( FRAP ) to assess the continuity of the forespore and mother cell membranes after degrading SpoIIIE in one or the other cell . Briefly , sporangia were stained with the membrane dye FM4-64 and the dye in the forespore photobleached with a laser . If the membrane of the mother cell and the forespore are connected to form a continuous unit , FM4-64 will diffuse from the mother cell to the forespore and fluorescence will completely recover ( Figure 6A , right ) . However , if the membranes are separated and physically disjointed , FM4-64 will not diffuse and forespore fluorescence will not recover ( Figure 6A , left ) . 10 . 7554/eLife . 06474 . 020Figure 6 . Role of forespore and mother cell SpoIIIE in membrane fission . ( A ) Membrane organization in wild type ( left ) and ΔSpoIIIE ( right ) strains . SpoIIIE ( green ) blocks the FM4-64 ( red ) diffusion from the mother cell to the forespore , which remains bleached ( pale red forespore ) . Arrows show diffusion of FM4-64 to the forespore . ( B ) Corrected fluorescence recovery ( cFR ) of individual sporangia without degradation ( JLG808 ) or after degradation in the forespore ( F , JLG821 ) , mother cell ( MC , JLG823 ) or both ( F and MC , JLG825 ) . When calculating the average cFR ( black lines ) we excluded curves ( pale lines ) that resembled those without degradation because they likely represent sporangia in which SpoIIIE is not yet degraded . ( C ) Average cFR in sporangia without SpoIIIE degradation ( green , n = 27 ) or degradation in the mother cell ( blue , n = 35 ) , the forespore ( orange , n = 33 ) or both ( red , n = 33 ) . Error bars , standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02010 . 7554/eLife . 06474 . 021Figure 6—figure supplement 1 . Cell-specific degradation of SpoIIIEATP− mutant . ( A ) Visualization of SpoIIIEATP−-GFP-SsrA* ( green ) by deconvolution fluorescence microscopy . SpoIIIE was not degraded ( strain JLG808 ) or selectively degraded in the mother cell ( strain JLG823 ) , in the forespore ( strain JLG821 ) or in both ( strain JLG825 ) . The SpoIIIE foci disappear when SpoIIIE is degraded simultaneously in forespore and mother cell , but not when degraded only in one cell . Membranes were stained with FM4-64 ( red ) . Scale bar is 1 μm . ( B ) Normalized fluorescence intensity of SpoIIIEATP−-GFP-SsrA* foci without degradation ( green; JLG808 ) , or after degradation in the forespore ( red; JLG821 ) or mother cell ( blue; JLG823 ) . Foci intensities were measured in sporangia with flat septum , curved septum and engulfing . Each dot represents the focus intensity from a single sporangium , and black dotted lines represent the mean . The mean foci intensity of sporangia with flat septa in the non-degradation strain was normalized to 100 . Similarly to wild type SpoIIIE ( Figure 3B ) , the foci intensity decreases to approximately half when SpoIIIEATP−-GFP is degraded either in the forespore shortly after polar septation ( curve septa ) . The mean foci intensities , standard deviations , and number ( n ) of analyzed sporangia are indicated in the table below the graph . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02110 . 7554/eLife . 06474 . 022Figure 6—figure supplement 2 . Fluorescence recovery after photobleaching ( FRAP ) of forespore membranes . Images show the FRAP the forespore membranes of representative sporangia when SpoIIIEATP− is not degraded ( strain JLG808 ) or selectively degraded in the forespore ( F; strain JLG821 ) , in the mother cell ( MC; strain JLG823 ) , or degraded in both ( MC and F; strain JLG825 ) . The first column ( −2 s ) is before photobleaching the forespore area ( indicated by yellow box ) . Snapshots after photobleaching were taken at the indicated times . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 022 Since SpoIIIE is only essential for septal membrane fission in the presence of trapped DNA , we used SpoIIIEATP− , which supports septal membrane fission ( Fleming et al . , 2010 ) but not chromosome translocation ( Sharp and Pogliano , 1999 ) . We detected the same number of molecules in the foci of SpoIIIEATP− as the wild type ( Table 1 ) , suggesting that the organization of the complex is similar . We constructed a GFP-SsrA* tagged version of SpoIIIEATP− , which was degraded similarly to the wild-type allele ( Figure 6—figure supplement 1 ) , and performed FRAP of engulfing sporangia . Fluorescence did not recover when SpoIIIE was not degraded , indicating that the membranes are separated ( Figure 6B , C , Figure 6—figure supplement 2 ) . However , fluorescence completely recovered in most sporangia within 60 s when SpoIIIE was degraded in the mother cell , the forespore or both ( Figure 6B , C , Figure 6—figure supplement 2 ) . These results indicate that SpoIIIE subcomplexes in both cells are required for septal membrane fission , supporting the model that assembly of a paired channel contributes to membrane fission ( Figure 7C ) . Furthermore , the observation that the separated membranes before degradation were converted to continuous membranes after degradation suggests that membrane fission is reversible during DNA translocation and depends on the integrity of the paired channel . 10 . 7554/eLife . 06474 . 023Figure 7 . Role of forespore and mother cell SpoIIIE in chromssome translocation . ( A ) Visualization of DAPI-stained DNA ( blue ) after cell-specific degradation of SpoIIIE-GFP-SsrA* ( green ) . Membranes are stained with FM4-64 ( red ) . Single arrowheads show forespores containing small amounts of DNA; double arrowhead shows anucleate forespores . Scale bar , 1 μm . ( B ) Forespore DAPI intensity normalized to the total intensity ( forespore plus mother cell ) of sporangia about to complete engulfment in the indicated degradation strains . Each dot shows the normalized DAPI intensity of one forespore . Median ( red line ) and the number of sporangia ( n ) are indicated . Sporangia containing <0 . 05 DNA in the forespore were considered anucleate . ( C ) Models for SpoIIIE and membrane organization before and after cell-specific SpoIIIE degradation . During division , SpoIIIE assembles channels that separate daughter cell membranes and mediate forward chromosome translocation . Selective degradation of SpoIIIE either in the mother cell or the forespore reverses membrane fission , with the remaining molecules exporting DNA from their respective cell . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02310 . 7554/eLife . 06474 . 024Figure 7—figure supplement 1 . Chromosome translocation at different stages of engulfment . The graph shows the normalized forespore DAPI intensity of wild type ( PY79 ) and SpoIIIEATP− mutant strain ( KP541 ) for sporangia at different stages of engulfment . Normalized forespore DAPI intensity was calculated as [Forespore DAPI intensity/ ( Foresore DAPI intensity + Mother cell DAPI intensity ) ] , for details see ‘Meterials and methods’ . As engulfment progresses , the normalized DAPI intensity of the wild type sporangia increases gradually until it reaches a plateau in cells undergoing engulfment . SpoIIIE degradation in either mother cell or forespore happens before engulfment starts ( Figure 3B ) , and therefore before chromosome translocation is completed . As expected , the DAPI intensity does not increase in the translocation-defective SpoIIIEATP− mutant sporangia . The average and standard deviation of at least 25 different sporangia is represented for every engulfment stage . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02410 . 7554/eLife . 06474 . 025Figure 7—figure supplement 2 . Effect of sspBEc expression on chromosome translocation . The normalized forespore DAPI intensity was calculated for sporangia about to complete engulfment , in strains not expressing sspBEc ( PY79 ) or expressing sspBEc in the forespore ( JLG170 ) , mother cell ( JLG180 ) , or both ( JLG323 ) . SpoIIIE is not tagged with SsrA* in those strains and therefore no degradation is expected . In these strains DNA translocation is not impaired . Each dot represents the normalized DAPI intensity of a sporangium . Median ( red line ) and the number of sporangia ( n ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02510 . 7554/eLife . 06474 . 026Figure 7—figure supplement 3 . Alternative approach to measure chromosome translocation upon cell-specific degradation of SpoIIIE . ( A ) We inserted a gene encoding CFP under the control of a σF-dependent promoter ( PspoIIQ ) in the cotC locus of Bacillus subtilis chromosome ( I ) . cotC ( 168° ) is close to the dif site ( 166° ) and therefore is one of the last regions of the chromosome to be translocated into the forespore . Since the CFP gene is expressed from a σF-dependent promoter , it will not be transcribed until it reaches the forespore . Therefore , we can use the presence of CFP in the forespore as an indicator that chromosome translocation has been completed ( II ) . If chromosome translocation is impaired and the cfp gene does not reach the forespore , no CFP signal will be detected ( III ) . ( B ) Deconvolution fluorescence microscopy of sporangia containing cfp under the control of spoIIQ promoter inserted in the cotC locus , in strains in which SpoIIIE is not degraded ( JLG981 ) , degraded in the forespore ( JLG1001 ) , degraded in the mother cell ( JLG1002 ) or degraded in both ( JLG1003 ) . When SpoIIIE is not degraded or degraded only in the forespore most sporangia show CFP signal in the forespore , indicating that translocation has been completed . However , when SpoIIIE is degraded in the mother cell or in both the forespore and mother cell simultaneously , many sporangia in late stages of engulfment do not show CFP signal in the forespore ( arrowheads ) , indicating that chromosome translocation is not completed . Cells were imaged at t2 . 5 . Scale bar is 1 μm . ( C ) Percentage of forespores showing CFP signal in strains containing SsrA* untagged SpoIIIE for sporangia in different stages of engulfment . The adaptor protein SspBEc was not expressed ( EBS606 ) or expressed in the forespore ( JLG978 ) , mother cell ( JLG979 ) , or both ( JLG980 ) . Contrary to the SsrA* tagged SpoIIIE strains ( from Firugre S6D and F ) all sporangia about to complete engulfment show GFP signal in the forespore , indicating that chromosome translocation is already completed at this stage . Expression of sspBEc does not affect the rate at which CFP signal appears in the forespore . Between 25 and 216 sporangia were analyzed , depending on the strain and cell type . ( D ) Percentage of sporangia in different stages of engulfment showing CFP signal in the forespore when SpoIIIE is not degraded ( JLG981 ) or degraded in the forespore ( JLG1001 ) , mother cell ( JLG1002 ) or both ( JLG1003 ) . Degradation of SpoIIIE in the mother cell or in both the mother cell and forespore simultaneously reduces the percentage of sporangia showing CFP signal in late stages of engulfment by ∼50% . Degradation of SpoIIIE in the forespore produces a delay in the appearance of CFP signal compared to the control , but >90% of the forespores are blue by the completion of engulfment . Between 32 and 306 sporangia were analyzed , depending on the strain and cell type . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 026 We next used the cell-specific degradation system to determine if both SpoIIIE subcomplexes were necessary for chromosome translocation , taking advantage of the observation that SpoIIIE degradation occurs before chromosome translocation is complete ( Figure 7—figure supplement 1 ) . We used the SpoIIIE-GFP-SsrA* cell-specific degradation strains and quantified the amount of DAPI-stained DNA in the forespore relative to the total DNA in sporangia that were about to complete engulfment ( Figure 7A , B ) . As previously reported ( Becker and Pogliano , 2007; Ptacin et al . , 2008 ) , when chromosome translocation is completed the normalized forespore DAPI intensity is just ∼30–40% ( of the total DNA ) rather than the expected 50% ( one full chromosome in the forespore and the other in the mother cell ) , likely due to self-quenching of DAPI in the smaller volume of the forespore . In the SpoIIIEATP− mutant , where chromosome translocation is blocked , and before DNA translocation starts in the wild type , the normalized forespore DAPI intensity is 10–15% ( Figure 7—figure supplement 1 ) . Without degradation , most sporangia finished chromosome translocation before the completion of engulfment showing an average normalized forespore DAPI intensity of 37% ( Figure 7B ) . Similarly , when SpoIIIE was degraded in the forespore , most sporangia completed chromosome translocation ( normalized DAPI intensity , 38% ) . However , 26% of sporangia showed a normalized DAPI intensity <30% ( Figure 7B ) , suggesting that in some sporangia the absence of the forespore complex impaired the ability of the mother cell complex to translocate DNA . In contrast , degradation of SpoIIIE in the mother cell blocked chromosome translocation ( normalized DAPI intensity , 20% ) in most sporangia and 20% of sporangia showed anucleate forespores ( Figure 7A , B ) , indicating that in the absence of the SpoIIIE mother cell subcomplex , the forespore subcomplex can ultimately translocate the chromosome in the reverse direction . As expected , degradation in both cells blocked chromosome translocation ( normalized DAPI intensity , 21% ) and generated 6% anucleate forespores , less than when SpoIIIE was degraded only in the mother cell ( Figure 7B ) . Expresison of sspBEc in strains in which SpoIIIE was not tagged with SsrA* did not show any defect in chromosome translocation , indicating that the observed phenotypes are indeed due to the degradation of SpoIIIE molecules ( Figure 7—figure supplement 2 ) . These results support the notion that SpoIIIE functions as a DNA exporter ( Sharp and Pogliano , 2002 ) , and that although it normally operates in the mother cell , it is capable of operating in the forespore in the absence of the mother cell subcomplex ( Figure 7C ) , generating anucleate forespores . Surprisingly , our results also provide evidence that the forespore subcomplex is necessary for maximal DNA translocation efficiency , a result confirmed by quantifying the frequency with which a CFP reporter of forespore gene expression that was positioned near the terminus entered the forespore ( Figure 7—figure supplement 3 ) . We investigate here the organization and function of the SpoIIIE DNA translocation channel during B . subtilis sporulation using cell-specific protein degradation , quantitative super-resolution microscopy , and FRAP . We were able to directly visualize the assembly of SpoIIIE into two subcomplexes consisting of subunits in each daughter cell . Cell-specific protein degradation and qPALM were used to determine the relative abundance of SpoIIIE in each daughter cell , and our results indicate that each contain approximately equal numbers of SpoIIIE molecules . These observations support a model in which each chromosome arm is transported through a paired channel that spans two lipid bilayers ( Figure 8B , Liu et al . , 2006; Burton et al . , 2007; Fleming et al . , 2010 ) . The distance between the clusters in each cell was found to be 55 nm , which is larger than the 40 nm width of the septum in the strains used here , providing some support for a channel architecture in which the motor domains of each complex are within the cytoplasm of the respective daughter cell ( Figure 8B , left ) rather than within the septum ( Figure 8B , right ) , although further experiments are required to address this point . 10 . 7554/eLife . 06474 . 027Figure 8 . Architecture of the SpoIIIE complex during DNA translocation . ( A and B ) Septal cross sections showing SpoIIIE ( red ) , DNA ( blue ) and lipids ( grey ) and displaying just one DNA strand traversing the septum for simplicity , although at least two arms of the circular chromosome must cross the septum . Our results indicate that the translocation channel has approximately equal numbers of SpoIIIE motor domains in each cell , consistent with the models shown in B , but not with the simple aqueous channel model in A . They further indicate that the SpoIIIE membrane domains are arranged in such a way that they form a continuous protein barrier at the septum that blocks diffusion within the lipid bilayer and that molecules in each cell are required to form this barrier . We envision two alternative organizations to achieve this . First , the transmembrane domains might form channels in each septal membrane that interact within the septal space and conduct the DNA across the septum , with the motor domains projecting into the cytoplasm ( left ) . Second , the transmembrane domains might form a contiguous external ring that blocks lipid diffusion , with motor domains projecting towards the lumen of an aqueous channel ( right ) . The distance between the centers of the forespore and mother cell clusters ( 55 nm ) is larger than the thickness of the septum ( 40 nm , Tocheva et al . , 2013 ) leading us to prefer the paired channel model ( left ) . ( C ) Each chromosome arm is translocated by a different SpoIIIE channel that is comprised of opposing motor domains that have the ability to export DNA from their respective cell . Translocation directionality is established by the interaction between SRS and the SpoIIIE γ domain , which activates the mother cell motor ( green ) , and inactivates the forespore motor ( grey ) . Consequently , the mother cell-SpoIIIE exports DNA until the final loop ( ter ) reaches the septum . We propose that translocation of the loop generates tension that destabilizes and opens the channels forming a larger pore to pass the loop . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02710 . 7554/eLife . 06474 . 028Figure 8—figure supplement 1 . Statistical analysis of qPALM data . A detailed analysis of qPALM data reveals the existence of two populations of sporangia with different numbers of SpoIIIE monomers at the septum . ( A and B ) Comparison between two different analyses—Gaussian Kernel Density Estimator ( GKDE ) and Variational Bayesian Gaussian Mixture Model ( VBGMM ) —for quantification of SpoIIIE in ( A ) the wild type strain and ( B ) the ΔσE strain . Both analyses lead to the finding of two populations with the peak values ∼24 and 50 ( red dotted lines ) . ( Top panels from A and B ) GKDE ( blue curves ) of the SpoIIIE counting data ( blue bars ) were fitted well by a mixture of two Gaussian functions ( red solid curves ) . See ‘Materials and methods’ for details on the analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 02810 . 7554/eLife . 06474 . 029Figure 8—figure supplement 2 . Model for chromosome translocation if four DNA strands cross the septum . Septum containing four arms of DNA still can translocate the entire chromosome toward the forepore . The orientation of SRSs along the chromosme and their interaction with SpoIIIE is such that three arms are translocated toward the forespore by the mother cell-SpoIIIE while one arm is translocated into the mother cell by the forespore-SpoIIIE . Therefore , the net translocation is toward the forespore . DOI: http://dx . doi . org/10 . 7554/eLife . 06474 . 029 The stability of the SpoIIIE complex during DNA translocation ( Fleming et al . , 2010 ) allowed us to use qPALM ( Lee et al . , 2012 ) to image and quantify SpoIIIE in living cells . However , it is important to note that imaging live rather than fixed cells would likely overestimate the dimensions of the complex if SpoIIIE moves during DNA translocation , which could explain why the measured width of the foci ( 80 nm ) is wider than expected for two side by side hexamers ( ∼24 nm ) . Furthermore , the spatial resolution of PALM ( ∼25 nm ) remains significantly larger than the size of most macromolecular assemblies , and it is larger than the distance between the two faces of the septum in wild type cells ( ∼20 nm ) . We partially overcame the current limitations of PALM by using two genetic tools , the thick septum mutant which increased the distance between the subcomplexes in each cell , and cell-specific protein degradation which improved our ability to discriminate between molecules in each cell . The method to extract absolute numbers using PALM may depend on other factors , such as label efficiency and inactivation ( Lee et al . , 2012; Puchner et al . , 2013; Durisic et al . , 2014 ) . Although our qPALM approach balances the over- and undercounting errors ( Lee et al . , 2012 ) , the numbers of SpoIIIE molecules determined here may represent a lower bound to their true number if some of the photoactivatable fusion domains did not fold properly at the foci . Nevertheless , it is tantalizing to note that our counting method ( Lee et al . , 2012 ) indicated that the SpoIIIE complex contains 31 ( ±18 ) SpoIIIE molecules , enough to form at least two dodecameric channels , one for each chromosome arm ( 24 molecules in total , Table 1 ) . Furthermore , a more detailed analysis of qPALM data revealed the existence of two populations of sporangia with different numbers of SpoIIIE monomers at the septum ( ∼25 and ∼50; Figure 8—figure supplement 1 ) , suggesting that some translocation complexes have enough molecules to assemble two channels while others could assemble four ( Figure 8—figure supplement 2 ) . Therefore , we speculate that SpoIIIE organizes around an even number of DNA arms ( two or four ) during sporulation ( Figure 8—figure supplement 2 ) , as would be expected if a circular chromosome crossed the septum one or two times , respectively . Cell-specific protein degradation was used both to quantify and to functionally dissect the role of SpoIIIE molecules in each cell . Our results demonstrate that SpoIIIE is required in both cells to maintain separate septal membranes during DNA translocation and that septal membrane fission is reversible in the presence of trapped DNA ( Figure 6 ) . The simplest explanation for this finding is that transmembrane domains in each cell assemble a structure that excludes lipids and blocks membrane diffusion between the two cells . This scenario could occur if during assembly of a translocation complex , SpoIIIE membrane domains in the mother cell and the forespore dock via their extracellular domains and the membrane domains of each subunit—within each bilayer—compact into their hexameric form , excluding lipids to form a paired central channel ( Almers , 2001; Peters et al . , 2001; Liu et al . , 2006; Fleming et al . , 2010 ) . It is possible that docking is indirect and requires additional , accessory proteins that connect the extracellular domains of SpoIIIE . In addition to the data presented here , the paired channel model for septal membrane fission is supported by our previous demonstration that septal membrane fission depends on the SpoIIIE transmembrane domain ( Sharp and Pogliano , 2003 ) and its large extracellular loop ( Liu et al . , 2006 ) and on the assembly of a compact focus ( Fleming et al . , 2010 ) . In contrast to membrane fission , which requires SpoIIIE in both cells , only mother cell SpoIIIE is essential for DNA segregation , demonstrating that SpoIIIE functions as a DNA exporter ( Sharp and Pogliano , 2002 ) . While the assembly of SpoIIIE in both sides of the septum provides a structural scaffold for membrane fission , it poses a topological challenge for directional chromosome translocation . Prior to the establishment of directional DNA translocation , both complexes could , in principle , export DNA out of its respective compartment ( Sharp and Pogliano , 2002; Becker and Pogliano , 2007; Ptacin et al . , 2008 ) ( Figure 7A , B ) , leading to a potential clash between complexes in each cell; as a result , these complexes would compete to translocate DNA either out of or into the forespore . However , our cell-specific degradation data show that chromosome translocation happens most efficiently when both complexes are present ( Figure 7B , Figure 7—figure supplement 3 ) , suggesting that rather than a competition , there is a potentiation between the two hexamers of a channel . We therefore propose that the hexamers within a dodecameric channel should be considered as an integral entity ( Figure 8C ) . In this model , docking of the transmembrane domains of hexamers in opposite septal membranes might contribute to the robustness of chromosome translocation in two ways . First , it might stabilize the translocation complex , distributing the tension produced by the movement of the chromosome between the two septal membranes and insulating the DNA from interactions with the membrane or periplasm . Second , it might facilitate the deactivation of the forespore motor domains after activation of the mother cell motor domains by encountering SRSs in the permissive orientation ( Besprozvannaya et al . , 2013; Cattoni et al . , 2013 ) . Some members of the SpoIIIE/FtsK/HerA superfamily , including SpoIIIE and FtsK in bacteria ( Burton and Dubnau , 2010 ) , and possibly HerA in archaea ( Iyer et al . , 2004 ) , are involved in the post-septational segregation of chromosomes when one of them is accidentally trapped in the septum during vegetative cell division . Since the trapped chromosome can belong to either daughter cell , a paired channel similar to the one described here could function as a bidirectional translocation machinery guided by SRS-like sequences that would allow the transport of the chromosome to the appropriate compartment . So , by analogy to SpoIIIE , we hypothesize that the transmembrane domain of these proteins might also form a paired channel crossing both septal membranes . All the strains used in this study are derivatives of B . subtilis PY79 . The strains , plasmids and oligonucleotides used in this study , can be found in Supplementary file 1A–C . The amino acid sequence of the linker between GFP-SsrA* and the target protein is provided in Supplementary file 1D . All the SpoIIIE fusion proteins used here support sporulation with wild type efficiency ( Supplementary file 1E ) . Sporulation was induced by resuspension ( Sterlini and Mandelstam , 1969 ) at 37°C , except the bacteria were grown in 20% LB prior to resuspension ( Becker and Pogliano , 2007 ) . Cells were sporulated in presence of 25 ng/ml membrane dye FM5-95 ( Life Technologies , Waltham , Massachusetts ) and harvested 1 hr and 45 min ( t1 . 75 ) after resuspension . 5 μl of cell suspension mixed with gold fiducial particles was spotted in a poly-L-Lys ( Sigma–Aldrich , St . Louis , Missouri ) coated coverslip and covered with a second coverslip to generate a cell monolayer ( Fleming et al . , 2010 ) . To minimize the over- and under- quantification of PA-FP molecules: first , the sample was illuminated with the excitation laser ( 561 nm , 22 mW/mm2 ) simultaneously with the activation laser ( 405 nm ) whose power varied ( from 0 to 9 . 1 mW/mm2 ) in time according to the Fermi activation scheme ( Lee et al . , 2012 ) with parameters tF = 3 . 2 min and T = 10 s; and second , we employed our previously developed optimal tc method ( Lee et al . , 2012 ) . PALM data was taken and analyzed using our custom built microscope and custom written Matlab program ( Fleming et al . , 2010; Lee et al . , 2012 ) . Single molecule counting was performed in strains with Dendra2 fusion protein . Quantitative single-molecule-counting PALM is based on the previously established method ( Lee et al . , 2012 ) and is carried out through the following steps: ( 1 ) obtain the photoactivation rate , blinking rate , and photobleaching rate of a PA-FP that will be used as a fluorescent marker , for instance , through in vitro single-molecule characterization , ( 2 ) acquire PALM raw movie data using Fermi-photoactivation scheme , which assures photoactivation of all the PA-FP molecules within a given experimental time at the almost constant rate ( Figure 5—figure supplement 2 ) , ( 3 ) analyze the raw data to localize the appearance of fluorescence bursts in space and time and render a reconstructed PALM image , ( 4 ) select a ROI , for example a focus of SpoIIIE , from the visual inspection of the PALM image , and ( 5 ) apply the optimal-τc counting ( Matlab code available in Source code 1 also at https://github . com/shyuklee/pafpcluster ) to the ROI to obtain the blinking-corrected number of molecules inside the ROI . The in vitro photophysics of Dendra2 was well described by four states: Nonactive ( N ) , Active ( A ) , Dark ( D ) , and photoBleach ( B ) ; and the rates: N to A ( ka ) , A to B ( kb ) , and A to D ( kd ) . The fluorescence recovery , D to A , was well explained by two different rates ( kr1 and kr2 ) with their population ratio ( α ) , which implies the existence of at least two dark states . The photoactivation rate ka can be externally changed by the intensity of UV light whereas all the other rates are constants fixed by the intrinsic molecular properties of Dendra2 . In order to optimize the temporal separation of molecules we changed ka in time such that a Dendra2 molecule gets photoactivated during a given experimental time , tF , with an almost uniform probability distribution . To account for the overcounting error induced by multiple transitions between the state A and the state D , we introduced a tolerance time of the dwell in the dark states , τc: if the fluorescence at the same location recovers within a given τc then the two events are considered due to a blink of an identical Dendra2 molecule rather than due to two independent photoactivation events of two different molecules . However , the introduction of τc also induces molecular undercounting error because it is probabilistically not impossible that two different Dendra2 molecules actually get photoactivated one after the other within a time interval shorter than the given value of τc . We can achieve the unbiased molecular counting by balancing the overcounting and undercounting through a careful selection of the value of τc . This optimal τc depends not only on all the rate constants but also on the actual number of Dendra2 molecules ( Nmol ) that we aim at counting by PALM . In the previous work , we obtained the optimal τc by stochastic simulations , but this method was computationally too costly to explore the extensive parameter space defined by the rate constants and Nmol ( Lee et al . , 2012 ) . We instead developed an approximate analytic solution of the optimal τc for instantaneous calculation . If we assume that Nmol molecules are independently photoactivated uniformly in time between 0 and tF , then the probability density function of the time lag , Δti , between the two successive ith and ( i + 1 ) th photoactivation events is given by the well known order statistics of the uniform distribution: ( 1 ) p ( Δti ) =Nmol ( 1−Δti/tF ) Nmol . In order to estimate the undercounting error , we apply a rule of counting for a given τc such that ith and ( i + 1 ) th photoactivations are combined together to be counted as one molecule if Δti < τc . Then the mean of undercounted number , denoted by ⟨Nu⟩ , can be calculated as follows: ( 2 ) ⟨Nu⟩=Nmol− ( 1+∑i=1Nmol−1 ∫0τc p ( Δti ) dΔti ) , ( 3 ) = ( Nmol−1 ) [1− ( 1−τc/tF ) Nmol] . The overcounting due to blinking can be estimated using the geometrically distributed probability of the number of transitions that a single Dendra2 molecule makes between the state A and the state D with the dwell in the state D being longer than τc before photobleaching ( Lee et al . , 2012 ) : ( 4 ) P{Nblink ( τc ) =n}=η¯n ( 1−η¯ ) , where ( 5 ) η¯=pτcη1−η+pτcη and pτc=e−kr1τc+αe−kr2τc1+α . Then the mean of total overcounted number of independently blinking Nmol molecules , denoted by ⟨No⟩ , is given by ( 6 ) ⟨No⟩=Nmol ∑n=0∞ P{Nblink ( τc ) =n} , ( 7 ) =Nmol η¯1−η¯ . Note that ⟨Nu⟩ and ⟨No⟩ are not the exact solutions but just approximate estimates of the undercounting and overcounting because they account for only the situation when the ith photoactivated molecule photobleaches much earlier than the next photoactivation of another molecule . Nevertheless , they are the dominant contributions to the counting error and we can expect to obtain the optimal τc rather accurately from the balance condition , ⟨Nu⟩ = ⟨No⟩ , which results in an analytic equation for the optimal τc: ( 8 ) h ( τc; Nmol , tF , kd , kb , kr1 , kr2 , α ) =0 , where ( 9 ) h ( τc; Nmol , tF , kd , kb , kr1 , kr2 , α ) ≡1− ( 1−τc/tF ) Nmol−pτckdkbNmolNmol−1 . This approximate analytic solution agrees well with the result from the stochastic simulation over experimentally realistic range of Nmol and tF when the rates are fixed to the rates of Dendra2 obtained from in vitro single-molecule experiments ( Figure 5—figure supplement 2 ) . To count the number of SpoIIIE-Dendra2 molecules inside a cluster , we iterated the feedback between τc and Nmol using the in vitro Dendra2 kinetic rates and the analytic solution of optimal τc ( Nmol ) until convergence ( Lee et al . , 2012 ) . Samples from sporulating cell cultures were taken 2 . 5 hr after resuspension ( t2 . 5 ) and added to agarose pads supplemented with 0 . 5 μg/ml of FM4-64 ( Life Technologies ) for membrane visualization and 40 ng/ml of DAPI ( Invitrogen , Waltham , Massachusetts ) for DNA visualization . Images were collected using an Applied Precision optical sectioning microscope equipped with a Photometrics CoolsnapHQ2 camera using identical exposure times for each sample . Images were deconvolved and analyzed with SoftWoRx version 5 . 5 ( Applied Precision , Issaquah , Washington ) . Sporangia with flat , curved , and engulfing septa ( as defined in Sharp and Pogliano , 1999 ) from 2–4 microscopy fields were quantified . To determine SpoIIIE-GFP-SsrA* foci intensities , GFP intensities of eight optical sections covering a total thickness of 1 . 05 μm within the cells were summed using SoftWoRx Z-projection tool . The intensity of each projected GFP focus was determined by drawing a ∼150 nm2 circumference centered in the middle of the focus , subtracting the average background intensity for each field . The average focus intensity after SpoIIIE-GFP degradation in the mother cell or in the forespore was normalized to the average intensity when SpoIIIE-GFP was not degraded . To determine the fluorescence intensities of GyrA-GFP-SsrA* and SigA-GFP-SsrA* in the forespore and mother cell , pixel intensities of four optical sections from deconvolved images covering a total thickness of 0 . 45 μm were summed . Mean GFP intensities of the forespore and the mother cell were determined separately by drawing a polygon encompassing the whole area of every cell . The ratios were calculated after subtracting the mean background intensity . For each graph , values were made relative to the average ratio of cells with flat septum in the non-degradation strain . To determine the degree of DNA translocation , DAPI pixel intensities of four optical sections covering a total thickness of 0 . 45 μm were summed . Samples were taken 3 hr after resuspension ( t3 ) , and DAPI intensities of forespore ( F ) and mother cell ( MC ) from sporangia about to complete engulfment were determined separately by drawing a polygon encompassing the whole DNA area of every compartment . After subtraction of the average background intensity , the normalized DAPI intensity [F intensity/ ( F intensity + MC intensity ) ] was determined for each sporangium . FRAP of forespore membranes was performed as described ( Fleming et al . , 2010 ) . Briefly , cells were sporulated in the presence of 2 μg/ml FM4-64 . After 2 . 5 hr , cells were washed three times with sporulation medium and placed onto 1 . 2% agarose pads . Forespore membranes were bleached with 0 . 3 s pulse from a 488-nm argon laser set to 30% power , and membrane images collected at appropriate time intervals . FRAP quantification was performed as described ( Fleming et al . , 2010 ) .	Bacillus subtilis is a bacterium that lives in the soil and is related to the bacteria that cause the diseases anthrax and botulism in humans . When nutrients are scarce , these bacteria can change into a dormant form called spores , which can withstand harsh environmental conditions . The spores can remain dormant for thousands of years until the conditions improve enough to allow the bacteria to grow again . During ‘sporulation’ , the membrane that surrounds the bacterium pinches inward near one end of the cell to produce a large mother cell and a smaller forespore . The spore DNA becomes trapped at the site of the division so that the forespore contains only about a third of the DNA of a normal cell . The remaining two thirds lie within the mother cell , and a protein called SpoIIIE is needed to pump this DNA into the forespore . Previous studies have shown that several SpoIIIE proteins team up to form a ‘complex’ in the membrane that moves the DNA and separates the two cells , but the precise arrangement of SpoIIIE inside cells remained unclear . Here , Shin , Lopez-Garrido , Lee et al . studied how SpoIIIE is organized in living B . subtilis cells , using fluorescent labels to observe the position of SpoIIIE proteins under a microscope . The experiments show that SpoIIIE is arranged as two smaller complexes , one in the mother cell and one in the forespore , each with an equal number of SpoIIIE proteins . This suggests that SpoIIIE assembles into a channel that connects the mother cell and forespore . To investigate the role of each complex , Shin , Lopez-Garrido , Lee et al . developed a technique called ‘cell-specific protein degradation’ , to destroy SpoIIIE proteins in either the mother cell or the forespore . These experiments show that only the mother SpoIIIE complex is required to move DNA into the forespore , although DNA moves more efficiently when both complexes are present . Furthermore , when SpoIIIE is only present in the forespore , DNA moved out of this cell and into the mother cell . In contrast , both the mother cell and forespore SpoIIIE are required to separate the membranes of the mother cell and forespore . Shin , Lopez-Garrido , Lee et al . 's findings suggest that SpoIIIE molecules in both cells cooperate to efficiently move DNA into the forespore and to separate the membranes . Further work is required to understand the nature of this cooperation and to determine if similar proteins in other organisms assemble in the same way .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Visualization and functional dissection of coaxial paired SpoIIIE channels across the sporulation septum
Fast directional growth is a necessity for the young seedling; after germination , it needs to quickly penetrate the soil to begin its autotrophic life . In most dicot plants , this rapid escape is due to the anisotropic elongation of the hypocotyl , the columnar organ between the root and the shoot meristems . Anisotropic growth is common in plant organs and is canonically attributed to cell wall anisotropy produced by oriented cellulose fibers . Recently , a mechanism based on asymmetric pectin-based cell wall elasticity has been proposed . Here we present a harmonizing model for anisotropic growth control in the dark-grown Arabidopsis thaliana hypocotyl: basic anisotropic information is provided by cellulose orientation ) and additive anisotropic information is provided by pectin-based elastic asymmetry in the epidermis . We quantitatively show that hypocotyl elongation is anisotropic starting at germination . We present experimental evidence for pectin biochemical differences and wall mechanics providing important growth regulation in the hypocotyl . Lastly , our in silico modelling experiments indicate an additive collaboration between pectin biochemistry and cellulose orientation in promoting anisotropic growth . In order for most dicot seedlings to emerge from the soil successfully , the hypocotyl must grow rapidly and anisotropically ( Baskin and Jensen , 2013 ) . Such tissue anisotropy is exhibited in many plant organs when directionality is key: roots moving through the soil , stems reaching upwards and climbing tendrils ( Baskin , 2005 ) . Anisotropy , in terms of differential growth , is defined as the relative change in principal dimensions over time; for example , in the young hypocotyl , there is an increase in length versus width . In the Arabidopsis hypocotyl , the direction of anisotropy ( upwards ) is relatively fixed but the magnitude of growth anisotropy ( how fast ) is presumed to change over time ( Gendreau et al . , 1997 ) . This presumption is based upon measurements of cell length over time which indicate that a ‘wave’ of elongation runs acropetally from the base of the organ towards the cotyledons ( Gendreau et al . , 1997 ) . Plant cells are contained within a stiff cell wall thus the cell wall must change to allow growth of cells and , ultimately , organs ( Braybrook and Jönsson , 2016 ) . With respect to cellular anisotropy , growth may be generated by a cell wall which yields to ( or resists ) forces in a spatially differential manner ( Baskin , 2005 ) . The cell wall is a complex material with a fibrillar cellulosic backbone within a pectin-rich matrix ( Cosgrove , 2016 ) . In the alga Nitella , cell wall structure has been proposed to regulate anisotropy through the coordinated orientation of cellulose fibers within the wall: circumferential wrapping of cellulose fibers restricts transverse growth and the passive ( or active ) separation of fibers allows axial growth leading to anisotropy ( Green , 1960; Probine and Preston , 1962 ) . This organization within a material , gives rise to directionally differential yielding to force and would make the wall material an anisotropic material . Material anisotropy can be tested by applying external force sequentially along two perpendicular directions and measuring the difference in yield: an anisotropic material would yield differently in the two directions . Consistent with this concept cellulose fiber orientation has been correlated with material anisotropy in Nitella ( Probine and Preston , 1962 ) and in epidermal cells of onion and Kalanchoe leaves ( Kerstens et al . , 2001 ) . It is attractive to imagine that every cell within an anisotropically growing organ would display cellulose orientation perpendicular to growth , like Nitella . Indeed , this has been demonstrated in maize and Arabidopsis roots , the wheat leaf epidermis , rice coleoptiles , soybean hypocotyls and onion scales ( Baskin et al . , 1999; Paolillo , 1995 , Paolillo , 2000; Verbelen and Kerstens , 2000; Pietra et al . , 2013 ) . However , there are many exceptions where the net cellulose orientation in the outer wall of the epidermis of elongating cells was not perpendicular to the axis of growth . These include rice and oat coleoptiles , Arabidopsis hypocotyls and roots , pea epicotyls and dandelion peduncles ( Paolillo , 2000; Verbelen and Kerstens , 2000; Iwata and Hogetsu , 1989; Roelofsen , 1966 ) . Cortical microtubule orientation may act as a proxy for newly-deposited cellulose orientation as in most cases they correlate strongly . Although some exceptions exist in root cells ( Himmelspach et al . , 2003; Sugimoto , 2003 ) , the correlation has been very well documented in the case of Arabidopsis hypocotyls where microtubules , cellulose-synthase complex movement and cellulose microfibrils orientation are correlated in epidermal cells ( Paredez et al . , 2006 ) . Most recently , transversely aligned microtubule orientation was observed in Arabidopsis hypocotyls on the inward facing epidermal cell walls and those of inner cortical tissues , while the outer face of the epidermis presented as unaligned ( Crowell et al . , 2011; Peaucelle et al . , 2015 ) . These data do not necessarily negate the hypothesis from Nitella , but instead underline the possibility that different cells in different tissues contribute to anisotropy differently . In complex multi-cellular organs like the hypocotyl it may not be necessary for each individual cell to mimic Nitella . Cortex cells exhibiting transverse cellulose or microtubule orientation could provide anisotropy to the epidermis through their physical connection . This sharing of information is consistent with the epidermal growth theory ( for examples and reviews see [Baskin and Jensen , 2013; Kutschera , 1992 , Kutschera , 2008; Kutschera and Niklas , 2007] ) : in growing plant organs , internal tissues can provide the force for growth but the act of growth only occurs once the epidermis , holding the tension , releases ( Baskin and Jensen , 2013 ) . To our knowledge , transverse microtubule or cellulose orientation in inner tissues alone has not been experimentally perturbed and so this hypothesis remains unconfirmed . While it is often assumed that cellulose orientation alone confers anisotropy , experimental evidence points to further complexity . Disruption of cellulose orientation has mixed effects on cell-shape anisotropy: treatment with cellulose synthesis inhibitors reduces cell anisotropy in roots and hypocotyls ( Desprez et al . , 2002; Heim et al . , 1991 ) with a developmentally stage-specific magnitude ( Refrégier et al . , 2004 ) ; the mutant botero/katanin has defects in microtubule orientation and shows reduced cell length but maintains some anisotropy ( Bichet et al . , 2001 ) ; mutations in cellulose synthase complex subunits cause a decrease in cell and organ length , but again some anisotropy is maintained ( Refrégier et al . , 2004; Chen et al . , 2003; Fagard , 2000; Fujita et al . , 2013 ) ; in some mutants early growth is normal when compared to wild-type ( prc1-1 [Refrégier et al . , 2004] ) . These subtleties strongly indicate that there may be more to tissue anisotropy than cellulose orientation alone ( Baskin , 2005 ) . The pectin matrix of the cell wall arises as a strong candidate for regulating anisotropic growth as the transition from slow to rapid growth has been hypothesized to involve changes in pectin chemistry ( Pelletier et al . , 2010 ) . It has recently been proposed that differential pectin rigidity , within individual epidermal cells , might dictate the onset of anisotropy ( in the absence of epidermal MT orientation ) ( Peaucelle et al . , 2015 ) . Still , a quantitative understanding of the contribution of wall anisotropy via cellulose fibers and wall heterogeneity via pectin biochemistry is lacking . Here , we use an interdisciplinary approach to address the question: how is anisotropic growth in the hypocotyl achieved through cell wall mechanics ? Historically , analysis of cell-level growth in dark-grown Arabidopsis hypocotyls focused on cell length alone ( Gendreau et al . , 1997 ) yet cell width over time is an important parameter for analysis of anisotropy . In order to deepen our understanding of anisotropic growth in the dark-grown hypocotyl , we undertook a detailed analysis of cell length and width from 0 hr post-germination ( HPG ) to 72HPG . Arabidopsis hypocotyls expressing a plasma-membrane YFP-tagged marker ( Willis et al . , 2016 ) were synchronized by selecting seeds at germination ( T0; radical emergence from endosperm ) . At 6 hr intervals , 20 seedlings were sampled for confocal imaging , with a new set being imaged at each timepoint since confocal imaging stopped dark-grown hypocotyl elongation . In order to focus on cell growth , analyses were restricted to epidermal files with no division; we observed that some files underwent transverse anticlinal divisions during the first 24HPG ( Figure 1—figure supplement 1; files lacking GL2 expression ) . Since there were no divisions in GL2-expressing files , cell indices were assigned by position along the hypocotyl . Non-dividing files had 17–19 cells starting from the collet ( the transition zone between the root and the hypocotyl ) and ending at the cotyledons base ( Figure 1A ) . Utilizing this system , cell geometry could be analyzed in time and by cell position , to approximate cell-level growth dynamics . Our analyses revealed that while cell length increased in an acropetal wave consistent with the literature ( Gendreau et al . , 1997; Peaucelle et al . , 2015; Refrégier et al . , 2004; Pelletier et al . , 2010 ) , cell width increased more slowly and evenly along the hypocotyl length ( Figure 1B , Figure 1—figure supplement 1 ) . These observations were consistent with differential regulation of axial and radial cell expansion . Also , in our system all hypocotyl cells were geometrically anisotropic from the time of germination , irrespective of position along the hypocotyl ( Figure 1C; shape anisotropy , ratio of cell length to width ) . Calculations of relative growth rates for each cell index , over 6H intervals , revealed that cell length was always increasing at a higher rate than cell width ( Figure 1D and Supplemental file 1; RGR ( L ) vs RGR ( W ) ) . Relative growth rate for cell length ( RGR ( L ) ) by position clearly demonstrated the movement of the acropetal wave , beginning around 24HPG ( Figure 1D , light blue arrows ) . Interestingly , at very early time intervals ( 0-12HPG ) there was a higher RGR ( L ) and a suppression of RGR ( W ) in basal hypocotyl cells , potentially as a holdover from germination ( Bassel et al . , 2014 ) ( Figure 1D; blue and red arrowheads respectively ) . The RGR ( L ) of cells within the acropetal wave was relatively constant ( 8 . 95% ± 0 . 56 per hour; Figure 1D , Supplementary file 1 ) , indicating that there was a transition from slow to rapid elongation within the wave but that growth rate did not increase over time as the wave moved . In contrast , the RGR ( W ) was , after the initial suppression , remarkably constant in time and space ( 2 . 6% ± 0 . 2 per hour; Figure 1D , Supplemental file 1 ) . Our data paint an accurate picture of dark-grown hypocotyl growth: cells are geometrically anisotropic at germination , their growth is always anisotropic , and the acropetal wave is only evident in the elongation of cells but not their expansion in width . Transverse cellulose orientation , or its proxy microtubule orientation , is commonly invoked to explain the mechanism of anisotropy . As hypocotyl cells displayed growth anisotropy from the time of germination , we examined whether they also exhibited transverse microtubule orientation . Microtubules were visualized by confocal microscopy imaging of hypocotyl basal epidermal cells . Imaging of 35S::GFP-MAP4 ( Marc et al . , 1998 ) in dark-grown hypocotyls was conducted in short periods after exposure to light to prevent light induced reorientation , MT rearrangements or rotary movements as previously reported ( Chan et al . , 2007; Lindeboom et al . , 2013; Sambade et al . , 2012 ) . MT images were recorded at 0HPG , 24HPG and 65HPG; representing the time of germination , the transition to rapid growth and the phase of rapid growth , respectively . MT angle was determined using MicroFilament Analyzer ( Jacques et al . , 2013 ) . Both the inner and outer epidermal faces of hypocotyl cells were imaged , when possible , as they have been shown to exhibit different patterns of MT angles ( Crowell et al . , 2011 ) . At germination , basal epidermal cells exhibited a wide range of MT angles on their outer epidermal face with a slight transverse tendency ( ~30% presenting transversely between 0 and 10°; Figure 2A , B ) . We were unable to image deeper at this stage , likely due to the dense cell contents scattering the excitation and emission light . By 24HPG , the average angle on the outer face was slightly axial with ~31% of MTs being oriented between 80–90° ( Figure 2A , B ) . These data correlate well with those recently reported for cells below the cotyledons before elongation in older hypocotyls ( Crowell et al . , 2011 ) . Similar patterns at 24HPG were observed with two other microtubule markers ( GFP-TUA6 and GFP-EB1 ) and GFP-CESA3 ( Figure 2—figure supplement 1; [Chan et al . , 2003; Desprez et al . , 2007; Mathur et al . , 2003; Ueda et al . , 1999] ) . At 24HPG , we could not observe MT signal at the inner epidermal face , but could at the adjacent cortex cell faces ( a phenomenon consistent across all three marker lines , at this early stage ) . MTs at cortex cell faces appeared more transversely aligned and presented ~41% between 0 and 10° ( Figure 2A , B ) . From these data we concluded that MTs at the outer epidermal face were weakly transverse at the time of germination and those in inner cortical tissues were more strongly transverse by 24HPG . By 65HPG , when hypocotyl cells were rapidly elongating , the outer epidermal face exhibited a MT angle trend towards axial alignment ( Figure 2A , B; ~32% between 80–90° ) . It is possible that this was due to the upcoming growth arrest these cells would soon experience . It is equally probable that this is the general angle trend seen on the outer epidermal face in the early stage of elongation at the hypocotyl base . The inner epidermal face of 65HPG cells did show MT signal and exhibited an transverse angle distribution ( Figure 2A , B; ~44% between 0 and 10° ) . These data led us to conclude that during anisotropic growth from the time of germination , MT-based anisotropy information likely came from cortex cells or inner epidermal faces . This conclusion is consistent with analyses in older hypocotyls ( Crowell et al . , 2011; Peaucelle et al . , 2015 ) . While it is difficult to compare values across experiments and imaging conditions , our values for percent microtubules presenting ‘transverse angles’ were weak compared to those reported previously ( Crowell et al . , 2011 ) . This may mean that at these early stages , anisotropy from MTs is weak and only consolidates later through mechanical feedback ( Hamant et al . , 2008; Sampathkumar et al . , 2014 ) ; however , we note that there is no quantitative functional data relating the degree of anisotropic growth , the degree of MT alignment , and the dynamic variability of these parameters . An attractive hypothesis is that in early dark-grown hypocotyl elongation , MT-based anisotropy and pectin-based elastic asymmetry work cooperatively to regulate anisotropy . Elastic asymmetry in hypocotyl epidermal cells is proposed to regulate anisotropic growth and has been attributed to the presence of more calcium cross-linked homogalacturonan ( HG ) epitopes in elongating walls ( Peaucelle et al . , 2015 ) . Since it was unclear how calcium cross-linked HG might facilitate either increased elasticity or increased growth , we undertook a broader examination of HG biochemistry on an expanded time frame . We performed cell-wall immunolocalizations on longitudinal sections of 4HPG and 24HPG hypocotyls to determine the distribution of methylated HG , de-methylated HG , and calcium cross-linked HG ( using LM20 , LM19 , and 2F4 antibodies respectively ) . At 4HPG , slower-growing epidermal transverse walls were marked by 2F4 and LM19 indicating the presence of de-methylated HG and calcium cross-linked HG ( Figure 3A , orange arrow heads; see inset for naming convention ) . The endosperm at this stage was also highly marked consistent with the literature ( Müller et al . , 2013 ) but without asymmetry ( Figure 3A , red asterisk ) . In epidermal cells , the faster growing axial walls were marked with LM20 indicating the presence of methylated pectin ( Figure 3A , gold arrow heads ) . At 24HPG , the asymmetry in 2F4 and LM20 antibodies was maintained , but the LM19 antibody marked both axial and transverse walls ( Figure 3A; immunolocalization controls can be found in Figure 3—figure supplement 1 ) . The lack of asymmetry in LM19 signal at 24HPG may indicate that as pectin is newly deposited it is de-esterified but only cross-linked in transverse walls; de-methylated HG in axial walls may be degraded ( Rui et al . , 2017; Xiao et al . , 2014 ) . When combined with the asymmetry in methylated HG signal ( LM20 ) , which could be due to differential delivery , these data hint at a complex cellular delivery mechanism . We concluded that slowly growing transverse walls had more de-methylated pectin which was calcium cross-linked , while faster growing axial walls had more methylated pectin . We detected asymmetry in pectin biochemistry at 4HPG , consistent with our growth data indicating that hypocotyl cells were always growing anisotropically ( Figure 1 ) . A recent study postulated that a reported transition from isotropic to anisotropic growth in dark-grown Arabidopsis hypocotyl cells was due to the appearance of a cell-level elastic asymmetry around 15HPG ( Peaucelle et al . , 2015 ) . To investigate cell wall elasticity under our conditions , where growth was anisotropic from germination onwards , we performed AFM-based nano-indentation on basal hypocotyl epidermides from dark-grown seedlings at 4HPG and 24HPG . At 4HPG , axial walls were more elastic ( lower indentation modulus ( IM ) ) when compared with transverse walls ( 18 . 4 MPa ± 1 . 9 vs 24 . 8 MPa ± 2 . 0; Figure 3B , Figure 3—figure supplement 1 ) . This correlated well with our observations that cells at this early time point were growing anisotropically and presented asymmetric pectin epitopes ( Figure 1BC , Figure 1—figure supplement 1 ) . This difference was still observed in basal cells at 24HPG coincident with an increase in overall elasticity when they were entering into the rapid growth phase ( 12 . 7 MPa ±0 . 4 vs 20 . 0 MPa ±0 . 8; Figure 3B; Figure 3—figure supplement 1 ) . At 24HPG , cell wall thickness was not significantly different between axial and transverse walls of basal hypocotyl cells , indicating that elasticity difference were underlain by biochemical and not geometrical differences ( Figure 3—figure supplement 1 ) . The elastic asymmetry increased to a ratio of 2 by 48HPG ( Figure 3—figure supplement 1 ) . It is possible that the increase in overall elasticity contributed to the shift to rapid growth observed at 24HPG and the start of the acropetal wave ( Figure 1 ) . From these data , it was apparent that a cell-level asymmetry in wall elasticity was present from the time of germination , coincident with growth anisotropy , and correlated with changes in pectin chemistry in dark-grown hypocotyl basal epidermal cells . It has been proposed that pectin asymmetry alone might account for a shift to anisotropic growth . To test whether pectin asymmetry could induce anisotropic growth we next performed an in silico test . We developed a finite element method ( FEM ) model of a hypocotyl epidermis ( based on methods in [Bozorg et al . , 2014 , Bozorg et al . , 2016] ) . The FEM model consisted of a 3D epidermal layer made up of individual cells whose individual walls could have separate mechanical properties proscribed ( Figure 4A; Appendix ) . Wall thickness was set according to the literature and our SEM observations ( Derbyshire et al . , 2007a ) . The epidermal layer was pressurized to provide the driving force for the growth of these cells and the space internal to the epidermis was also pressurized to simulate internal tissue force . Growth was implemented using a Lockhart model ( Lockhart , 1965 ) where strain above a yield threshold set the growth rates relative to principal strain directions ( Bozorg et al . , 2016 ) . Expanded details of the model can be found in the Appendix . When the axial and transverse anticlinal walls had the same elasticity ( no asymmetry ) and when no material anisotropy was specified ( cellulose orientation was not coordinated ) , the pressure forces caused the maximal strain ( and stress ) to be transverse ( Figure 4A ) . This result would lead to a radial swelling of the organ and was consistent with basic mechanical theories of hoop stress ( Baskin and Jensen , 2013 ) . In order to drive axial anisotropy , in the absence of cellulose-based material anisotropy , a 100-fold elastic asymmetry had to be invoked ( Figure 4A ) . These results led us to hypothesize that a 2-fold elastic asymmetry alone , as measured in our experiments ( Figure 3 ) , would be insufficient to drive anisotropic growth . Based on the literature , and our own observations of MT angle at early growth stages , we next added material anisotropy to our simulations . When material anisotropy favoring axial strain was specified at the inner epidermal wall ( as measured in [Crowell et al . , 2011] ) , we obtained axial growth anisotropy ( Figure 4B ) ; strikingly , addition of a 2-fold elastic asymmetry , consistent with our experiments ( Figure 3 ) , enhanced the magnitude of growth anisotropy ( Figure 4B , C ) . Since it was also possible that internal tissue provided anisotropic information ( i . e . the cortex , [Hejnowicz et al . , 2000] ) , we simulated this situation by specifying axial pressure in the inner-epidermal space and also recovered axial anisotropy ( Figure 4—figure supplement 1 ) . The addition of 2-fold elastic asymmetry in the epidermis was again able to enhance the magnitude of axial anisotropy ( Figure 4—figure supplement 1 ) . A sensitivity analysis of the two cases with anisotropic information and 2-fold elastic asymmetry indicated that they were most sensitive to variation in the degree of anisotropy and that increasing elastic asymmetry showed a positive correlation with growth anisotropy ( Figure 4—figure supplement 1 ) . In conclusion , our finite element mechanical model led us to propose that while epidermal elastic asymmetry alone was insufficient to drive axial growth anisotropy , it was able to contribute by increasing the anisotropy achieved when anisotropic information was provided by inner tissues or the inner epidermal wall . Our experimental and computational results led us to believe that pectin biochemistry could have an impact on growth anisotropy; however , our observations were correlative . In order to test a causal relationship , we altered pectin methylation in dark-growing hypocotyls and observed any subsequent changes in the cell shape . In Arabidopsis , the methylation of HG can be controlled by the antagonistic activity of two protein families , PECTIN METHYLESTERASE ( PME ) and PECTIN METHYLESTERASE INHIBITOR ( PMEI ) ; PME activity leads to de-esterification and likely to calcium cross-linking and increased rigidity , while PMEI would have the opposite effect ( Caffall and Mohnen , 2009; Levesque-Tremblay et al . , 2015a , Levesque-Tremblay et al . , 2015b ) . Note that PME activity could also lead to HG degradation by POLYGALACTURONASE ( PG ) , whose activity is also important for proper hypocotyl growth ( Rui et al . , 2017; Xiao et al . , 2014 ) . To alter pectin methylation in the hypocotyl , we utilized transgenic lines expressing either PECTIN METHYLESTERASE5 ( PME5 ) or PECTIN METHYLESTERASE INHIBITOR3 ( PMEI3 ) under ethanol induction ( Peaucelle et al . , 2008 ) ( Verification of induction in Figure 5—figure supplement 1 ) . We used AFM-based nano-indentation to examine basal cell wall elasticity in induced hypocotyls ( non-transgenic ( NT ) , PME , and PMEI ) . We observed that both PME5 and PMEI3 induction abolished cell-level elastic asymmetry: PME5 increased the rigidity in both axial and transverse anticlinal walls , while PMEI3 decreased the rigidity in both ( Figure 5A; Figure 5—figure supplement 1 ) . These changes in cell wall elastic asymmetry were accompanied by changes in cell shape anisotropy: induction of PMEI3 led to more anisotropic cells within the elongation wave and PME5 induction to less anisotropic cells ( Figure 5B; Figure 5—figure supplement 1 ) . Note that the position of the wave was not altered with PMEI3 induction , indicating that ectopically altering pectin chemistry could alter growth rate but not the position of the acropetal wave . Also , anisotropy was not lost in either transgenic induction indicating that loss of cell wall asymmetry alone is not enough to abolish anisotropic cell shape , or presumably growth . Altogether , it appears that pectin asymmetry has a contributory , not sole regulatory , role in anisotropy supporting our computational results . When NT , PME5 and PMEI3 hypocotyls were exposed to the inducer the following changes in growth were observed at the organ level: when compared to NT , PME5 induction abolished the rapid elongation phase and PMEI3 induction increased early elongation essentially flattening out the difference between the slow and rapid phases ( Figure 5C ) . These results were consistent with a promotive role for pectin methylation in rapid hypocotyl elongation . To confirm that the expected changes in pectin chemistry were occurring , we performed immunolocalizations on transverse sections of hypocotyls; namely , that PME5 induction yielded more de-methylated pectin signal ( LM19 antibody; Figure 5—figure supplement 1 ) and that PMEI3 induction yielded more methylated pectin signal ( LM20 antibody; Figure 5—figure supplement 1 ) . These data are thoroughly consistent with an increase in the relative amount of pectin methylation contributing to the transition from slow to rapid elongation , a point we will revisit once again at the end of this report . As we were primarily interested in anisotropic growth , we also examined how hypocotyl width was altered with changes in pectin biochemistry . Commensurate with the change in cell-level anisotropy , PME5 induction resulted in a reduced hypocotyl length ( Figure 5C ) and also a reduction in hypocotyl width ( Figures 5D , 24 and 48HPG ) . Conversely , induction of PMEI3 led to an increase in hypocotyl length ( Figure 5C ) and an increase in hypocotyl width ( Figures 5D , 24 and 48HPG ) . Taken together these data hint at a role for pectin chemistry in cell , and organ , growth anisotropy; however , in no case was anisotropy completely abolished indicating a more complex regulation of anisotropy than pectin asymmetry alone could provide , further supporting our additive model for anisotropy in the hypocotyl . Our observations of weak MT transverse alignment and pectin asymmetry , and our computational modelling , strongly indicated an additive role for these two mechanical factors . Since we had observed that ectopic alteration of pectin biochemistry could not fully abolish cell-level anisotropy , we next asked whether loss of MT-based anisotropy could be affected by altering pectin . Oryzalin , a drug that blocks the polymerization of MT , is known to affect the trajectories , distribution and densities of cellulose synthase complexes ( Paredez et al . , 2006; Chan et al . , 2010 ) , to change the organization in cellulose microfibril orientations , and to induce cell swelling ( a trend towards isotropy ) ( Baskin et al . , 2004 , Baskin et al . , 1994; Lucas et al . , 2011 ) . We treated seedlings with 5 µM oryzalin while inducing either PME5 or PMEI3 . NT control hypocotyls , treated with the inducer , showed a reduction in cell shape anisotropy when treated with oryzalin indicating typical cell swelling ( Figure 6A , B ) . This effect was reduced in induced PME5 plants and enhanced in induced PMEI3 plants ( Figure 6A , B ) . While there was a response to oryzalin in PME5-induced hypocotyls , the cell swelling was reduced indicating a compensatory cell wall strength in these cells ( Figure 6B ) . The opposite was true in oryzalin-treated PMEI3-induced hypocotyls where the cell swelling was more dramatic than either PME5 or NT , despite induced-PMEI3 cells being the most anisotropic without oryzalin treatment ( Figure 6A , B ) . These data indicated that changes in pectin biochemistry could modulate the effect of MT-derived cell swelling and isotropy , but again this modulation was never complete . Pharmacological treatments would be unlikely to affect existing cellulose fiber alignment from before the time of treatment , and so it is likely that treated hypocotyl cell walls maintained some pre-treatment cellulose-based anisotropy . We observed that the slowly-growing upper regions of dark-grown hypocotyls in these experiments exhibited less swelling upon oryzalin treatment ( Figure 6B , insets ) . Taken together with our earlier observations of decreasing pectin-based wall IM over time ( Figure 3B; Figure 3—figure supplement 1 ) and the effect of ectopic pectin alteration on hypocotyl elongation ( Figure 5C ) , we wondered if there might be endogenous differences in pectin chemistry along the hypocotyl length . To examine pectin biochemistry in this context , we performed immunolocalizations on transverse sections of dark-grown hypocotyl at 0HPG , when all parts of the hypocotyl were slowly growing , and 24HPG , when basal cells were entering the rapid elongation phase , using antibodies for methylated HG ( LM20 ) and de-methylated HG ( LM19 ) . At the time of germination ( 0HPG ) we observed strong signal for de-methylated HG ( LM19 ) in basal and apical transverse sections ( Figure 6D , E ) and weak signal for methylated HG ( LM20; Figure 6D , E ) . At 24HPG , the de-methylated HG signal remained high in the slow-growing apical sections but was low in basal sections ( Figure 6F , G ) . The calcium cross-linked HG antibody 2F4 showed a similar pattern ( Figure 6—figure supplement 1 ) . Methylated HG exhibited a complementary pattern with lower signal in sections from slow growing apical regions and higher signal in sections from rapidly growing basal regions ( Figure 6F , G ) . We could not discern any difference between tissue layers in our sections , indicating an organ-wide change in HG methylation state . These data led us to hypothesize that de-methylated pectin kept the hypocotyl in a slowly growing state , while methylated pectin allowed it to grow rapidly . An attractive hypothesis is that maintenance of pectin methylation allows for the onset of the acropetal growth wave; however , other growth-related parameters might be involved , such as vacuolar structure and resulting water uptake ability ( Scheuring et al . , 2016 ) . In multicellular anisotropically growing organs there is no reason stricto senso for every cell to have anisotropic wall properties ( Baskin and Jensen , 2013 ) . Indeed , our data presented here and those of others suggest that this is not the case in the dark-growing Arabidopsis hypocotyl ( Peaucelle et al . , 2015; Pelletier et al . , 2010; Derbyshire et al . , 2007a ) . Instead , it appears that anisotropic information originates at the inner face of epidermal cells and/or within cortical cells . Here we present a harmonious model of cell-wall controlled anisotropic growth: pectin asymmetry in the epidermis enhances anisotropic growth controlled by cellulose anisotropy . Our experimental analysis of native pectin biochemistry and manipulations of pectin biochemistry support a role for pectin asymmetry within the epidermis as a contributor to anisotropic growth; axial epidermal walls present markers of a more elastic pectin matrix and these walls grow faster , while the slow-growing transverse cell walls present markers of more rigid pectin . These biochemical observations are backed up by parsimonious observations of cell wall elasticity . When native pectin biochemistry was over-ridden by ectopic expression of PME5 or PMEI3 , native pectin asymmetry was an important component of anisotropic growth , but not the sole regulator . Our modelling experiments support an additive role for pectin asymmetry to anisotropic hypocotyl elongation , when combined with cellulose-based material anisotropy . There is an elephant in the room: although a strong correlation between MT orientation , CESA track movements and/or cellulose microfibril orientation have been reported in the literature in several systems ( Crowell et al . , 2011; Mueller and Brown , 1982; Takeda and Shibaoka , 1981 ) there are also reports that show no correlation ( Sugimoto , 2003; Emons et al . , 1992 , Emons et al . , 2007; Fujita et al . , 2011 ) . In the latter , cellulose synthase complex ( CSC ) movements have been shown to persist even in the absence of MTs . In the mor1-1 mutant , temperature-induced microtubule disorganization had no effect on cellulose microfibril orientation on inner epidermal walls ( Fujita et al . , 2011 ) . It is also prudent to note that cellulose microfibrils may undergo passive alignment once deposited within the apoplast and as such MT and CESA orientations may not accurately reflect cellulose fiber orientation ( Braybrook , 2017 ) . Given the observations and hypotheses above , it is possible that the outer epidermal wall of young dark-grown hypocotyls does contain transversely aligned cellulose fibrils in spite of the disperse orientation in both MT and CESA markers; however , direct imaging of cellulose fibers in hypocotyls just after germination is technically impossible at this time . In spite of this limitation , we believe the conclusion that pectin asymmetry contributes to anisotropic growth remains strong . We demonstrate that hypocotyl epidermal cells are anisotropic from the time of germination , an observation made possible by measuring both cell length and cell width . These observations build upon work describing the changes in cell length alone during elongation and the definition of the acropetal wave ( Gendreau et al . , 1997 ) . By assessing cell width and length in time we have been able to proxy each cell’s anisotropic growth . We further hypothesize that this early anisotropic growth is likely directed by internal wall material anisotropy ( inner epidermal face , cortex walls or their combined weak material anisotropy ) and is enhanced by cell-level elastic asymmetry . Our in silico modelling approaches have allowed us to further explore our hypotheses and provided some insight into their validity . First , the measured elastic asymmetry ( 2-fold , consistent with that reported recently ( Peaucelle et al . , 2015 ) was insufficient to drive axial anisotropic growth in our epidermal model , unless accompanied by internally provided anisotropic force or anisotropic properties of the inner epidermal wall . When combined , a 2-fold elastic asymmetry enhanced the anisotropic growth directed by internal tissues , leading us to hypothesize that elastic asymmetry aids in growth anisotropy . It is possible that the difference measured by our indentation tests underestimated the effective elasticity of the hypocotyl cell walls . It may also be that there are internal mechanical pectin asymmetries which contribute to anisotropy; our current indentation methods are restricted to epidermal cells . Cell wall elasticity , as measured here , is also an immediate property of cell walls which is only correlated with growth; future work must focus on uncovering how cell wall elasticity might relate to cell wall growth ( Braybrook and Jönsson , 2016; Braybrook , 2015 ) . Our data supports a role for pectin methylation in rapid cell elongation in the dark-grown Arabidopsis hypocotyl: pectin methylation is high in rapidly elongating hypocotyl cells as is wall elasticity; when pectin methylation is enhanced , walls are more elastic and rapid elongation starts early; when pectin de-methylation is induced , walls are less elastic and the rapid elongation phase is suppressed . An attractive hypothesis would be that maintenance of newly deposited pectin in a methylated state allows cell walls to expand more rapidly , and the conversion to a de-methylated state slows growth in the dark-grown Arabidopsis hypocotyl . Pectin with low methylation has been associated with non-growing areas in different species ( Fenwick et al . , 1997; Fujino and Itoh , 1998; Liberman , 1999 ) . However , the literature is complex: PMEI activity ( maintenance of methylated pectin ) has been associated with increased cell elongation in the Arabidopsis root ( PMEI1 and PMEI2 overexpression [Lionetti et al . , 2007] ) , but expression of another PMEI ( PMEI4 ) in the hypocotyl delayed the transition to rapid growth and had no effect on length growth rate ( Pelletier et al . , 2010 ) ; PMEI3 induction generated stiffer walls in Arabidopsis shoot meristems ( Braybrook and Peaucelle , 2013; Peaucelle et al . , 2011 ) . There are several possible explanations for these differences in phenotype: since the PMEI family is large and diverse ( Wang et al . , 2013 ) , it is likely that different proteins have different activities due to structure and environment ( Bou Daher and Braybrook , 2015; Giovannoni , 1989; Jolie et al . , 2010; Tian et al . , 2006 ) ; as we have demonstrated here , analysis of cell or organ length alone may obscure changes in width and it is possible that PMEI4 induced greater but more isotropic growth . When PMEI5 was overexpressed in adult Arabidopsis plants , stems displayed twice the diameter compared to controls further supporting the need to examine both width and length of organs ( Müller et al . , 2013 ) . The PME over-expression literature is less complicated ( possibly due to being slimmer ) : ectopic PME expression has been shown to reduce methylation and hypocotyl length previously , consistent with our study ( Derbyshire et al . , 2007b ) . However , a hypocotyl-expressed PME ( At3G49220 ) was found to be highly expressed after 30HPG and in elongating cell regions ( Pelletier et al . , 2010 ) . Our data show that both methylated and de-methylated HG can be observed in rapidly elongating hypocotyls ( Figures 3 and 6 ) indicating that PME activity is converting newly deposited methylated pectin into a de-methylated state during elongation . However , we must note that the available antibodies do not discriminate between patterns or degree of de-methylation . Different PMEs may have different activities ( Wolf and Greiner , 2012 ) resulting in HG chains more likely to cross-link or be targeted for polygalacturonase-mediated degradation , fates which may be linked to the pattern of de-methylation ( Wakabayashi et al . , 2000 , Wakabayashi et al . , 2003 ) . As an example , knockdown or silencing of a pollen PME in tobacco and Arabidopsis led to reduced pollen tube elongation ( Bosch and Hepler , 2006; Jiang , 2005 ) but treatment of the pollen tubes with orange-peel PME also reduced growth ( Marc et al . , 1998 ) ; it is not clear whether these apparently contradictory results are due to differential PME activity but the situation is clearly complex . PMEs also exhibit differential activities due to pH ( Jolie et al . , 2010; Hocq et al . , 2017 ) . While PMEs with alkaline pI remove the methyl groups in blocks , acid pI PMEs do so in a random fashion ( Jolie et al . , 2010 ) . Most of the Arabidopsis PMEs have an alkaline pI but there are some with acidic pI ( Tian et al . , 2006 ) . The pattern of de-methylation has an impact on the fate of HG with block-wise de-methylation leading to Ca-cross linking and random leading to degradation by PG ( Willats et al . , 2001 ) . While some data exists for transcriptional changes in PME and PMEI genes ( Pelletier et al . , 2010 ) it remains to be seen whether these changes result in changes in wall biochemistry and mechanics given the complexities of their post-translational activities . Perhaps the most puzzling contradiction to our data is the opposite phenotype shown recently for cell shape and rigidity for the same transgenic lines ( Peaucelle et al . , 2015 ) ; in our work we see a full ( 100% ) penetrance of phenotype upon induction ( Figure 2—figure supplement 1 ) , whereas the earlier study reported only a 10–20% penetrance of their desired phenotype – seedlings which were carried through for further analyses ( Peaucelle et al . , 2015 ) . It is possible that the amount of induction stimulus , or delivery method , had an effect on the phenotype presented; however , we must note that our immunolocalizations and rigidity data are consistent with the predicted biochemical activity of both PME and PMEIs . We used two markers for opposing states of pectin biochemistry , whereas only looking at de-methylated pectin alone might have given a reduced picture . Our analyses of cell shape upon microtubule disruption ( oryzalin treatment ) combined with pectin biochemistry manipulation further support our mechanical interpretation of the pectin-mechanics relationship put forward here: ectopic PME5 expression leads to stronger cell walls and partially suppresses oryzalin-induced cell swelling while ectopic PMEI3 expression enhances cell swelling and isotropy . Taken together , our data suggest an important role for pectin methylation in hypocotyl anisotropy; they also highlight the complexity of the experiments and the field as a whole . Lastly , feedback between cell-wall integrity and wall biochemistry/structure may add more complexity to the system as the oligogalacturonides generated by the lysis of the HG can act as signaling molecules and affect plant development ( for more references and reviews see [Wolf and Greiner , 2012; Wolf et al . , 2009] ) . In fact , these oligogalacturonides have been recently shown to be responsible for sustaining cell elongation in dark grown hypocotyls ( Sinclair et al . , 2017 ) . Due to the overexpression system used here , an ethanol induced transcriptional system , there is a time lag between induction and response . During this time and growth-time itself , we cannot discount that changes in wall biochemistry and mechanics induced by our PME and PMEI might have fed back through this system resulting in altered growth or further alterations in mechanics . Again though , we must stress the parsimonious nature of the predicted role of de-methylation on pectin gel mechanics , the observed mechanical changes in our system , the co-incident changes in wall biochemistry , and changes in cell shape and growth . We will return , lastly , to the question we began with: How does a seedling elongate upwards rapidly ? The data presented here make a strong case that changes in pectin chemistry , and resultant wall rigidity , are important for the anisotropic growth that is critical for the hypocotyl . However , changes in pectin alone are likely insufficient to direct anisotropy: we observed that anisotropy of internal tissues is likely to be required for anisotropic growth which is aided by elastic asymmetry in the epidermis . We therefore present a harmonious model of dark-grown hypocotyl elongation where the anisotropy provided by cellulose is enhanced by epidermal elastic asymmetry . Our experiments and conclusions also leave us with several new questions: how does altered elasticity actually affect growth ? How might elastic asymmetry be established in the first place ? How is the acropetal wave , and change in pectin chemistry , instructed ? These are questions which we look forward to investigating in the future . Transgenic lines sourced as indicated in Key Resources Table . Seeds were germinated on ½ MS plates containing Gamborg’s B5-vitamins but no sucrose . Germination was defined as the time when the radicle broke through the endosperm ( 0HPG ) . At this time , seedlings were selected and aligned with the radicle pointing downwards on ½ MS+B vitamins with 1 . 5% sucrose . The plates were wrapped with two layers of foil to simulate constant darkness . For PME5 and PMEI3 ( Sampathkumar et al . , 2014 ) induction , 0HPG seedlings placed in the middle of petri dishes flanked by two 500 µl microfuge tubes were placed at each side containing 200 µl of 100% ethanol each; this treatment achieved 100% penetrance of phenotype ( Figure 5—figure supplement 1 ) . Orzyalin treatment was as follows: Oryzalin was dissolved in DMSO and added to cooled media before pouring , to a final concentration of 5 µM . Mock treatment consisted of DMSO; 0HPG seeds ( genotype PMEI3/M-YFP , PME5/M-YFP or M-YFP alone; F3 homozygous lines generated by crossing ) were transferred to media with oryzalin or mock ( DMSO ) and grown in the dark for 48H prior to confocal imaging . Immunolocalizations were performed on 0 . 5 µm thick sections of LR White embedded hypocotyls . LM19 and LM20 antibodies ( PlantProbes , UK ) were diluted 200 times in PBS with 2% BSA . DyLight 488 goat anti-rat ( Cambridge Bioscence/Bethyl ) secondary antibody was diluted 400 times . 2F4 ( P . van Cutsem , gift ) immunolabelling was performed as in ( Liners et al . , 1992 ) Briefly , the primary antibody was diluted 100 times in TCN ( 20 mM Tris , 0 . 5 mM CaCl2 and 150 mM NaCl ) with 1% w/v skim milk . Alexa Fluor 488 goat anti-mouse ( Invitrogen , UK ) secondary antibody , was diluted 200 times . Images were acquired using a Leica TCS SP8 confocal microscope . Ratios were obtained using ImageJ by drawing a line along the walls in question and using the average fluorescence intensity of the line . Sample numbers were: Transverse 4HPG , n = 7 ( from 2 hypocotyls ) ; Transverse 24HPG , n = 12 ( from 3 hypocotyls ) ; ratio calculations LM19 , n = 72 from 7 sections; ratio calculations LM20 , n = 41 from 7 sections; transverse 0HPG , n = 6 ( from 2 hypocotyls ) ; transverse 24HPG , n = 14 ( from 4 hypocotyls ) ; control immunos , n = 9 ( from 2 to 4 seedlings each ) ; transverse 48HPG for NT/PME/PMEI , n = 9 each ( from 4 hypocotyls ) . Seedlings were incubated for 6 hr at 37°C in a solution of 50% water and 50% 2x GUS stain ( 50 mM KPO4 , 0 . 1% triton X-100 , 0 . 3 mg/mL X-GlcA ( 5-Bromo-4-chloro-3-indolyl-β-D-glucuronic acid , sodium salt ) , 1 mM K4Fe ( CN ) 6 , 1 mM K3Fe ( CN ) 6 , 0 . 1 v/v 1M KPO4 pH 7 ( 61 . 5mL 1M K2PO4 and 38 . 5 mL 1M KH2PO4 in 100 mL ) . samples were washed three times in 70% ethanol and one time in water and mounted in 50% glycerol under a coverslip and sealed with nailpolish . Images were acquired from 35S::GFP-MAP4 , 35S::GFP-TUA6 , 35S::GFP-EB1 and CESA3::CESA3-GFP seedlings using a Leica TCS SP8 confocal microscope using a 63X oil objective ( 1 . 4 numerical aperture ) . For microtubule orientation , we used the MicroFilament Analyzer ( MFA ) tool ( Jacques et al . , 2013 ) . Sample numbers were: 0H: n = 65 ( from 4 to 5 hypocotyls ) ; 24H: n = 30 ( from 9 hypocotyls ) ; 65HPG: n = 13 ( from 6 hypocotyls ) . For cortex analysis , n = 36 ( from 5 hypocotyls ) . 20 seedlings of Arabidopsis thaliana expressing a myristoylated-YFP were imaged for each time point , as confocal imaging stopped dark-grown hypocotyl elongation . Images were acquired using a Leica TCS SP8 confocal microscope . For cells , length and width were measured in Fiji ( Liners et al . , 1992 ) ; data were collected from 2 to 3 non-dividing files per hypocotyl . Cell diameter was recorded at the level of the central length of each cell . For cell shape analyses in induced NT , PME , and PMEI plants 10 seedlings of each were analysed for 48HPG and 24HPG . Hypocotyl widths and lengths at 24HPG and 48HPG were measured in 6–12 hypocotyls per treatment . Dividing cell characterization was conducted on 20 hypocotyls for each time point , and 4 seedlings were screened for GL2::GFP expression pattern . For oryzalin treated seedlings , a total of 12 seedlings for each treatment were imaged and the dimensions 8 cells per seedling , from the base , were measured . For imaging dark grown hypocotyls , a custom IR imaging setup was used , design available upon request . Images were acquired at 10 min intervals over 5 days . Images for selected time points were extracted and hypocotyl length was measured in Fiji ( Schindelin et al . , 2012 ) . Further discussion of AFM methods and interpretation can be found in the Technical Supplement . AFM-based nano-indentation experiments were designed and performed according to ( Braybrook , 2015 ) . Briefly , dissected and plasmolyzed ( 0 . 55M mannitol; minimum 15 min ) hypocotyls were indented using a Nano Wizard 3 AFM ( JPK Instruments , DE ) mounted with a 5 nm diameter pyramidal indenter ( Windsor Scientific , UK ) on a cantilever of 45 . 5 N/m stiffness; cantilever stiffness was calibrated by thermal tuning . For each hypocotyl , two areas of 50 × 100 µm were indented with 16 × 32 points: an area just before the collet and one slightly higher , to encompass basal cells . Indentations were performed with 500nN of force yielding an indentation depth range of 250–500 nm . Sample numbers were as follows: 4HPG , n = 24 cells ( from 6 hypocotyls ) ; 24HPG , n = 18 cells ( from 6 hypocotyls ) ; PME/PMEI/NT at 48HPG , n = 18 cells ( from 6 hypocotyls each ) . Force indentation approach curves were analyzed using JPK SPM Data Processing software ( JPK Instruments , DE; v . spm 5 . 0 . 69 ) using a Hertzian indentation model and a pyramidal tip shape . We have chosen to adopt the term ‘indentation modulus’ in place of ‘Young’s Modulus’ or ‘Apparent Young’s Modulus’ in order to distinguish these tests from those designed to assess Young’s modulus in materials science ( Cosgrove , 2016 ) . Indentation modulus maps were then imported into MatLab ( MATLAB 2016a , MathWorks , Inc . , USA ) and values were selected from anticlinal cell walls . For each grid area , 10–50 points were chosen from anticlinal walls and used for subsequent analyses , representing data from 3 to 10 cells depending on cell length in the scan area . Rations of IM were calculated by straightforward division of averages and propagation of SEM . Mann-Whitney tests for significant differences were performed in R as distributions were non-normal . A technical discussion on AFM-based analyses may be found in the Appendix . Brass stubs were covered with 50% lanolin solution in water that was preheated to 50°C and vigorously vortexed prior to applying the seedlings . 24HPG seedlings were placed on the lanolin coat and immediately plunge frozen in liquid nitrogen under vacuum . Frozen samples were then transferred under vacuum to a prep chamber of a PT3010T cryo-apparatus ( Quorum Technologies , Lewes , UK ) and maintained at −145°C . For cryo fracture a level semi-rotary cryo knife was used to randomly fracture the hypocotyls . All samples were sputter coated with a 3 nm platinum coat . Samples were then transferred and maintained cold under vacuum into the chamber of a Zeiss EVO HD15 SEM fitted with a cryo-stage . Images were taken on the SEM using a gun voltage of 6 kV , I probe size of 460 pA , a SE detector and a working distance of 4 mm . Details of the modelling can be found in the Appendix . In brief , a 3D finite element methods mechanical model was developed to evaluate mechanical signals and growth for cell walls of the epidermal cell layer of a hypocotyl . We used a 3D template and where prisms with six walls were utilized to represent individual cells ( Figure 4 ) . Each wall was triangulated from its centroid into triangular ( planar ) elements . The dimensions of the cell walls are proportional to the average values of those seen in experiments , for example Figure 1 . The thickness of the walls was included by adjusting their corresponding Young’s moduli assuming the material strength is proportional to the amount of material in a unit area . Individual cells were assembled into a 3D structure representing an epidermal cell layer . A wall in between two cells was divided into two adjacent walls and connected via the corner nodes . In this set up each pair of adjacent walls experienced the same deformation while they could hold individual mechanical properties . The two ends of the template were closed and the tissue was pressurized on the outer surface . The mechanical signals of cells close to each end were excluded from the analysis to avoid artefacts caused by boundary conditions ( see simulation edges in Figure 4 ) . In order to reduce the boundary effects , vertices at the two ends of the cylinder were constrained to stay in a plane parallel to the XY plane while allowed to move freely in the X and Y directions . For all of our analyses we did not exclude any data points . For AFM-based experiments samples sizes were low due to technical difficulty in experimentation: sample mounting was very difficult and often of 10 mounted samples only 2 remained fixed at 4HPG and 24HPG . AFM-based data was non-normally distributed so a Wilcoxan rank-sum test ( aka Mann-Whitney-Wilcoxan ) was used to see if the data from two independent samples were equivalent: this nonparametric test follows the null hypothesis that a random value selected from group 1 is equally likely to be greater or lesser than a random member of group 2 . For normally distributed data , such as growth and cell dimensions , t-tests were used ( singly or pair-wise comparisons ) ; for multi-sample comparisons ( e . g . NT vs . PME vs . PMEI ) pair-wise t-tests are shown but ANOVA gave similar results . All raw data produced and utilized in this study can be downloaded from the DRYAD data repository ( doi:10 . 5061/dryad . 4s4b3nf ) . Modeling code can be accessed through the Sainsbury Laboratory’s GitLab page ( https://gitlab . com/slcu/teamHJ/behruz/3Dhypocotyl; copy archived at https://github . com/elifesciences-publications/3Dhypocotyl ) .	Unlike animal cells , plant cells are surrounded by a stiff shell called the cell wall . Cell walls are composed of two main types of material: cellulose , the strong fibers that make up paper , and a pectin gel , which holds everything together . In order for plants to grow , the cell wall has to yield to the pressure inside the cell and allow stretching . The direction of individual cell growth in plants is thought to be controlled by the direction of cellulose fibers in the wall; if they wrap around the cell like hoops on a barrel , the cell can only grow ‘up’ and not ‘out’ . Cellulose direction is dictated by the orientation of tracks inside the cell called microtubules . Another recent idea says that the pectin gel can control growth direction; if the side walls of a cell have less gelling they can elongate more , increasing upward growth . What had not been examined is whether cellulose and pectin might both contribute to directional growth . Young seedlings emerge from the soil through the directional growth of the young stem , or hypocotyl . Using advanced microscopy , nano-materials testing , genetics techniques and computational models Bou Daher et al . studied the hypocotyl of a commonly studied plant called Arabidopsis thaliana . The results demonstrate that not only do both components of the cell wall control growth , but they work together from different tissues within the plant . The orientation of microtubules ( and hence cellulose fibers ) in cells in the inner tissues of the hypocotyl combines with pectin gelling in the outer tissue layer to produce fast , directional growth . Understanding how directional growth is achieved could enable us to change it in useful ways . This could lead to a number of agricultural improvements . For example , many seedlings are lost as they first grow through the soil to reach the light , so improving directional growth could increase crop yields . In order to do this , researchers would need to explore how common the co-operative mechanism Bou Daher et al . have discovered is in other plant species ( such as soybean , corn and wheat ) and in other plant organs ( like the adult stem and the roots ) .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology" ]	2018	Anisotropic growth is achieved through the additive mechanical effect of material anisotropy and elastic asymmetry
Research in the field of human immunology is restricted by the lack of a system that reconstitutes the in-situactivation dynamics of quiescent human antigen-specific T-cells interacting with dendritic cells . Here we report a tissue-like system that recapitulates the dynamics of engineered primary human immune cell . Our approach facilitates real-time single-cell manipulations , tracking of interactions and functional responses complemented by population-based measurements of cytokines , activation status and proliferation . As a proof of concept , we recapitulate immunological phenomenon such as CD4 T-cells' help to CD8 T-cells through enhanced maturation of DCs and the effect of PD-1 checkpoint blockades . In addition , we characterise unique dynamics of T-cell/DC interactions as a function of antigen affinity . The field of human immunology has gained much attention in the last few decades with the increasing realisation that despite the tremendous utility of mouse models , they can only provide limited insight for advancing translational research ( Wagar et al . , 2018; Council MRC , 2008 ) . Most of the research in human immunology is based on observational studies , whereby , for example , data sets of gene expression profiles are used to identify disease biomarkers . Although these studies promote our understanding and help in developing hypotheses about disease mechanisms , they fall short , and experimental research is needed ( Davis , 2008 ) . The classical tools in experimental human immunology rely heavily on immortalised cell lines , stimulated clonal lines , peripheral blood , and more rarely humanised mouse models . Each of these tools has its limitation . Cell lines have diverged tremendously from primary cells , making their signalling machinery and response thresholds no longer representative ( Colin-York et al . , 2019 ) ; bulk peripheral blood assays lack the spatial organisation of tissues; and humanised mouse models have only a portion of their system ‘humanised’ , rendering the remaining parts confounding factors ( Theocharides and Manz , 2018 ) . To address these limitations , we introduce a holistic approach to allow researchers to manipulate and study human immune cells in a tissue-like environment with minimal aberrations to their natural behaviours . We report a toolkit that achieves two critical goals for engineering T-cells: ( 1 ) high efficiency TCR expression and maintenance of T-cell motility , activation dynamics and viability using mRNA electroporation and ( 2 ) a tissue-like 3D culture ( Figure 1 ) . Successful engineering of T-cell specificity requires the stable expression of a functional exogenous TCR , as well as the preservation of important cellular characteristics . One such characteristic is cell motility , which is essential for efficient antigen search in a physiological 3D setting . Here , we have optimised the procedures to engineer naïve human CD4 and CD8 T-cells using mRNA electroporation . The common approach to render T-cells antigen-specific relies on lentiviral transduction of the TCR , which requires prior expansion of T-cells to achieve high expression efficiency and hinders the study of natural quiescent populations , namely naïve and memory T-cells ( Dai et al . , 2009 ) . To avoid the need for cell expansion , it is possible to use electroporation to deliver the genetic materials . However , reported approaches using electroporation suffer from low efficiency of expression , low viability , and most importantly impaired cell motility , a critical consideration in the context of delivery to target tissues . Using a square-wave electroporation apparatus , we are able to express the 1G4 TCR , specific to the NY-ESO tumour associated antigen ( Chen et al . , 2000 ) in naïve ( Figure 2A ) and memory ( Figure 2—figure supplement 1A ) CD8 T-cells as well as the 868 TCR specific for the HIV gag protein ( Varela-Rohena et al . , 2008 ) ( Figure 2—figure supplement 1B ) , both presented by HLA-A0201 . Successful pairing of the alpha and beta heterodimer of the TCR is important for the expression and assembly at the cell surface , this was achieved by introducing cysteine modifications in the TCR chains ( Figure 2—figure supplement 1B and Table 1 ) ( Cohen et al . , 2007 ) , as well as co-transfection with the human CD3ζ ( Figure 2—figure supplement 1C ) , as the endogenous protein is most likely sequestered by the endogenous TCR limiting the expression of the introduced TCR . The functionality of naïve CD8 T-cells following TCR electroporation was confirmed by the formation of a mature immunological synapse in naïve and activated CD8 T-cells ( IS , Figure 2B and Figure 2—figure supplement 1D–E ) , and the upregulation of the activation marker CD69 ( Figure 2—figure supplement 2A ) . The efficiency of TCR expression varied among donors with an average value of 81 ± 7% ( mean ± SD , n = 13 ) , with more than 90% cell viability and 80–90% cell recovery , both of which were severely reduced using alternative electroporation approaches ( Figure 2—figure supplement 3 ) . The induction of expression of an exogenous TCR in CD4 T-cells is considerably harder than in CD8 T-cells ( Dai et al . , 2009 ) . Using natural sequences or the introduction of cysteine modifications used for 1G4 and 868 failed to induce expression of TCRs in naïve CD4 T-cells . However , replacing the constant region of the human TCR with that of mouse TCR ( Table 2 ) ( Cohen et al . , 2006 ) allowed for successful chain pairing and we were able to achieve expression in 72 ± 8 . 9% ( mean ± SD , n = 15 ) of the cells detected using H57 mAb against the mouse constant region of the β chain . We have successful expressed three different HLA-DPB104 restricted TCRs in naïve CD4s: 6F9 ( Figure 2C ) and R12C9 , both specific to the MAGE-A3 protein ( Yao et al . , 2016 ) and SG6 specific for NY-ESO ( Zhao et al . , 2006a ) ( Figure 2—figure supplement 4A ) . The expression of SG6 was consistently lower than 6F9 and R12C9 in naïve CD4 T-cells . However , when the same TCRs were electroporated into recently activated CD4 T-cells , high expression levels were achieved in 97 . 8 ± 1% ( mean ± SD , n = 15 ) of the cells ( Figure 2—figure supplement 4B ) . The functionality of the naïve CD4 T-cells upon TCR electroporation was confirmed by IS formation ( Figure 2D and Figure 2—figure supplement 4C–D ) and T-cell activation ( Figure 2—figure supplement 2B ) . The enhanced expression of the TCR in activated CD4 T-cells resulted in enhanced pMHC accumulation at the IS ( Figure 2—figure supplement 4D ) . Using the square-wave electroporator we were able to maintain the motile behaviour and interaction dynamics of naïve CD8 T-cells expressing 1G4 , at similar levels to those of untouched cells , on 2D stimulatory ‘spots’ ( Figure 2—figure supplement 5 , Video 1 and Mayya et al . , 2018 ) . Naïve T-cells get activated by interacting with professional antigen presenting cells ( APCs ) ; the most potent are dendritic cells ( DCs ) . The closest model for bona fide DCs is monocyte-derived DCs ( moDCs ) with well-established protocols for their generation . The caveat with those protocols is their extended duration ( 7–10 days ) ( Zhou and Tedder , 1995 ) , a time-frame which may alter autologous T-cells’ phenotype . We therefore optimised a 48 hr protocol ( Obermaier et al . , 2003 ) to generate fully matured moDCs . During the first 24 hr monocyte differentiation into DC was induced , followed by 24 hr maturation to get activated DCs ( acDCs ) ( Figure 3 ) . The ‘express’ moDCs are able to upregulate MHC and other costimulatory molecules and produce a similar profile of soluble factors as the ‘classical’ moDC ( Figure 3—figure supplement 1A–B ) . We have also quantified peptide loading on these cells and achieved levels similar to those previously reported with other APCs ( Figure 3—figure supplement 1C ) ( Zehn et al . , 2006 ) . T-cells–DC interactions have been extensively studied in mice using explants and intravital imaging ( Miller et al . , 2002 ) ; however , such studies are practically impossible in humans . Furthermore , accurate control and manipulation of parameters such as antigen dose and cell ratios is limited . Having successfully engineered naïve human T-cells we set out to establish a flexible 3D platform to interrogate their dynamics and interactions with APCs using a multitude of correlated functional readouts . The desired platform should: ( 1 ) support the culture of T-cells and APCs for multiple days , ( 2 ) support the motility of the cells , ( 3 ) allow live imaging , ( 4 ) enable downstream analysis and ( 5 ) be easy to use . To achieve these goals we chose collagen I -based 3D matrices ( Gunzer et al . , 2000 ) , which we optimised to support the culture of human immune cells . Culturing cells in the presence of human serum results in better basal motility than FBS ( Figure 4A , Figure 4—figure supplement 1A and Video 2 , Table 3 ) , which is significantly enhanced by the addition of homeostatic chemokines such as CCL19 ( Video 3 ) , CCL21 and CXCL12 ( Video 4 ) . However , only CXCL12 was able to maintain high motility in cultures with FBS . We have explored different sources of collagen I ( undigested and trypsin digested bovine and human collagen ) and other complex extracellular matrices , ECM ( Figure 4—figure supplement 1B ) . All collagens were equally good in supporting motility and interactions , while more advanced ECM may have a marginal advantage for motility . Yet , the larger batch variation , higher background activation and the complexity in extracting cells for downstream analysis weighed against adopting them for our assays . In our view , the use of bovine collagen-I as the 3D scaffold , with human serum , and CXCL12 to enhance motility provides an optimal 3D system with similar movement parameters to T-cells in mouse lymph nodes or explanted human tissues ( Bougherara et al . , 2015; Murooka et al . , 2012; Salmon et al . , 2012; Woolf et al . , 2007 ) . We have also re-evaluated two additional electroporation methods , namely Lonza-Amaxa and ThermoFisher-Neon . In addition to their lower performance in cell recovery and protein expression ( Figure 2—figure supplement 3 ) , they further resulted in poor motility dynamics and hence activation of cells within the 3D environment , where the search for antigen loaded APCs becomes a confounding factor ( Figure 4—figure supplement 1C–F ) . We present here a proof of concept of the usefulness of our set-up by interrogating naïve CD8 T-cells activation . We used live cell imaging to follow the interactions in real-time by upregulation of CD69 ( Figure 4B and Video 5 ) as well as calcium flux ( Figure 4C and Video 6 ) . We observed major changes in interaction dynamics as a function of peptide affinity ( Aleksic et al . , 2010 ) ( Figure 4D ) . A high affinity peptide , NY-ESO-9V , ( Video 7 ) induced T-cell arrest , whereas a lower affinity peptide , NY-ESO-9L , ( Video 8 ) , led to similar dynamics non-peptide loaded ( Video 9 ) where no stable contacts are formed . Since our 3D culture offers a relatively simple way to extract cells for flow cytometry and downstream analyses , we were able to correlate the interaction dynamics with the activation state of the cells and observe marginal differences in the response relative to the major differences in dynamics ( Figure 4E , Figure 4—figure supplement 1A ) . Using our system , we were also able to monitor cytokine production ( Figure 4—figure supplement 1B ) , proliferation ( Figure 4—figure supplement 1C ) and intracellular markers ( Figure 4—figure supplement 1D ) . In order to study the interaction dynamics of T-cells at different time points we could pre-load the DCs prior to or following gel polymerisation to mimic the delivery of antigen to lymph nodes ( Figure 4—figure supplement 1E and Videos 10–11 , respectively ) . Our ability to engineer both CD4 and CD8 T-cells independently with different TCRs enables us to interrogate complex dynamics of a trinary cellular system whereby both types of T-cells can interact with similar or different APCs . We show here that using our system we can interrogate the dynamics of CD4 T-cells help to CD8 T-cells ( Castellino et al . , 2006 ) ( Figure 5A , Video 12 ) where the activation of CD4 T-cells ( Figure 5B , top ) coincides with enhanced maturation of the DCs by upregulation of CD86 ( Figure 5B , middle ) and enhanced activation of the CD8 T-cells ( Figure 5B , bottom ) . Furthermore , we utilised our system to mimic one of the most successful immune checkpoint blockades currently in clinical use; PD-1 blockade , which is known to act by enhancing CD8 T-cell responses towards target cells . To that end , we used memory CD8 T-cells and treated them with two clones of blocking antibodies against PD1; Nivolumab and EH12 . Despite the lack of any change in the motility dynamics of the cells ( Figure 5—figure supplement 1A ) , there was a clear evidence that the presence of the checkpoint blockade enhances the clustering of CD8 T-cells around their targets ( Figure 5C ) . This has also been corroborated by enhanced activation , although only observable for late readouts such as 41BB expression and target cell killing ( Figure 5D ) . Interestingly , we saw an enhanced expression of PDL-1 ( Figure 5—figure supplement 1B ) , the ligand for PD-1 , on CD8 T-cells , which could lead to cis interactions that act as a cell intrinsic regulation mechanism ( Sugiura et al . , 2019 ) . Surprisingly , we noted differences between the pathways enhanced by the two blocking antibody clones: EH12 seemed superior at enhancing surface markers , whereas Nivolumab outperformed EH12 in the killing assays . In this report , we present , to the best of our knowledge , a unique and first of its kind holistic approach to engineer quiescent antigen-specific human T-cells and study them in a context akin to physiological settings . Our tools are widely applicable to generate T-cells expressing different receptors and other proteins with high efficiency and minimal adverse effects associated with other engineering methods , which are essential for implementation of a successful translational approach . We provide a proof of concept where our system is used to probe the dynamics of T-cell/APC interactions as a function of peptide affinity , recapitulate CD4 help to CD8 T-cells and interrogate PD1 checkpoint blockade . Our experimental approach can be further extended towards reconstitution of more complex biology such as tumour infiltrating lymphocytes and the role of regulatory T-cells in cancer . We believe that the mechanical and biophysical features introduced by our platform provide greater fidelity to in-vivo conditions and will improve predictive power of pre-clinical studies in all human cell-based systems . We propose that our system should be integrated with other classical and computational approaches to enhance translational research . RPMI 1640 ( 31870074 ) was purchased from ThermoFisher Scientific . Hyclone fetal bovine serum ( FBS ) was obtained from Fisher Scientific . The following anti-human antibodies were purchased from Biolegend: CD62L ( DREG-56 ) , CD3 ( OKT3 ) , CD8 ( HIT8a ) , HLA-A2 ( BB7 . 2 ) , CD80 ( 2D10 ) , CD86 ( IT2 . 2 ) , CD40 ( 5C3 ) , CD69 ( FN50 ) , 4-1BB ( 4B4-1 ) , CD107 ( H4A3 ) , TNFα ( Mab11 ) and anti-mouse TCRβ ( Η57–97 ) . The following anti-human antibodies were purchased from BD Bioscience: CD45RA ( HI100 ) , CD4 ( RPA-T4 ) , and CD11c ( B-ly6 ) , CD14 ( MΦP9 ) and PD-1 ( EH12 . 2H7 ) . Nivolumab ( Ab00791-13 . 12 ) was obtained from Absolute Antibody . The following anti-human antibodies were purchased from eBioscience: CD83 ( HB15e ) , IFNγ ( 4S . B3 ) and PDL-1 ( MIH1 ) . HLA-A0201 with 9V was generated and folded in house . HLA-DPB104:01 biotinylated monomers were obtained from the NIH tetramer facility . NY-ESO-9V157-165 ( SLLMWITQV ) , NY-ESO-4D157-165 ( SLLDWITQV ) , NY-ESO-6T157-165 ( SLLMWTTQV ) and NY-ESO-9L157-165 ( SLLMWITQL ) , MAGE-A3243-258 ( KKLLTQHFVQENYLEY ) , NY-ESO161-180 ( WITQCFLPVFLAQPPSGQRA ) and HIV p17 GAG77-85 ( SLYNTVATL ) were purchased from GeneScript with >95% purity and resuspended in DMF at 1 mg/ml and stored at −20°C . The sequences for the 1G4 TCR were provided by Vincenzo Cerundolo , the 868 TCR was provided by Andrew Sewell , and all the MHC-II restricted TCRs ( SG6 , R12C9 and 6F9 ) were shared by Steven Rosenberg . All TCR constructs were synthesised using GeneArt services from ThermoFisher; a Kozak sequence was added and codon optimisation for human expression was performed . The 1G4 and 868 constructs included a cysteine modification in the alpha and beta chains ( see Supplementary table for sequences ) . Constructs for the MHC-II restricted TCRs were synthesised so that the human constant region was replaced with that of mouse ( see Supplementary table for sequences ) . The alpha and beta chains were cloned separately into a pGEM-GFP-64A plasmid ( gift of James Riley ) between HindIII and NotI for MHC-I restricted TCRs and for human CD3ζ and between AgeI and NotI for MHC-II restricted TCRs . Human T-cells were isolated from anonymised leukopoiesis products obtained from the NHS at Oxford University Hospitals ( REC 11/H0711/7 ) . Resting human T-cell subsets were isolated using negative selection kits ( Stemcell Technologies ) , total CD4 or CD8 T-cells were enriched using RosetteSep kit from Stemcell Technologies , followed by EasySep kit for the corresponding naïve and memory sub-population , following the manufacturer’s protocol . Cell purity was assayed with anti-CD4 , anti-CD8 , anti-CD62L and anti-CD45RA and all cells used had >95% purity . CD4 and CD8 T-cells were cultured at 37°C , 5% CO2 , in RPMI 1640 ( Roswell Park Memorial Institute ) medium supplemented with 10% FBS ( Gibco ) , 5% penicillin-streptomycin ( PenStrep , Gibco ) , 1x MEM non-essential emino acids solution , 20 mM HEPES , 1 mM sodium pyruvate , 2 mM Glutamax and 50 μM 2-mercaptoethanol ( Sigma ) ( all from Thermo Fisher unless stated otherwise ) . For T-cell activation , 1:1 anti-CD3/anti-CD28- coated T-cell activation beads ( 11132D ) were added at a 1:1 ratio with 50 U/ml of IL-2 for two days , followed by removal of the beads and propagation of the culture in IL-2 until day 7 . Monocytes were enriched from the same leukopoiesis products as T-cells using RosetteSep kit ( Stemcell Technologies ) , and were then cultured at 1−2 × 106/ml in 12-well plates with 1 ml of differentiation medium containing 50 ng/ml Interleukin 4 ( IL-4 , 200–04 A , Peprotech ) and 100 ng/ml granulocyte-monocyte colony stimulating factor ( GM-CSF , 11343125 , Immunotools ) for 24 hr . For maturation the following cytokines were added for an additional 24 hr: 1 μM prostaglandin E2 ( PGE2 , P6532 , Sigma ) , 10 ng/ml interleukin one beta ( IL1β , 201-LB-025/CF , Biotechne ) , 50 ng/ml tumour necrosis factor alpha ( TNFα , 300-01A , Peprotech ) and 20 ng/ml interferon gamma ( IFNγ , 285-IF-100/CF , Bio-techne ) . Cell purity of the monocyte population was assessed using antibodies against CD14 and CD11c and typically was above 80% . Any non-monocyte contaminants should be removed by adhering the cells for 2–3 hr immediately after isolation followed by washing off any unbound cells and applying the differentiation media . Peptide loading on DC was quantified using an AlexaFluor 647 conjugated high affinity soluble TCR ( c113 against NY-ESO ) ( Zhao et al . , 2007 ) with a known dye ratio and Quantum MESF AlexaFluor 647 beads as calibration ( 647 , Bangs Laboratories ) . Secretion of soluble factors was assessed using 0 . 5 ml of culture medium , and the corresponding controls of media containing the differentiation and activation cytokines , following the manufacturer’s protocol ( Proteome Profiler Human XL Cytokine Array kit , ARY022B , R &D Systems ) . For antigen-specific experiments , the donors were assessed for the relevant HLA . For CD8-based experiments , 50 μl of whole blood was taken for flow cytometry analysis using an anti-HLA-A2 mAb ( clone BB7 . 2 ) . For CD4 -based experiments , 50 μl of blood was used to extract DNA with the QIAamp DNA Mini kit ( 51304 , Qiagen ) . The extracted DNA was used for PCR analysis using the AllSet+ Gold HLA DPB1 High-Resolution Kit ( 54070D , VH Bio ) to determine whether or not the suitable haplotype was expressed . We were also able to induce the expression of a single chain dimer of HLA-A02 in moDCs from donors which were HLA-A02-negative donors using the mRNA electroporation approach . To express TCR constructs within resting naïve and memory T-cells , we used mRNA electroporation . mRNA for the relevant TCRα , TCRβ and CD3ζ chains was prepared from a linearised pGEM-64A vector or a T7 containing PCR product was done using mMESSAGE mMACHINE T7 ULTRA Transcription kit as per manufacturer’s protocol ( ThermoFisher , AM1345 ) . The mRNA was purified using MegaClear kit ( ThermoFisher , AM1908 ) and was aliquoted and kept at −80°C ( Zhao et al . , 2006b ) . To achieve high efficiency of expression , mRNA quality was assessed using agarose gels and samples showing any sign of degradation were discarded . Any mRNA preparation with yields bellow 40 μg mRNA per 1 μg of DNA was considered of low quality; finally , all mRNA aliquots were stored at concentrations > 1 μg/μl , which was achieved by concentrating the mRNA product using ammonium acetate precipitation or by combining >3 reactions of in vitro transcription during the MegaClear clean-up step . T-cells were harvested and washed three times with Opti-MEM ( LifeTechnologies ) . The cells were resuspended at 25 × 106/ml and 2 . 5−5 × 106 cells were mixed with the desired mRNA products and aliquoted into 100–200 μl per electroporation cuvette ( Cuvette Plus , 2 mm gap , BTX ) . For 106 cells CD8 T-cells , 2 μg of each TCRα , TCRβ and CD3ζ RNA was used . For 106 CD4 T-cells , 4 μg of TCRα and TCRβ was used . Cells were electroporated at 300 V , 2 ms in an ECM 830 Square Wave Electroporation System ( BTX ) . The cells were then collected from the cuvette and cultured in 1 ml pre-warmed media . Amaxa and Neon electroporation systems were also tested but failed in either achieving efficient TCR expression or affected T-cell motility . For Amaxa electroporation , the human T-cell nucleofector kit ( VAPA-1002 ) was used and the manufacturer's protocol was followed . Three different settings were tested: V-24 , U-14 and T-23 . For Neon electroporation , 2 . 5 × 106 cells were resuspended in a 100 μl tip , and two settings were tested ( 1700 V 10 ms , 4 pulses and 2150 V , 20 ms one pulse ) ( Aksoy et al . , 2019 ) . Note , extended culture ( >5 days ) prior to electroporation also results in marginal reduction in TCR expression efficiency , although no reduction is observed for control GFP . The exogenous TCR can be detected up to 96 hr post electroporation . SLBs were prepared as previously described ( Dustin et al . , 2007 ) . In brief , piranha and plasma cleaned coverslips were mounted on sticky-Slide VI0 . 4 chamber ( ibidi ) . Small unilamellar liposomes ( LUVs ) were prepared using 4 mM 18:1 DGS-NTA ( Ni ) , ‘NTA-lipids’ , ( 790404C-AVL , Avanti Polar Lipids ) , 4 mM CapBio 18:1 Biotinyl Cap PE ( 870282C-AVL , Avanti Polar Lipids ) , ‘CapBio-lipids’ , and 4 mM 18:1 ( Δ9-Cis ) PC , ‘DOPC-lipids’ , ( 850375C-AVL , Avanti Polar Lipids , ) . NTA-lipids were used at a final concentration of 2 mM and CapBio lipids were used at a dilution resulting in site density of 100 molecules/μm ( Council MRC , 2008 ) . SLBs were allowed to form by incubating the coverslips with the appropriate LUVs for 20 min at room temperature ( RT ) followed by a wash with HEPES-buffered saline ( HBS ) supplemented with 1 mM CaCl2 and 2 mM MgCl2 , and human serum albumin ( HBS/HSA ) . The SLBs were loaded with saturating amounts of AlexaFluor568 labelled streptavidin , 5 μg/ml ( S11226 , ThermoFisher ) for 10 min at RT followed by pMHC at 100 molecules/μm ( Council MRC , 2008 ) and ICAM-1 200 molecules/μm ( Council MRC , 2008 ) for 20 min at RT . Micro-patterned surfaces presenting pMHC molecules were prepared using micro-contact printing by modifying the procedures described previously ( Mayya et al . , 2018 ) . Topological masters with repetitive circle patterns of 10 μm in diameter , spaced 30 μm centre-to-centre on a square grid were used to cover the entire channel of a sticky-Slide VI0 . 4 chamber ( ibidi ) . The master was used for casting of polydimethylsiloxane ( PDMS ) elastomer stamps . Sylgard 184 ( Dow Corning ) was used by mixing 1/7 ( v/v ) of curing agent to the elastomer . Rectangular stamps of PDMS were coated with biotinylated , AlexaFluor 674 labelled Fc IgG ( AG714 , Merck Millipore ) at 2 μg/ml in 150 μl of PBS for one hour . The blocks were then rinsed extensively in PBS , PBS with 0 . 05% Tween-20 and finally in MilliQ-grade water followed by gentle drying with N2 to remove droplets of water . Fc coated PDMS blocks were stamped onto the cleaned coverslips for 5 min under ~20 g of load for efficient transfer . Attempts to directly stamp pMHC , avidin or streptavidin resulted in partially or completely non-functional proteins . Streptavidin stamps were also non-uniform and eroded with subsequent washes . The patterned coverslip was then affixed to the sticky-Slide VI0 . 4 and washed sequentially with MilliQ-grade water and PBS , then coated with 13 . 5 μg/ml of CCL21 in 30 μl for one hour , followed by 3 μg/ml of ICAM1 in 180 μl for 3 hr at 37°C . Coverslips were then used immediately for the following steps or kept in PBS at 4°C overnight . The stamped and protein coated channels were blocked with 10% dialysed FCS ( to remove any free biotin ) for 30 min at RT . Then unlabelled streptavidin was introduced into the channel at 2 μg/ml for 10 min at RT , followed by washing and introduction of 9V/A2 pMHC at 1 μg/ml for 20 min at RT . DCs were loaded with the peptide specified in the figures for 90 min at 37°C . For imaging experiments , T-cells were labelled with a volumetric dye as described below ( Imaging ) or loaded with cell trace violet ( C34557 , ThermoFisher ) for proliferation assays , or otherwise kept unlabelled . Collagen mix was made by modifying the approach of Gunzer et al . ( 2000 ) using 5 μl 7 . 5% sodium bicarbonate , 10 μl 10x MEM , 75 μl 3 mg/ml Bovine Collagen I ( PureCol ) and a chemokine , 1 μg/ml of CCL21 ( 300-35-100 , Peprotech ) or CCL19 ( 300-29B , Peprotech ) or 0 . 3 μg/ml of SDF1-a/CXCL12 ( 300-28A , Peprotech ) . All the work with collagen was done on ice to prevent polymerisation . Other matrices tested: VitroCol Human Collagen I ( 5007-A , CellSystems ) , Matrigel ( at final conc . of ~4 mg/ml , two batches tested with similar results ) , and GelTrex ( A1413301 , ThermoFisher , at final conc . n of ~8 mg/ml , two batches tested with similar results ) . All 3D cultures were made with bovine collagen I , except for the comparison shown in Figure 4—figure supplement 1B , and mainly with CXCL12 unless otherwise stated . Cell-containing media ( 15 μl ) supplemented with 10% Human serum ( S-101B-FR , APS ) ( or 10% FBS ) wasadded to 35 μl of the collagen mix so that the final culture contained 200 , 000 CD8 , 200 , 000 CD4s T-cells and DCs at varying ratios ( typically 1:1 , 1:5 and 1:10 to T-cells ) . Lower cell numbers can be used for functional experiment but the reported numbers here are ideal for good density for imaging . Collagen mix is then added to a μ-slide angiogenesis chamber or sticky-Slide VI0 . 4 chamber ( ibidi ) and put upside down in a 37°C incubator with 5% CO2 for ~60 min ( to entrap the cells in the gel as it polymerises ) . The wells were then topped up with 30 μl media containing the same concentration of chemokines as the gel ( for μ-slide angiogenesis ) or 100 μl per well ( for sticky-Slide VI0 . 4 ) . For in gel loading experiments , the media contained a peptide at 2x the reported concentration . Samples were either taken for imaging or kept in the incubator for functional analysis . For PD-1 blockade , the antibody or isotype control was added to the T-cells for 15 min prior to incorporation in the gel so the final antibody concentration was 10 μg/ml . Extended culture in the 3D system leads to ~50% reduction in cell speed due to chemokine desensitisation and can be overcome by the addition of fresh chemokine after 18 hr . 1G4 TCR expression was assessed using 9V-HLA-A2 tetramers , 868 TCR expression was assessed using SL9-HLA-A2 streptamers ( 6-7004-001 , iba ) and 6F9 , SG6 and R12C9 expression was assessed with anti-mouse TCR β ( H57 ) antibody . The MHC II restricted TCRs seem to have very low affinity towards their antigen , preventing detection using tetramers ( attempts to measure affinity using SPR suggest a KD above 500 μM ) . In brief , 50 , 000 cells were stained with 2 . 5 μg/ml tetramers ( 9V/A2 ) or 5 μg/ml streptamers ( SL9/A2 ) or 1 μg/ml antibody for 20 min at 4°C , washed with PBS containing 2% FBS and 2 mM EDTA and taken for analysis on a flow cytometr ( FORTESSA X-20 , BD Biosciences ) . To analyse samples from collagen gels , the gels were first digested into solution using collagenase VII ( Sigma #C0773 ) at 100 U/ml for 1 hr at 37°C . The supernatant was subsequently removed , and the cells were stained with the desired antibody combination for 20 min 4°C . For intracellular staining , the cells were cultured in collagen gels and brefeldin A or monensin was added , after 20 hr for 2 hr to prevent cytokine secretion to the supernatant and retain them inside the cells , along with stimulation of phorbol 12-myristate 13-acetate ( PMA , P8139 , Sigma ) at 1 μg/ml and ionomycin ( I0634 , Sigma ) at 1 μM , the cells were then stained using the FoxP3 intracellular staining kit ( 00-5523-00 , eBioscience ) and the desired antibodies . For experiments measuring LAMP1 expression , an antibody against it was added at the start of the culture . LDH Cytotoxicity Detection kit ( MK401 , Takara Bio ) was used as per manufacturer instructions to detect cell killing . Cells were labelled with one of the volumetric dyes ( all from ThermoFisher ) : CMFDA ( 250 nM ) ) , CMRA ( 500 nM ) or Deep Red ( 100 nM ) for 20 min in complete media . For imaging CD69 activation , the collagen mixture described above was supplemented with a BV421 labelled anti-CD69 at 1 μg/ml ( adjustments in volume are made to the media to maintain the same final collagen concentration ) . For calcium imaging , T-cells were loaded with 1 μM of Fluo4-AM ( ThermoFisher ) for 20 min at 37°C . Cells were imaged using either a Dragonfly Spinning Disk system , a Perkin Elmer Spinning disk or an Olympus FluoView FV1200 confocal microscope using a 30x Super Apochromat silicone oil immersion objective with 1 . 05 NA . All microscopes included an environmental chamber to maintain temperature at 37°C and CO2 at 5% . Time-lapse images were acquired every 30 or 60 s . z-Scans were acquired every 3 μm . SLB imaging was performed on an Olympus IX83 inverted microscope equipped with a TIRF module and Photomertrics Evolve delta EMCCD camera using an Olympus UApON 150x TIRF N . A 1 . 45 objective . Microscopy data from collagen gels was rendered and analysed using built-in tools in IMARIS software ( Bitplane ) . Speeds bellow 2 μm/min were considered to correspond to stationary cells . Synapse images were analysed using Fiji . Image analysis for micro-contact printing data was done using TIAM ( Mayya et al . , 2015 ) as described in Mayya et al . ( 2018 ) . In brief , cells were tracked using DIC images . IRM and fluorescence signal-positive segments of the tracks were used to calculate the attachment rate and the number of cells interacting with stimulatory spots to extract the half-life of interactions where a simple first order rate was used dA/dt=-k[A]; where A is cell number and k is the off-rate constant representing the exit from the spots by dissolution of the IS . Data from IMARIS and FlowJo were plotted and analysed in Prism ( GraphPad ) , where all statistical tests were performed .	The human immune system protects the body from infection or cancer by detecting foreign and abnormal elements , known as antigens , and initiating a response to clear them . It relies on a type of white blood cell called a T-cell to distinguish which substances are the body’s own , and which are infectious . Each T-cell is designed to attack a specific antigen , which they recognise using a unique 'T-cell receptor' . During an infection , immune cells called 'antigen presenting cells' hold out antigens for the T-cells to look at . Only when an antigen matches the T-cell's receptor can the T-cell get activated and trigger an immune response . There are still gaps in our understanding about how human T-cells interact with antigen presenting cells . Since only a small number of T-cells in the human blood have the same receptor , it is difficult to collect the large number of identical T-cells needed to study this interaction . In addition , it is impractical to image how these interactions occur in a living human body . Now , Abu-Shah et al . have developed a new system that engineers human T-cells to have the same specific receptor . T-cells collected from human blood received the genetic information for identical receptors via a technique called electroporation . This involves mixing the cells with a single-strand copy of the receptor gene and then applying electric pulses to make the cell membranes leaky so the code for the receptor can get inside the cells . To study the interaction between these genetically engineered T-cells and antigen presenting cells , Abu-Shah et al . created a three-dimensional system that mimics the environment T-cells normally experience inside the body . T-cells cultured using this system behaved similarly to immune cells in the human body , and displayed the characteristics needed to trigger an immune response . With this new system , researchers could recreate other aspects of the human immune response outside of the body , incorporating different types of immune cells and different genetic modifications . Not only could this improve our understanding of the human immune system , it could also be used as a way to screen specific drugs during pre-clinical studies .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "tools", "and", "resources", "immunology", "and", "inflammation" ]	2019	A tissue-like platform for studying engineered quiescent human T-cells’ interactions with dendritic cells
Although pain is a common symptom of various diseases and disorders , its contribution to disease pathogenesis is not well understood . Here we show using murine experimental autoimmune encephalomyelitis ( EAE ) , a model for multiple sclerosis ( MS ) , that pain induces EAE relapse . Mechanistic analysis showed that pain induction activates a sensory-sympathetic signal followed by a chemokine-mediated accumulation of MHC class II+CD11b+ cells that showed antigen-presentation activity at specific ventral vessels in the fifth lumbar cord of EAE-recovered mice . Following this accumulation , various immune cells including pathogenic CD4+ T cells recruited in the spinal cord in a manner dependent on a local chemokine inducer in endothelial cells , resulting in EAE relapse . Our results demonstrate that a pain-mediated neural signal can be transformed into an inflammation reaction at specific vessels to induce disease relapse , thus making this signal a potential therapeutic target . Multiple sclerosis ( MS ) is genetically a T helper cell ( CD4+ T cell ) -mediated autoimmune disease ( International Multiple Sclerosis Genetics Consortium et al . , 2011 ) that is characterized by chronic inflammation in the central nervous system ( CNS ) and relapse in over two thirds of patients ( Steinman , 2009 ) . Such relapse is marked by inflammatory lesions consisting of various immune cells including CD4+ T cells , CD8+ T cells , B cells , and macrophages followed by a loss of neurological function in various regions of the CNS ( Steinman , 2014 ) . To understand the molecular mechanisms involved in the relapse , disease model animals including experimental autoimmune encephalomyelitis ( EAE ) mice have been designed to study the corresponding molecular mechanism ( Miller et al . , 1995; Steinman , 2009 ) . However , a detailed explanation for the relapse is still lacking . Pain induction is a symptom of MS ( Thompson et al . , 2010 ) and sometimes a determinant of MS activity ( Ehde et al . , 2003 ) . Pain induction is commonly interpreted as a biomarker of inflammation , although inflammation can sometimes occur in the CNS without it ( Watson et al . , 1991 ) . Yet a correlation between pain intensity and the pathogenesis of inflammation is poorly established and still debated ( Beiske et al . , 2004; Kalia and O'Connor , 2005; Solaro et al . , 2004; Ehde et al . , 2006 ) . More specifically , it has not been demonstrated whether pain induction is involved in MS relapse . One critical machinery for inflammation development is a local chemokine inducer in non-immune cells termed the inflammation amplifier ( formerly IL-6 amplifier ) , which we originally discovered in several disease models including EAE ( Ogura et al . , 2008; Hirano , 2010 ) . In this system , massive chemokine expression caused by the simultaneous stimulation of NFkB and STAT3 is followed by disruption of local homeostasis due to the local accumulation of various immune cells ( Murakami and Hirano , 2011; Murakami et al . , 2011; Lee et al . , 2012 , 2013 ) . Activation of the inflammation amplifier is associated with many human diseases and disorders such as autoimmune and neurodegenerative diseases including MS ( Murakami et al . , 2013; Atsumi et al . , 2014 ) . Gateways for pathogenic CD4+ T cells in an EAE model have been established by activation of the inflammation amplifier via an anti-gravity-mediated regional neural signal in the fifth lumbar cord ( L5 ) ( Arima et al . , 2012; Mori et al . , 2014 ) . Soleus muscle-mediated sensory-sympathetic neural signals enhance inflammation amplifier activation in L5 dorsal vessels via local norepinephrine expression derived from sympathetic neurons to create gates through which immune cells reach the CNS and cause inflammation there ( Arima et al . , 2012 ) . These gates can be artificially manipulated via the activation of local neural signals by electrically stimulating muscles , resulting in an accumulation of immune cells in the intended regions ( Arima et al . , 2012 ) . Gating blood vessels by regional neural stimulations , a phenomenon termed the gateway reflex , is an example of the neural signaling-mediated regulation of inflammation ( Tracey , 2012; Sabharwal et al . , 2014 ) and has potential therapeutic value not only in preventing autoimmunity , but also in augmenting the effects of immunotherapies against infections and cancers ( Ogura et al . , 2013; Atsumi et al . , 2014 ) . Because it is known that neural activations are induced by various situations including pain ( Delmas et al . , 2011; Palazzo et al . , 2012; LaMotte et al . , 2014 ) , we considered whether neural signals mediated by pain induction plays a role in the pathogenesis of EAE relapse via the gating of certain blood vessels . In the present paper , we induced a pain stimulus in EAE-recovered mice that had minimal EAE symptoms . The result was EAE relapse . Mechanistic analysis demonstrated that a neural signal via sensory-mediated sympathetic activation leads to an accumulation of MHC class II+CD11b+ cells at ventral vessels of the L5 cord ( L5 ventral vessels ) in a manner dependent on CX3CL1 chemokine expression followed by an accumulation of various immune cells including pathogenic CD4+ T cells . These results demonstrate that pain-mediated neural signals risk reactivating inflammation via the accumulation of immune cells at specific sites and therefore may offer a new therapeutic target for relapse in chronic inflammatory diseases like MS . We have previously employed a passive transfer method for EAE induction ( Ogura et al . , 2008; Arima et al . , 2012 ) . We isolated myelin oligodendrocyte glycoprotein ( MOG ) specific CD4+ T cells from C57BL/6 mice having an immunization with MOG plus CFA and stimulated the resulting CD4+ T cells with the MOG peptide and antigen presenting cells in vitro . The activated MOG-specific CD4+ T cells ( pathogenic CD4+ T cells ) , which include Th1 and Th17 cells , were intravenously injected into wild type C57BL/6 mice . EAE was induced within 1 week and disappeared at about 2–3 weeks after the transfer ( Figure 1A ) . We refer to this disappearance as the remission phase and these mice as EAE-recovered . In contrast , we found no relapse even more than 300 days after the pathogenic T cell transfer ( Figure 1A ) . Consistent with these results , it is known that the lack of natural relapses in the C57BL/6 mouse strain . Thus , our EAE model develops a transient clinical symptom upon the transfusion of pathogenic CD4+ T cells that is followed by the remission phase . 10 . 7554/eLife . 08733 . 003Figure 1 . Pain induction causes EAE relapse . Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . ( A ) EAE development without pain induction ( n = 8–10 per group ) . ( B ) EAE development with or without pain induction on the same day of T cell transfer ( day 0 , thick arrow ) ( n = 3–5 per group ) . ( C ) Immunohistochemical staining for cfos in the somatosensory area of wild type mice and EAE-recovered mice and EAE-recovered mice with pain induction ( n = 2–3 per group ) ( top ) . Quantification of the histological analysis of the ACC ( bottom ) . ( D ) EAE development with or without pain induction 20 days ( thick arrow ) after T cell transfer in EAE-recovered mice ( n = 3–5 per group ) . ( E ) EAE development in an EAE relapse induced by pain model in the presence or absence of Gabapentin ( day 22–34 , thin arrows ) . Pain was induced 22 days after the T cell transfer in EAE-recovered mice ( thick arrow ) ( n = 3–5 per group ) . ( F ) EAE development with or without capsaicin treatment ( day 21–23 , thin arrows ) in the whiskers or forefeet ( n = 3–5 per group ) . ( G ) EAE development in a relapsing-remittent model in the presence Gabapentin ( red ) , Pregabalin ( blue ) , or saline ( black ) everyday from day 15 after immunization ( thin arrows ) ( n = 4–5 per group ) . Mean scores ± SEM are shown . * , p < 0 . 05 , , p < 0 . 01 , * , p < 0 . 001 , n . s . , not significant . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 00310 . 7554/eLife . 08733 . 004Figure 1—figure supplement 1 . Ligation of the middle branch of trigeminal neurons increased von Frey test values . Von Frey test scores in the presence or absence of ligation ( n = 8–10 per group ) . Mean scores ± SEM are shown . * , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 00410 . 7554/eLife . 08733 . 005Figure 1—figure supplement 2 . Mice treated with capsaicin enhanced the accumulation of MHC class II+CD11b+ cells and cfos expression in neurons of the somatosensory area without pain induction . Immunohistochemical staining for MHC class Ⅱ in the L5 cord with administration of capsaicin in the whiskers or forefeet region ( n = 4 per group ) . Dotted circles show accumulated cells at the ventral vessels ( left ) . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( middle ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right ) . Immunohistochemical staining for cfos in the somatosensory area of EAE-recovered mice with capsaicin treatment ( 8 hr ) in the whiskers or forefeet ( n = 3–5 per group ) . Quantification of the histological analysis of the somatosensory area based on cfos MFI ( right bottom ) . Mean scores ± SEM are shown . , p < 0 . 01; *** , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 005 To show whether pain induction is involved in the development of EAE , we ligated the middle branch of the trigeminal nerves , which is composed of only sensory neurons ( Thygesen et al . , 2009 ) . Although this pain induction did not accelerate disease development , particularly during the onset period ( until 10 days after the pathogenic T cell transfer ) , the pain persisted and maintained symptoms at a high level from 15 days post T cell transfer compared with sham mice ( Figure 1B ) . We found that pain induction itself ( without pathogenic T cell transfer ) did not induce the development of EAE ( Figure 1B ) and that pain induced in mice with pathogenic T cells gradually caused a low clinical score over 20 days after the transfer ( Figure 1B ) . We found negligible cfos expression in neurons of the somatosensory area , which is a key area involved in pain sensation ( Li et al . , 2010 ) , in the brain of EAE-recovered mice and negligible pain sensation according to the von Frey test of the extremity , trunk of the body , face , head ( Figure 1C and data not shown ) , suggesting that remittent mice have minimal pain . Then we investigated whether pain induction was involved in the relapse of EAE . We induced pain in EAE-recovered mice that had developed transient EAE . Pain induction resulted in a relapse of EAE ( Figure 1D ) . Finally , it is important to investigate whether pain induction itself was involved in the EAE relapse . Pain medicines such as Gabapentin and/or Pregabalin , which reduced cfos expression in neurons of the somatosensory area or von Frey test values after pain-induction ( data not shown ) , suppressed the development of EAE relapse ( Figure 1E ) . On the other hand , capsaicin , a painful agent , injected into the whiskers or forefeet caused EAE relapse ( Figure 1F ) . A relapsing-remittent model of SJL/PLP mice was also suppressed by pain medicines ( Figure 1G ) . Thus , pain induction developed relapse in EAE models . We further investigated whether the sensory pathway is involved in the pain-induced EAE relapse . Activated neurons , as defined by NeuN and cfos expressions , expressed the sensory markers TRPV1 and/or Nav1 . 8 in trigeminal ganglions . We found negligible cfos expression , however , in the subtypes of these markers in the absence of ligation in the remittent phase ( Figure 2A ) . Immunohistochemical experiments using serial sections showed that cfos+-activated neurons also expressed TRPV1 and/or Nav1 . 8 ( Figure 2B ) . Treatment with a Nav1 . 8 blocker , A803467 , suppressed the development of the EAE relapse ( Figure 2C ) , and TRPV1-deficient hosts showed suppressed relapse ( Figure 2D ) . We also found that animals treated with A803467 and TRPV1-deficient hosts showed reduced cfos expression in neurons of the somatosensory area or low von Frey values after pain-induction ( Figure 2—figure supplement 1 , right panels and data not shown ) . Moreover , ligation of the facial nerves , which mainly contain motor nerves and the vagus nerves , which mainly contain efferent fibers such as parasympathetic and motor nerves to various viscera induced negligible development of EAE relapse ( Figure 2—figure supplement 2 and data not shown ) . Therefore , we concluded that the activation of sensory pathways are involved in the pain-induced relapse . 10 . 7554/eLife . 08733 . 006Figure 2 . Sensory activation is involved in EAE relapse . ( A ) Percentages of cfos+ cells among NeuN+ cells in the presence or absence of ligation in the trigeminal ganglions of wild type mice ( n = 2–3 per group ) . ( B ) Percentages of cfos+ cells with sensory markers TRPV1 and/or Nav1 . 8 in the presence or absence of ligation of the trigeminal ganglions of wild type mice ( n = 2–3 per group ) . ( C ) EAE development with or without A803467 treatment ( day 22–30 , thin arrows ) in the presence of pain induction 22 days ( thick arrow ) after T cell transfer in EAE-recovered mice ( n = 4–5 per group ) . ( D ) EAE development in TRPV1 deficient mice in the presence of pain induction 21 days ( arrow ) after T cell transfer in EAE-recovered mice ( n = 4–5 ) . Mean scores ± SEM are shown . * , p < 0 . 05 , * , p < 0 . 001 , n . s . , not significant . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 00610 . 7554/eLife . 08733 . 007Figure 2—figure supplement 1 . Mice treated with A803467 and TRPV1-deficient hosts suppressed the accumulation of MHC class II+CD11b+ cells and cfos expression in neurons of the somatosensory area after pain induction . Immunohistochemical staining for MHC class II in the L5 cord with pain induction ( 8 hr ) and treatment with A803467 in wild type or TRPV1-deficient EAE-recovered mice ( left top ) ( n = 3–5 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( bottom right ) . Immunohistochemical staining for cfos in the somatosensory area of EAE-recovered mice with pain induction ( 8 hr ) and treatment with A803467 administration in wild type or TRPV1-deficient EAE-recovered mice ( top right ) ( n = 3–5 per group ) . Quantification of the histological analysis of the somatosensory area based on cfos MFI ( right bottom ) . Mean scores ± SEM are shown . , p < 0 . 01; *** , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 00710 . 7554/eLife . 08733 . 008Figure 2—figure supplement 2 . Mice treated with ligation via vagus nerves did not induce the accumulation of MHC class II+CD11b+ cells . Immunohistochemical staining for MHC class II in the L5 cord with ligation ( 8 hr ) via vagus or trigeminal nerves 21 days after T cell transfer in EAE-recovered mice ( n = 4–5 per group ) . Dotted circles show accumulated cells at the ventral vessels ( top ) . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right bottom ) . Mean scores ± SEM are shown . * , p < 0 . 05 , , p < 0 . 01 , * , p < 0 . 001 , n . s . , no significance . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 008 We then investigated whether sympathetic activation is also involved in EAE relapse after pain induction . We first investigated blood flow speed , which is controlled by autonomic neurons including sympathetic ones , in the dorsal vessels of L5 and L1 , the bottoms of the forefeet , the hindlimb , femoral vessels , and brain surface vessels in pain-induced mice . Blood flow speed was faster in pain-induced than sham-operated mice ( Figure 3A ) . These results suggested that sympathetic activation affects blood vessels in the CNS after pain induction . Consistent with these results , neurons of the somatosensory area , particularly the anterior cingulate cortex ( ACC ) , where sensory nerves are localized with autonomic nerves including sympathetic ones ( Ikemoto et al . , 1999; Critchley , 2005 ) , were activated according to cfos expression ( Figure 1C ) . Moreover , sympathetic ganglions were activated after pain induction , with the activation status in L5 being higher than in other cords ( Figure 3B ) . Furthermore , treatment with atenolol , an inhibitor of the norepinephrine β1 receptor , significantly suppressed the relapse development ( Figure 3C ) , and treatment with 6-OHDA , which induces sympathectomy , suppressed the EAE relapse ( Figure 3D ) , while mice treated with atenolol and 6-OHDA did not suppress cfos expression in neurons of the somatosensory area after pain induction ( Figure 3—figure supplement 1 ) . Moreover , EAE mice under immobilization stress or forced swimming did not show EAE relapse , although these conditions did increase serum corticosterone , norepinephrine , and epinephrine similarly to pain induction ( Figure 3E , F ) . Therefore , these data suggest specific sympathetic pathways triggered by pain induction , but not pain-mediated systemic hormonal responses and/or stress-mediated events , play a role in the development of EAE relapse . 10 . 7554/eLife . 08733 . 009Figure 3 . Sympathetic activation is involved in pain-mediated EAE relapse . Pain was induced in wild type C57BL/6 mice or EAE-recovered mice 20 days after pathogenic T cell transfer . ( A ) Blood flow speeds in blood vessels of various organs were measured 2 days after pain induction ( closed bars ) or in sham operated mice ( open bars ) ( n = 3–5 per group ) . ( B ) Activation of sympathetic neurons in the first or fifth lumbar sympathetic ganglia was evaluated by cfos expression . Immunohistochemical staining for tyrosine hydroxylase ( TH ) is also shown ( red ) ( n = 3 per group ) . Quantification of the histological analysis of the somatosensory area based on cfos MFI ( right ) . ( C ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . Pain was induced 22 days later ( thick arrow ) , and EAE development was evaluated with or without atenolol treatment every day from day 22 in EAE-recovered mice ( thin arrows ) ( n = 3–5 per group ) . ( D ) EAE development with or without 6-OHDA treatment at day 19 and day 21 ( thin arrows ) upon pain induction at day 23 ( thick arrow ) after T cell transfer in EAE-recovered mice ( n = 3–5 per group ) . ( E ) EAE development with pain induction ( diamonds ) , immobilization stress ( circles ) , or forced swimming ( FST ) ( triangles ) every day from 21 days after T cell transfer in EAE-recovered mice ( thin arrows ) ( n = 4–5 per group ) . ( F ) Serum concentrations of corticosterone , norepinephrine , and epinephrine with pain , immobilization stress or forced swimming 21 days after T cell transfer in EAE-recovered mice ( n = 4–5 per group ) . Mean scores ± SEM are shown . * , p < 0 . 05 , , p < 0 . 01 , * , p < 0 . 001 , n . s . , not significant . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 00910 . 7554/eLife . 08733 . 010Figure 3—figure supplement 1 . Mice treated with atenolol and 6-OHDA did not suppress cfos expression in neurons of the somatosensory area after pain induction . Immunohistochemical staining for cfos in the somatosensory area of EAE-recovered mice with pain induction and treatment with atenolol or 6-OHDA administration ( top ) ( n = 3–5 per group ) . Quantification of the histological analysis of the somatosensory area ( bottom ) . Mean scores ± SEM are shown . n . s . , not significant . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 010 We hypothesized that inflammation could be induced in the L5 cord after pain induction , because the disease symptoms of the EAE relapse after pain induction were similar to the symptoms developed by primary EAE whose inflammation was initially developed in the L5 cord after pathogenic CD4+ T cell transfer ( Arima et al . , 2012 ) . Many cell populations increased in the L5 cord after pain induction ( Figure 4A ) , especially those of CD11b+ cells and particularly MHC class II+ ones , in EAE-recovered mice . Furthermore , the accumulated CD4+ T cells in early phase ( day 1–2 ) after pain induction were mainly Th17 cells , while in later time points ( day 5 and later ) after pain induction , we found not only other helper subset in L5 cord and lymphocytes accumulations in several regions of brain just like primary EAE responses ( data not shown ) . 10 . 7554/eLife . 08733 . 011Figure 4 . Pain induction accumulates MHC class II+CD11b+ cells at the ventral vessels of L5 via the anterior cingulate cortex in EAE-recovered mice . ( A ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . Pain was induced 20 days later in EAE-recovered mice ( EAE-recovered mice ) . The L1 or L5 cord was isolated at days 0 , 1 , 2 , and 4 after pain induction ( n = 3–4 per group ) , and the corresponding number of cells was evaluated using a flow cytometer . Because L1 spinal cord has about sevenfold bigger volume than L5 cord according to an 11 . 7 tesla MRI ( Mori et al . , 2014 ) ( L1 average volume , 7 . 1 mm3 and L5 average volume , 1 . 0 mm3 ) , immune cell numbers of L1 cord were divided by 7 . 1 . Upper Figures , CD11b+ cells ( blue ) ; CD11b+MHC class II+ cells ( red ) . Lower Figures , CD4+TCR+ cells ( blue ) ; CD8+TCR+ cells ( red ) ; B220+ cells ( green ) ; NK1 . 1+TCR- cells ( violet ) ; NK1 . 1+TCR+ cells ( sky blue ) ; γδTCR+ cells ( orange ) ; Gr1+ cells ( light green ) ; CD11c+ ( light violet ) . Cell numbers ± SEM are shown . Statistical comparisons were made with wild type . ( B ) Relative infiltration of MHC class II+CD11b+ cells in the L5 cord of C57BL/6-SJL mice based on parabiosis experiments using C57BL/6 mice ( CD45 . 2+ ) and C57BL/6-SJL mice ( CD45 . 1+ ) in the presence ( EAE-recovered ) or absence ( WT ) of EAE development induced by MOG-specific pathogenic CD4+ T cell transfer ( n = 4–5 per group ) . 30 days after the T cell transfer , L5 cords of EAE-recovered C57BL/6-SJL hosts were evaluated . CD45 . 2+ cells are peripheral derived cells from another body . ( C ) Immunohistochemical staining for MHC class II in the L5 cord with or without pain induction and treatment with clodronate liposome administration in the cisterna magna ( CM ) or the peritoneal cavity ( ip ) ( top ) ( n = 3 per group ) . Dotted circles show accumulated cells at the ventral vessels ( top ) . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right bottom ) . ( D ) EAE development was evaluated with clodronate liposome administration into the cisterna magna ( CM ) or the peritoneal cavity ( ip ) day 21 after pathogenic T cell transfer ( thin arrows ) in the presence of pain induction in EAE-recovered mice ( n = 3–5 per group ) . Pain was induced 23 days later ( thick arrow ) . ( E ) Immunohistochemical staining for cfos in the somatosensory area of EAE-recovered mice with pain induction and MK801 administration to the somatosensory area ( left top ) ( n = 3 per group ) . Quantification of the histological analysis of the somatosensory area based on cfos MFI ( bottom ) . Immunohistochemical staining for MHC class II in the L5 cord with pain induction and MK801 administration to the somatosensory area ( right top ) ( n = 3 per group ) . Dotted circles show accumulated cells at the ventral vessels ( top ) . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( bottom left ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( bottom right ) . ( F ) Immunohistochemical staining for cfos in the somatosensory area of EAE-recovered mice with L-Homocysteic acid administration to the somatosensory area ( left top ) ( n = 3 per group ) . Quantification of the histological analysis of the somatosensory area based on cfos MFI ( left bottom ) . Immunohistochemical staining for MHC class II in the L5 cord with L-Homocysteic acid administration to the somatosensory area ( right top ) ( n = 3 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right bottom ) . Mean scores ± SEM are shown . * , p < 0 . 05; , p < 0 . 01; * , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 011 We performed parabiosis experiments using CD45 . 1 and CD45 . 2 hosts to investigate whether MHC class II+CD11b+ cells in the L5 of EAE-recovered mice originated from resident microglial cells or peripheral monocytes . We found similar percentages of CD45 . 1+ cells and peripheral derived CD45 . 2+ cells in MHC class II+CD11b+ cells in the L5 cord of EAE remittent parabiosis hosts , while MHC class II+CD11b+ cells in the L5 cord of wild type parabiosis mice had only a low number of CD45 . 1+ cells ( Figure 4B ) . These results suggested that the majority of MHC class II+CD11b+ cells in the L5 cord of EAE remittent hosts originated from peripheral monocytes but not resident microglial cells . Pain induction resulted in a strong accumulation of MHC class II+CD11b+ cells in the L5 cord , particularly at the ventral vessels ( Figure 4C , ligation ) . We measured MFI and the number of MHC class II+ cells in the dotted circles . Moreover , depletion of MHC class II+CD11b+ cells in the cisterna magna but not in the peritoneal cavity suppressed the accumulation of MHC class II+CD11b+ cells at L5 ventral vessels and the relapse development after pain induction ( Figure 4C , D ) . These results demonstrated that the accumulation of MHC class II+CD11b+ cells , which were likely derived from the CNS side , at the L5 ventral vessels is important for EAE relapse . We next investigated the connection between the sensory and sympathetic pathways , which is critical for the accumulation of MHC class II+CD11b+ cells at the L5 ventral vessels . Specific sensory afferents begin at the brainstem and travel via sympathetic efferent pathways through ACC . Neurons in the ACC are activated by NMDA receptors . We found an antagonist of NMDA receptors , MK801 , injected to the ACC not only suppressed the expression of cfos but also the accumulation of MHC class II+CD11b+ cells at the ventral vessels in the L5 cord after pain induction ( Figure 4E ) , while an agonist of NMDA receptors , L-Homocysteic acid , injected to the ACC enhanced the expression of cfos and the accumulation of MHC class II+CD11b+ cells at the ventral vessels in the L5 cord without ligation of the trigeminal nerves ( Figure 4F ) . These results suggested that the pain-mediated neural activation was at least in part dependent on the ACC . Thus , it is possible that specific sensory afferents to the brainstem are important for sympathetic efferent pathways , which are critical for the accumulation of MHC class II+CD11b+ cells . We next investigated how pain stimulation accumulates MHC class II+CD11b+ cells at L5 ventral vessels . We hypothesized that sympathetic-mediated chemokine expressions might be involved . TH + sympathetic nerves were innervated and Creb molecules , a transcriptional factor of norepinephrine , were activated around L5 ventral vessels ( Figure 5A , B ) . We also found chemokine expression around the ventral vessels ( Figure 5C ) . Treatment with atenolol and 6-OHDA-mediated sympathectomy suppressed the accumulation of MHC class II+CD11b+ cells but not cfos expression in neurons of the somatosensory area after pain induction ( Figure 5D , E , Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 08733 . 012Figure 5 . Pain mediated CX3CL1 expression via the sympathetic pathway is critical for the accumulation of activated MHC class II+CD11b+ cells at L5 ventral vessels in EAE-recovered mice . Immunohistochemical staining for ( A ) TH ( tyrosine hydroxylase ) , phospho-CREB ( B ) , merging of phospho-CREB with nuclei ( B′ ) or CX3CL1 ( C ) in the ventral side of the L5 cord using serial sections ( n = 3–5 per group ) . White dotted polygons indicate the shape of the L5 ventral vessel . Arrows show TH , phosphor-CREB , and CX3CL1 staining around the ventral vessels . ( D and E ) Immunohistochemical staining for MHC class II in the L5 cord with pain induction and treatment with atenolol ( D ) or 6-OHDA ( E ) ( top ) ( n = 3 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right bottom ) . ( F ) The expression of CX3CR1 on CD11b+MHC class II+ cells of the spinal cord in healthy wild type mice ( WT , left and red in right ) and EAE-recovered mice ( EAE-recovered , middle and blue in right ) ( n = 4 per group ) . ( G ) CX3CL1 mRNA expression in the ventral vessels was investigated by real time PCR with or without pain induction in EAE-recovered mice ( n = 3–5 per group ) . ( H ) CX3CL1 expression was enhanced in the presence of norepinephrine . Primary CD11b+ cells in wild type mice ( n = 2 ) were stimulated with norepinephrine for 24 hr . Culture supernatants were collected and assessed using an ELISA specific for mouse CX3CL1 . ( I ) Immunohistochemical staining for MHC class II in the L5 cord with pain induction and treatment with an anti-CX3CL1 antibody , which was injected into the cisterna magna ( top ) ( n = 3–5 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( right bottom ) . ( J ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . Pain was induced 23 days later ( thick arrow ) , and EAE development was evaluated with or without anti-CX3CL1 antibody injection into the cisterna magna in EAE-recovered mice on day 23 ( thin arrow ) ( n = 4–5 per group ) . Mean scores ± SEM are shown . * , p < 0 . 05; *** , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 01210 . 7554/eLife . 08733 . 013Figure 5—figure supplement 1 . CX3CL1 expression in CD11b+ cells was enhanced in the presence of norepinephrine in a manner dependent on β1 and β2 receptors . Primary CD11b+ cells isolated from β1 and β2 receptor deficient mice ( n = 2 per group ) were stimulated with norepinephrine for 24 hr . Culture supernatants were collected and assessed using an ELISA specific for mouse CX3CL1 . Mean scores ± SEM are shown . * , p < 0 . 05 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 013 To identify which specific chemokines are involved in the sympathetic-mediated accumulation of MHC class II+CD11b+ cells in the L5 cord , we first investigated the expression of chemokine receptors on MHC class II+CD11b+ cells in EAE-recovered mice , finding that the expression of CX3CR1 molecules and their ligand , CX3CL1 , were increased particularly after pain induction ( Figure 5F , G ) . Immunohistochemistry experiments confirmed that CX3CL1 molecules were found around the L5 ventral vessels ( Figure 5C ) . Norepinephrine treatment increased CX3CL1 expression in primary CD11b+ cells ( Figure 5H ) , while norepinephrine-mediated CX3CL1 expression was suppressed in CD11b+ cells derived from β1 and β2 receptor-deficient mice ( Figure 5—figure supplement 1 ) . Indeed , blockade of CX3CL1 by a neutralizing antibody in the CNS significantly suppressed the L5 accumulation of MHC class II+CD11b+ cells as well as the relapse development after pain induction ( Figure 5I , J ) . These results demonstrated that pain-mediated CX3CL1 chemokine expression via the sympathetic pathway is critical for the accumulation of MHC class II+CD11b+ cells at the L5 ventral vessels of EAE-recovered mice . MHC class II+CD11b+ cells in EAE-recovered mice expressed not only MHC class II molecules but also costimulatory molecules ( Figure 6A ) and had the ability to stimulate pathogenic CD4+ T cells without the addition of MOG peptide ( Figure 6B ) , suggesting MHC class II+CD11b+ cells presented self-antigen peptides that stimulated pathogenic CD4+ T cells . We then investigated the relationship between MHC class II+CD11b+ cells and pathogenic CD4+ T cells in EAE relapse after pain induction . Anti-CD4 antibody treatment significantly suppressed the development of the EAE relapse ( Figure 6C ) despite the accumulation of MHC class II+CD11b+ cells ( Figure 6D ) . Blockades of sympathetic activation and CX3CL1 expression suppressed the accumulation of not only MHC class II+CD11b+ cells , but also pathogenic CD4+ T cells at L5 ventral vessels ( Figure 6E ) . These results are consistent with the idea that pain induction first causes an accumulation of MHC class II+CD11b+ cells at the ventral vessels of the L5 cord and then a MHC class II+CD11b+ cell-mediated accumulation and activation of pathogenic CD4+ T cells , the latter being also critical for the development of EAE relapse . 10 . 7554/eLife . 08733 . 014Figure 6 . Pain-mediated accumulation of MHC class II+CD11b+ cells and CD4+ T cells at L5 ventral vessels is important for the development of EAE relapse . ( A ) The expression of MHC class II , CD80 , and CD86 on CD11b+ cells in the spinal cord of wild type mice ( red ) and EAE-recovered mice ( blue ) ( top ) ( n = 4 per group ) . Mean Fluorescence Intensity ( bottom ) . ( B ) CD11b+ cells isolated from EAE-recovered mice but not wild type mice have the potential of antigen presentation to CD4+ T cells without peptide addition ( n = 2 per group ) . Culture supernatants were collected and assessed using an ELISA specific for mouse IL-2 . ( C ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . EAE development was evaluated 19 days later with or without intra-peritoneal injection of an anti-CD4 antibody ( days 19 , 21 , and 23 , thin arrows ) and pain induction ( day 19 , thick arrow ) in EAE-recovered mice ( n = 4–5 per group ) . ( D ) Immunohistochemical staining for MHC class II in the L5 cord with pain induction ( 8 hr ) and treatment with anti-CD4 antibody ( using the same hosts as ( C ) ) ( top ) ( n = 4–5 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( bottom left ) and cell number of the accumulated MHC class II+ cells in dotted circles by using serial frozen sections stained with anti-MHC class II antibody and Hoechst ( bottom right ) . ( E ) Immunohistochemical staining for CD4 in the L5 cord with or without pain induction ( 8 hr ) and treatment with anti-CX3CL1 antibody or atenolol ( top ) ( n = 3–5 per group ) . Dotted circles show accumulated cells at the ventral vessels . Quantification of the histological analysis around the ventral vessels based on MFI in dotted circles ( left bottom ) and cell number of the accumulated CD4+ cells in dotted circles by using serial frozen sections stained with anti-CD4 antibody and Hoechst ( right bottom ) . ( F ) CCL20 and CCL5 mRNA expressions at L5 ventral vessels was evaluated with or without pain induction in EAE-recovered mice ( left and middle ) ( n = 3 per group ) . Quantification of the histological analysis for p65+pSTAT3+ endothelial cells around ventral vessels based on serial frozen sections in EAE-recovered mice with or without pain induction ( right ) ( n = 3 per group ) . ( G ) Immunohistochemical staining for CD4 and MHC class II in the L5 cord with pain induction and treatment with anti-IL-6Rα antibody or anti-IL-17A antibody ( n = 4–5 per group ) . Quantifications of the histological analysis around ventral vessels based on serial frozen sections ( CD4 [left] and MHC class II [right] ) are shown . ( H ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . EAE development was evaluated from day 0–32 after pathogenic T cell transfer with or without anti-IL-6 receptor α antibody treatment ( days 20 , 22 , and 24 , thin arrows ) upon pain induction ( day 20 , thick arrow ) in EAE-recovered mice ( n = 4–5 per group ) . ( I ) Pathogenic CD4+ T cells isolated from EAE mice were intravenously transferred into wild type C57BL/6 mice . EAE development was evaluated 15 days later with or without intra-peritoneal injection of an anti-IL-17A antibody treatment ( days 15 , 17 , and 19 , thin arrows ) or pain induction ( day 15 , thick arrow ) in EAE-recovered mice ( n = 4–5 per group ) . Mean scores ± SEM are shown . * , p < 0 . 05; , p < 0 . 01; * , p < 0 . 001 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 01410 . 7554/eLife . 08733 . 015Figure 6—figure supplement 1 . MHC class II+CD11b+ cells in the CNS of EAE-recovered mice expressed TNFα and IL-1β . Expression of TNFα , and IL-1β in MHC class II+CD11b+ cells in the L5 cord of EAE-recovered mice ( n = 3 per group ) . Mean scores ± SEM are shown . * , p < 0 . 05; ** , p < 0 . 01 . Experiments were performed at least 3 times; representative data are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08733 . 015 Because activated pathogenic CD4+ T cells express various cytokines including NFkB and STAT stimulators like IL-17 and IL-6 , we considered whether chemokines were induced in the L5 ventral vessels via activation of the inflammation amplifier , a local chemokine inducer in endothelial cells . L5 ventral vessels indeed had increased expressions of various chemokines , including CCL20 and CCL5 , in EAE-recovered mice particularly after pain induction ( Figure 6F ) . NFkB and STAT3 molecules were simultaneously activated around L5 vessels after pain induction ( Figure 6F ) . The blockade of IL-6 , IL-17A , or CCL20 signaling suppressed the pain-mediated accumulation of pathogenic CD4+ T cells as well as the development of the EAE relapse without affecting MHC class II+CD11b+ cell accumulation ( Figure 6G–I and data not shown ) . These results are consistent with the idea that excess expression of chemokines at L5 ventral vessels is critical for the development of EAE relapse via pathogenic CD4+ T cell accumulation . We here demonstrated that pain-mediated sympathetic pathways induce relapse following sensory activation in EAE models . To induce pain , we mainly used nerve injury via the trigeminal nerves and painful agents including capsaicin injected into the whiskers or forefeet . To clarify if the relapse responses were due to neural signals and not pain-mediated systemic hormonal responses or stress-mediated events , we first investigated sensory and sympathetic pathways , finding that sensory neurons expressing TRPV1 and/or Nav1 . 8 , sensory-sympathetic connections in the ACC , and sympathetic neurons were involved in the response . On the other hand , EAE mice under stress conditions did not develop EAE relapse , although the stress inductions did increase serum corticosterone , norepinephrine , and epinephrine similarly to pain induction ( Figure 3E , F ) . These data suggest specific sensory-sympathetic signals triggered by pain induction plays a role in the development of EAE relapse . We could divide the relapse responses into at least four steps . The first step is pain-mediated sensory activation followed by sympathetic activation . We found various sensory neurons are activated after pain induction ( Figure 2B ) . TRPV1 deficiency or treatment with a Nav1 . 8 blocker , A803467 , suppressed the development of the EAE relapse and its related events including the accumulation of MHC class II+CD11b+ cells around L5 ventral vessels and cfos expression at the neurons of the somatosensory area after pain induction ( Figure 2C , D , and Figure 2—figure supplement 1 ) . Consistent with these results , we showed that the injections of a TRPV1 agonist , capsaicin in EAE-recovered mice induced the development of EAE relapse and related events , phenotypes that are consistent with those induced by pain in the trigeminal nerves ( Figure 1F and Figure 1—figure supplement 2 ) . The second step is pain-mediated sympathetic activation . This activation triggers the regional expression of the chemokine CX3CL1 followed by the accumulation of MHC class II+CD11b+ cells at the L5 ventral vessels . Because Creb was activated around the L5 ventral vessels and atenolol and 6-OHDA-mediated sympathectomy suppressed the accumulation of MHC class II+CD11b+ cells ( Figure 5B , D , E ) , we further concluded that the norepinephrine pathway plays a role . CD11b+ cells in the CNS are likely to express CX3CL1 , because these cells induced CX3CL1 after norepinephrine stimulation , at least in vitro ( Figure 5H and Figure 5—figure supplement 1 ) . The number of MHC class II+CD11b+ cells was higher in the L5 cord than other cords , particularly in EAE-recovered mice , even without pain induction ( Figure 4A ) . The L5 specificity of the relapse development may be because the L5 cord is the site of initial inflammation and is the tissue most damaged during the primary development of EAE . Moreover , it is known that a certain level of the sympathetic pathway is activated by anti-gravity responses even at steady state ( Arima et al . , 2012 ) . Therefore , we hypothesized that a low but sufficient level of sympathetic activation may maintain a low level of the inflammation state in the L5 cord to accumulate MHC class II+CD11b+ cells in the L5 cord of EAE-recovered mice and develop EAE relapse after pain induction . The third step for EAE relapse is the accumulation of pathogenic CD4+ T cells at the L5 ventral vessels . This step is most likely dependent on self-antigen presentation via MHC class II+CD11b+ cells , because these cells stimulated MOG-specific pathogenic CD4+ T cells ( Figure 6B ) . The fourth and final step is mediated by excessive chemokine expressions and is most likely triggered by the activation of the accumulated pathogenic CD4+ T cells in the third step . Pathogenic CD4+ T cells transferred into mice include Th17 and Th1 cells and express various cytokines beyond IL-17 and IFNγ , including IL-6 , TNFα , etc . , all of which are stimulators of NFkB and STATs ( Ogura et al . , 2008; Lee et al . , 2012; Murakami et al . , 2013; Atsumi et al . , 2014 ) . Moreover , we have previously shown that endothelial cells express IL-6 after norepinephrine stimulation ( Arima et al . , 2012 ) and MHC class II+CD11b+ cells in the CNS of EAE-recovered mice expressed various cytokines ( Figure 6—figure supplement 1 ) . Therefore , we suggest that the local chemokine expression at L5 ventral vessels via various cytokines subsequently enhances an accumulation of immune cells , as shown in Figure 4A . Indeed , we found blockades of IL-6 , IL-17A , or CCL20 signals suppressed EAE relapse ( Figure 6H , I , and data not shown ) . The excessive chemokine induction results in excess immune cell migration , which compromises local homeostasis and can trigger inflammatory diseases . Thus , accumulation of both MHC class II+CD11b+ cells and pathogenic CD4+ T cells at the ventral vessels of the L5 cord stimulate the excessive expression of chemokines in non-immune cells to trigger the EAE relapse . Although the flow cytometry data in Figure 4A showed that all cell types were recruited at similar rates and did not accumulate in distinct waves , it should be pointed out that the sensitivity of the FACS analysis is low with regards to the MHC class II+CD11b+ cells and CD4+ T cells found in the immunohistochemistry . Therefore , FACS analysis will not reveal sequential steps . The results of the blocking experiments , however , do show obvious four steps: Regarding the first step , blockades of sensory pathways by using A803467 , TRPV1-deficient hosts , and MK801 at the ACC region suppressed somatosensory activations , sympathetic activations , and MHC class II+CD11b+ cell accumulation in the ventral vessels , and EAE-relapse . Regarding the second step , we used atenolol and 6-OHDA to suppress sympathetic activation and MHC class II+CD11b+ cell accumulation in the ventral vessels and/or EAE-relapse . However , this treatment did not affect sensory-mediated somatosensory activations , suggesting that sympathetic activation is a downstream event after sensory activation . Also regarding the second step , blockade of CX3CL1 did suppress MHC class II+CD11b+ cell accumulation in the ventral vessels and EAE-relapse but not somatosensory activations and sympathetic activations , suggesting that CX3CL1 expression is a downstream event after sensory and sympathetic activation . Regarding the third step , anti-CD4 antibody application suppressed CD4+ T cell accumulation in the ventral vessels and EAE-relapse but not somatosensory activations , sympathetic activations , or MHC class II+CD11b+ cell accumulation in the ventral vessels , suggesting that CD4+ T cell accumulation is a downstream event of sensory and sympathetic activation and of MHC class II+CD11b+ cell accumulation in the ventral vessels . MHC class II+CD11b+ cell accumulation in the ventral vessels as observed within 8 hr . Regarding the fourth step , blockades of IL-17 and IL-6 signal by neutralizing antibodies suppressed EAE-relapse but not somatosensory activations , sympathetic activations , MHC class II+CD11b+ cell accumulation in the ventral vessels , or CD4+ T cell accumulation , suggesting that the inflammation amplifier activation by cytokines including IL-17 and IL-6 is a downstream event after sensory and sympathetic activation , MHC class II+CD11b+ cell accumulation in the ventral vessels , and pathogenic Th17 cell accumulation in the ventral vessels . These results clearly suggested there are the four-steps that occur sequentially and are critical for the development of pain-mediated EAE relapse in our EAE models . Regarding the induction of different chemokine-mediated axes by different sympathetic neurons ( pain in this paper vs anti-gravity in the previous paper [Arima et al . , 2012] ) , we should point out that we used different hosts in these two cases . We used normal mice for anti-gravity experiments , but EAE-recovered mice for pain-mediated ones . In other words , the numbers of MHC class II+CD11b+ cells are completely different: normal mice have almost no MHC class II+CD11b+ cells in the CNS , while EAE-recovered mice have many MHC class II+CD11b+ cells in the CNS particularly in the L5 cord . In EAE-recovered mice , a regional norepinephrine output via pain inductions functions on MHC class II+CD11b+ cells as well as endothelial cells around the ventral vessels of the spinal cords , causing the expression of CX3CL1 and accumulation of MHC class II+CD11b+CX3CR1+ cells in a manner dependent on the CX3CL1-CX3CR1 axis . These accumulated MHC class II+CD11b+CX3CR1+ cells present autoantigens to pathogenic Th17 cells located in the blood stream . On the other hand , in normal mice , a regional norepinephrine output via anti-gravity responses functions just on endothelial cells because of there are almost no MHC class II+CD11b+ cells in the CNS . This case results in CCL20 expression , which causes the accumulation pathogenic Th17 cells that express the CCR6 receptor , IL-17 and IL-6 . We hypothesized pain-mediated sympathetic signaling acts via a high concentration of norepinephrine at the sympathetic neurovascular connection , but not through the systemic induction of hormones such as epinephrine and norepinephrine , which play a major role in the development of EAE relapse , because neither an accumulation of MHC class II+CD11b+ cells at L5 ventral vessels nor EAE relapse in remittent mice that had immobilization stress or forced swimming were observed , even though serum norepinephrine and epinephrine increased just like in the pain-induction condition , which did induce MHC class II+CD11b+ cell accumulation as well as EAE relapse . However , epinephrine and norepinephrine induced CX3CL1 expression dose-dependently in MHC class II+CD11b+ cells , which are also present in spleen and lymphonodes , and endothelial cells and fibroblasts induced various chemokines , including CCL20 , after stimulation by epinephrine and norepinephrine , particularly in the presence of NFkB and STAT3 activation ( Arima et al . , 2012 ) . Therefore , we hypothesized that systemic epinephrine/norepinephrine molecules induced a systemic pro-inflammatory state via chemokine expression . MS GWAS analysis in 2011 showed a clear link between MS patients and the MHC class II-CD4+ T cell axis , but also suppressive effects on the MHC class I-CD8+ T cell axis ( Consortium , 2011 ) , suggesting that MS is a CD4 genetic disease but not CD8 one . On the other hand , it is known that MS active lesions contain many CD8+ T cells sometimes having features of local antigen reactivity ( Tsuchida et al . , 1994; Dressel et al . , 1997; Crawford et al . , 2004 ) . We therefore hypothesized that CD8+ T cell accumulation is mediated by inflammation responses primarily induced by autoreactive CD4+ T cells . Consistently , our results from transfer EAE also showed the accumulations of various immune cells including CD8+ T cells in L5 cord . It is well known that many patients with MS experience pain ( Thompson et al . , 2010 ) . Kalia and O'Connor found that the proportion of patients with a progressive course of MS trend higher among patients with chronic types of pain ( Kalia and O'Connor , 2005 ) . It was also reported that central pain is the first and sole symptom of MS relapse in some patients ( 1 . 6% of all patients , 6% of patients with central pain ) ( Osterberg et al . , 2005 ) . At the same time , several reports have failed to show a clear relationship between pain and disease duration and/or course ( Beiske et al . , 2004; Kalia and O'Connor , 2005; Michalski et al . , 2011 ) . Thus , there does not seem to be a direct correlation clinically between pain intensity and MS disability or lesional load . We hypothesize that there are several reasons why we did not observe a clear relationship between pain and MS progression , which we explain in view of the above four steps . In the first step , differences in pain sensitivity and sensory activation among individuals complicate the investigation . It is known that loss of pain sensitivity may vary with the region or size of the affected site between patients , such that a bigger affected site correlates with less sensitivity in MS patients ( Huber et al . , 1988 ) . In addition , depression occurs with high frequency in MS patients ( Chwastiak et al . , 2002; Siegert and Abernethy , 2005 ) , and this might also affect sensation . In the second step , the degree of sympathetic activation should vary among individual patients following the first step , as too might the connection between the sensory and sympathetic pathways . Moreover , depression in MS patients also affects the sympathetic pathway ( Guinjoan et al . , 1995; Rumsfeld and Ho , 2005 ) . In the third and fourth steps , which are mediated by autoreactive CD4+ T cell-mediated inflammation amplifier activation following MHC class II+CD11b+ cell accumulation , different affected regions in quantity and/or quality between MS patients due to differences in the autoantigen distribution recognized by autoreactive CD4+ T cells and different neural activations in each patient should disrupt the relationship between pain and MS progression . Neural activations , which are critical for the establishment of autoreactive CD4+ T cell gateways in the vessels , might also be affected by various factors in each patient . These reasons might explain the disrupted relationship between pain sensation and MS progression . In our experiments , however , we managed to control the majority of these variations in MS mouse models . Thus , based on our analysis , we concluded that pain-mediated neural signal induces EAE relapse . We propose at least three factors as important for determining the sites of relapse in MS patients: ( 1 ) the existence of excess MHC class II+CD11b+ cells at a particular site , ( 2 ) activation of a sympathetic pathway that accumulates MHC class II+CD11b+ cells in certain blood vessels , and ( 3 ) the presence of organ-specific autoreactive CD4+ T cells in the periphery . Even when MHC class II+CD11b+ cells are accumulated in a specific vessels , the inflammation is not induced without autoreactive CD4+ T cells , which recognized some autoantigens on the accumulated MHC class II+CD11b+ cells . We assume the sites of excess MHC class II+CD11b+ cells and sympathetic activations vary among patients as too the antigens recognized by autoreactive T cells , which would explain the variation in MS relapse regions . In other words , if we correctly detect the site of MHC class II+CD11b+ cell accumulation , the sites of sympathetic activation , and the presence of autoreactive CD4+ T cells in the periphery , we might be able to discern where the MS relapse begins and how to regulate it . C57BL/6 mice were purchased from Japan SLC ( Tokyo , Japan ) , Adrb1-deficient mice , Adrb2-deficient mice and TRPV1-deficient mice were purchased from The Jackson Laboratory ( Bar Harbor , ME ) , C57BL/6-SJL mice were purchased from Taconic ( Germantown , NY ) , and SJL/J mice were purchased from Charles River ( Yokohama , Japan ) . All mice were maintained under specific pathogen-free conditions according to the protocols of the Hokkaido University and the Osaka University Medical Schools . EAE induction was performed as described previously ( Ogura et al . , 2008; Arima et al . , 2012 ) . Briefly , C57BL/6 mice were injected with a MOG ( 35–55 ) peptide ( Sigma–Aldrich , Tokyo ) in complete Freund's adjuvant ( Sigma–Aldrich ) at the base of the tail on day 0 followed by intravenous injection of pertussis toxin ( Sigma–Aldrich ) on days 0 , 2 , and 7 . On day 9 , CD4+ T cells from the resulting mice were sorted using anti-CD4 microbeads ( Miltenyi Biotec , Tokyo ) . The resulting CD4+ T cell-enriched population ( 4 × 106 cells ) was cocultured with rIL-23 ( 10 ng/ml; R&D Systems , Minneapolis , MN ) in the presence of MOG peptide-pulsed irradiated splenocytes ( 1 × 107 cells ) for 2 days . Cells ( 1 . 5 × 107 cells ) were then injected intravenously into wild type mice . Clinical scores were measured as described previously ( Ogura et al . , 2008; Arima et al . , 2012 ) . For the induction of relapsing-remitting EAE , SJL mice were injected with a proteolipid protein ( PLP ) ( 139–151 ) peptide ( TOCRIS , Tokyo ) in complete Freund's adjuvant ( Sigma–Aldrich ) at the base of the tail on day 0 followed by intravenous injection of pertussis toxin ( Sigma–Aldrich ) on days 0 , 2 , and 7 . The severity of EAE was evaluated in a blinded fashion using following scale as previously described ( Moriya et al . , 2008 ) . 0 , normal; 1 , limp tail; 2 , mild paraparesis of the hind limbs with unsteady gait; 3 , moderate paraparesis with preservation of voluntary movement; 4 , paraplegia; 5 , moribund . Animal treatments were performed according to the Guidelines of the International Association for the Study of Pain ( Zimmermann , 1983 ) . Before the experiments , the animals were allowed to habituate to the housing facility for 3 days . Ligation of the middle branch of trigeminal neurons and the von Frey test were performed as described previously ( Figure 1—figure supplement 1 ) ( Krzyzanowska et al . , 2011 ) . In brief , surgical procedures were performed under anesthesia . The right trigeminal neuron was exposed and loosely ligated around the distal part of the nerve with a polyglycolic acid suture ( Akiyama MEDICAL CO . , Ltd , 6-0 , Sapporo , Japan ) , as previously described in rats ( Vos et al . , 1994 ) . Care was taken not to cause excessive compression of the nerve , as this could hamper the emergence of allodynia ( Martin and Avendano , 2009 ) . In sham surgeries ( n = 5 ) , the trigeminal neuron was exposed but left untouched . Tactile allodynia was measured using von Frey filaments . After perfusion with ice cold PBS , spinal cords were dissected and enzymatically digested using the Neural Tissue Dissociation Kit ( P ) ( Miltenyi Biotec , Tokyo ) . CD11b+ cells were isolated by suspending them in MACS buffer and staining them with anti-CD11b microbeads ( Miltenyi Biotec ) followed by separation in a magnetic field using an MS column ( Miltenyi Biotec ) . Spines were harvested and embedded in SCEM compound ( SECTION-LAB Co . Ltd . , Hiroshima , Japan ) and prepared as sections using the microtome device CM3050 ( Leica Microsystems , Tokyo ) and macrotome device CM3600XP ( Leica Microsystems ) with Cryofilm type IIC9 ( SECTION-LAB Co . Ltd . ) . The resulting sections were stained with hematoxylin/eosin or immunohistochemical staining and analyzed with a BZ-9000 microscope ( KEYENCE , Osaka , Japan ) . Analysis was performed by HS ALL software in one fluorescence microscope BZ-II analyzer ( KEYENCE ) . Frozen sections ( 10 μm ) were prepared according to a published method ( Kawamoto , 2003; Arima et al . , 2012 ) . The following antibodies were used for the flow cytometry analysis: FITC-conjugated anti-CD19 ( eBioscience , Tokyo ) , anti-Gr1 ( eBioscience ) , anti-CD80 ( eBioscience ) , anti-CD45 . 2 ( eBioscience ) , PE-conjugated anti-TCRβ ( eBioscience ) , anti-NK1 . 1 ( eBioscience ) , anti-I-A/I-E ( BioLegend , Tokyo ) , anti-CD86 ( eBioscience ) , anti-CD193 ( CCR3 ) ( BioLegend ) , anti-CMKLR1 ( eBioscience ) , PE-Cy7-conjugated anti-CD8 ( eBioscience ) , anti-CD3 ( eBioscience ) , anti-CD45 . 1 ( eBioscience ) , eFluor450-conjugated anti-CD45 ( eBioscience ) , anti-CD4 ( eBioscience ) , APC-conjugated anti-CD4 ( BioLegend ) , anti-γδTCR ( eBioscience ) , anti-CD11c ( eBioscience ) , anti-I-A/I-E ( BioLegend ) , anti-CD45 . 2 ( eBioscience ) , biotin-conjugated anti-CD11b ( eBioscience ) , anti-CX3CR1 ( Abcam , Tokyo ) , anti-CD195 ( CCR5 ) ( eBioscience ) , anti-CD197 ( CCR7 ) ( eBioscience ) , anti-CD183 ( CXCR3 ) ( eBioscience ) , anti-CD184 ( CXCR4 ) ( eBioscience ) , and anti-CD185 ( CXCR5 ) ( eBioscience ) . The following antibodies were used for immunohistochemistry: anti-phospho-STAT3 ( Tyr705 , D3A7 ) , anti-phospho-NFkB anti-phospho-p65 , anti-phospho-CREB ( Cell Signaling , Tokyo ) , anti-tyrosine hydroxylase ( Abcam ) , anti-cFos ( Sigma–Aldrich ) , control rabbit IgG ( DA1E ) ( Cell Signaling ) , anti-CX3CL1 ( Abcam ) , anti-Nav1 . 8 antibody ( Abcam ) , anti-VR1 antibody ( Abcam ) , anti-NeuN antibody ( Millipore , Tokyo ) , biotin-conjugated anti-CD4 ( BioLegend ) , anti-CD11b ( eBioscience ) , anti-I-A/I-E ( BioLegend ) , anti-CD86 ( BioLegend ) , Alexa Fluor 488 goat anti-rabbit IgG ( H + L ) , Alexa Fluor 546 goat anti-rabbit IgG ( H + L ) , Alexa Fluor 647 goat anti-chicken IgG ( Invitrogen , Tokyo ) , and Streptavidin Alexa Fluor 546 conjugate ( Invitrogen ) . The following antibodies were used for in vivo neutralization: purified anti-mouse CCL20 mAb , anti-mouse IL-17 Ab , and anti-CX3CL1 Ab ( R&D Systems ) . The anti-CD4 antibody was purified as described previously ( Ueda et al . , 2006 ) . The anti-IL-6 receptor antibody was obtained from Chugai Pharmaceutical Co ( Tokyo , Japan ) . Atenolol , capsaicin , 6-Hydroxydopamin hydrochloride , A-803467 , Norepinephrine , MK801 , and L-Homocysteic acid were purchased from Sigma–Aldrich . Gapapentin was purchased from Tokyo Chemical Industry ( Tokyo ) . Pregabalin was purchased from Taconic ( Tokyo ) . The VECTASTAIN Elite ABC Rabbit IgG Kit and the DAB Peroxidase Substrate Kit were purchased from Vector Laboratories ( Burlingame , CA ) . CX3CL1 and IL-2 levels in cell culture supernatants were determined using ELISA kits from R&D Systems and eBiosciences , respectively . Norepinephrine and epinephrine levels in serum were determined using EIA kits from Labor Diagnostika Nord ( Nordhorn , Germany ) and corticosterone levels in serum using EIA kits from Abnova ( Taipei , Taiwan ) . To generate single cell suspension , spinal cords were dissected and enzymatically digested using the Neural Tissue Dissection Kit ( Miltenyi Biotec ) , and 106 cells were incubated with fluorescence-conjugated antibodies for 30 min on ice for cell surface labeling . The cells were then analyzed with cyan flow cytometers ( Beckman Coulter , Tokyo ) . The collected data were analyzed using Summit software ( Beckman Coulter ) and/or Flowjo software ( Tree Star , Ashland , OR ) . Immunohistochemistry was performed as described previously with slight modifications ( Lee et al . , 2012 ) . Approximately 100 frozen sections ( 15 μm ) were fixed with acetic acid/ethyl alcohol ( 1:19 ) for 15 min followed by PBS-washing for 10 min . Tissues around the ventral vessels in the sections were collected by a laser micro-dissection device , DM6000B ( Leica Microsystems ) , and prepared for total RNA measurements by the GenElute Mammalian Total RNA Kit ( Sigma–Aldrich ) and Ethachinmate ( Nippon Gene , Tokyo ) . A GeneAmp 5700 sequence detection system ( ABI , Tokyo ) , KAPA PROBE FAST ABI Prism qPCR Kit ( Kapa Biosystems , Boston , MA ) , and KAPA SYBR FAST ABI Prism qPCR Kit ( Kapa Biosystems ) were used to quantify the levels of CCL20 mRNA , CCL5 mRNA , CX3CL1 mRNA , IL-1β mRNA , TNFα mRNA , and HPRT mRNA . The PCR primer pairs used for real-time PCRs using KAPA PROBE FAST ABI Prism qPCR Kit were as follows: mouse HPRT primers , 5′-AGCCCCAAAATGGTTAAGGTTG-3′ and 5′-CAAGGGCATATCCAACAACAAAC-3′ , probe , 5′-ATCCAACAAAGTCTGGCCTGTATCCAACAC-3′; mouse CCL20 primers , 5′-ACGAAGAAAAGAAAATCTGTGTGC-3′ and 5′-TCTTCTTGACTCTTAGGCTGAGG-3′ , probe , AGCCCTTTTCACCCAGTTCTGCTTTGGA; mouse CX3CL1 primers , 5′-CGTTCTTCCATTTGTGTACTCTGC-3′ and 5′-AGCTGATAGCGGATGAGCAAAG-3′ , probe , 5′-TCAGCACCTCGGCATGACGAAATGCG-3′; and mouse CCL5 primers , 5′-CTCCCTGCTGCTTTGCCTAC-3′ and 5′-CGGTTCCTTCGAGTGACAAACA-3′ , probe , 5′-TGCCTCGTGCCCACGTCAAGGAGTATT-3′ . The PCR primer pairs used for real-time PCRs using the KAPA SYBR FAST ABI Prism qPCR Kit were as follows: mouse HPRT primers , 5′-GATTAGCGATGATGAACCAGGTT-3′ and 5′-CCTCCCATCTCCTTCATGACA-3′; mouse IL-1β primers , 5′-TTGACGGACCCCAAAAGATG-3′ and 5′-TGGACAGCCCAGGTCAAAG-3′; and mouse TNFα primers , 5′-TACTGAACTTCGGGGTGATCGGTCC-3′ and 5′-CAGCCTTGTCCCTTGAAGAGAACC-3′ . The conditions for real-time PCR were 40 cycles at 95°C for 3 s followed by 40 cycles at 60°C for 30 s . The relative mRNA expression levels were normalized to the levels of HPRT mRNA . In some experiments , one of anti-CCL20 antibody ( 200 μg/mouse ) , anti-IL-17A antibody ( 100 μg/mouse ) , anti-IL-6 receptor-α antibody ( 500 μg/mouse ) , anti-CD4 antibody ( 200 μg/mouse ) , or atenolol ( 500 μg/mouse ) along with clodronate liposomes 300 ( 1 mg/mouse ) , Gabapentin ( 1 mg/mouse ) , and A-803467 ( 200 μg/mouse ) were intraperitoneally injected into EAE-recovered mice that had pain induced at the same time . Gabapentin ( 1 mg/mouse ) and Pregabalin ( 250 μg/mouse ) were intraperitoneally injected into relapsing-remitting EAE induced SJL/J mice . Capsaicin ( 5 μg/mouse ) was subcutaneously injected into the whiskers or forefeet of EAE-recovered mice . Anti-CX3CL1 antibody ( 10 μg/mouse ) , atenolol ( 10 μg/mouse ) , and clodronate liposomes 300 ( 100 μg/mouse ) ( Katayama Chemical Industries , Osaka ) were infused into the subarachnoid CSF from the cisterna magna ( Xie et al . , 2013 ) . 6-OHDA ( 5 mg/mouse ) was twice intraperitoneally injected into EAE-recovered mice before ligation ( Seeley et al . , 2013 ) . The head of an anesthetized mouse was placed in a stereotaxic device . Fur above the skull was shaved , and the skin was cleaned with 70% ethanol . A 30-gauge needle was lowered toward the ACC ( AP 0 . 7 mm; ML 0 . 3 mm; VD 1 . 75 mm ) , and MK801 or L-Homocysteic acid ( 3 μg/μl and 100 mM , respectively , 0 . 5 μl each delivered over 90 s ) were injected as described previously ( Kim et al . , 2011 ) . Blood flow volume was measured in blood vessels from each tissue in C57BL/6 mice ( 6–8 weeks old ) in the presence or absence of ligation using Omegazone OZ-1 ( Omegawave , Tokyo ) . Naïve CD4 T cells from 2D2 mice and CD11b+ cells from EAE-recovered mice were sorted using anti-CD4 microbeads and anti-CD11b microbeads , respectively ( Miltenyi Biotec ) . The resulting CD4+ T cell-enriched population ( 1 × 105 cells ) was cocultured with the isolated CD11b+ cells ( 5 × 104 cells ) without MOG-peptide addition in a 96 well plate for 3 days . IL-2 levels in cell culture supernatants were determined using ELISA kits ( eBioscience ) . Parabiosis was performed as described previously ( Duyverman et al . , 2012 ) . In brief , the right side of C57BL/6 ( CD45 . 2+ ) mice and the left side of C57BL6-SJL ( CD45 . 1+ ) mice were shaved and sterilized . Matching skin incisions were made from the base of the foreleg to the base of the hind leg of each mouse . The skin flaps of the parabiotic pair were attached by 6-0 polyglycolic acid suture , and the skin was closed with clips to finalize the parabiosis surgery . After 2 weeks , pathogenic CD4 T cells from EAE induced CD57BL/6 ( CD45 . 2+ ) mice were transferred to each mouse . The infiltrated monocyte/macrophage was analyzed after mice recovered from the EAE symptoms . EAE-recovered mice were subjected to immobilization stress in a plastic tube for 30 min/day over 7 days ( Yoshihara and Yawaka , 2013 ) . EAE-recovered mice were subjected to the forced swim model in a tub for 15 min/day over 4 days as described previously ( Stone and Lin , 2011 ) . Student's t tests ( two-tailed ) and ANOVA tests were used for the statistical analysis of differences between two groups and that of differences between more than two groups , respectively . All animal experiments were performed following the guidelines of the Institutional Animal Care and Use Committees of the Institute for Genetic Medicine , the Graduate School of Medicine , Hokkaido University ( Sapporo , Japan ) and the Graduate School of Frontier Bioscience and Graduate School of Medicine , Osaka University ( Osaka , Japan ) .	Multiple sclerosis ( or MS for short ) is a disease in which the insulating covers of nerve cells in the brain and spinal cord become inflamed and damaged . Depending on which nerves are affected , this disease can cause a wide range of symptoms , ranging from numbness and muscle spasms to visual disturbances and chronic pain . Many other diseases and disorders also have pain as a symptom , but it is not well understood if pain itself can directly contribute to the development of disease . Most people with MS will , initially , experience periods when their symptoms get worse ( called ‘relapses’ ) , which are then followed by periods of improvement . Arima , Kamimura et al . investigated whether the sensation of pain itself could trigger a relapse in a mouse model of MS . The experiments showed that a painful sensation could trigger a relapse in the mice via the so-called ‘gateway reflex’ . This reflex describes the phenomenon whereby nerve impulses lead to the release of signaling molecules that cause the walls of nearby blood vessels to open and allow immune cells to move from the bloodstream to the central nervous system . This in turn stimulates the development of inflammation , which causes an imbalance in the affected sites of the central nervous system . These findings demonstrate that pain itself triggers a signal—sent via nerve impulses followed by the release of signaling molecules—that can lead to a relapse; and suggest that interfering with this signal could potentially help to treat to protect against relapses in MS . Following on from this work , it will be important to confirm if the gateway reflex exists in humans , and whether it is linked to other diseases that don't involve the central nervous system .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2015	A pain-mediated neural signal induces relapse in murine autoimmune encephalomyelitis, a multiple sclerosis model
Bacterial biofilms can generate micro-heterogeneity in terms of surface structures . However , little is known about the associated changes in the physics of cell–cell interaction and its impact on the architecture of biofilms . In this study , we used the type IV pilus of Neisseria gonorrhoeae to test whether variation of surface structures induces cell-sorting . We show that the rupture forces between pili are fine-tuned by post-translational modification . Bacterial sorting was dependent on pilus post-translational modification and pilus density . Active force generation was necessary for defined morphologies of mixed microcolonies . The observed morphotypes were in remarkable agreement with the differential strength of adhesion hypothesis proposing that a tug-of-war among surface structures of different cells governs cell sorting . We conclude that in early biofilms the density and rupture force of bacterial surface structures can trigger cell sorting based on similar physical principles as in developing embryos . Physical interactions and in particular mechanical forces are involved in sorting of different cell types . These requirements have been established in the field of embryonic development ( Gonzalez-Rodriguez et al . , 2012 ) . Cells must divide , change shape , adhere to each other , migrate , disperse , or cluster . To complete these processes , they modulate their physical properties including rigidity , position , motility , and strength of adhesion ( Fagotto , 2014 ) . The development of bacterial biofilms resembles eukaryotic development in various aspects including genetic programs that are controlled by master regulators ( Monds and O'Toole , 2009 ) . However , it is unclear to which extent physical forces are involved in determining the architecture of biofilms . An early approach for understanding cell sorting on the basis of physical interactions was the differential adhesion hypothesis ( Steinberg , 1963 ) . This hypothesis states that a population of cells will sort into spatially separated subpopulations , if the strength of cell–cell adhesion differs between cell types . The biophysical explanation for this sorting is the tendency of cell clusters to minimize their surface , similar to the minimization of surface tension in liquids . Depending on the strength of cohesion between cells of the same kind and the strength of adhesion between different kinds of cells , the model predicts perfect mixing , encasement , partial segregation , and full segregation of cells ( Steinberg , 1963; Graner and Glazier , 1992 ) . Some of these morphotypes have been observed experimentally ( Steinberg and Takeichi , 1994; Foty and Steinberg , 2004 ) . Recent experiments indicate that equilibrium descriptions are not sufficient for describing the process of cell sorting during development; they provide evidence that active force generation by cells is involved ( Landsberg et al . , 2009; Gonzalez-Rodriguez et al . , 2012; Kashef and Franz , 2015; Mao and Baum , 2015 ) . In particular , the differential interfacial tension hypothesis introduces cortical contractility in addition to adhesion ( Brodland and Chen , 2000 ) . Interestingly , Harris proposed another hypothesis , the differential strength of adhesion hypothesis ( DSAH ) ( Harris , 1976 ) . He envisions a tug-of-war of cells actively moving along each other by retractions of cellular extensions pulling the cell body into the direction of the strongest and least breakable bonds . Cells with lower bond strength would be squeezed outward to the peripheral layer . Sorting of cells also occurs during bacterial biofilm development . Cell–cell interactions and biofilm architecture are often governed by cell appendages with polymeric organization . One of the most ubiquitous appendages is the type IV pilus ( T4P ) ( Berry and Pelicic , 2015 ) . It mediates surface attachment , surface motility , and its retraction can generate high mechanical force ( Maier and Wong , 2015; Maier , 2013 ) . The T4P is a helical polymer consisting mainly of the major subunit PilE ( Craig et al . , 2004 ) . The major subunit can be post-translationally modified by O-linked glycosylation or by phosphoform-modifications . Cell sorting based on T4P has been observed in different experimental setups . Vibrio cholerae can express a T4P-like system and generate toxin co-regulated pili . Cells lacking the major pilin of the toxin co-regulated pilus did not integrate into wt microcolonies ( Kirn et al . , 2000 ) . Mutations in the major pilin subunits induced different pilus morphology and affected the size of pilus bundles . However , cell sorting based on these mutations was not reported ( Kirn et al . , 2000 ) . Differential fluorescence labeling of Pseudomonas aeruginosa showed that cells with T4P-dependent motility form a cap on top of non-motile stalks formed by cells that lacked the gene for the major pilin subunit ( Klausen et al . , 2003 ) . Swarming of P . aeruginosa is generated by a single flagellum , but moderated by polar T4P ( Anyan et al . , 2014 ) . When T4P production was suppressed , swarming motility increased considerably , most likely due to reduced cellular clustering mediated by pili . When piliated and non-piliated cells were co-cultured , non-piliated cells dominated the swarm edge . Neisseria meningitidis upregulates expression of a phosphotransferase that is required for phospho-form modification of the major pilin upon attachment to endothelial host cells ( Chamot-Rooke et al . , 2011 ) . Simulations of pilus bundling suggest post-translational modifications affecting the physical interactions between T4P ( Chamot-Rooke et al . , 2011 ) . All of these studies imply that T4P can generate a rich diversity of sorting phenomena . However , in contrast to embryonic development , the biophysical basis underlying cell sorting in bacterial biofilms remains elusive . Here , we test the hypothesis that the physical interactions between bacteria govern the morphology of mixed bacterial cell clusters . The T4P of Neisseria gonorrhoeae ( gonococcus ) was used as a model system for systematic variation of cell–cell interactions . We genetically engineered strains with different densities of pili and their abilities to generate force . Moreover , the breakage-force between pili of different cells could be fine-tuned by replacing a gene responsible for pilin post-translational modification . We found that on agar plates , where the dynamics are mostly determined by steric interactions caused by cell division , cells with the lowest pilus density and cells with the lowest pilus-breakage force sorted towards the colony boundary . Sorting was prominent at the front of expanding colonies but within the bulk of the colony sorting was incomplete . In liquid environment where cells actively move by pilus retractions , sorting correlating with pilus density and pilus-breakage force was nearly complete and the resulting morphotypes followed the DSAH . Active pilus retraction was essential for cell sorting , suggesting that the cells sort by a tug-of-war between cells . N . gonorrhoeae and their type IV pili served as a model system for addressing the hypothesis that mechanical interactions control sorting and morphology of mixed clusters ( aka microcolonies ) . Gonococci are peritrichously piliated , that is , they generate pili at random locations ( Holz et al . , 2010 ) . Because T4P are necessary for microcolony formation , variation in the breakage force between pili directly affects cell–cell interactions . There is preliminary evidence that gonococcal microcolonies show surface tension-like behavior . Microcolonies formed by gonococci with retractile pili are spherical ( Higashi et al . , 2009 ) . Upon fusion , two microcolonies rapidly form a sphere with larger radius as expected for liquid drops ( Dewenter et al . , 2015 ) . Importantly , the retraction of T4P generates mechanical force ( Maier et al . , 2002 ) . Multiple pili can coordinate through a tug-of-war mechanism leading to surface motility ( Marathe et al . , 2014 ) . Hence , cell movement inside clusters is conceivable . These properties could support cell sorting as proposed by the DSAH ( Harris , 1976 ) . To test the hypothesis of segregation of cells depending on receptor–ligand pair densities , we engineered gonococcal strains with different pilus densities . To start with , we generated a gonococcal strain that had the gene for the major pilin subunit , pilE , replaced by gfpmut3 and kan ( pilE::gfpmut3 kan ) , henceforth called P− green strain ( Table 1 ) . This strain did not form clusters nor did it interact with piliated bacteria . Thus , within our detection limit , the P− green bacteria show no interactions . Next , we used a strain with three copies of the pilE gene , P++ ( igA1::pilE pilE ermC ) . This strain contains two copies of the pilE gene in addition to the native copy , increasing the pilus density compared to the wild type ( Holz et al . , 2010 ) . Since the pilins are unaltered , the rupture forces between individual pili are expected to be equal to the wt forces . However , since the density of pili is higher , the total attractive force between two bacteria is higher . 10 . 7554/eLife . 10811 . 003Table 1 . Strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 003StrainRelevant genotypeSource/ReferenceVD300wild type , opa- selected–Ng105 P+ greenigA1::PpilE gfpmut3 ermCThis studyNg106 P+ redlctP: PpilE mcherry aadA:aspCThis studyNg081 P− greenpilE::PpilE gfpmut3 kanThis studyNg095 G− greenpglF::PpilE gfpmut3 kanThis studyNg109 P++igA1::pilE pilE ermCThis studyNg110 P++ redigA1::pilE pilE ermC; lctP: PpilE mcherry aadA:aspCThis studyNg118 Q− greenpilQ::m-Tn3cm; igA1::PpilE gfpmut3 ermC; recA6ind ( tetM ) ;This studyNg116 P+ redlctP: PpilE mcherry aadA:aspC; recA6ind ( tetM ) ;This studyNg119 T− greenpilT::m-Tn3cm; igA1::PpilE gfpmut3 ermCThis studyNg120 T− redpilT::m-Tn3cm; lctP: PpilE mcherry aadA:aspCThis studyNg121 T−G− greenpilT::m-Tn3cm; pglF::PpilE gfpmut3 kanThis study In the next step , we investigated the effect of pilin post-translational modification on the breakage force between pili . Wt gonococci bear a disaccharide composed of a hexose residue linked to a proximal 2 , 4-diacetamido-2 , 4 , 6-trideoxyhexose ( HexDATDH ) at serine 63 ( Hegge et al . , 2004 ) . PglF is involved in membrane translocation of lipid-attached carbohydrates ( Hegge et al . , 2004 ) . Deletion of pglF strongly reduces O-linked pilin glycosylation . Hence , we generated strain G− green by replacing pglF with gfpmut3 ( Table 1 ) . When interfering with the post-translational modification of the pilins , we expect that the rupture force between pili is affected . We used laser tweezers for measuring the rupture forces , employing a previously established assay ( Dewenter et al . , 2015 ) . The surface was coated with a crude pilus preparation extracted from the strain of interest . Subsequently , single bacteria ( monococci ) were caught in a laser trap and held close to the T4P-coated surface ( Figure 1A ) . The forces applied by the optical trap kept the cell body at the center of the trap . Type IV pili can generate very large force by retraction when they bind to surfaces ( Maier et al . , 2002 ) . Binding of T4P to the pilus-coated surface and subsequent retraction led to a deflection of the cell body . Upon breakage of the bond between the retracting pilus and the pili on the surface , the cell body is pulled back into the center of the laser trap . Thus , the maximum force between pili attached to the bacterium and the pili attached to the surface can be measured . Henceforth , we will call this force rupture force . The distribution of rupture forces of wt bacteria pulling on wt pili was in agreement with a Gaussian distribution with a maximum at Fwt/wt = ( 39 ± 1 ) pN ( Figure 1B ) . The distribution of rupture forces generated by G− green was shifted to higher values of FG−/G− = ( 46 ± 1 ) pN . Most interestingly , the force generated by G− green onto the surface coated with wt pili , FG−/wt = ( 25 ± 1 ) pN , was considerably lower than the interaction forces between pili of the same kind . Moreover , we monitored the frequency at which bacteria pulled onto the pilus-coated surface and found no significant difference between the strains ( Figure 1C ) , confirming that pilus–pilus rupture force and not the frequency of pilus retractions was responsible for the altered cell–cell interaction in response to the loss of pilin glycosylation . Taken together , we found that FG−/G− > Fwt/wt > FG−/wt . 10 . 7554/eLife . 10811 . 004Figure 1 . Rupture forces between T4P . ( A ) Principle of force measurement . The surface was coated with pili . A single monococcus was trapped in an optical trap . One or multiple pili can bind to pili at the surface . When they retract they deflect the bacterium from the center of the laser trap . The deflection is proportional to the force acting on the bond between the bacterium-associated pili and the pili at the surface . When the bond between the pili is ruptured , the bacterium moved back to the center of the trap . For each retraction event , the maximum force generated prior to rupture was registered as the rupture force . ( B ) Distribution of rupture forces . Full lines: Gaussian fit for F < 65 pN . For F > 65 pN , the linear regime is exceeded and forces are overestimated . ( C ) Distribution of retraction frequencies . Gray bars: wt bacteria on surface coated with wt pili . Red line: G− green bacteria on surface coated with G− green pili . Green line: G− green bacteria on surface coated with wt pili . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 004 In summary , the breakage force between gonococci is fine-tuned by post-translational modification of pilin . Together with mutants of varying T4P density , our T4P toolbox will enable us to test how mechanical forces between cells affect cell sorting . We developed a tool for directly visualizing the spatio-temporal dynamics of segregation between different strains . Gonococci were inoculated onto agar plates at a density low enough to ensure microcolonies arising from individual bacteria . Time-lapse microscopy was used for following the dynamics of their off-spring . This way , we ensured that the population was clonal at the start of the experiment . New clones inside these populations with modifications in T4P were generated as follows . N . gonorrhoeae are naturally competent for transformation , which is the import and inheritable integration of DNA from the environment ( Burton and Dubnau , 2010 ) . Transformation allows for fluorescent labeling in conjunction with manipulation of genes affecting T4P . A control strain was generated which became fluorescent without affecting piliation by integrating the fluorescence reporter into a non-essential gene ( igA1::gfpmut3 ermC ) , henceforth called P+ green strain ( Figure 2A ) . Genomic DNA ( gDNA ) from this strain was isolated and applied onto an agar plate . Each cell of the expanding wt colony can integrate gfpmut3 ermC into the igA1 locus and become fluorescent with a certain probability . Since the fluorescent marker was inheritably integrated , all progeny of a single transformant was fluorescent ( Video 1 ) . In Figure 2B , the first P+ green cell was detected at 5 hr . As expected for range expansion experiments ( Hallatschek et al . , 2007 ) , little mixing between P+ green and wt was observed . Instead , P+ green formed a sector . This observation suggests that the mobility of the bacteria in this assay was mostly determined by steric interactions caused by cell division . We note that T4P-mediated surface motility was observed only within 1–3 bacterial diameters at the expanding front ( Video 2 ) . 10 . 7554/eLife . 10811 . 005Figure 2 . Assay for direct visualization of spatio-temporal dynamics of a new clone and its progeny within an expanding microcolony . ( A ) Strategy for visualization of a new clone . Genomic DNA ( gDNA ) from strain P+ green in which gfpmut3 recombines into igA1 is spread on the agar surface . When a single bacterium imports and integrates the DNA , then the bacterium becomes fluorescent . Type IV pilus ( T4P ) is unaffected . ( B ) Time-lapse of de novo occurring P+ green clone ( orange ) within an expanding wt colony . Scale bar: 10 µm . The arrow denotes the time point when fluorescence is detectable . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 00510 . 7554/eLife . 10811 . 006Video 1 . Spatio-temporal dynamics of de novo occurrence of a P+ clone and its offspring . Chromosomal DNA from strain P+ green is spread on the agar surface . Upon import and integration of DNA into a single bacterium , pili are lost and the bacterium becomes fluorescent . Time-lapse ( Δt = 15 min ) of de novo occurring P+ green clone within an expanding colony . Brightfield image and fluorescence images were merged . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 00610 . 7554/eLife . 10811 . 007Video 2 . Mobility of bacteria on agar plate . P+ green bacteria were inoculated onto an agar plate . Images were acquired at Δt = 10 s . Bacteria residing within the bulk of the microcolony were immobile at this time scale . At the front , individual bacteria are motile . This motility depends on active T4P retraction and was not observed for non-piliated bacteria . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 007 Next , we investigated whether loss of T4P caused segregation . gDNA from the P− green strain was isolated and applied onto an agar plate . The replacement of the pilin gene by gfpmut3 resulted in loss of pili and gain of fluorescence ( Figure 3A ) . P− bacteria arising close to the expanding front were expelled from the bulk of the colony and their offspring spread rapidly along the front , encircling the entire colony ( Figure 3C , D , Video 3 ) . 10 . 7554/eLife . 10811 . 008Figure 3 . Spatio-temporal dynamics of de novo occurrence of clones with reduced pilus density and their offspring . ( A ) Strategy for visualization of pilus-loss in a single bacterium . gDNA from strain P− green in which pilE is replaced by gfpmut3 is spread on the agar surface . Upon import and integration of DNA into a single bacterium , pili are lost and the bacterium becomes fluorescent . ( B ) Strategy for visualization of reduction of pilus-density . gDNA from P+ green strain is spread on the agar surface . Integration of DNA into a single hyperpiliated P++ bacterium carrying two additional copies of pilE in the igA1 locus led to the pilus density decreasing to wt level and to the acquisition of fluorescence . ( C ) Time-lapse of de novo occurring P− green clone ( orange ) within an expanding colony . Scale bar: 50 µm . ( D ) Detail of ( C ) . Scale bar: 10 µm . ( E ) Time-lapse of de novo occurring P+ green clone ( orange ) within an expanding colony of P++ . Scale bar: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 00810 . 7554/eLife . 10811 . 009Video 3 . Spatio-temporal dynamics of de novo occurrence of a P− clone and its offspring . Chromosomal DNA from strain P− green in which pilE is replaced by gfpmut3 is spread on the agar surface . Upon import and integration of DNA into a single bacterium , pili are lost and the bacterium becomes fluorescent . Time-lapse ( Δt = 15 min ) of de novo occurring P− green clone within an expanding colony . Brightfield image and fluorescence images were merged . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 009 Furthermore , we assessed whether reduction of pilus density ( i . e . , the average number of pili per cell ) leads to a segregation of variants . DNA from P+ green cells was isolated and spread onto an agar plate ( Figure 3B ) . Hyperpiliated P++ gonococci were seeded onto the plate . The integration of DNA from P+ green into this strain replaces the two additional pilE copies by a gfpmut3 gene , and as a consequence , the pilus density reduces to wt level . We found that the P+ green cells residing close to the expanding front were expelled from the colony and encased the P++ colony ( Figure 3E ) . So far , the experiments showed that cell sorting occurred at the front of the expanding microcolonies but sorting was incomplete when considering the entire microcolony . We analyzed how the probability that P− green gonococci moved towards the front was dependent on their location within the colony . For systematic analysis of the dynamics of a de novo occurring P− green clone and its offspring , we automatically detected the contour of a sector of fluorescent bacteria generated by a single variant ( Figure 5—figure supplement 1 ) . The location of the fluorescent pixel closest to the expanding front was determined as a function of time ( Figure 4A ) . P− green bacteria residing directly at the front remained at the front and tended to form blebs ( Figure 3C ) . P− green bacteria closer than 3 µm to the front showed a probability of more than 50% to move towards the front ( Figure 4C ) . At times , hopping of gonococci to the front was observed ( e . g . , red line in Figure 4A ) . This probability decreased as a function of distance . In contrast , the control strain P+ green only showed a probability of 50% of residing and staying at the front ( Figure 4B , C ) , which decreased as a function of distance and stayed significantly below the P− green probability for reaching the front . 10 . 7554/eLife . 10811 . 010Figure 4 . Probability that the offspring of a newly arisen P− gonococcus is moving towards the expanding front . ( A ) Location of the bacterium closest to the front of the P− green subpopulation as a function of time . ( 7 out of 148 traces are shown . ) ( B ) Location of the bacterium closest to the front of the P+ green subpopulation as a function of time . ( 6 out of 307 traces are shown . ) ( C ) Probability that the offspring moves towards the front as a function of its distance from the front for P− green ( red ) and P+ green ( black ) . Error bars: standard deviation of three independent experiments weighted with the number of traces per experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 010 Our data show a strongly increased probability of reaching the front for non-piliated bacteria . T4P are responsible for cell–cell interaction in gonococcal colonies . Thus , the mobility of P− gonococci within a P+ colony might be increased . To this end , we measured the spatial variance σ2 of all offspring of a single P− green transformant compared to the variance of the offspring of a single P+ green transformant ( Figure 5 ) inside growing microcolonies . The variance is expected to increase because bacteria are both mobile , but also increase in number within the colony . By rescaling to the size of the microcolony , we found that the variance still increased ( Figure 5—figure supplement 1C , D ) , suggesting that bacteria are mobile . 10 . 7554/eLife . 10811 . 011Figure 5 . Spreading of new clones within an expanding colony . gDNA from P− green was spread on the agar plate and wt cells were seeded . ( A ) Fluorescence time lapse . Yellow lines: front of the expanding colony . Red outlines: boundaries of sectors formed by the offspring of a single transformant . Scale bar: 10 µm . Spatial variance σ2=σx2+σy2 of three sectors as a function of ( B ) time and ( C ) number of offspring ( fluorescent bacteria per sector ) . The colors correspond to the colors of the numbers at 8 . 75 hr . ( D ) Average variance of P− green ( red ) and P+ green ( black ) within the colony as a function of the number of offspring N . Error bars: standard error as obtained from >50 sectors for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 01110 . 7554/eLife . 10811 . 012Figure 5—figure supplement 1 . Analysis of the dynamics of the offspring of a single transformant . ( A ) Time-lapse in brightfield of a single cell growing into a colony . ( B ) Local intensity variance of the same cells . ( C ) Time-lapse of the same colony in brightfield keeping the proportion of colony radius to sub-image constant . Contour lines of emerging fluorescent patches are labeled in red . The last column shows all contour lines normalized by the colony radius and superimposed onto a unit-circle . ( D ) Fluorescence images showing contour of segmentation . The last column shows a superposition of all preceding contour lines ( green ) onto the fluorescence image of the final time-point . ( E ) Increase of fluorescence Itot over time normalized by the fit parameter of single cell fluorescence I0 . The solid line is a fit onto the data . The dashed line gives the inferred number of fluorescent bacteria . The inset shows Itot normalized by I0 and N , that is , the normalized intensity per fluorescent bacterium . All scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 01210 . 7554/eLife . 10811 . 013Figure 5—figure supplement 2 . Segregation dynamics of non-piliated PQ− green from P+ red . ( A ) pilQ deletion strain PQ− green was inoculated at low density with a higher density of P+ red* . ( B ) Zoom of ( A ) . Scale bar: 20 µm . Arrows depict PQ− green bacteria that later spread ring-like along the expanding front . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 013 Next , we tested whether the mobility of P− green within a wt microcolony was increased . For this analysis , only variants that are not touching the front are analyzed ( Figure 5A–C ) . Colony 2 ( green ) shows an example of a sector touching the front at t = 7 . 5 hr . Later time points of such sectors were automatically removed from the analysis . As expected , the variance increased as a function of time ( Figure 5B ) . Different growth rates might obscure the effect of differential physical interactions on cell . Hence , the generation time of gonococci expressing different amounts of the major pilin pilE , which results in varying densities of pili per cell ( Long et al . , 2001; Holz et al . , 2010 ) ( Table 2 , ‘Materials and methods’ ) , was measured . The level of pilin expression strongly affected the growth rate . The P+ green strain had a generation time of tg = 49 ± 3 min . The non-piliated P− green strain had a significantly lower generation time ( tg = 41 ± 1 min ) , whereas the growth rate of the hyperpiliated P++ red strain showed an increased generation time ( tg = 56 ± 4 min ) . In our mobility analysis , we compensated for different generation times , by comparing the variance of P− green with P+ green as a function of the number of offspring ( Figure 5C , D ) . We found that the variance of the P− green cells increases more strongly as a function of the number of offspring than the variance of P+ green cells , indicating an increased mobility of P− green within a colony of wt bacteria . 10 . 7554/eLife . 10811 . 014Table 2 . Generation timesDOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 014StrainGeneration timeTotal number of evaluated coloniesig:A1:PpilE gfpmut3 ermC ( P+ green ) 48 . 6§ ± 2 . 7† min76igA1::pilE pilE ermC; lctP: PpilE mcherry aadA:aspC ( P++ red ) 55 . 6 ± 3 . 7 min33pilE::gfpmut3 kan ( P− green ) 41 . 1 ± 0 . 8 min84pglF::PpilE gfpmut3 kan ( G− green ) 49 . 7 ± 2 . 0 min14pilQ::m-Tn3cm; igA1::PpilE gfpmut3 ermC; recA6ind ( tetM ) ( PQ− green ) 51 . 1 ± 0 . 3 min280lctP: PpilE mcherry aadA:aspC; recA6ind ( tetM ) ( P+ red* ) 49 . 4 ± 0 . 6 min211§Average value of results obtained from at least three independent experiments . Each experiment comprises the analysis of several individual colonies . †Corrected sample standard deviation from at least three independent experiments . Furthermore , segregation of strains with different pilus densities but similar growth rates was assessed . We generated the gonococcal strain PQ− green with a deletion of pilQ . Since the PilQ proteins form the pore through which the pilus is exported , this results in a P− phenotype . The generation time of PQ− green was slightly higher than the piliated strain P+ red* . We inoculated non-piliated PQ− green bacteria at a low density together with a higher density of piliated P+ red* bacteria ( Figure 5—figure supplement 2A , B ) . Similar to P− green , the PQ− green close to the front of the expanding colony was expelled from P+ red* ( Figure 5—figure supplement 2A ) . Once reaching the front , PQ− green expanded along the front and encircled the bulk of P+ red* cells . This control experiment confirms that loss of pili induces segregation independently of decreased generation time . In summary , bacteria with the lowest pilus density segregate to the front of mixed expanding microcolonies . The sorting efficiency strongly depends on the location of the less piliated clone , with highly efficient sorting close to the front . Pilin glycosylation reduces the rupture force between pili ( Figure 1 ) . In the context of the DSAH , we predict that bacteria with glycosylated pili segregate to the front of an expanding microcolony . In strain G− green , the gene encoding for the flippase , pglF ( Aas et al . , 2007 ) , was replaced by a gene encoding for a green fluorescent protein ( pglF::gfpmut3 kan ) . Deletion of pglF reduces pilin glycosylation severely ( Aas et al . , 2007 ) . Importantly , the generation time of tg = 50 ± 2 min of this strain was comparable to the wild-type strain with reporter ( Table 2 ) . If cell sorting was observed , then we could attribute it to altered pilus–pilus interactions . gDNA of G− green was spread onto an agar plate and individual wt cells were inoculated . Wt gonococci that integrated the fluorescence reporter simultaneously lost the glycosylation of their pili . The sectors formed by G− green bacteria could not be discerned from sectors formed by P+ green . Therefore , we inverted the scenario by inoculating G− green gonococci together with gDNA from wt bacteria onto an agar plate ( Figure 6A ) . G− green gonococci that incorporated the extracellular DNA simultaneously lost their fluorescence and gained the ability to glycosylate their pili . Here , glycosylated ( wt ) bacteria segregated at the front of the expanding colony ( Figure 6B ) . 10 . 7554/eLife . 10811 . 015Figure 6 . Spatio-temporal dynamics of de novo occurrence of clones with the ability to glycosylate pili . ( A ) Strategy for visualization of gain of glycosylation in a single bacterium . gDNA from wt is spread on the agar surface . Upon import and integration of wt DNA into a single G− green bacterium , both the fluorescence is lost and pilus glycosylation is gained . ( B ) Time-lapse of de novo occurring wt clone within an expanding G− green colony . Full yellow line: front of the microcolony obtained from the brightfield image . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 015 We conclude that gain of O-linked pilin glycosylation causes segregation of gonococci . The bacteria with glycosylated pili dominate the front of the expanding colony . The underlying cause of sorting is most likely the difference in pilus–pilus rupture forces . The agar plate assay described above is highly useful for visualizing the spatio-temporal dynamics of mixed microcolonies . However , cell sorting is restricted to the expanding front , most likely because the cell mobility is low within the bulk of the colony . We used a complementary assay for characterizing the degree of cell sorting . In liquid environment , T4P retractions ensure rapid cell movement and microcolonies assemble , disassemble , and change their shape within minutes ( Dewenter et al . , 2015 ) . Therefore , cell sorting based on different physical interactions can be investigated neglecting any potential segregation resulting from different growth rates . We mixed bacteria with different T4P and fluorescent markers in liquid and used confocal microscopy for imaging . As control , a suspension of P+ red and P+ green was used . They formed spherical and well-mixed microcolonies ( Figure 7A , Video 4 ) . 10 . 7554/eLife . 10811 . 016Figure 7 . Loss of pili or gain of post-translational modification causes segregation in aqueous environment . Confocal stacks of gonococci mixed in a liquid and incubated on a glass surface for ( 3–5 ) hr . ( A ) P+ red was mixed with P+ green , ( B ) P+ red and P− green , ( C ) P++ red and P+ green , ( D ) P+ red and G− green . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 01610 . 7554/eLife . 10811 . 017Video 4 . Confocal reconstruction of P+ green and P+ red . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 017 When P+ red and P− green were mixed , the wt bacteria formed spherical microcolonies ( Figure 7B , Video 5 ) , while the P− green bacteria did not show interaction amongst themselves or with wt bacteria . 10 . 7554/eLife . 10811 . 018Video 5 . Confocal reconstruction of P+ red and P− green . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 018 Both the DAH and the DSAH predict that if cells with different densities of the same adhesins are mixed , then the cells with higher density will be surrounded by cells with lower density . To test this hypothesis in our bacterial model system , P++ red and P+ green were mixed . We found that P+ green formed a shell surrounding P++ red ( Figure 7C , Video 6 ) . This behavior is consistent with stronger physical interactions among P++ bacteria due to their high-pilus density , lower interactions among P+ bacteria due to lower pilus density , and intermediate interactions at the interface between P++ red and P+ green . 10 . 7554/eLife . 10811 . 019Video 6 . Confocal reconstruction of P++ red and P+ green . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 019 The pglF-deletion strain G− green showed differential rupture forces , namely FG−/G− > Fwt/wt > Fwt/G− ( Figure 1 ) . In a mix of P+ red and G− green , P+ red formed crescents surrounding G− green ( Figure 7D , Video 7 ) . This morphotype agrees remarkably well with the prediction of the DAH ( Steinberg , 1963 ) . 10 . 7554/eLife . 10811 . 020Video 7 . Confocal reconstruction of P+ red and G− green . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 020 We conclude that bacteria sort dependent on pilus density and pilin post-translational modification . The morphologies of the mixed colonies can be inferred from the rupture forces between the pili . All experimental results show that bacterial cell sorting can be explained on the basis of pilus–pilus rupture forces . Cell sorting on agar plates where motility is low hints on the importance of cell mobility for sorting . The observation of almost complete sorting in liquid environment is in agreement with both the passive DAH and the DSAH , the latter includes cellular force generation . We addressed the role of force generation in cell sorting , by transferring T4P-related mutations into a pilT− background ( T− ) . pilT encodes for the T4P retraction ATPase PilT and is essential for pilus retraction ( Wolfgang et al . , 1998 ) . In T− cells , T4P can form , but cannot retract , nor generate force . As a consequence , bacteria are unable to perform active movement . T− red and T− green were mixed and imaged with confocal microscopy ( Figure 8A ) . Unlike wt gonococci , the clusters did not round up to form spherical microcolonies . Furthermore , red and green bacteria did not mix , but only grew into separate clusters . Similarly , inhibition of glycosylation in one strain did not change the morphologies in a mix of T−G− green and T− red cells ( Figure 8B ) nor did they resemble the shape predicted by the DAH and the DSAH . 10 . 7554/eLife . 10811 . 021Figure 8 . Active T4P retraction affects colony morphology . Confocal stacks of gonococci deficient in active force generation by T4P retraction were mixed in a liquid and incubated on a glass surface for ( 3–5 ) hr . ( A ) T− red was mixed with T− green , ( B ) T− red , and T−G− green . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 021 We conclude that active force generation by T4P is required for the cluster morphology predicted by the DAH and the DSAH . The differential adhesion hypothesis , by Malcom Steinberg , predicts the morphologies of cell-clusters formed by two types of cells with different surface adhesins ( Steinberg , 1963 ) . The morphologies we observed in liquid culture are in accordance with the predicted ones . However , active pilus retraction was required for generating these well-defined morphologies . Recently , we provided evidence that gonococci move over surfaces by a tug-of-war of multiple pili ( Marathe et al . , 2014 ) . Therefore , it is likely that bacteria within microcolonies employ a tug-of-war mechanism for segregation ( Figure 9 ) . Based on the pilus–pilus rupture forces , all four morphotyes shown in Figure 7 can be explained . Bacteria with the same composition and number of pili exhibit equal net rupture forces between all cells . Accordingly , we found well-mixed aggregates . Bacteria without pili show no interaction amongst themselves or with wt bacteria , in agreement with Fwt/P− ≈ FP−/P− ≈ 0 . Co-incubation of hyperpiliated P++ and wt led to the formation of a sphere of P++ in the center and a fully surrounding shell of P+ in agreement with FP++/P++ > FP++/wt > Fwt/wt . Finally , a mix of P+ and glycosylation-inhibited G− bacteria with FG−/G− > Fwt/wt > FG−/wt formed a sphere of G− bacteria and with a crescent of P+ partially surrounding it . To conclude , the mixed morphologies coincide with the prediction of the DSAH ( Harris , 1976 ) . 10 . 7554/eLife . 10811 . 022Figure 9 . Model for tug-of-war mechanism of cell sorting . Pili form contacts between bacteria . For simplicity , only single pilus–pilus bonds between bacteria are shown . Pili continuously elongate and retract . During retraction , they generate force on an attached object . When pili bind to other pili and retract they generate force on each other . The probability of bond rupture increases with force . Since the rupture force between G− and G− pili is larger than between wt and G− pili , the latter bond is more likely to rupture . Pilus retraction pulls the cell body in the direction in which the strongest , least breakable bonds are formed . Cells whose bonds are most easily broken are squeezed outward to the peripheral layer . DOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 022 Adding to the above , active forces have been shown to be important for cell sorting during embryonic development . In developing tissues , cortical contractile forces and cell adhesion act antagonistically , opposing each other with contraction tending to minimize and adhesion to maximize the contact area ( Fagotto , 2014 ) . In comparison , the cell body in our bacterial system can be considered as a point-object and the forces of pilus retraction are always attractive . Cell sorting as a consequence of variations in density and post-translational modification is very likely to affect the biology of pathogenic Neisseria species . Neisserial populations generate micro-heterogeneity via phase and antigenic variation of T4P-associated genes ( Rotman and Seifert , 2014 ) . The detection of multiple antigenic variants in the sequence of the major pilin after 2 days of biofilm growth highlights this ( Kouzel et al . , 2015 ) . Apart from the pilin gene , other genes affecting post-translational modification and pilus density are phase-variable ( Marri et al . , 2010 ) . We propose that phase and antigenic variation trigger cell sorting and thereby affect the architecture of biofilms . We have tested the hypothesis that differential physical interactions between bacteria cause cell sorting in early biofilms . We generated a toolbox of gonococcal type IV pili with varying pilus density and pilus–pilus rupture forces . The morphotypes of mixed microcolonies were in remarkable agreement with the predictions of the DSAH proposing that cells sort on the basis of differential net rupture forces . Likewise to embryonic development , our findings suggest mechanical forces govern cell sorting in early biofilm formation . N . gonorrrhoeae was grown overnight at 37°C and 5% CO2 on agar plates containing gonococcal base agar ( 10 g/l Bacto agar [BD Biosciences , Bedford , MA , USA] , 5 g/l NaCl [Roth , Darmstadt , Germany] , 4 g/l K2HPO4 [Roth] , 1 g/l KH2PO4 [Roth] , 15 g/l Bacto Proteose Peptone No . 3 [BD] , 0 . 5 g/l soluble starch [Sigma–Aldrich , St . Louis , MO , USA] ) , and the following supplements: 1 g/l D-Glucose ( Roth ) , 0 . 1 g/l L-glutamine ( Roth ) , 0 . 289 g/l L-cysteine-HCL × H20 ( Roth ) , 1 mg/l thiamine pyrophosphate ( Sigma–Aldrich ) , 0 . 2 mg/l Fe ( NO3 ) 3 ( Sigma–Aldrich ) , 0 . 03 mg/l thiamine HCl ( Roth ) , 0 . 13 mg/l 4-aminobenzoic acid ( Sigma–Aldrich ) , 2 . 5 mg/l β-nicotinamide adenine dinucleotide ( Roth ) , and 0 . 1 mg/l vitamin B12 ( Sigma–Aldrich ) . Before each experiment , gonococcal colonies were resuspended in GC-medium . All strains in this study were generated into the same VD300 background by transformation with either newly created plasmids or gDNA of existing strains . The strains with the fluorescent reporter in the igA1 locus were constructed as follows: the promoter region of pilE ( PpilE ) with a SacI restriction site was amplified from gDNA of gonococcal strain MS11 using primers NK83 and NK135 . gfpmut3 with a SacI restriction site was amplified from the pAM239 using primers NK134 and NK133 . A PCR fusion between PpilE and gfpmut3 was generated using NK83 and NK133 . The product was inserted into SacI site of p2/16/1 ( Wolfgang et al . , 2000 ) , resulting in pIga::PpilE gfpmut3 . This plasmid was used to introduce the gfpmut3 ermC alleles into the iga locus of strain VD300 , generating strain Ng105 iga::PpilE gfpmut3 ermC . The mcherry fluorescent reporter strains were constructed using an integration vector for insertion of alleles between lctP and aspC on the gonococcal chromosome , following a previously reported approach ( Mehr and Seifert , 1997 ) . aspC was amplified from chromosomal DNA using NK63 and NK64 introducing an engineered MCS at 5′ end and a HindIII at 3′ end . lctP was amplified using NK61 and NK62 introducing a HindIII restriction site at the 5′ end and a MCS at 3′ end . The fragments were fused with primers NK61 and NK64 and inserted into HindIII site of pUP6 , resulting in pLA . The aminoglycoside-3′ adenyltransferase ( aadA ) gene , encoding for a streptomycin/spectinomycin adenyltransferase , was amplified from R100 . 1 plasmid with primers NK 59 and NK 60 and cloned into a unique SacI site in the MCS of pLA , resulting in pLAS . PpilE was amplified using NK23 and NK46 generating a FseI restriction site . mcherry was amplified from pSP64 with NK19 and NK51 generating a PacI restriction site . The PCR fusion between PpilE and mcherry , amplified with NK46 and NK51 , was inserted between FseI and PacI sites of pLAS resulting in pLAS::PpilE mcherry . This plasmid was used to introduce the mcherry aadA alleles between lctP and aspC of strain VD300 , generating strain Ng106 lctP:mcherry aadA:aspC . Furthermore , this plasmid was used to introduce mcherry into strain N400 ( Freitag et al . , 1995 ) , finally yielding strain Ng116 recA6ind ( tetM ) ; lctP: PpilE mcherry aadA:aspC . Finally , the plasmid was introduced in Ng109 igA1::pilE pilE ermC , which in turn was generated by transformation with gDNA of N400 pilE+++ ( Holz et al . , 2010 ) , yielding strain Ng110 igA1::pilE pilE ermC; lctp: PpilE mcherry aadA:aspC . The Ng081 pilE::PpilE gfpmut3 kan strain was constructed using a PCR fragment that contains fluorescent gene reporter with a selectable drug marker flanked by ends of the target gene . Individual fragments were amplified using the primers listed in Table 3 by PCR either from chromosomal DNA of gonococcal strain MS11 or from plasmid pIga::PpilE gfpmut3 . The 5′ flanking region of pilE and the region of gfp were amplified from DNA from plasmid pIga::PpilE gfpPmut3 with primers KH1a and KH4 . The kanamycin selectable marker was amplified from plasmid pUP6 using KH5 and KH6 . The 3′ flanking region of pilE was amplified from chromosomal DNA using KH7 and KH8 . The amplified fragments were fused yielding the final fragment pilE::gfpmut3 kan , which was used for transformation . 10 . 7554/eLife . 10811 . 023Table 3 . Primers used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 10811 . 023PrimersSequence 5′–3′KH1aATGCCGTCTGAATTCCGACCCAATCAACACACCKH4GTTCAATCATATGTGACCTCCTCTATTTGTATAGTTCATCCKH5TAGAGGAGGTCACATATGATTGAACAAGATGGATTGCKH6TCACTTACCGTCAGAAGAACTCGTCAAGAAGGKH7TTCTTCTGACGGTAAGTGATTTCCCACGGKH8ATGCCGTCTGAACGCACCGATATAGGGTTTGKH34AAAAAGAATTCATGCCGTCTGAAGCAAAATCGACCTGCACCATCTGATKH35CGGGTGTGTTGATTGGGTCGGTTTTGATGTCCGGTCGGCGGCKH36GCCGCCGACCGGACATCAAAACCGACCCAATCAACACACCCGKH37GAACCGACATAGAAGTAGTCAGGATGATTTTCAGAAGAACTCGTCAAGAAGGCGKH38CCTTCTTGACGAGTTCTTCTGAAAATCATCCTGACTACTTCTATGTCGGTTCKH39AAAGGATCCATGCCGTCTGAACATCAAAAGCGGGCGGGGGNK83TTGAGTCTTCCGACCCAATCAACACACCCGATACN135AGTTCTTCTCCTTTACGCATAAAATTACTCCTAATTGAAAGGGN134TTTCAATTAGGAGTAATTTTATGCGTAAAGGAGAAGAACTTTTCACNK133TTGAGCTCCTATTTGTATAGTTCATCCATGCCNK19TTAGGAGTAATTTTATGGTGAGCAAGGGNK23TCGCCCTTGCTCACCATAAAATTACTCNK46CATTGGCCGGCCTTCCGACCCAATCAACACACCNK51CATTAATTAATTACTTGTACAGCTCGTCCATGCCNK60TTCGGTCTCCACGCATCGTCAGNK61TTTAAGCTTATGGCACTTTTCCTCAGCATATTCCCNK62GAGCTCTTAATTAAATGCATGGCCGGCCCTAGAGGAAGAAAATCATTGCCGCGACNK63GGCCGGCCATGCATTTAATTAAGAGCTCATGTTCTTCAAGCACATCGAAGCCNK64TTTAAGCTTTTACAAGACTTTCACGATGCTTTCGCNK65TTTTAATTAAATGCGTAAAGGAGAAGAACT TTTCACTGGNK66GTAAGGCCGGCCCTATTTGTATAGTTCATCCATGCCATGTGTAATC The Ng095 pglF::PpilE gfpmut3 kan strain was constructed using a PCR fragment that contains fluorescent gene reporter with a selectable drug marker flanked by ends of the target gene . Individual fragments were amplified using the primers listed in Table 3 by PCR either from chromosomal DNA of gonococcal strain VD300 or from plasmid pIga::PpilE gfpmut3 . The 5′ 2flanking region of pglF was amplified from chromosomal DNA with primers KH34 and KH35 . PpilE was generated using the PCR from chromosomal DNA with KH36 and N135 . The gfpmut3 reporter gene was amplified from plasmid pIga::PpilE gfpmut3 with N134 and KH4 . The kanamycin selectable marker was amplified from plasmid pIga::PpilE gfpmut3 using KH5 and KH37 . The 3′ flanking region of pglF including start of pglB was amplified from chromosomal DNA using KH38 and KH39 . The amplified fragments were fused yielding the final fragment pglF::gfpPmut3-kan , which was used for transformation . Strains Ng119 , Ng120 , and Ng121 were generated as follows . gDNA of GT17 pilT::m-Tn3cm was used to transform P+ green , P+ red , and G− green , respectively , yielding T− green , T− red , and T−G− green . Strains Ng118 and Ng116 were generated as follows . The plasmid pIga::PpilE gfpmut3 was used to transform N400 ( Freitag et al . , 1995 ) , yielding P+ red* . The resulting N400 igA1::PpilE gfpmut3 ermC was then transformed with gDNA of Ng055 pilQ::m-Tn3cm , yielding PQ− green . Direct DNA sequencing of PCR products derived from gonococcal transformants was performed by GATC Biotech AG ( Konstanz , Germany ) to verify the insertions and the absence of any other alterations . gDNA was isolated with the DNeasy Blood & Tissue kit ( Qiagen , Hilden , Germany ) from strains that already carry the genes for the fluorescent protein in the locus of interest . Individual colonies of non-fluorescent and fluorescent bacteria were picked from overnight plates and highly diluted in GC-medium yielding a gonococcal suspension with mainly dark and a few fluorescent bacteria . The latter served as a control to analyze the generation time via the rate of fluorescence increase . 4 µl gDNA solution and 1 µl gonococcal suspension were mixed and immediately used to inoculate a drop onto the center of a GC-plate . The drop was allowed to dry and the plate was mounted onto an inverted epi-fluorescence microscope ( Nikon Ti-E , Japan ) equipped with a motorized stage and a custom-built thermo-box . The plate is held in a custom-built chamber with a glass bottom to allow for imaging and CO2 supply , keeping the plate under an atmosphere of 37°C and 5 . 9% CO2 . Time-lapse imaging was done employing a 40×/0 . 6 long-working distance air objective ( Nikon ) capturing images every 15 min over a grid of 120 different positions yielding an imaging area of 2 . 3 by 2 . 5 mm . P+ red and PQ− green colonies were grown overnight on GC-plates . A small suspension with mostly P+ red and a few PQ− green was mixed in GC-medium . A spot of 5 µl of the suspension was used to inoculate GC-plates , yielding a fully close front of bacteria . The range expansion of growing P+ red competing with PQ− green was monitored using the time-lapse microscopy described for the single-cell transformation assay on agar . It should be noted that the higher number of cells in a range expansion inevitably leads to the presence of stochastically varied non-piliated cells . In order to decrease this effect and focus on piliated red cells and non-piliated green cells , strains in a recAind background were employed , which highly reduces the chance of spontaneously occurring P− cells ( Criss et al . , 2005 ) . Time-lapse epi-fluorescence microscopy under an atmosphere with enriched CO2 at 37°C allowed to take videos of gonococcal micro-colonies growing from individual cells ( Figure 5—figure supplement 1A ) . The transformation of gDNA could be visualized by detection of fluorescence resulting from the expression of fluorescent proteins ( Figure 5—figure supplement 1D ) . The recorded videos were processed as follows: Images were taken over a grid of 120 positions . A normalized , smoothed , average of the images of all positions at the beginning of the experiment is used to correct for inhomogeneous illumination . At the beginning only single cells are present , hence only the average illumination is visible in the average image . All positions of a single time-point were stitched together . High local intensity variance in windows of 5× 5 pixels served as indicator for the presence of a growing microcolony , that is , there is a high variation of dark and bright pixels for colonies , but only small variance for pixels without colonies ( Figure 5—figure supplement 1B ) . Thresholding the local variance image gives a binary image indicating the growing colonies . The binary image was further processed by binary dilations , to ensure connected segmentations of colonies , which was followed by binary erosions to reduce any segments due to noise , but , also , to shrink the contour of the segment of a colony , such that the contour lies more on top of the bacteria at the front . After filling the holes in all binary segments , the contour pixels were defined to be the pixels that are affected by a further binary erosion . A circle fit on the contour pixels of each individual segment gives the information of center and radius of the colony . The centers of all circles were tracked to find regions of interests ( ROIs ) of growing microcolonies that have not yet grown into another colony . Fluorescence was segmented as follows: the stitched fluorescence image was convolved with a gaussian kernel of 11 × 11 px with σ = 1 . 5 px . The local intensity background of each ROI was found and subtracted . Thresholding of the background corrected fluorescence inside every ROI allowed to segment the patches of fluorescent bacteria ( Figure 5—figure supplement 1 ) . The radii of the colony turned out to grow exponentially for at least 10 hr . Further image analysis was restricted to this time-period . Given the center and radius of the growing colony at any time , all Cartesian coordinates can be transformed into polar coordinates . The exponential growth of the radius of the colony implies that all distances inside the colony should , also , grow in the same exponential fashion . This can be illustrated by normalizing the polar coordinates by the radius . Figure 5—figure supplement 1C , D show a row of brightfield and fluorescence images , with the circle fit of the colony to fill always the same proportion of the sub-image . As a result , the scale bar shrinks in time , while the apparent colony size stays constant . Due to this normalization , the apparent position of segmented patches , also , stays roughly constant . The last column in ( b ) shows a superposition of all time-points of the normalized coordinates onto a unit-circle at the top , and a superposition of all time-points without normalization at the bottom onto the final fluorescence image . Due to the clear separation of clonal normalized coordinates , a manual selection of areas inside the unit-circle allowed to easily define all coordinates resulting from the same clone . The spatial variance σ2 can now be calculated using the information of background-corrected intensities and the Cartesian and polar coordinates for each individual clone for all time-points . To this end , the total variance σ2 , the orthogonal variances σx2 and σy2 in Cartesian coordinates are expressed as follows:σ2=σx2+σy2 , σx2=∑i=1Nxi2IiItot− ( ∑i=1NxiIiItot ) 2 , σy2=∑i=1Nyi2IiItot− ( ∑i=1NyiIiItot ) 2 , Itot=∑i=1NIi . Our experiment runs during the exponential growth phase . Following a transformation event , the number of fluorescent proteins per cell generated by the offspring of a single transformant is expected to increase and eventually saturate at an equilibrium concentration . Thus , before saturation , the overall fluorescence initially increases more quickly , but then tends to increase exponentially ( Figure 5—figure supplement 1E ) . In our experiments , a fraction of the growing colonies originated from initially fluorescent cells . These colonies expressed the fluorescent reporter at its equilibrium concentration right from the start of the experiment and the rate of fluorescence increase was used to measure the generation time . The development of the total fluorescence Itot of each clone is used to deduce the number of bacteria within a clonal patch . To this end , the rate of fluorescence increase Itot ( t ) can be modeled assuming a constant rate increasing the fluorescence per cell , kprod , and degradation of fluorescence , kdeg , with N being the number of fluorescent cells:dItotdt=kprodN−kdegItot . Thus , the total fluorescence with I0 being the equilibrium intensity of a single cell can be fitted as follows:Itot ( t ) =I0N ( t−t0 ) ( 1−e−2kdeg ( t−t0 ) ) , where: N ( t ) =2t/τgrowth I0 , t0 , and kdeg are fit parameters . The value for τgrowth is taken from the rate of fluorescence increase of those growing colonies , which were fluorescent right from the beginning of the experiment . These colonies expressed the fluorescent reporter at or close to its equilibrium concentration and the rate of fluorescence increase was used to measure the generation time , τgrowth . The model Itot ( t ) fits well onto the data ( Figure 5—figure supplement 1E ) , such that the time at which the clone arose , t0 , together with the generation time , τgrowth , allows to estimate the number of fluorescent bacteria N ( t − t0 ) . Gonococcal microcolonies were cultivated at 37°C in ibidi µ-Slides I0 . 8 Luer . For microcolony development in a continuous-flow chamber , GC-medium with 1% IsoVitaleX was used . Bacteria from overnight plates of each strain were re-suspended in GC to an optical density at 600 nm ( OD600 ) of 0 . 1 and two hundred microliters of each culture were inoculated into flow chambers and left for 1 hr at 37°C to allow attachment to the glass surface . After 1 hr , the flow was resumed and pumped through the flow cells at a flow rate of 3 ml/hr for each channel ( 0 . 2 mm/s ) by using a peristaltic pump ( model 205U; Watson Marlow , Calmouth , UK ) . Gonococcal microcolonies in aqueous environment were visualized by using a Confocal Laser Scanning Microscope ( Nikon Ti-E C1 ) . Images were obtained using a 60×/1 . 2 water objective ( Nikon ) ( pinhole size of 60 µm ) . Dual fluorescence emission was observed using an argon-ion laser with 488 nm for green fluorescent protein and a 543-nm filter of HeNe laser for mCherry fluorescence . To avoid cross-talk between colors , the images were acquired sequentially , each only with corresponding laser excitation . From each fluidic channel , five to six image stacks were acquired randomly down through the channel . 3D visualization of image was done using ImageJ . The optical tweezers were assembled on a Zeiss Axiovert 200 microscope as described previously ( Clausen et al . , 2009; Anderson et al . , 2014 ) . The assay was modified for measuring the force generated by a single surface-attached bacterium as a function of time ( Anderson et al . , 2014; Dewenter et al . , 2015 ) . The trap was calibrated by power spectrum analysis of the Brownian motion of individual monococci assuming a bacterial diameter of 1 µm and was found to have a stiffness of 0 . 14 pN/nm ± 10% . The position of trapped bacteria was detected at 20 kHz using a quadrant photodiode and the sample stage was movable via a combined piezo and electric motor . We selected for monococci , that is , round bacteria since force measurement could be performed with higher accuracy than with diplococci . Using the drag force method , we found that the linear force range of the optical trap was d = 450 nm when gonococci were trapped . Therefore , the maximum force we detected was F = 65 pN . The retraction experiments were performed in a sealed chamber using a low-density suspension of gonococci , allowing for undisturbed measurement of single bacteria over a period of minutes . During the experiment , a single bacterium was trapped near the surface using the optical tweezers . Retraction of surface-bound pili resulted in deflection of the bacterium from the center of the optical trap as measured by the four-quadrant photodiode . The force acting on a bacterium is proportional to its deflection and was calculated using the calibration described above . After several minutes , a detrimental effect of the optical trap on the bacterium became apparent and it was abandoned .	Communities of bacterial cells can live together embedded within a slime-like molecular matrix as a biofilm . This allows the bacteria to hide from external stresses . A single bacterium can replicate itself and develop into a biofilm , and over time the bacterial cells in specific regions of the biofilm will start to interact with their neighbors in different ways . These interactions occur via structures on the surface of the bacterial cells , and the differences in these interactions resemble those that occur as cells specialize during the development of animal embryos . Previous research into embryonic development has shown how differences in the physical interactions between embryonic cells are essential for sorting the cells into their correct locations and shaping the embryo . However , little is known about which processes govern the development of biofilms . Now , Oldewurtel et al . have asked whether differences in the physical interactions between bacteria trigger cell sorting during the early stages of biofilm development . The experiments involved measuring the force required to break the cell–cell connections ( called the ‘rupture force’ ) in biofilms of a bacterium called Neisseria gonorrhoeae . Oldewurtel et al . found that , in agreement with previous predictions , physical interactions were important for sorting bacterial cells into clusters based on the structures on their surfaces . Bacterial cells actively pull on the surface structures of their neighbors , which allows the cells to sort themselves in a tug-of-war fashion . This means that a cell will move in the direction where it can pull the strongest ( i . e . , in the direction where the rupture force is highest ) . While bacteria and embryos use different molecules to generate these pulling forces , these findings indicate that the basic physical principles are similar in both systems . One of the next challenges will be to evaluate how biofilms might benefit from the structures that develop due to cell sorting .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2015	Differential interaction forces govern bacterial sorting in early biofilms
Affinity and dose of T cell receptor ( TCR ) interaction with antigens govern the magnitude of CD4+ T cell responses , but questions remain regarding the quantitative translation of TCR engagement into downstream signals . We find that while the response of mouse CD4+ T cells to antigenic stimulation is bimodal , activated cells exhibit analog responses proportional to signal strength . Gene expression output reflects TCR signal strength , providing a signature of T cell activation . Expression changes rely on a pre-established enhancer landscape and quantitative acetylation at AP-1 binding sites . Finally , we show that graded expression of activation genes depends on ERK pathway activation , suggesting that an ERK-AP-1 axis plays an important role in translating TCR signal strength into proportional activation of enhancers and genes essential for T cell function . The question of how the T cell receptors ( TCRs ) of CD4+ T cells respond to ligands of differing affinities and concentrations with such remarkable specificity is of great interest to the study of immunity . The TCR binds to antigen presented by the molecules encoded by the Major Histocompatibility Complex ( MHC ) such that strength of the TCR signal in response to a particular peptide-MHC complex ( pMHC ) is dependent on all three components– the antigenic peptide , the MHC , and the TCR itself ( Heber-Katz et al . , 1982 ) . Variations in signal efficiency are thus caused by the generated TCR sequence ( Hedrick et al . , 1984; Jerne , 1971 ) , genetic differences in MHC ( Heber-Katz et al . , 1982; Hedrick et al . , 1982 ) , and the peptide being presented ( Solinger et al . , 1979 ) . Even small differences in the number or affinity of these pMHC-TCR interactions are read by the TCR and have important consequences for the nature and extent CD4+ T cell activation; high-affinity interactions lead to inflammatory responses at a lower concentration of antigen , increased Interleukin 2 ( IL2 ) and IFNγ production , and increased proliferation ( Heber-Katz et al . , 1982; Hedrick et al . , 1982; Solinger et al . , 1979; Alexander , 1993; Rogers and Peptide dose , 1999; Rogers et al . , 1998; Sloan-Lancaster et al . , 1993; Tubo et al . , 2013 ) , whereas lower affinity interactions can lead to incomplete phosphorylation of downstream signaling complexes ( Kersh et al . , 1998b; Sloan-Lancaster et al . , 1994 ) , anergy ( Sloan-Lancaster et al . , 1993 , 1994 ) , or TCR antagonism ( Alexander , 1993; Kersh et al . , 1998b ) . The precise result of low-affinity engagement varies with experimental conditions , but in each case , a cellular phenotype distinct from high-affinity engagement is produced . There exists a well-characterized model system for studying the effects of ligand affinity on CD4+ T cell activation: the AND mouse is a strain with a transgenic CD4+ T cell TCR ( Kaye et al . , 1989 ) . This TCR recognizes pigeon cytochrome c ( PCC ) along with synthetic and species-variant cytochrome c oligopeptides ( Solinger et al . , 1979; Rogers and Peptide dose , 1999; Rogers et al . , 1998 ) . Notably , though many of the peptides differ from PCC by a single amino acid , the effects of TCR recognition of the peptides vary greatly . Kinetic parameters and cytokine output of the interaction with many cytochrome c peptides and their analogues have been described ( Rogers and Peptide dose , 1999; Rogers et al . , 1998 ) . Differences in microcluster formation at the membrane have likewise been described ( Varma et al . , 2013 ) . These variable responses to ligands of differing affinity are especially interesting in the context of the digital TCR response . TCR responses have been characterized as digital ( Coward et al . , 2010 ) —that is , signaling downstream of the TCR is either all-on or all-off , such that a given T cell must either be committed to a full response or to no response . Previous work has established this switch-like behavior as observable in terms of extracellular markers such as CD69 ( Das et al . , 2009; Daniels et al . , 2006 ) , ERK pathway component localization ( Das et al . , 2009; Daniels et al . , 2006; Prasad et al . , 2009 ) , NF-κB activation ( Kingeter et al . , 2010 ) , NFAT localization ( Marangoni et al . , 2013; Podtschaske et al . , 2007 ) , cell-cycle entry ( Au-Yeung et al . , 2014 ) , and cytokine production ( Podtschaske et al . , 2007; Huang et al . , 2013 ) . As a result , differences in the magnitude of responses to ligands of varying affinity would be attributed to greater frequencies of T cells responding at the population level , rather than per-cell variability ( Au-Yeung et al . , 2014; Huang et al . , 2013; Zikherman and Au-Yeung , 2015; Butler et al . , 2013 ) . Still , some aspects of the TCR response have been described as analog , or varying in proportion to the strength of signaling: CD3ζ chain phosphorylation ( Kersh et al . , 1998a; Sloan-Lancaster et al . , 1994; Daniels et al . , 2006; Kersh et al . , 1999; Kersh et al . , 1998a ) ; Zap70 activation ( Daniels et al . , 2006; Prasad et al . , 2009 ) ; intracellular calcium concentrations ( Irvine et al . , 2002 ) ; expression of the transcription factor IRF4 ( Man et al . , 2013; Nayar , 2014 ) ; and cell division time ( Marchingo , 2014 ) . It is unclear how these analog components of the TCR response fit in to a digital model . Both the ability of the TCR to discriminate with high resolution between ligands and the digital nature of TCR signaling have been extensively studied at the level of signaling . Downstream of the TCR , a number of signaling pathways govern the molecular response to engagement , allowing T cells to grow , divide , and acquire immune effector functions consistent with the inciting stimulus ( Murphy and Blenis , 2006; O'Sullivan and Immunology , 2015; Proud , 2007; Santamaria and Ortega , 2006; Wang and Green , 2012 ) . AKT and PKCθ interact at the cell membrane and jointly serve to induce nuclear translocation of the pro-inflammatory transcription factor NF-κB , which in turn is able to activate target genes ( Huang and Wange , 2004 ) . In particular , AP-1 , which comprises homo- or heterodimers assembled from proteins of the Fos , Jun , and ATF transcription factor families ( Murphy et al . , 2013 ) , requires both TCR and co-stimulatory signaling ( Rincón and Flavell , 1994 ) , and it is usually activated by the Ras/Raf/Mek/Erk pathway ( Murphy and Blenis , 2006; Schade and Cutting edge , 2004 ) . At least four feedback loops have been identified in thymocytes and peripheral T cells downstream of the TCR ( Coward et al . , 2010; Feinerman et al . , 2008 ) . Collectively , these feedback loops serve to enforce a digital response by either dampening sub-threshold signaling or amplifying above-threshold signaling , resulting in T cell responses that are all-off or all-on , respectively . Despite these insights into the signaling pathways downstream of TCR activation , there is little known about the transcriptional programs that determine the distinct phenotypes resulting from high- versus low-affinity stimulation . In this study , we address the question of affinity at the level of the chromatin . We take advantage of the PCC system to assess the effects of varying the dose and affinity of peptide presentation to CD4+ T cells on enhancer formation and gene expression , giving us a genome-wide picture of how TCR signaling is able to achieve such highly specific responses despite its apparent digital signaling pattern . We find first that the digital/analog dichotomy is too simple , and instead CD4+ T cells respond to ligands of varying dose and affinity by modulating both the frequency of responding cells and the level of activation of responding cells at a single cell level . In other words , activation markers are analog with respect to the strength of TCR signaling when comparing across doses and affinities , but for any single dose and affinity , the overall signaling pattern is digital for the population of cells . We next show that at the population level , the combined effects of analog precision and increasing frequency of responder cells produce gene expression patterns that directly reflect the strength of TCR signaling for a set of activation signature genes . These gene expression patterns can be used to assess CD4+ T cell activation status , and we develop a tool for ranking arbitrary CD4+ T cell populations by activation score . Underlying these gene expression patterns , we find that the enhancer landscape is largely pre-existing , such that TCR engagement results in activation of primed enhancers rather than through selection of new enhancers . Finally , we show that the graded activation score and the expression of activation signature genes are dependent on the amount of phosphorylated ERK activity downstream of the TCR . Together , these results suggest that the degree of ERK activation translates the analog TCR signal resulting from varying the dose and affinity of TCR engagement into downstream gene expression programs . In order to understand the effects of the digital TCR response on the transcriptional landscapes of CD4+ T cells , we first sought to characterize the 'on-state' of the TCR response . CD4+ T cells and CD11c+ antigen presenting cells ( APCs ) were isolated from AND transgenic mice ( Figure 1—figure supplement 1A , B ) and co-cultured for 24 hr with one of a panel of previously described ( Rogers and Peptide dose , 1999; Rogers et al . , 1998 ) peptides at several different doses . Cell activation was then measured at a single cell level using flow cytometry . As expected , for each given peptide and dose , CD4+ T cells followed a digital pattern , appearing either all-on or all-off according to extracellular activation markers such as CD69 ( Figure 1A ) and CD25 ( Figure 1B ) . Increasing the peptide dose or affinity significantly increased the percent of activated cells in the population ( Figure 1C , D ) . However , when we compared across peptides and concentrations , it was clear that the activation level of the on-state cells was not 'all on . ' Gating on CD69+ cells , each different peptide and different dose of a peptide achieved a different amount of CD69 per cell ( Figure 1E ) . Gating on CD25+ cells yields similar results , with varying amounts of CD25 expressed per cell dependent on both the dose and the affinity of the stimulation ( Figure 1F ) . Thus , while under a given condition the CD4+ T cells were either on or off as per classical digital signaling , when comparing across a panel of conditions , the activation level of the on-state cells is analog with respect to the strength of the TCR signal . 10 . 7554/eLife . 10134 . 003Figure 1 . Both frequency of responding cells and per-cell activation levels increase with increasing signal strength . ( A ) Purified AND T cells and CD11c+ APCs were co-cultured in the presence of indicated peptides at the indicated concentrations for 24 hr . Flow cytometry was then used to phenotype the CD4+ T cells . The histograms show CD69 expression of CD4+ T cells . The annotated bar indicates the gate used to identify CD69+ CD4 cells in subsequent figures . ( B ) Gating on CD4+ cells as in 1A , there is a bimodal distribution of CD25 expression resulting from activating levels of high-affinity ( PCC ) or low-affinity ( K99A ) peptides . The annotated bar indicates the gate used to identify CD25+ cells . ( C ) The percent of CD4+ cells that are CD69+ ( using gate shown in 1A ) varies with the peptide presented and concentration of the indicated peptide . ( D ) The percent of CD4+ cells that are positive for the activation marker CD25 varies with both peptide and dose . ( E ) Gating on CD4+ CD69+ cells ( as shown in 1A ) , the geometric mean fluorescence intensity ( MFI ) of CD69 per cell in each condition varies . ( F ) The geometric MFI of CD25 , gated on CD4+ CD25+ cells ( as shown in 1C ) , varies with peptide and dose . ( P-values based on Student’s t test; p<0 . 05 , p<0 . 01 , *p<0 . 001 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 00310 . 7554/eLife . 10134 . 004Figure 1—figure supplement 1 . CD4+ T cells and APCs were purified from AND mice . ( A ) The AND mouse TCR includes the V alpha 11 chain and the V beta 3 chain . The great majority of naïve CD4+ T cells extracted from AND mouse spleens expressed this pair of TCR chains . Naïve CD4+ T cells were isolated using negative selection with the Miltenyi MACS system , as described in the methods section . ( B ) APCs for peptide presentation were extracted from mouse spleens using positive selection for CD11c . The extracted cells were largely positive for both CD11c and MHC class II molecules . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 004 In order to understand the effect of this variability genome-wide , we selected two peptides—the low-affinity K99A and the high-affinity PCC—and sequenced mRNA from CD4+ T cells exposed to both a low and a high concentration of each peptide . We compared the gene expression profiles at 24 hr across five conditions ( no-peptide; low-dose , low-affinity ( 10 μM K99A ) ; high-dose , low-affinity ( 100 μM K99A ) ; low-dose , high-affinity ( 0 . 1 μM PCC ) ; and high-dose , high-affinity ( 10 μM PCC ) ) , four out of five of which displayed some degree of activation as measured by extracellular markers such as CD69 or CD25 . We used principal component analysis ( PCA ) to determine the primary axes of variation across the approximately 3 , 000 genes that were expressed above ten reads per kilobase per million ( RPKM ) and at least two-fold different between any two conditions . A single principal component explained more than 99% of the variance in gene expression changes ( Figure 2—figure supplement 1A ) . The first principal component ranks the samples according to what would be expected based on TCR signal strength and extracellular markers such as CD69 and CD25 ( Figure 2A ) . As PC1 captured the gene expression changes concomitant with increasing activation , we extracted the most positive 10% and most negative 10% of the genes along PC1 to determine which genes were important for the axis . The 10% of genes contributing the most positive signal to PC1 were increasing in a generally graded manner with TCR signal strength across the samples , and the 10% of genes contributing the most negative signal were decreasing ( Figure 2B; two-tailed p-values based on permutation testing of 2 . 9e-11 and 1 . 8e-28 , respectively ) . 10 . 7554/eLife . 10134 . 005Figure 2 . RNA-Sequencing reveals graded expression of activation signature genes . ( A ) Principal component analysis ( PCA ) of the approximately 3 , 200 genes that changed between any two samples reveals that the primary axis of variation ( PC1 , shown along the x-axis ) orders the five conditions by increasing TCR signal strength: No Peptide; low-dose , low-affinity ( 10 μM K99A ) ; high-dose , low-affinity ( 100 μM K99A ) ; low-dose , high-affinity ( 0 . 1 μM PCC ) ; and high-dose , high-affinity ( 10 μM PCC ) . ( B ) After ordering the ~3200 genes used for PCA by their contribution to PC1 , we extracted the top 10%—that is , the ~320 genes contributing most positively to a sample’s PC1 value—and the bottom 10%—that is , the ~320 genes contributing most negatively to a sample’s PC1 value . Each group displays a clear trend , with the top 10% increasing in expression as signal strength increases , and the bottom 10% decreasing in expression . Each blue line represents a gene , with reads per kilobase per million ( RPKM ) normalized from 0 to 1 across the five conditions . Significance was determined using permutation testing , where the mean difference between genes in the No Peptide sample as compared to 10 μM PCC was normally distributed over randomly generated groups of genes . This normal distribution was compared to the top 10% and bottom 10% genes to generate a p-value . ( C ) Genes in the top 10% of PC1 , termed activation signature genes , include many genes previously identified as important to CD4+ T cell activation , such as Tbx21 ( Tbet ) , Stat1 , and Tnf . Reads per kilobase per million ( RPKM ) for each increases with increasing signaling strength . ( D ) Irf4 , a transcription factor previously shown to be more highly expressed with increasing TCR affinity , shows graded expression across the five conditions . ( E ) Gene Ontology ( GO ) analysis of activation signature genes shows enrichment for protein biosynthesis and molecular chaperone genes . P-values shown are Benjamini-Hochberg adjusted p-values . ( F ) As measured by flow cytometry , the geometric MFI of Tbet in CD4+ cells increases on a per-cell basis with increasing signal strength . Note that MFI of the entire population of CD4+ cells is shown , as Tbet distribution is unimodal . ( G ) Similarly , per-cell protein levels of IRF4 increase with increasing signal strength when measured with flow cytometry . ( p-values based on Student’s t test; p<0 . 05 , p<0 . 01 , p<0 . 001 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 00510 . 7554/eLife . 10134 . 006Figure 2—figure supplement 1 . RNA-Sequencing reveals graded expression of activation signature genes . ( A ) 99 . 5% of the variance between the five conditions could be explained with a single principal component . ( B ) Eif3a , a protein biosynthesis gene , increased in expression with increasing TCR signaling strength . ( C ) Cct2 , a molecular chaperone gene , increased in expression with increasing TCR signaling strength . ( D ) Not all gene expression changes are reflected in extracellular protein levels . Il2rb , an activation signature gene , does not significantly change when measured with extracellular flow cytometry . ( E ) Cd200 , an activation signature gene , shows a graded gene expression pattern across the five conditions , and this is reflected by extracellular presentation of the CD200 molcule as measured by flow cytometry . This plot shows the geometric MFI of CD4+ CD200+ cells across the conditions . ( F ) Similarly , Ly6a increases in both population gene expression levels and single-cell protein levels as measured by flow cytometry . This plot shows the geometric MFI of CD4+ Ly6a+ cells across the conditions . ( G ) The receptor TNFSF11 ( RANKL ) shows a graded increase in gene expression that is reflected in the per-cell levels of extracellular expression of the protein . This plot shows the geometric MFI of CD4+ RANKL+ cells across the conditions . ( p-values based on Student’s t test; p<0 . 05 , p<0 . 01 , p<0 . 001 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 006 Collectively , the most indicative 10% of genes for PC1 provide a multidimensional signal for ranking the samples in one dimension according to TCR signal strength and activation state of the CD4+ T cell; we therefore call these genes “activation signature genes . ” Looking more closely at the genes in this group , we see many well-documented immune response genes such as Tbx21 ( Tbet ) , Stat1 , and Tnf ( Figure 2C ) , all of which increase in a graded manner along with TCR signal strength at the population level . Irf4 , previously reported to increase in expression in an analog manner downstream of the TCR on a per-cell basis ( Man et al . , 2013; Nayar , 2014 ) , was also among the activation signature genes , and showed the same graded response pattern at the population level across the conditions ( Figure 2D ) . Gene Ontology analysis yielded several enriched gene categories among the activation signature , including Myc targets , post transcriptional regulation of gene expression , MTORC1 signaling and response to cytokine and regulation of translational initiation ( Figure 2E ) exemplified by Eif3a ( Figure 2—figure supplement 1B ) . Protein biosynthesis has been previously shown to be increased upon T cell activation , and here we see that many of the genes involved in increasing translational activity are themselves up-regulated in a manner that is proportional to the level of activation across the population ( Beretta , 2004; Bjur et al . , 2013 ) . Another enriched ontological category was molecular chaperone genes that are responsible for protein folding and unfolding , including six of the Cct family of chaperones ( for example , Cct2: Figure 2—figure supplement 1C ) and five heat shock family members ( for example , Hsph1 ) that increased with TCR signal strength . The graded increase of activation signature genes at the population level corresponded with single-cell increases in CD69 and CD25 protein , but it was unclear whether the graded response of proteins associated with activation signature genes was generalizable . In order to determine whether proteins derived from activation signature genes increased on a per-cell basis in more cases , we determined the per-cell protein levels of a panel of genes from the activation signature set with flow cytometry . Not all increases in mRNA levels were reflected at the level of protein ( Figure 2—figure supplement 1D ) , but for those that were , the increases in mRNA resulted in increases both in the frequency of cells responding ( data not shown ) and in protein levels on a per-cell basis . This is exemplified by the expression of Tbet ( Figure 2F ) , IRF4 ( Figure 2G ) , CD200 ( Figure 2—figure supplement 1E ) , Ly6a ( Figure 2—figure supplement 1F ) , and Tnsf11 ( RANKL; Figure 2—figure supplement 1G ) . To investigate whether graded levels of expression of activation signature genes seen at the RNA level was a reflection of per-cell activation levels or just the frequency of responder cells we conducted both RNA flow cytometry and RNA-seq for cells sorted by extracellular CD69 expression . RNA flow cytometry involves the hybridization of fluorescent probes to RNA targets within a cell , followed by analysis of per-cell fluorescence via flow cytometry . Thus , observed fluorescence intensity corresponds directly to single-cell expression levels of mRNAs of interest . Using this technique , we observed that mRNA expression levels of Irf4 increased across the five conditions on a per-cell basis in the same manner as seen at the population level and at the protein level ( Figure 3—figure supplement 1A ) . For each population , the mean fluorescence intensities of Irf4 mRNA increased in the same manner as seen with population-level mRNA expression and single-cell protein levels ( Figure 3A , Figure 3—figure supplement 1B ) . In contrast , beta actin mRNA as measured by this technique was constant under each treatment condition ( Figure 3—figure supplement 1C , D ) . Thus , for some mRNAs , in addition to modulating the frequency of responding cells , peptide dose and affinity tune the strength of response to TCR signaling on a per-cell basis . 10 . 7554/eLife . 10134 . 007Figure 3 . Single-cell RNA levels reflect whole-population levels for a subset of genes . ( A ) RNA flow cytometry was used to determine the single-cell levels of RNA transcripts for Irf4 across the five peptide conditions . The mean fluorescence intensity of the Irf4 RNA probe increases across the five conditions assayed via RNA flow cytometry . Data is representative of two technical and two biological replicates . ( B ) Treated cells were sorted , and RNA-sequencing was performed on the CD69+ cells from each peptide condition such that the effect of responder frequency was controlled . PCA of the genes that changed between any two samples reveals that the primary axis of variation ( PC1 , shown along the x-axis ) orders the four conditions that had any retrievable CD69+ cells by increasing TCR signal strength: low-dose , low-affinity ( 10 μM K99A ) ; high-dose , low-affinity ( 100 μM K99A ) ; low-dose , high-affinity ( 0 . 1 μM PCC ) ; and high-dose , high-affinity ( 10 μM PCC ) . ( C ) As with the whole-population RNA-sequencing data , looking at the top 10% of genes along PC1 revealed an expression profile that reflects the analog signal seen with external activation markers such as CD25 and CD69 . Each blue line represents a gene , with reads per kilobase per million ( RPKM ) normalized from 0 to 1 across the five conditions . ( D ) Gene Ontology ( GO ) analysis of the genes in the top 10% of PC1 shows enrichment for terms related to metabolic processes such as translation and RNA biosynthesis . ( E ) There was additionally a digital cluster of genes identified that showed threshholding behavior upon cell activation . For the digital cluster , expression levels were low in the control CD69- cells , and universally higher across all CD69+ samples . The peptide conditions are sorted along the x-axis , and the normalized RPKM for the 100 most highly induced/most consistently expressed genes in the cluster is shown along the y-axis . ( F ) GO analysis of the genes in the digital cluster shows an enrichment for terms related to immune signaling pathways . ( G ) RPKM levels for several genes classified as 'translation initiation factors' are shown . For most of the genes , expression is low for the CD69- control , and increases in an analog fashion across the CD69+ cells from each treated condition . The peptide conditions are sorted along the x-axis , and RPKM for each gene is shown along the y-axis . ( H ) RPKM levels for several genes in the 'ribosome biogenesis' GO category are shown . Expression is low for the CD69- control , and increases in an analog fashion across the CD69+ cells from each treated condition . The peptide conditions are sorted along the x-axis , and RPKM for each gene is shown along the y-axis . ( I ) RPKM levels for several genes indicative of the digital cluster are shown . Expression is low for the CD69- control , but relatively level across all CD69+ samples . The peptide conditions are sorted along the x-axis , and RPKM for each gene is shown along the y-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 00710 . 7554/eLife . 10134 . 008Figure 3—figure supplement 1 . RNA flow cytometry for Irf4 shows graded increases . ( A ) RNA flow cytometry was used to determine the single-cell levels of RNA transcripts for Irf4 across the five peptide conditions . On the x-axis , the fluorescence intensity of CD69 protein is shown , and on the y-axis , the fluorescence intensity of an RNA-hybridizing probe targeted at Irf4 mRNA . Both the frequency and the per-cell intensity of Irf4 increase with increasing TCR signaling . Data is representative of two technical and two biological replicates . ( B ) RNA flow cumulative distribution plots for the Irf4 probe in CD69+ cells under the indicated treatment conditions . ( C ) RNA flow cumulative distribution plots for the beta actin probe in CD69+ cells under the indicated treatment conditions . ( D ) Mean Fluorescence Intensity of the beta actin probe in CD69+ cells under the indicated treatment conditions . ( E ) FPKM values for Rpl8 , Rsp9 and Irf4 in CD69+ cells under the indicated treatment conditions . ( F ) FPKM values for Cd69 , Tnf and Il2 in CD69+ cells under the indicated treatment conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 008 In order to look genome-wide , we conducted RNA-sequencing on CD4+ T cells treated as before , but sorted by the activation marker CD69 , such that we selectively analyzed responder cells . In this way , we were able to control for the effect of frequency—the observed RNA-seq levels for each condition reflected only the activated cells , and thus were not diluted by non-responder cells . As with the whole-population RNA-seq , we used PCA to determine the primary axes of variation across the four samples for which CD69+ cells could be collected . ( Insufficient responder cells could be retrieved from the no-peptide condition for sequencing . ) Across the four treated samples , the first principal component again sorted the conditions according to the relative strength of TCR signaling ( Figure 3B ) . Looking at the genes composing the top 10% of PC1 , we saw , as with the whole-population RNA-seq , that gene expression patterns largely reflected the same trend , increasing with increasing TCR signal ( Figure 3C , Figure 3—figure supplement 1E ) . To assess what genes made up the activation signature seen in PC1 , we performed GO analysis on the genes in the top 10% of PC1 . As with the whole-population data , genes associated with cell metabolic processes like translation and RNA biosynthesis were enriched ( Figure 3D ) , indicating that the expression of genes in key cellular activation pathways increases with increasing TCR signaling on a per-cell basis as well as at the population level . In contrast to the whole-population data , there was a group of genes that were near maximally activated in CD69+ cells treated with the low concentration of K99A and exhibited similar expression across all treatment conditions . This ‘digital’ pattern is illustrated in Figure 3E for the 100 most highly induced/most consistently expressed genes and for representative genes in Figure 3—figure supplement 1F . GO analysis demonstrated that this digital cluster of genes was enriched for immune signaling terms ( Figure 3F ) . Expression profiles for representative mRNAs in the translation initiation factor and the ribosome biogenesis groups are illustrated in Figure 3G , H and Figure 3—figure supplement 1E , and representative mRNAs in the digital group are illustrated in Figure 3I and Figure 3—figure supplement 1F . Thus , the graded increase in expression of activation signature genes at the population level is a reflection of both frequency of responding cells and incremental increases in expression levels on a single-cell level for at least a subset of genes , including key cellular metabolism genes . Given that PC1 was able to distinguish between the five conditions according to activation state , we extracted the genes from the top and bottom of the whole-population PC1 that were consistent across replicates to use as a general-purpose activation score able to correctly rank the five conditions by TCR signal ( Figure 4A ) . We compared samples from several publicly available datasets , and the activation score was able to quantitatively rank conditions within a given experiment set such that activated and naïve CD4+ T cells could be distinguished and further that the effects of various genetic perturbations of CD4+ T cell responses could be observed . For example , the activation score correctly recapitulated the findings that polarized helper subsets of CD4+ T cells were more pro-inflammatory than unstimulated cells or induced and natural regulatory T cells ( Wei et al . , 2009 ) ( Figure 4B ) ; that at the population level plate-bound anti-CD3 and anti-CD28 induced stronger activation signals than APCs plus antigen ( Tan et al . , 2014 ) ( Figure 4C ) ; that costimulation was important for achieving higher activation states but checkpoint inhibitors could block this effect ( Wakamatsu et al . , 2013 ) ( Figure 4D ) ; that knockout of Trim28 , a molecule necessary for optimal IL2 production , diminished CD4+ T cell activation status ( Chikuma et al . , 2012 ) ( Figure 4—figure supplement 1A ) ; and that acute LCMV infection produced more robust activation in CD4+ T cells than chronic infection ( Doering et al . , 2012 ) ( Figure 4—figure supplement 1B ) . 10 . 7554/eLife . 10134 . 009Figure 4 . PC1 can be used to rank arbitrary CD4+ T cell data sets . ( A ) An activation score derived from the top and bottom genes along PC1 ranks the five conditions according to TCR signaling strength . The score correctly captures that 100 μM K99A and 0 . 1 μM PCC are very similar in activation status . Note that the score ranks samples within an experiment , but is not an absolute metric for comparing across experiment groups . ( B ) The activation score can be used to compare arbitrary CD4+ T cell data sets . Here , activation scores were calculated for microarrays from helper T cell subsets , and they demonstrate that naïve cells and native regulatory T cells ( nTregs ) are less classically activated than Th1 , Th2 , and Th17 polarized subsets . ( C ) The activation score captures the fact that plate-bound anti-CD3 and anti-CD28 stimulation of helper T cell subsets results in stronger signaling than antigen as presented by APCs . ( D ) CD4+ T cells were subjected to a variety of stimulatory or inhibitory treatments: anti-CD3 alone , or anti-CD3 with anti-CD28 , anti-BTLA , anti-CD80 , anti-CTLA4 , anti-ICOS , or anti-PD1 . Gene expression profiles at early ( 1 hr and 4 hr ) and late ( 20 hr and 48 hr ) time points yield activation scores in line with the characterization of CD28 , BTLA , CD80 , and ICOS as co-stimulatory , and CTLA4 and PD1 as inhibitory . Although it might be expected that anti-CD80 would have an inhibitory effect , these results are in line with the conclusions from the originally published analysis . ( E ) Naïve CD4+ splenocytes were isolated from 39 mouse strains . Using the PC1-derived activation score , we can rank the CD4+ cells from each strain as either more or less activated under basal conditions . Using the activation score , we recapitulate the finding that C57Bl6 mice have more pro-inflammatory cells than BALB/c mice . The highest scoring strain , DBA/2 , shows top-quartile expression of immune effectors as well as protein biosynthesis genes . ( P-values based on Student’s t test; p<0 . 05 , p<0 . 01 , p<0 . 001 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 00910 . 7554/eLife . 10134 . 010Figure 4—figure supplement 1 . PC1 can be used to rank arbitrary CD4+ T cell data sets . ( A ) The activation signature score was used to quantify CD4+ T cell activation status under Trim28 knockout conditions . Loss of Trim28 resulted in a lowering of the activation signature score , corroborating the previously reported results that the Trim28-deficient cells produced less IL2 . ( B ) Acute LCMV infection results in a higher activation signature score for CD4+ T cells at early time-points , marking the peak of infection before cells begin to turn off activation programs . The activation signature score does not reach such a high level in a chronic infection model . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 010 In order to test the value of the activation score , we used it to rank naïve CD4+ T cells from 39 inbred mouse strains ( Mostafavi , 2014 ) ( Figure 4E ) . The activation score quantified the variability in the isolated CD4+ T cells according to activation status , revealing that the genetic differences between the strains yielded different levels of activity even under unstimulated conditions . As would be predicted by known strain phenotypes , C57BL/6 cells were more activated than most strains , while BALB/c mice were less activated than most strains . The lupus-prone MRL strain and the type 1 diabetes-prone NOD strain had cells that ranked as relatively activated , whereas the type 1 diabetes-resistant NON strain had a relatively low activation score . The strain with the highest activation score , DBA/2 , had top-quartile expression of more than half of the activation signature genes ( p=7 . 4e-30 by chi-squared test ) . These included a number of immune effectors such as Irf4 , Cd25 , Il2rb , Nfkb1 ( p105/p50 ) , and Nr4a1 ( Nur77 ) , as well as 17 of 32 genes from the protein biosynthesis group and 5 of 12 genes from the molecular chaperone group . Differences in the immune phenotypes of the DBA/2 strain , such as resistance to malaria , have been largely attributed to B cell-dependent mechanisms ( Bakir et al . , 2006 ) , but the activation score here indicates that naïve CD4 T cells from DBA/2 mice are skewed toward an activated phenotype . Three of four wild-derived strains had low activation scores: CAST , MSM , and WSB mice . All three of these wild-derived strains had bottom-quartile expression of the immune effectors Irf1 , Irf8 , Stat1 , Nfkb1 , and Tnf , indicating that these CD4+ T cells possess a less inflammatory gene expression profile under homeostatic conditions . Thus , the activation score serves as a widely applicable and quantitative measure of CD4+ T cell activity , and can be used to assess the relative activation status of a variety of CD4+ T cell samples . We have developed a publically available , open source tool ( see Materials and methods ) to facilitate the scoring and ranking of datasets by interested parties . In order to better understand the changes in genome-wide expression patterns that occurred with TCR stimulation , we compared enhancer landscapes with and without stimulation . We first performed ChIP-sequencing for H3K4me2 , a marker of primed and active promoters and enhancers ( He et al . , 2010; Kaikkonen et al . , 2013 ) , across the five conditions . By and large , the H3K4me2-marked regions across the five conditions were very similar ( Table 1 ) , with a comparison of tag counts associated with specific genomic regions under no peptide or 1 µM PCC illustrated in Figure 5A . Though there were some enhancers showing at least two-fold change in H3K4me2 tag counts across conditions , these regions were at the lower end of the tag count range and therefore differences were not significant ( Figure 5A , red points ) . Thus , the gene expression and phenotypic changes seen after activation were not due to selection of new signal-dependent enhancers . 10 . 7554/eLife . 10134 . 011Table 1 . Pairwise overlap between the H3K4me2 peaks . Each cell contains the count of overlapping peaks where each condition shown has at least 40 tags ( normalized ) . In the diagonal is the total number of peaks with at least 40 tags for the given condition . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 011No Peptide10 μM K99A100 μM K99A0 . 1 μM PCC10 μM PCCNo Peptide160121495112487147831310510 μM K99A15616125491483813192100 μM K99A1269512603122470 . 1 μM PCC155691328110 μM PCC1344410 . 7554/eLife . 10134 . 012Figure 5 . Primed enhancers are pre-existing , but gain activation markers with treatment . ( A ) Comparing primed enhancers marked by H3K4me2 peaks reveals strong correlation between untreated and treated samples . Normalized tag counts in the No Peptide condition are plotted against those in a 1 μM PCC condition , with red dots coloring those that are more than two-fold up-regulated in the 1 μM PCC condition . The up-regulated enhancers are both few in number and low in tag count . ( B ) De novo motif finding identifies lineage-determining transcription factor ( LDTF ) motifs among primed enhancers shared by the five conditions . An ETS motif is most prominent , and a RUNX motif is likewise highly enriched over the randomly selected background . Both ETS and RUNX factors play important roles in T cell development . ( C ) Among primed enhancers shared by all five conditions , including the untreated condition , pro-inflammatory transcription factor motifs are enriched . An IRF family motif , AP-1 family motif ( represented by BATF ) , and NF-κB motif ( represented by REL ) are all significantly enriched among shared enhancers marked by H3K4me2 . ( D ) . Comparing H3K27Ac tag counts at enhancers in No Peptide as compared to 1 μM PCC treatment reveals that many enhancers see increasing H3K27Ac deposition upon stimulation . Points in red indicate greater than two-fold increase in tags upon treatment . ( E ) Enhancers that are more active upon stimulation , as determined by greater than two-fold H3K27Ac tags in 1 μM PCC treatment as compared to No Peptide , are enriched for pro-inflammatory transcription factor motifs . BATF , an AP-1 family member , and NF-κB are most prominent . ( F ) Enhancers that are more active with stimulation are enriched near activation signature genes , as can be seen with this enhancer upstream of the activation signature gene Il2ra ( CD25 ) . ( G ) Enhancers upstream of the activation signature gene Cd69 show an increase in H3K27Ac deposition upon treatment with 1 μM PCC . ( H ) Genome-wide , deposition of H3K27Ac , a marker of transcription factor activity , reflects increasing TCR signal strength at the binding sites of AP-1 family members , including BATF . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 01210 . 7554/eLife . 10134 . 013Figure 5—figure supplement 1 . Primed enhancers are pre-existing , but gain activation markers with treatment . ( A ) Despite the similarity of enhancer profiles across conditions , the H3K4me2 mark 'spreads' from the transcription start site along the body of genes at a subset of genes . This subset is significantly enriched for activation signature genes such as Cd69 , shown here . ( B ) Irf4 , another activation signature gene , shows a similar spreading of dimethyl along the body of the gene after treatment . ( C ) Normalized tag counts along the first 4000 bp of all activation signature genes show that , overall , the No Peptide condition shows a narrow peak at the transcription start site , but this peak is smoothed out as the dimethyl mark spreads along the body of the gene in the treated samples . ( D ) The spread of dimethyl seen at activation signature genes is not present at genes in the bottom 10% of PC1 , where all five conditions show a similar pattern of dimethyl deposition across the first 4000 bp of the genes . ( E ) The Interferon Response Family ( IRF ) motif is enriched in enhancers from many cells in the T cell lineage , with motif frequency peaking in thymocytes ( DN1 through DP ) . The grey portion of the bars represents mean background enrichment of the motif , and the colored section of the bar shows the difference in enrichment between the cell type indicated and the background . From top to bottom , the cells indicated are: Embryonic stem cells; Lin−Sca-1+c-Kit+ ( LSK ) hematopoietic progenitor cells; fetal liver derived double negative 1 thymocytes; fetal liver derived double negative 2a thymocytes; fetal liver derived double negative 2b thymocytes; double negative 3 thymocytes; double positive thymocytes; Th1 polarized CD4+ T cells; Th2 polarized CD4+ T cells; Th1 polarized CD4+ T cells from a Stat1 knockout model; thioglycollate-elicited macrophages; and the adipocyte-derived 3T3L1 cell line . ( F ) De novo motif finding identifies lineage-determining transcription factor ( LDTF ) motifs among H3K27Ac-marked enhancers shared by the five conditions . As with the primed enhancers , the ETS family motif and a RUNX motif are highly enriched . ( G ) Among activated enhancers shared by all five conditions , pro-inflammatory transcription factor motifs are enriched . As in primed enhancers , IRF , NF-κB , and AP-1 motifs are all enriched in the shared H3K27Ac peaks . ( H ) As seen with BATF , deposition of H3K27Ac , a marker of transcription factor activity , reflects increasing TCR signal strength at the genome-wide binding sites of the AP-1 family member JunB . I . Similarly , deposition of H3K27Ac reflects increasing TCR signal strength at the genome-wide binding sites of the AP-1 family member cJun . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 013 At promoters , H3K4me2 marks were shared across the five conditions , but activation signature genes showed spreading of the H3K4me2 mark along the body of the gene after TCR stimulation . In contrast , H3K4me2 peaks were narrow and focal for the untreated condition at many of these genes . This effect can be seen at Cd69 ( Figure 5—figure supplement 1A ) and Irf4 ( Figure 5—figure supplement 1B ) , resulting in a global increase of the ratio of gene body tags to promoter tags at activation signature genes ( Figure 5—figure supplement 1C ) but not genes in the bottom 10% of PC1 ( Figure 5—figure supplement 1D ) . This implies that the process of activating these genes subsequent to TCR stimulation induces deposition of the dimethyl mark along the body of the genes as they are transcribed . We used de novo motif finding ( Heinz et al . , 2010 ) to identify lineage-determining transcription factors ( LDTFs ) , also known as pioneer factors or master regulators , which establish cell-type-specific enhancer landscapes , and determine the available open chromatin for subsequent binding of signal-dependent transcription factors ( SDTFs ) ( Garber et al . , 2012; Heinz et al . , 2013; Mullen et al . , 2011; Soufi et al . , 2012; Trompouki et al . , 2011 ) . The top motif was an ETS motif ( Figure 5B ) , capable of being bound by a number of ETS factors that are expressed in CD4+ T cells , including Ets1 , Ets2 , and Elf1 ( Anderson et al . , 1999 ) . These enhancers tend to be shared across similar cells as well as thymic T cell precursors ( Heinz et al . , 2010; Zhang et al . , 2012 ) . Similarly , Runx factors play an important role in T cell development ( Wong et al . , 2011 ) , and correspondingly the Runx family motif was highly enriched among primed enhancers . Several known motifs for SDTFs were also enriched among the H3K4me2-marked enhancers ( Figure 5C ) , including an Interferon Regulatory Factor ( IRF ) motif . Although IRFs respond to interferon signaling ( Ozato et al . , 2007 ) , and would not be expected to be active in unstimulated cells ( Murphy et al . , 2013; Glasmacher , 2012; Li et al . , 2012 ) , it is possible that the IRF motif is a 'memory' of states of activation during the development of CD4+ T cells , and indeed IRF motifs can be found in several related cell types and multiple stages of thymocyte development ( Figure 5—figure supplement 1E ) , suggesting that the primed enhancers in naïve CD4+ T cells are predisposed to act as binding sites for key SDTFs ( Heinz et al . , 2010; Zhang et al . , 2012; Buecker et al . , 2014; Mikkelsen et al . , 2010; Mishra et al . , 2014; Vahedi et al . , 2012 ) . Similarly , an AP-1 motif and an NF-κB motif were significantly enriched in primed enhancers ( Figure 5C ) , corresponding with the fact that TCR signaling greatly increases activity of both of these transcription factors ( Huang and Wange , 2004; Rincón and Flavell , 1994 ) . Given that H3K4me2-marked regions were not substantially changed across the five conditions , we next performed ChIP-sequencing for H3K27Ac , a marker for active enhancers ( Creyghton et al . , 2010 ) , under a stimulated condition ( 1 μM PCC ) and the unstimulated condition . In contrast to H3K4me2 , a substantial portion of enhancers exhibited increases in the H3K27Ac activation mark ( Figure 5D , Table 2 ) . The union set of enhancers was enriched for a similar set of motifs as the primed enhancers ( Figure 5—figure supplement 1F , G ) . Enhancers that became more active with TCR engagement were highly enriched for both AP-1 and NF-κB motifs ( Figure 5E ) . Further , this group of enhancers was more likely to be proximal to activation signature genes than would be expected at random ( p-value = 2 . 0e-20 by chi-squared test ) . These enhancers included , for example , those upstream of Il2ra ( CD25; Figure 5F ) and Cd69 ( Figure 5G ) . 10 . 7554/eLife . 10134 . 014Table 2 . Pairwise overlap between the H3K27Ac peaks . Each cell contains the count of overlapping peaks where each condition shown has at least 40 tags ( normalized ) . In the diagonal is the total number of peaks with at least 40 tags for the given condition . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 014No Peptide10 μM K99A100 μM K99A0 . 1 μM PCC10 μM PCCNo Peptide6410359157915988597710 μM K99A3671355735943584100 μM K99A8877869987890 . 1 μM PCC9867936810 μM PCC10021 To investigate whether there was a quantitative relationship between TCR signal strength and enhancer activation , we performed independent ChIP-Seq for H3K27Ac in response to both peptides at low and high concentrations . Given the prevalence of the AP-1 motif in the signal-responsive enhancers , we analyzed H3K27Ac tag counts at AP-1 binding sites genome-wide using publicly available ChIP-Sequencing data from in vitro activated TH17 cells ( Li et al . , 2012 ) . There was an increase in H3K27Ac deposition at AP-1 binding sites that reflected the graded strength of TCR signaling ( BATF shown in Figure 4H; other AP-1 family members in Figure 5—figure supplement 1H , I ) , indicating that AP-1 binding sites became more active in a graded manner corresponding to increasing TCR signaling . Graded changes in H3K27Ac were much less pronounced at CTCF binding sites , which occur at enhancers but are also more broadly distributed and play roles in establishing boundary elements . We next looked at changes in super-enhancers ( Hnisz et al . , 2013; Lovén et al . , 2013; Whyte et al . , 2013 ) upon activation using the H3K27Ac mark . Most super-enhancers ( approximately 450 out of 700 total ) identified were shared by both the unstimulated and stimulated conditions . GO analysis of genes nearby the shared super-enhancers showed enrichment for leukocyte activation genes as well as Pleckstrin homology genes , which are critical components of a number of kinase signaling pathways downstream of the TCR ( Figure 6A ) , indicating that super-enhancers in CD4+ T cells prime genes important for inflammatory signaling . These basally-primed super-enhancers included regions near key T cell genes such as Ets1 ( Figure 6B ) , Runx1 ( Figure 6—figure supplement 1A ) , Ctla4/Icos/CD28 ( Figure 6C ) , and Irf4 ( Figure 6—figure supplement 1B ) . Notably , even though many of the super-enhancers exist prior to stimulation , super-enhancers near activation signature genes show an increase in H3K27Ac signal subsequent to TCR signaling ( Figure 6D ) . 10 . 7554/eLife . 10134 . 015Figure 6 . Super-enhancers prime T cell activation genes . ( A ) Gene Ontology ( GO ) analysis of genes nearest to the ~450 super-enhancers shared by treated and untreated conditions show enrichment for T cell activation genes . P-values shown are Benjamini-Hochberg adjusted p-values . ( B ) H3K27Ac marks a large super-enhancer around the lineage-determining transcription factor Ets1 in both the No Peptide and 1 μM PCC conditions . The super-enhancer spans the ~600 kbp region shown . ( C ) The ~400 kbp super-enhancer region encompassing Cd28 , Ctla4 , and Icos is marked by H3K27Ac in both treated and untreated conditions . ( D ) Despite being heavily marked by H3K27Ac in both untreated and treated conditions , shared super-enhancers near activation signature genes show a significant gain in H3K27Ac tags in response to stimulation as compared to the shared super-enhancers not proximal to activation signature genes . In other words , basally primed super-enhancers near activation signature genes see significant increases in activity upon stimulation , correlating with increased gene expression at the activation signature genes . ( E ) Some regions of H3K27Ac deposition required TCR stimulation to pass the super-enhancer threshold , as can be seen here at the ~60 kbp region encompassing BATF , an AP-1 family member . While H3K27Ac is clearly present under basal conditions , there is a substantial increase in enhancer activity upon treatment with 10 μM PCC . ( F ) Il2ra ( CD25 ) shows increased enhancer activity and formation of a super-enhancer in the treated condition . ( G ) Similarly , the region surrounding Tbx21 ( Tbet ) shows substantial increases in activity subsequent to stimulation , resulting in the formation of a super-enhancer . ( p-values based on Student’s t test; p<0 . 05 , p<0 . 01 , p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 01510 . 7554/eLife . 10134 . 016Figure 6—figure supplement 1 . Super-enhancers prime T cell activation genes . ( A ) A basally-primed super-enhancer encompasses the Runx1 locus . ( B ) Despite being up-regulated in response to treatment , the Irf4 locus has a super-enhancer even under untreated conditions . ( C ) Lag3 , a negative regulator of T cell signaling that is up-regulated with treatment , sees increased enhancer activity and formation of a super-enhancer in treated conditions . ( D ) Stat5b , a transcription factor important for T cell signaling , shows increased enhancer activity and formation of a super-enhancer in the treated condition . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 016 118 of the 568 super-enhancers identified after TCR stimulation were not identified as super-enhancers in the unstimulated condition . The super-enhancers that required TCR signaling were enriched for leukocyte activation genes ( Benjamini-Hochberg adjusted p-value = 4 . 6e-7 ) crucial for T cell activation , including Batf ( Figure 6E ) , Il2ra ( CD25 , Figure 6F ) , Tbx21 ( Tbet , Figure 6G ) , Lag3 ( Figure 6—figure supplement 1C ) , and Stat5b ( Figure 6—figure supplement 1D ) . The Ras/Raf/Mek/Erk pathway downstream of the TCR activates the AP-1 transcription factor family through a series of phosphorylation events and transcriptional induction of immediate-early genes ( Murphy et al . , 2013; Rincón and Flavell , 1994 ) . Given the fact that the AP-1 motif was enriched at enhancers showing increasing activity and the fact that AP-1 binding sites saw increasing H3K27Ac deposition , we sought to determine whether AP-1 and the ERK pathway were relevant to the increasing expression of activation signature genes across the conditions . We first compared the level of phosphorylated ERK ( p-ERK ) in each condition using flow cytometry , and found that , like CD69 and CD25 , the amount of p-ERK in the p-ERK+ cells varied on a per-cell basis in each condition , increasing with TCR signal strength ( Figure 7A ) . Frequency of pERK+ cells also differed across peptide conditions ( Figure 7—figure supplement 1A ) , indicating that both frequency and single-cell response level play a role in the effect of MEK inhibitors . As the increasing levels of p-ERK paralleled the general pattern of expression of the activation signature genes , we assessed binding frequencies of AP-1 factors in the gene promoters of the activation signature genes as compared to the bottom 10% of PC1 genes , and found that activation signature genes showed a significantly higher frequency of AP-1 binding ( Figure 7B ) . This enrichment for AP-1 binding was not dependent on the expression level of the genes in each group , as segmenting the genes by RPKM showed the same pattern ( Figure 7—figure supplement 2 ) . 10 . 7554/eLife . 10134 . 017Figure 7 . ERK signaling translates TCR signal strength into graded gene expression . ( A ) ERK phosphorylation is a measure of ERK pathway activity . Flow cytometry for phospho-ERK after 3 . 5 hr of co-culturing shows that , on a per-cell basis , increasing signal strength yields increasing levels of phospho-ERK among CD4+ phospho-ERK+ cells . ( B ) ERK pathway activation is upstream of the transcription factor AP-1 . ChIP-sequencing tags for four AP-1 family members ( BATF , cJun , JunB , and JunD ) in Th17 cells shows that there is an enrichment for AP-1 binding near the promoters ( plus or minus 1 , 000 bp from the TSS ) of activation signature genes ( top 10% of PC1 ) as compared to all genes or the genes in the bottom 10% of PC1 . ( C ) A MEK inhibitor that dampens signaling upstream of the ERK pathway preferentially diminishes expression of activation signature genes , as seen in the fact that the RPKM of genes in the top 10% of PC1 is significantly reduced with treatment . ( D ) The reduction of RPKM seen with the activation signature genes is not a general effect , as the RPKM of the genes in the bottom 10% of PC1 are not significantly affected by MEK inhibitor treatment . ( E ) We quantified the effect of MEK inhibitor treatment using the activation signature score . Treatment with the MEK inhibitor reduces the activation signature score for all samples . ( F ) Schematic of the ERK-AP-1 axis . See text for details . ( p-values based on Student’s t test; p<0 . 05 , p<0 . 01 , *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 01710 . 7554/eLife . 10134 . 018Figure 7—figure supplement 1 . ERK signaling translates TCR signal strength into graded gene expression . ( A ) ERK phosphorylation is a measure of ERK pathway activity . Flow cytometry for phospho-ERK after 3 . 5 hr of co-culturing shows that the frequency of pERK+ cells is dependent on both the peptide treatment and the dose of MEK inhibitor used , with increased strength of signaling and decreased MEKi leading to greater percentages of pERK+ cells . ( B ) Per-cell levels of phospho-ERK as measured by flow cytometry are analog with respect to the dosage of MEK inhibitor treatment . For each peptide condition , increasing the concentration of the MEK inhibitor gradually reduces the geometric MFI of phospho-ERK in the treated cells . ( C ) MEK inhibitor treatment at 0 . 5 μM ( IC50 ) results in the preferential reduction of activation signature genes . CD69 , one of the activation signature genes , reflects this decrease in expression level at the protein level , as measured by extracellular flow cytometry . ( D ) In contrast to CD69 , CD4 does not show a change in expression level upon treatment with the MEK inhibitor . ( E ) For the genes in the top 10% of PC1 , MEK inhibitor treatment ( green lines ) results in a decrease of expression as compared to the uninhibited samples ( blue lines ) . Each line plots the normalized RPKM of one gene across the five samples , either with the MEK inhibitor ( green ) or without ( blue ) . ( F ) In contrast to the top 10% , the genes in the bottom 10% of PC1 are not significantly affected by MEK inhibitor treatment , and , if anything , increase in expression level rather than decrease . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 01810 . 7554/eLife . 10134 . 019Figure 7—figure supplement 2 . AP-1 enrichment is independent of expression level . ( A ) The enrichment for AP-1 in promoters of genes in the top 10% of PC1 ( shown in Figure 7B ) is not dependent on the expression level of the genes . Here , we subset the genes in the top 10% of PC1 by RPKM , and the increasing levels of AP-1 across the different groups of genes holds . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 01910 . 7554/eLife . 10134 . 020Figure 7—figure supplement 3 . AP-1 enrichment is independent of expression level . ( A ) The decrease in RPKM for MEKi treated samples for genes in the top 10% of PC1 ( shown in Figure 7C ) is not dependent on the expression level of the genes . Here , we subset the genes in the top 10% of PC1 by RPKM , and the affect of MEK inhibitor treatment is consistent . DOI: http://dx . doi . org/10 . 7554/eLife . 10134 . 020 ERK pathway activation and AP-1 binding therefore seemed to correlate well with the graded profile of activation signature genes and the increasing activation score across the samples . In order to determine whether the gradual increase in ERK pathway activation was causal in translating TCR signaling into gradual increases in the expression of activation signature genes , we pretreated the CD4+ T cells with a low-dose MEK inhibitor ( MEKi ) . MEK inhibition upstream of ERK was capable of suppressing p-ERK activity entirely , and titration of MEKi yielded intermediate levels of p-ERK on a per-cell basis ( Figure 7—figure supplement 1B ) . Low-dose MEK inhibition decreased the levels of the extracellular signaling marker CD69 ( Figure 7—figure supplement 1C ) , which was in the top 10% of PC1 , but this suppression was not universal , as CD4 , an example of a gene not in the top 10% of PC1 , was not significantly affected ( Figure 7—figure supplement 1D ) . In order to see if this preferential suppression of activation signature genes was widespread , we performed RNA-seq on the five conditions after pretreatment with MEKi at IC50 ( 0 . 5 μM ) . If TCR signaling strength upstream of pERK yields graded levels of ERK that are in turn essential for the graded levels of activation response genes , then reduction of pERK levels should move each sample downwards in activation score , such that the high-dose , high-affinity case looks like the low-dose , high-affinity; the low-dose , high-affinity looks like the high-dose , low-affinity; and so on . Accordingly , MEK inhibition at IC50 decreased expression of activation signature genes ( Figure 7C , Figure 7—figure supplement 1E ) , but , as with extracellular expression of CD69 , this effect was selective; expression of genes in the bottom 10% was increased or unchanged ( Figure 7D , Figure 7—figure supplement 1F ) . The decrease in expression among genes in the top 10% of PC1 upon MEKi treatment was not dependent on expression level of the genes , as the effect was consistent when genes were segmented by RPKM level ( Figure 7—figure supplement 3 ) . Using the activation score to assess total T cell activation status , we found that MEKi shifted each sample down in score ( Figure 7E ) , as would be expected if the graded levels of pERK seen with each condition were prescriptive of the activation status of the condition . Thus , graded levels of pERK downstream of the TCR help to translate the analog activation signal to graded levels of enhancer activity and gene expression genome-wide ( Figure 7F ) . Understanding how CD4+ T cells respond to ligands of different doses and affinities is critical to understanding the nature of the adaptive immune response to both pathogens and self . Here , we have shown that the traditional model of a purely digital TCR response is too simple; on a per-cell basis , stronger TCR signals result in higher levels of phosphorylated ERK , a proportional increase in enhancer acetylation , and quantitative increases in activation markers such as CD69 and CD25 ( Figure 7F ) . As a result of both these single-cell differences and the increasing frequency of respondent cells , varying the dose or the affinity of the pMHC-TCR interaction results in a gene expression profile that is graded corresponding to increasing strength of TCR signaling . We observed no evidence for Th1/Th2 skewing as a function of peptide dose/affinity , which may reflect the specific peptides used for stimulation . Notably , the predominance of PC1 and the graded gene expression patterns together indicate that dose and affinity are not interpreted separately downstream of the TCR , but rather overall signaling strength sets the level of activation across the population . A similar conclusion has been drawn by looking at the sum of activation parameters induced by antigen- , co-stimulatory- and cytokine-receptors ( Marchingo , 2014 ) . Prior analysis of T cell responses at the single cell level showed that single T cell engagement of different numbers of identical peptide agonists ( titration of signal strength ) resulted in digital responses read out as increasing numbers of T cells producing the same amount of TNF or IL2 protein ( Huang et al . , 2013 ) . Intriguingly , different levels of responses were seen when naïve cells were compared with blast and memory cell responses , such that the rate of TNF and IL2 synthesis in blasts and memory cells was nearly ten-fold higher than in naïve cells . Our findings are consistent with these earlier studies in that TNF and IL2 are members of the ‘digital’ class of RNAs identified in CD69+ cells ( Figure 3—figure supplement 1E ) . Furthermore , the increase in expression of genes involved in protein expression revealed by the present studies could at least partly explain the observation that the rates of TNF and IL2 protein production are increased in blasts and memory cells in comparison to naïve cells . The analogue increase in protein biosynthetic machinery could also at least partly explain the observation that CD69 expression is digital at the level of mRNA but analogue at the level of protein expression ( Figure 1A , Figure 3—figure supplement 1F ) . Ranking genes along a primary axis of variation allowed us to extract a set of activation signature genes that increase in a graded fashion at the population level proportionally to TCR signal strength , and further to establish an activation score that can rank arbitrary CD4+ T cell samples by the strength of signaling . The data presented here therefore gives us a greater understanding of the CD4+ T cell response to ligands of varying concentrations and affinities , and informs our understanding of CD4+ T cell activation under diverse conditions . Significant differences in primed enhancers have been demonstrated under several stimulating conditions in macrophages , and demonstrate the ability of cells to quickly remodel chromatin to initiate particular gene expression programs ( Kaikkonen et al . , 2013; Ostuni et al . , 2013 ) . Surprisingly , we did not find significant changes in the primed enhancer landscape upon TCR activation in CD4+ T cells . Similarly , even the more labile activation mark H3K27Ac was largely similar across conditions , with many activation genes marked as super-enhancers even before stimulation . While it remains to be seen whether non-TCR signaling pathways or polarizing conditions induce more dramatic changes , the data presented here indicates that the CD4+ T cell enhancer landscape is largely pre-established , with subsets of H3K4me2-marked enhancers increasing in activity , but little in the way of de novo enhancer establishment . This finding helps to explain the speed and plasticity of the CD4+ T cell response ( Zhou et al . , 2009 ) —if all enhancers are primed basally , and many are even activated basally , then pro-inflammatory transcription factors can bind at established enhancers and initiate new gene expression programs with minimal additional transcriptional machinery . Both the frequency of AP-1 binding and the level of pERK correlate with the strength of TCR signaling and the graded expression of activation signature genes . At least one of the feedback loops leading to digital TCR signaling , the son of sevenless ( SOS ) positive feedback loop , exists upstream of ERK , and it has been shown in thymocytes that pERK signaling is digital ( Das et al . , 2009; Daniels et al . , 2006; Prasad et al . , 2009 ) . However , studies focused on EGFR signaling upstream of ERK indicated that discrete pulses of ERK activity result in quantitative levels of downstream signaling activity ( Albeck et al . , 2013 ) . Our analog results for pERK can be interpreted to support the notion that TCR signaling in thymocytes functions differently than TCR signaling in mature T cells . Notably , both Themis and SOS , two key components of digital signaling in thymocytes , do not seem to be critical to mature T cell signaling ( Fu et al . , 2013; Warnecke et al . , 2012 ) . The graded levels of pERK in CD4+ T cells prove important for downstream enhancer and gene activity . We have here established a mechanistic link between the level of ERK signaling and the expression patterns of activation signature genes by using an inhibitor of MEK , upstream of ERK . Low-dose MEK inhibition selectively decreased expression of the activation signature genes such that the activation score under the inhibited conditions was incrementally decreased . This indicates that the analog levels of pERK seen on a per-cell basis are translated at a population level into increased enhancer and gene activity , and that 'turning down' pERK levels selectively diminishes the activation status of the cells . This finding is of particular interest in light of the clinical availability of numerous RAF , MEK , and ERK inhibitors ( Zhao and Adjei , 2014; Samatar and Poulikakos , 2014 ) . Our findings suggest that low-dose ERK pathway inhibition could be used to selectively decrease the activity of activation signature genes in CD4+ T cells , achieving low-level immunosuppression without killing T cells or completely removing their ability to respond to TCR signaling . Further , the effect of MEK inhibitors on CD4+ T cells raises questions about the immunosuppressive effects of using MEK inhibitors in cancer treatment , especially as current clinical trials combine MEK inhibitors with checkpoint-blockade inhibitors ( Zhao and Adjei , 2014; Vella et al . , 2014 ) . Notably , NFκB is one of many transcription factors known to play important roles in T cell biology , and indeed we find a κB motif enriched among enhancers that are responsive to stimulation ( Figure 5E ) . Further research as to the relationship between the strength of TCR signaling and other signal-dependent transcription factors such as NF-κB and NFAT is warranted . In sum , this study makes use of a unique model system to dissect the transcriptional responses of CD4+ T cells to increasing strength of signaling , and demonstrates that analog levels of pERK within the context of digital TCR signaling flow downstream to result in graded gene expression profiles and enhancer landscapes that can be used to characterize CD4+ T cell signaling at large . AND mice on a B10 . BR background were received from Dr . Michael Croft ( Rogers and Peptide dose , 1999; Rogers et al . , 1998 ) and bred in a specific pathogen free facility . All animal experiments were in compliance with the ethical standards set forth by UC San Diego’s Institutional Annual Care and Use Committee ( IUCAC ) . Spleens were extracted and manually digested . CD11c+ cells were isolated using Miltenyi Biotec Inc . ( San Diego , CA ) MACS magnetic cell separation with positive selection for CD11c ( CD11c , Biolegend , cat . no . 117304 ) . Subsequently , the CD11c- splenic fraction was used to negatively select for naïve CD4+ T cells using the Miltenyi MACS system with the following antibodies: CD11c ( Biolegend , cat . no . 117304 ) ; CD45R ( eBioscience , cat . no . 13-0452-86 ) ; CD11b ( eBioscience , cat . no . 13-0112-86 ) ; CD25 ( eBioscience , cat . no . 36-0251-85 ) ; CD49b ( eBioscience , cat . no . 13-5971-85 ) ; CD69 ( eBioscience , cat . no . 13-0691-85 ) ; CD8a ( eBioscience , cat . no . 13-0081-86 ) ; Ly-6G ( eBioscience , cat . no . 13-5931-86 ) ; MHC class II ( eBioscience , cat . no . 13-5321-85 ) ; TER-119 ( eBioscience , cat . no . 13-5921-85 ) . CD11c+ and CD4+ cells were cultured at a ratio of 1:2 in 96-well round-bottom plates for 24 hr , 108 hr ( for proliferation assay ) , or 3 . 5 hr ( for ERK phosphorylation staining ) . Peptides were added at indicated concentrations with the CD11c+ and CD4+ cells in DMEM supplemented with 10% Fetal Bovine Serum . For sequencing experiments , CD4+ cells were re-isolated from the culture using the Miltenyi MACS system and the same set of antibodies as above less CD25 and CD69 . For phospho-ERK staining , whole splenic cells were used , rather than purified CD11c+ and CD4+ cells . Peptides were ordered from Peptide 2 . 0 ( Chantilly , VA ) with the following amino acid sequences ( Rogers and Peptide dose , 1999; Rogers et al . , 1998 ) : PCC – KAERADLIAYLKQATAK K99A – KAERADLIAYLAQATAK Y97K – ANERADLIAKLKQATK K99E – ANERADLIAYLEQATK MCC – ANERADLIAYLKQATK Lyophilized peptides were resuspended in water , and added at the indicated concentrations to the cell cultures . Unstimulated CD4+ cells received an equivalent amount of water alone . Flow cytometry was performed on a LSR II and LSR Fortessa , both from BD Biosciences ( San Jose , CA ) . Cells were stained as per manufacturers’ protocols with the following antibodies: CD4-APC ( eBioscience , cat . no . 17-0042-83 ) ; CD4-PE-Cyanine7 ( BioLegend , cat . no . 116016 ) ; CD69-FITC ( eBioscience , cat . no . 11-0691-82 ) ; CD25-PE ( eBioscience , cat . no . 12-0251-82 ) ; Valpha11-FITC ( BD Pharmingen , cat . no . 553222 ) ; Vbeta3-PE ( BD Pharmingen , cat . no . 553209 ) ; CD11c-PE-Cyanine7 ( eBioscience , cat . no . 25-0114-82 ) ; IRF4-PerCP-Cy5 ( eBioscience , cat . no . 46-9858-80 ) ; Tbet-PE ( Santa Cruz Biotechnology , cat . no . SC-21749 ) ; CD122-PE ( BioLegend , cat . no . 105905 ) ; Ly6a-APC-Cyanine7 ( BioLegend , cat . no . 108125 ) ; CD200-APC ( BioLegend , cat . no . 123809 ) ; TNFSF11-APC ( BD Biosciences , cat . no . 560296 ) ; phospho-ERK-Alexa Fluor 488 ( Cell Signaling , cat . no . 4344S ) . Live/dead staining was performed using Fixable Aqua ( Life Technologies , cat . no . L34957; or Biolegend , cat . no . 423102 ) . Cells were gated on CD4+ , Aqua- cells . For phosphor-ERK staining , permeabilization was performed using BD Phosflow Perm Buffer III ( cat . no . 558050 ) and BD Fix Buffer I ( cat . no . 557870 ) . Analysis was performed with FlowJo v10 . 6 ( Tree Star; Ashland , OR ) . All flow cytometry results shown are representative of at least two biological replicates , and the results shown in Figure 1 were reflected with samples held out of each high-throughput sequencing assay . CD11c+ cells and naïve CD4+T cells were isolated form spleens of AND mice using magnetic separation as described above . CD11c+ cells and CD4+ T cells were cultured at a ratio of 1:2 in 96 –well round bottom plates with various concentrations of K99A ( 10 μM and 100 μM ) and PCC ( 0 . 1 μM and 10 μM ) . After 22 hr , cells were harvested and mRNA expression of Irf4 was analyzed on single cell level by flow cytometry in combination with CD4 , and CD69 protein staining , using FlowRNA II Assay kit ( Affymetrix eBioscience ) according to manufacturer’s protocols ( Porichis et al . , 2014 ) . Cells were analyzed with a BD AriaII flow cytometer . Data were analyzed using FlowJo v887software ( Tree Star ) . Two technical and two biological replicates were obtained across all conditions using the two different peptides at two different concentrations to stimulate naïve CD4+ T cells . Prior to sequencing , CD4+ T cells were separated from the co-cultured cells using Miltenyi MACS negative selection as described above for the initial culturing . ChIP-sequencing for H3K4me2 and H3K27Ac in the 1 μM peptide treatments was performed as described ( Gilfillan et al . , 2012 ) , with the following modifications: sodium butyrate was used to inhibit de-acetylation; and three RIPA and three LiCl washes were performed instead of five and one . ChIP-sequencing for H3K4me2 and H3K27Ac across the five conditions was performed as described ( Gosselin et al . , 2014 ) . RNA-sequencing was performed as described , with minor modifications ( Wang et al . , 2011 ) . Two replicates of the H3K4me2 ChIP-seq across all five conditions were obtained; three replicates across three conditions and one across all five conditions were obtained for the H3K27Ac ChIP-seq; and two replicates across all five conditions of the RNA-seq were obtained . ChIP-sequencing antibodies used were: H3K4me2 ( Millipore , cat . no . 07–030 ) and H3K27Ac ( Abcam , cat . no . ab4729 and Active Motif , cat . no . 39135 ) . To analyze the transcriptome of activated CD4+ T cells , we sorted CD69 positive cells . CD11c+ and CD4+ T cells were cultured for 22 hr as described above using two peptides at different concentrations K99A ( 10 μM and 100 μM ) and PCC ( 0 . 1 μM and 10 μM ) . After harvesting , cells were stained with Zombie Aqua live/dead stain ( Biolegend ) and with CD4-PE ( clone RM4-5 , eBioscience ) and CD69-PE Cy7 ( clone H1 . 2F3 , eBioscience ) conjugated antibodies . Cells were sorted with a BD AriaII cell sorter using a 70 μm nozzle . Live CD4+ cells were sorted into two populations according to the expression of the CD69 activation marker . CD69+ cells from the various culture conditions were used for RNA extraction . Unstimulated CD69- cells were used as controls . After extraction with Trizol , RNA was PolyA-selected ( MicroPoly ( A ) Purist kit , Ambion ) and libraries were generated as previously described ( Heinz et al . , 2013 ) . Samples were sequenced using NextSeq2 ( Illumina ) according to manufacturer recommended protocols . CD4+ T cells , isolated as described above , were pre-treated with 0 . 5 μM Promega U0126 ( cat . no . V1121 ) for thirty minutes at 37°C . CD11c+ cells and peptides were subsequently as indicated and cultured in the presence of the inhibitor for 24 hr . ChIP-sequencing reads were mapped to the mm10 genome using Bowtie2 ( Langmead and Salzberg , 2012 ) , and RNA-sequencing reads were mapped using STAR ( Dobin et al . , 2013 ) . Default allowed error rates were used , and only uniquely mapping reads were used in downstream analysis . Initial processing of aligned data and peak calling was performed using Homer ( Heinz et al . , 2010 ) . IDR analysis ( Landt et al . , 2012 ) for ChIP-sequencing replicates was performed using the homer-idr package ( Allison , 2015 ) . Vespucci ( Allison et al . , 2014 ) was used for counting AP-1 tags in gene regions . Gene Ontology analysis was performed using the Metascape Gene Annotation and Analysis Resource tool ( http://metascape . org/gp/index . html#/main/step1 ) using the express analysis settings ( Tripathi et al . , 2015 ) . Super-enhancers were called using Homer ( Heinz et al . , 2010 ) , which follows the published procedure ( Hnisz et al . , 2013; Lovén et al . , 2013; Whyte et al . , 2013 ) by first stitching together peaks into larger regions and then sorting regions by normalized H3K27Ac tag count . Region scores are plotted against rank , and a threshold is defined by finding the point at which the tangent to the plotted rank-scores is one . Regions past that threshold are called super-enhancers . Underlying data sets , including RPKM values and peaks , as well as code for all analyses described is publicly available at https://github . com/karmel/and-tcr-affinity . Analyses were performed using iPython Notebook ( Perez and IPython , 2007 ) . Clustering and PCA was performed using the scikit-learn package ( Pedregosa , 2011 ) . For full execution details and parameters , please see the code in the Github repository linked here . To generate the list of genes used in the activation signature score , we separately ran Principal Component Analysis on two replicates of RNA-Seq data and also the combined expression data from both replicates . Genes with an RPKM less than 100 in the No Peptide condition or a standard deviation greater than 20% of the No Peptide expression level across replicates were then omitted from the target set of genes . Remaining genes were sorted along PC1 , and genes that were in the top ten percent in all three data sets ( 215 ) or the bottom ten percent in all three data sets ( 137 ) were included in the set of activation signature score genes used in analysis . To compute the activation signature score , we take the dot product of the values of genes along PC1 in the combined RNA-Seq data set and the mean-centered expression levels for those genes for each sample in an experimental data set , yielding a single scalar score for each experiment . The scores across samples are then scaled by the max score , ensuring values are in the range of [-1 , 1] . The activation signature score tool is described and downloadable here: https://github . com/karmel/and-tcr-affinity/tree/master/andtcr/rna/activationscore . Publicly available datasets used for the analyses in Figure 3 and Figure 3—figure supplement 3 is available from GEO with the following Accession Codes: GSE14308 , GSE32224 , GSE41866 , GSE42276 , GSE54938 , and GSE60337 . AP-1 binding data is from GSE39756 . For Figure 5E , the following datasets were used: GSE56456 , GSE31233 , GSE40463 , GSE21365 , GSE56098 , GSE21512 . Raw and processed data are provided in the Gene Expression Omnibus ( GEO ) under accession number GSE69545 .	T helper cells recognize and respond to bacteria , viruses and other invading microbes and thus play a central role in the adaptive immune system . These cells have a receptor on their surface that binds to fragments of proteins – known as oligopeptides – from the microbes that have been digested and presented on the surfaces of other immune cells . Once active , T helper cells multiply , grow and release signals that regulate genes in other cells to promote immune responses . Previous studies suggest that a T helper cell’s response is binary – that is , either on or off . However , this does not explain how the strength of the T cell response to infection can vary . Allison et al . used a technique called high-throughput sequencing to examine the activity of genes in T helper cells from mice that had been genetically engineered to only produce one type of T cell receptor . For the experiments , the T cells were exposed to various concentrations of different peptides known to bind either well or poorly to the receptor . Allison et al . found that , once activated , the response of an individual T cell was not binary , but instead was related to the strength of the signal it received through its receptor . Further experiments showed that although a subset of the genes activated in T helper cells do respond in a binary fashion , the activities of many other genes involved in immune responses and cell metabolism were related to the strength of the signal from the receptor . This “analog” gene activation depends on the level of activity of the MAP kinase signaling pathway . Together , Allison et al . ’s findings help us to understand how T cells are able to fine-tune immune responses to invading microbes . The next challenge will be to investigate the mechanisms underlying binary and analog gene activity in T cells .	[ "Abstract", "Introduction", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "immunology", "and", "inflammation" ]	2016	Affinity and dose of TCR engagement yield proportional enhancer and gene activity in CD4+ T cells
The rice seedling blight fungus Rhizopus microsporus and its endosymbiont Burkholderia rhizoxinica form an unusual , highly specific alliance to produce the highly potent antimitotic phytotoxin rhizoxin . Yet , it has remained a riddle how bacteria invade the fungal cells . Genome mining for potential symbiosis factors and functional analyses revealed that a type 2 secretion system ( T2SS ) of the bacterial endosymbiont is required for the formation of the endosymbiosis . Comparative proteome analyses show that the T2SS releases chitinolytic enzymes ( chitinase , chitosanase ) and chitin-binding proteins . The genes responsible for chitinolytic proteins and T2SS components are highly expressed during infection . Through targeted gene knock-outs , sporulation assays and microscopic investigations we found that chitinase is essential for bacteria to enter hyphae . Unprecedented snapshots of the traceless bacterial intrusion were obtained using cryo-electron microscopy . Beyond unveiling the pivotal role of chitinolytic enzymes in the active invasion of a fungus by bacteria , these findings grant unprecedented insight into the fungal cell wall penetration and symbiosis formation . Interactions between bacteria and fungi are widespread in nature and play pivotal roles in ecological and medicinal processes ( Frey-Klett et al . , 2011 ) . Moreover , fungal-bacterial associations are widely used for the preservation of the environment ( e . g . , mycorrhizae in reforestation ) , agriculture ( e . g . , food processing ) , and biotechnology ( e . g . , pharmaceutical research ) ( Scherlach et al . , 2013 ) . Beyond the most commonly observed microbial cell–cell interactions , there is a growing number of known endosymbioses where bacteria dwell within fungal hyphae ( Bonfante and Anca , 2009; Kobayashi and Crouch , 2009; Lackner et al . , 2009b; Frey-Klett et al . , 2011 ) . Symbioses with endofungal bacteria are often overlooked , yet they may have a profound effect on the host's lifestyle . Bacterial endosymbionts of arbuscular mycorrhizal fungi , for example , might be implicated in the vitamin B12 provision for the fungus ( Ghignone et al . , 2012 ) . Endobacteria Rhizobium radiobacter , isolated from the mycorrhizal fungus , exhibit the same growth promoting effects and induce systemic resistance to plant pathogenic fungi in the same way that the fungus harboring the endobacteria does . Thus , it was proposed that the beneficial effects for the plant result directly from the presence of bacteria ( Sharma et al . , 2008 ) . The rice seedling blight fungus , Rhizopus microsporus , and its endosymbiont bacterium , Burkholderia rhizoxinica represent a particularly noteworthy example of a bacterial-fungal endosymbiosis ( Partida-Martinez and Hertweck , 2005; Lackner and Hertweck , 2011 ) . The fungus harbors endosymbionts of the genus Burkholderia , which reside within the fungal cytosol , as shown by confocal laser scanning microscopy , transmission electron microscopy ( EM ) and freeze–fracture EM ( Partida-Martinez et al . , 2007a , 2007b , 2007c ) . The bacteria are harnessed by the fungus as producers of highly potent antimitotic macrolides ( Scherlach et al . , 2006 ) , which are then further processed by the host into the phytotoxin rhizoxin ( Scherlach et al . , 2012 ) . The toxin represents the causative agent of rice seedling blight , which weakens or kills the rice plants ( Lackner et al . , 2009b ) . Both the saprotrophic fungus and the endofungal bacteria benefit from the nutrients released , and R . microsporus provides a protective shelter for the bacterial partner . The Rhizopus-Burkholderia association also stands out as it employs an elegant mechanism that allows the persistence and spreading of the symbiosis through spores containing the endosymbionts ( Partida-Martinez et al . , 2007c ) ( Figure 1 ) . Yet it is unknown how the vegetative reproduction of the fungus has become totally dependent upon the presence of the endobacteria ( Partida-Martinez et al . , 2007c ) . Insights into the genome of B . rhizoxinica and mutational studies have unveiled several symbiosis factors ( Leone et al . , 2010; Lackner et al . , 2011a , 2011b ) . 10 . 7554/eLife . 03007 . 003Figure 1 . Microscopic image of Burkholderia rhizoxinica ( green ) residing in the cytosol of Rhizopus microsporus . The GFP encoding B . rhizoxinica cells can re-colonize the sterile R . microsporus , then induce fungal sporulation . The endobacterium is transmitted via fungal vegetative spores ( upper right corner ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 003 A plausible scenario for the evolution of the symbiosis is a shift from antibiosis or antagonism to mutualism . The rhizoxin complex secreted by the bacteria arrests mitosis in almost all eukaryotic cells . Yet , Rhizopus , amongst other zygomycetes , has gained resistance to this toxin due to a mutation at the β-tubulin binding site ( Schmitt et al . , 2008 ) . Furthermore , phylogenetic analyses point to host switching events during evolution ( Lackner et al . , 2009a ) , which is also supported by the engagement of an hrp locus of B . rhizoxinica ( Lackner et al . , 2011a ) . In addition to this , the LPS layer of the B . rhizoxinica is known to be unique to its niche , due to high resemblance to fungal sugar content ( Leone et al . , 2010 ) . Although there is ample knowledge on the persistence of the symbiosis , it has remained fully enigmatic how the bacteria enter the fungal cells . Interestingly , there is no sign of endo-/phagocytosis , which rules out a major avenue of bacterial colonization ( Partida-Martinez and Hertweck , 2005; Partida-Martinez et al . , 2007a , 2007c ) . Bacterial invasion of eukaryotic cells is a major area of research in infection biology , and a large body of knowledge has been gathered on the pathogen's strategies to invade host cells ( Cossart and Sansonetti , 2004 ) . In addition to host driven endocytosis , a number of enzymes have been described that act locally to damage host cells and to facilitate the entry of the pathogen into the tissue ( Harrison , 1999; King et al . , 2003 ) . Yet , this knowledge is limited to the invasion of human , animal and plant cells . It has been reported that some bacteria employ extracellular enzymes for mycophagy ( Leveau and Preston , 2008 ) . However , despite a growing number of described fungal endobacteria ( Lackner et al . , 2009b; Frey-Klett et al . , 2011 ) , there is a striking lack of knowledge about the avenues and active mechanisms that permit fusion with or entry into fungal hyphae , where the fungus is left intact to serve as a host for the endobacteria . Here we report the genomics- and proteomics-driven discovery of a new bacterial invasion process that involves the secretion of chitinolytic enzymes . Furthermore , we present the first electron microscopic snapshots of the actual infection process . Both pathogens ( or antagonists ) and mutualists often employ the same mechanisms during the infection process ( Dale and Moran , 2006 ) . Thus , we mined the gene repertoire of B . rhizoxinica ( Lackner , 2011b ) for potential molecular infection mechanisms known from pathogenic bacteria . A type 2 secretion system ( T2SS ) , also called general secretion pathway ( gsp ) , encoded by a 12 kb gene cluster on the B . rhizoxinica chromosome ( Table 1 ) seemed to be a promising candidate to enable the bacterium to enter the host . T2SS are typically involved in the secretion of various toxins and lytic enzymes ( Cianciotto , 2005; Korotkov et al . , 2012 ) and the overall organization of T2SS gene clusters is well conserved between related species ( Figure 2 ) . We generated targeted deletion mutants to investigate the role of the T2SS in the infection process . Specifically , we selected gspC and gspD , since their gene products are essential proteins of the type 2 secretion machinery in related bacteria ( DeShazer et al . , 1999; Korotkov et al . , 2011 ) . The outer membrane pore is formed presumably by the multimeric secretin GspD , and GspC appears to link the inner and outer membranes by providing the contact to GspD via a homology region ( Korotkov et al . , 2011 ) . Although it is notoriously difficult to genetically modify the symbiotic bacteria , we succeeded in generating ΔgspC and ΔgspD mutants using a double-crossover strategy . 10 . 7554/eLife . 03007 . 015Table 1 . Annotation of the B . rhizoxinica T2SS gene clusterDOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 015GeneProposed function of encoded proteinHomolog in Klebsiella oxytocaPercent identity ( BlastP ) Homolog in Burkholderia pseudomalleiPercent identity ( BlastP ) gspCConnecting inner and outer membrane complexpulC–gspC61gspDSecretin , outer membrane pore formationpulD39gspD64gspECytoplasmitc ATPase , energy for translocation of pseudopilinspulE54gspE89gspFAnchoring protein , inner membrane platform for pseudopilinspulF46gspF83gspGMajor prepilin-like protein , pilus-like structure formationpulG56gspG87gspHPseudopilin subunit , pilus-like structure formationpulH70gspH66gspIPseudopilin subunit , pilus-like structure formationpulI44gspI66gspJPseudopilin subunit , pilus-like structure formationpulJ43gspJ55gspKPseudopilin subunit , pilus-like structure formationpulK26gspK62gspLAnchoring protein , inner membrane platform for pseudopilinspulL27gspL57gspMAnchoring protein , inner membrane platform for pseudopilinspulM24gspM56gspNConnecting inner and outer membrane complexpulN29gspN63gspOPrepilin , inner membrane peptidasepulO43gspO6010 . 7554/eLife . 03007 . 004Figure 2 . Schematic view of the organization of the type 2 secretion system ( T2SS ) gene clusters from various bacterial species . The T2SS gene loci of the two R . microsporus endosymbiotic bacterial strains B . rhizoxinica and Burkholderia endofungorum , as well as the squid endosymbiont Vibrio fischeri , Escherichia coli , and the human pathogens Burkholderia pseudomallei and Pseudomonas aeruginosa are displayed here . The spaces between the arrows represent non-adjacent genes ( single genes are located further away on the genome ) , while lines indicate closely linked genes . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 004 To begin , we addressed the proteolytic potential of B . rhizoxinica wild type ( wt ) and the T2SS defective mutants to evaluate the effect of the knock-outs ( Figure 3 , Figure 3—figure supplement 1 ) . Using a skim milk plate assay we detected strong proteolytic activity in the wt supernatant , while the T2SS mutants showed no activity ( Figure 3—figure supplement 3 ) . The ability of the isolated endobacteria to re-infect the fungus and to control fungal sporulation was examined using a sporulation bioassay . The appearance of mature sporangia that form sporangiospores is seen as an indication of a successful establishment of the symbiosis . In co-cultures of wt B . rhizoxinica and the cured fungal host , sporulation is visible after 2–3 days . In contrast , there was absolutely no visible spore formation upon co-cultivation with B . rhizoxinica ΔgspD::Kanr or ΔgspC::Kanr ( Figure 3 ) . Furthermore , fluorescence microscopy proved to be most helpful to distinguish between mutants defective in colonization or induction of fungal sporulation . A constitutive GFP-expressing strain allowed monitoring of the invasion of bacteria into the fungal hyphae . While fluorescent bacteria with an intact T2SS were able to enter the fungal cells , there was no detection of any endobacteria when either of the T2SS mutants was co-cultured with the cured fungus ( Figure 3 , Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 03007 . 005Figure 3 . Photographs and microscopy images of R . microsporus hyphae several days after inoculation with B . rhizoxinica wt or mutant strains in 6-well plates . The pictures present the infection of R . microsporus with wt or mutant B . rhizoxinica in the following order: ( A ) B . rhizoxinica wt , ( B ) B . rhizoxinica Δchi::Kanr , ( C ) B . rhizoxinica ΔgspC::Kanr , ( D ) —B . rhizoxinica ΔgspD::Kanr , ( E ) —control ( no bacteria added ) . Spore formation is visible in ( A ) , while spore formation was not detected in ( B–E ) even after 5 days of co-incubation . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 00510 . 7554/eLife . 03007 . 006Figure 3—figure supplement 1 . Knock out strategy of site directed mutagenesis to B . rhizoxinica . ( A ) —Illustration of PCR strategy and primers used to confirm mutant genotypes . ( B ) —Confirmation of B . rhizoxinica ΔgspD::Kanr by gel electrophoresis of PCR products amplified from genomic DNA of B . rhizoxinica ΔgspD::Kanr or wt , respectively . Primers were combined in individual PCR reactions as indicated in above . ( C ) Confirmation of B . rhizoxinica ΔgspC::Kanr by PCR . ( D ) Confirmation of B . rhizoxinica Δchi::Kanr by PCR . ( E ) Confirmation of B . rhizoxinica Δcbp::Kanr by PCR . ( F ) Confirmation of B . rhizoxinica Δchts::Kanr by PCR . ( G ) Confirmation of B . rhizoxinica Δchi::Kanr by PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 00610 . 7554/eLife . 03007 . 007Figure 3—figure supplement 2 . Fluorescence microscopy carried out with B . rhizoxinica wt and mutant strains B . rhizoxinica ΔgspD , B . rhizoxinica ΔgspC and B . rhizoxinica Δchi in a 3 day co-culture with sterile R . microsporus . Whereas wt bacteria clearly localize within the fungal hyphae , no endobacteria carrying mutations in either the chitinase gene or the components of the T2SS are detectable within the living hyphae ( upper part ) . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 00710 . 7554/eLife . 03007 . 009Figure 3—figure supplement 3 . Lytic potential of B . rhizoxinica secretome . ( A ) —Proteolytic assay using skim milk agar plates . Aliquots of the supernatants of B . rhizoxinica wt , B . rhizoxinica ΔgspD::Kanr , B . rhizoxinica ΔgspC::Kanr , respectively are incubated on filter disks on the plate . ( B ) —Chitinolytic plate assay using B . rhizoxinica culture supernatant on agar plate supplemented with 0 . 05% colloidal chitin and stained with Calcofluor . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 009 In order to identify the secreted factors that could play a role in the bacterial fungal interaction we performed comparative 2-D gel electrophoresis of the exoproteomes ( secretomes ) of wt and mutant bacteria ( Figure 4 ) . The secretome analysis of the T2SS mutants showed a substantial reduction in the total protein yield ( 0 . 1% ± 0–015% of the wt secretome ) despite an intense protein precipitation with fivefold TCA and the requirement of 100 µg of protein from each samples to be loaded on 2-D gel electrophoresis . In the wt secretome we identified surprisingly few proteins , although they were present in great abundance . Using MALDI-TOF we were able to detect the majority of the secreted proteins as chitin-binding protein ( Cbp ) and chitosanase ( Chts ) , which are encoded in the bacterial genome . The chitin-binding protein belongs to the non-catalytic carbohydrate-binding proteins of the CBM33 family ( Henrissat and Davies , 2000 ) , which can bind to chitin and facilitate the action of chitinases . The chitosanase is part of the glycoside hydrolase family 46 with a specific hydrolytic activity on chitosan ( Henrissat and Davies , 2000 ) . 10 . 7554/eLife . 03007 . 011Figure 4 . 2D gel analysis of the secretomes of wild type and mutants . ( A ) B . rhizoxinica wt , ( B ) B . rhizoxinica ΔgspC::Kanr , ( C ) B . rhizoxinica ΔgspD::Kanr and ( D ) B . rhizoxinica wt in co-culture with R . microsporus . Chitin-binding protein ( I-5 , III-14 , III-15 , IV-6 , II-5 ) , chitinase ( I-6 , III-9 ) and chitosanase ( I-3 , I-4 , III-10 , III-11 , IV-3 , IV-4 , IV-5 , II-3 ) were identified . List of identified proteins is shown in Table 2 . In general , each gel was loaded with 100 μg of TCA-precipitated proteins from the culture supernatant . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 01110 . 7554/eLife . 03007 . 016Table 2 . Secretome proteins identified by MALDI-TOFDOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 016Spot no . Accesion no . ( NCBI ) Protein indetificationTotal Mass0 ( kDA ) Total pIMascot scoreMatching peptidesSequence coverage ( % ) SignalP prediction ( y/n ) I-1gi│312169059Glutamate/aspartate-binding protein32 . 49 . 664 . 21340 . 1yI-2gi│312169059Glutamate/aspartate-binding protein32 . 49 . 6114 . 01849 . 5yI-3gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 588 . 31038 . 9yI-4gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 5124 . 01547 . 7yI-5gi│312168534Chitin-binding protein27 . 07 . 785 . 2849 . 2yI-6gi│312168091Chitinase ( EC 3 . 2 . 1 . 14 ) 42 . 18 . 730 . 8623 . 1yII-1gi│312169059Glutamate/aspartate-binding protein32 . 49 . 6107 . 02051 . 5yII-2gi│312169059Glutamate/aspartate-binding protein32 . 49 . 6114 . 01846 . 8yII-3gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 5169 . 01560 . 3yII-4gi│312168534Chitin-binding protein27 . 07 . 780 . 4729 . 1yIII-1gi│312169059Glutamate/aspartate-binding protein32 . 49 . 6133 . 02051 . 9yIII-2gi│312169059Glutamate/aspartate-binding protein32 . 49 . 696 . 21951 . 2yIII-3gi│312167620Toluene transport system Ttg2d protein23 . 79 . 569 . 0530 . 5yIII-4gi│312168022Ribosome recycling factor ( RRF ) 21 . 09 . 074 . 0944 . 1nIII-5gi│312169310Adenosylhomocysteinase ( EC 3 . 3 . 1 . 1 ) 52 . 75 . 9247 . 02969 . 2nIII-6gi│312166966Chaperone protein DnaK69 . 74 . 9222 . 03048 . 2nIII-7gi│31216888860 kDa chaperonin GroEL57 . 45 . 197 . 71641 . 0nIII-8gi│312169323S-adenosylmethionine synthetase ( EC 2 . 5 . 1 . 6 ) 42 . 84 . 8182 . 02063 . 8nIII-9gi│312167529Protein translation elongation factor Tu ( EF-Tu ) 43 . 15 . 2148 . 02266 . 9nIII-10gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 5156 . 01763 . 6yIII-11gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 5127 . 01654 . 0yIII-12gi│312168091Chitinase ( EC 3 . 2 . 1 . 14 ) 42 . 18 . 742 . 8829 . 1yIII-13gi│312168185Peptidyl-prolyl cis-trans isomerase ( EC 5 . 2 . 1 . 8 ) 28 . 78 . 991 . 01240 . 8yIII-14gi│312168534Chitin-binding protein27 . 07 . 7101 . 01259 . 8yIII-15gi│312168534Chitin-binding protein27 . 07 . 760 . 6534 . 8yIII-16gi│312167051Superoxide dismutase ( EC 1 . 15 . 1 . 1 ) 23 . 55 . 9128 . 0958 . 7nIII-17gi│312168185Peptidyl-prolyl cis-trans isomerase ( EC 5 . 2 . 1 . 8 ) 28 . 78 . 984 . 41337 . 6yIII-18gi│312167149Inorganic pyrophosphatase ( EC 3 . 6 . 1 . 1 ) 19 . 44 . 8117 . 01284 . 1nIII-19gi│31216779534 kDa membrane antigen precursor21 . 86 . 885 . 51360 . 7yIV-1gi│312169059Glutamate/aspartate-binding protein32 . 49 . 6151 . 02362 . 6yIV-2gi│312169059Glutamate/aspartate-binding protein32 . 49 . 698 . 61545 . 5yIV-3gi│312167773Chitosanase ( EC 3 . 2 . 1 . 132 ) 39 . 66 . 571 . 3823 . 0yIV-4gi│312168534Chitin-binding protein27 . 07 . 795 . 3737 . 3y Chitin and chitosan are well known as major structural components of the fungal cell wall ( Gooday , 1990; Adams , 2004 ) . Chitosan is a dominant component of the Zygomycete cell wall , but chitin is also abundant , as we could show by calcofluor staining of the R . microsporus cell wall ( Figure 5B ) . While screening the B . rhizoxinica genome for genes for chitinolytic enzymes , we also detected a gene for a chitinase that contains a signal sequence for secretion . The corresponding gene product could also in fact be detected by 2-D gel electrophoresis and MALDI analysis , albeit in lower abundance than Cbp and Chts . From the structure-based alignment and phylogenetic information ( Figure 5C ) we can conclude that the B . rhizoxinica chitinase ( Chi ) belongs to the family 18 chitinases in subfamily B . The closest structural homolog is PF-ChiA from Pyrococcus furiosus that has endochitinase activity ( Nakamura et al . , 2007 ) . 10 . 7554/eLife . 03007 . 008Figure 5 . Functional analyses of chitinolytic enzymes . ( A ) Chitinase activity in cell-free culture supernatants of an E . coli harboring recombinant chitinase from B . rhizoxinica after incubation for 30 , 60 , and 90 min , in comparison to the activity of a recombinant S . lividans enzyme . Error bars indicate the standard deviation of three individual experiments . ( B ) Calcofluor staining of R . microsporus hyphae . ( C ) Phylogenetic analysis of chitinases from the GH family 18; protein sequences were retrieved from NCBI and comprise subfamily A and B sequences . ( D ) Chitin-binding assay performed using acid released crab shell chitin and the supernatant of B . rhizoxinica wt . SDS page of the non-bound fraction ( S ) , the bound protein fraction ( F1 ) and the pelleted chitin with the rest of the bound protein ( F2 ) . The three indicated proteins were identified using MALDI-TOF . ( E ) Gene expression assay for T2SS and chitinolytic proteins . The expression of T2SS genes gspC and gspD as well as chi , cbp and chts in B . rhizoxinica were monitored using RT qPCR in pure culture , in co-cultivation with a cured host ( R . microsporus ) and after re-infection of the cured host . The gene rpoB was used as an internal standard for the calculation of expression levels and normalization . The expression of all five genes is substantially increased in the wt during co-cultivation , while expression levels after re-infection decreased to nearly the level in pure wt culture . Error bars indicate standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 00810 . 7554/eLife . 03007 . 010Figure 5—figure supplement 1 . Multiple alignment of the chitinase protein sequences was performed based on the three-dimensional structure of the ChiB sequence from S . marcescens ( pdb1e15 ) and Chi from P . furiosus ( 2dsk ) . Red and blue characters indicate α-helices and β-strands , respectively . The DXDXE motif is highlighted in yellow . Alignment was created using PROMALS3D . The CID motif of subfamily A chitinases in S . marcescens in underlined in orange , it is absent in B . rhizoxinica and P . furiosus sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 010 A chitin-binding assay verified that all three proteins bind to chitin ( Figure 5D ) . This finding is fully in line with the observation that the amount of chitinolytic proteins in the 2-D gel is greatly reduced when samples of fungal-bacterial co-cultures are applied ( Figure 4D ) , most likely because the proteins bind to the fungal cell wall . Despite the total decrease in secreted protein amount , analysis of the T2SS mutants , ΔgspD::Kanr and ΔgspC::Kanr , revealed that all three chitinolytic proteins are substantially reduced in the secretome . Chitinase could not be detected in the secretome of ΔgspD::Kanr , even when a 10-fold amount of precipitated secretome was loaded on the gel ( Figure 4C ) . This indicates that the detected proteins are all collected from the dead cells rather than being excreted , and proving the selective secretion of chitinolytic proteins through T2SS . In order to investigate which chitinolytic proteins are essential for hydrolyzing the fungal cell wall we individually deleted the corresponding genes . Using the gene deletion strategy described above , we successfully obtained the mutants Δchi::Kanr , Δcbp::Kanr and Δchts::Kanr . All three mutants were tested in the previously described sporulation assay . The chitosanase and the chitin-binding protein null mutants restored the symbiosis and retained their ability to illicit spore formation . In contrast , the chitinase deletion strain was no longer able to induce spore formation even after 1 week of extended co-culture ( Figure 3 ) . By using fluorescence microscopy , we found that bacteria that were incapable of producing chitinase could not invade fungal cells . The culture supernatant of a pure wild-type B . rhizoxinica culture showed weak activity in a chitinolytic plate assay . Also no significant activity within the B . rhizoxinica secretomes could be detected in an assay with the aqueous substrate CM-chitin-RBV . To unequivocally prove its ability for chitinolysis , the chi gene was cloned and chitinase was heterologously produced in Escherichia coli . The B . rhizoxinica chitinase-enriched secretome was tested in an assay with the aqueous substrate CM-chitin-RBV . A high chitinolytic activity could thus be observed that remained stable over several hours , while the E . coli expression host ( negative control ) showed no activity ( Figure 5A ) . Next , we wanted to address the question whether or not the production of chitinolytic proteins is constitutive or dependent upon the presence of the fungal host . Therefore , we monitored the expression of genes coding for chitinolytic proteins ( chi , cbp and chts ) and components of the T2SS ( gspD and gspC ) . All genes are 30- to 160-fold up-regulated in bacterial-fungal co-cultures compared to the pure bacterial culture ( Figure 5E ) . Surprisingly , the expression levels after re-infection nearly decrease to the level found in pure culture . These results strongly indicate that all of the tested genes play a crucial role during the infection process although they are not required for the maintenance of the symbiosis . According to our functional analyses , bacteria produce and secrete chitinolytic enzymes during infection . We reasoned that the bacteria employ these enzymes to locally digest the fungal cell wall , as a means of entering the cells . To monitor the bacterial invasion of the fungus , we performed several microscopic investigations using GFP-labeled bacteria . As early as 1 day after the infection , when spore formation is not yet visible , confocal laser scanning microscopy revealed that the bacteria were inside the fungal hyphae ( Figure 1 ) . We then used cryo-electron microscopy to capture the symbionts in the act of infection . This technique allows for a relatively low disturbance of the sample and few artifacts . The micrographs permitted an image of clearly distinguishable fungal hyphae ( Rm ) and a large number of bacteria ( Br ) surrounding or attaching to them . Fungal hyphae can be seen with a very smooth surface where single bacteria or bacterial colonies are attaching to it ( Figure 6A–E ) . Yet at this point , the attachment seems to be purely superficial , and both organisms can still be clearly distinguished from one another . A tight attachment occurs as soon as 1 hr after the co-incubation of the bacteria and fungus . We observed fibrillar structures connecting the bacteria to the fungal surface ( Figure 6D ) . In addition to this , we noted pleomorphism of the bacteria ( irregular shapes in Figure 6C inset and in Figure 6D , E ) . At a later stage ( Figure 6E–H ) , the bacteria seem to lose their sharp , pronounced form and enter the fungal cells by fusing with their cell wall ( Figure 6E–H ) . After 20 hr of co-culture fungal hyphae appear to lose some of their form and structural integrity . The fusion sites are still clearly visible even though some of the structures can only be vaguely identified as bacteria ( Figure 6C small image and Figure 6H ) . In some cases , bacteria were caught sticking halfway through the cell wall ( Figure 6E , F ) . By arranging the single steps in sequence we obtained snapshots of the complete course of the infection . 10 . 7554/eLife . 03007 . 012Figure 6 . Course of infection of B . rhizoxinica ( Br ) to R . microsporus ( Rm ) observed by scanning cryo-electron microscopy . ( A–D ) Attachment/adherence of bacteria to fungal hyphae after 1 hr and ( F–H ) 20 hr of co-cultivation; ( C–D , F–H ) bacterial and fungal cell walls start to merge; ( D ) fibrillar structures connecting a bacterial cell to the hyphal surface; ( E–H ) fusion of cell walls and the intrusion of bacterial cells into the fungal hyphae . White arrows mark areas of particular interest . ( I–J ) Scanning electron microscopy of a co-culture of sterile R . microsporus with B . rhizoxinica ΔgspD ( I ) and B . rhizoxinica Δchi ( J ) . Attachment of both mutant strains to the hyphal surface is visible , however , no intimate contact or fusion events could be observed . Scale bars represent 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 012 Notably , no active engulfment of the bacterium by the fungus has been observed by cryo EM . To further rule out a scenario involving endocytosis we employed fluorescent staining that would permit visualizing endocytosis and vesicular traffic of the fungus . We stained the fungal membrane with styryl dye FM4-64 ( Invitrogen , Carlsbad , USA ) and the bacterial symbionts by bacteria-specific dye Syto 9 ( Invitrogen , Carlsbad , USA ) . After 1 hr of co-incubation , we observed bacteria attached to the fungus but no fungal membrane surrounding the bacterium . After 5 hr , we detected bacteria within the hyphae , and again no fungal membrane was visible ( Figure 7 ) . 10 . 7554/eLife . 03007 . 013Figure 7 . Evidence for the lack of active engulfment of B . rhizoxinica by R . microsporus . ( A ) Bacteria ( green ) attaches to fungal hyphae within 1 hr . ( B ) The endocytotic activity of the fungus can be observed by the red vesicles that are present all around the hyphae and highly accumulated at the apical tip . After the infection , no fungal membrane is visible around the bacterium . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 013 To evaluate the effect of the genes crucial for hyphal entry , we conducted scanning electron microscopy studies of B . rhizoxinica ΔgspD::Kanr and Δchi::Kanr . We screened several co-cultures and detected bacterial cells attached to the surface of fungal hyphae . Nevertheless , it should be pointed out that the bacteria appeared to be only loosely attached , and even after extended co-incubation no fusion events could be detected ( Figure 6I , J ) . B . rhizoxinica and R . microsporus form a phytopathogenic alliance that jointly produces and secretes the highly potent phytotoxin rhizoxin , the virulence factor inducing rice seedling blight ( Scherlach et al . , 2012 ) . Symbiosis factors such as the hrp locus of B . rhizoxinica and the LPS layer contribute to the persistence of the tight association of the host fungus and its specific bacterial endosymbiont ( Leone et al . , 2010; Lackner et al . , 2011a ) . Although B . rhizoxinica has undergone significant genome reduction ( Moran et al . , 2008; Lackner et al . , 2011b ) it still retains the ability to grow in pure culture and to re-infect the sterile fungal host . During this process the bacteria have to penetrate the fungal cell wall barrier . Transmission electron microscopy and freeze-fracture electron microscopy have showed that the endobacteria are not surrounded by a fungal membrane ( Partida-Martinez and Hertweck , 2005; Partida-Martinez et al . , 2007a , 2007c ) , which rules out an phagocytosis-like vesicular uptake , as seen in the Nostoc punctiforme—Geosiphon pyriformis symbiosis ( Mollenhauer et al . , 1996 ) . In this paper we have unveiled an alternative avenue for an active bacterial invasion of fungal hyphae involving the secretion of chitinolytic enzymes . Based on genomic and proteomic analyses we have discovered a type 2 secretion system in the fungal endosymbiont B . rhizoxinica that is central to the Burkholderia-Rhizopus interaction . Core components of the T2SS were targeted for deletion and corresponding mutants were incapable of forming a symbiosis . Previous mutational studies of various T2SS have provided evidence of their involvement in pathogenesis ( DeShazer et al . , 1999; Ali et al . , 2000; Roy Chowdhury and Heinemann , 2006 ) . T2SS may also be absent in pathogens , and several T2SS of mutualists have already been described ( Filloux , 2004; Cianciotto , 2005 ) . However , they are absent in some well-studied aphid and insect symbionts ( Cianciotto , 2005 ) . In a recent report on the T2SS of the obligate arbuscular mycorrhizal fungus symbiont Candidatus Glomeribacter gigasporarum , it was shown that the expression of the gene coding for GspD was up-regulated in the obligate symbiont ( Ghignone et al . , 2012 ) . Here , we demonstrate for the first time that a T2SS is crucial for a bacterial-fungal symbiosis . We also elucidate the key role of the T2SS in secreting chitinolytic enzymes and chitin-binding proteins . Chitin is well known as one of the major structural components of the fungal cell wall ( Gooday , 1990; Adams , 2004 ) , and chitinases are secreted by bacteria primarily during mycophagy ( de Boer et al . , 2004; Leveau and Preston , 2008 ) and pathogenesis ( Chernin et al . , 1995; Connell et al . , 1998; Francetic et al . , 2000 ) . In this study we found that the deletion of the chitinase ( chi ) gene completely abolished the bacteria's ability to enter the fungal hyphae and thus rendering it incapable of establishing a functional symbiosis . The deletion of two additional genes coding for a chitin-binding protein and a chitosanase showed no effect on the sporulation assay . However , these two enzymes are present in great abundance in the B . rhizoxinica wt secretome and likely support the action of chitinase . This idea of these enzymes' function is further supported by the fact that the chitosan is highly abundant in the cell wall of zygomycetes ( and fungi in general ) and that secretion of both proteins is highly reduced in the T2SS mutants , which are unable to intrude the fungus . Expression levels of these proteins are highly increased in co-culture with the fungus similar to the chitinase gene , suggesting a co-regulation of the transcription of the three genes . Chitin binding protein ( Cbp ) is the most abundant protein in the cell-free supernatant of B . rhizoxinica , and its expression levels are higher in co-culture with the host fungus . We therefore assume that Cbp facilitates bacterial attachment to the fungal hyphae and renders the chitin matrix more accessible to chitinase degradation as it was proposed for family 33 Cbps . Small Cbps also promote the recognition and degradation of chitin by streptomycetes ( Schrempf , 2001 ) . Chitin-binding proteins could play a role in various close interactions where bacteria attach to the hyphal surface and form fibrillar structures , as in specific Streptomyces-Aspergillus ( Siemieniewicz and Schrempf , 2007 ) and Paenibacillus-Fusarium co-cultures ( Dijksterhuis et al . , 1999 ) . Chitin binding may also set the stage for intrusion , as observed in Burkholderia spp . and AM fungal spores ( Levy et al . , 2003 ) and Rhizopus hyphae , as reported in this study . In light of the fact that chitosan is the dominant component of the Zygomycete cell wall ( Bartnicki-Garcia and Nickerson , 1962 ) it is surprising that only the chitinolytic enzyme plays a crucial role in the active invasion of bacteria into fungal cells . However , chitosanase likely supports the invasion process . To establish an intimate association , the physical contact must occur at the right time and the right place and may be dependent upon many factors ( Bright and Bulgheresi , 2010 ) . The microscopic images and the gene expression studies indicate that the bacteria attach themselves to the fungus even before the chitinolytic enzymes are produced and secreted . In this way low concentrations of lytic enzymes would be sufficient for local activity . Thus , fungal cell wall penetration is a more melting-like , mild process without damaging the hyphae . A similar scenario has been described in the context of plant infection , where the precise and highly localized cellulolytic activity of cellulase CelC2 from Rhizobium leguminosarum bv . trifolii degrades the host plant cell wall during penetration ( Robledo et al . , 2008 ) . The model of topical cell wall lysis in the Burkholderia-Rhizopus interaction is supported by the microscopic snapshots of the progress of hyphal colonization and intrusion . As early as 1 hr after co-incubation with fungal hyphae , a close attachment of the bacterial cells can be observed , followed by fusion with the fungal cell wall . This process is observed very locally for every bacterium , even when several bacteria form fusion structures close to each other on the fungal cell wall . Although parts of the penetrated cell wall may appear a bit irregular ( Figure 6F ) , there are no visible signs of cell lysis or loss of integrity surrounding the intrusion sites . Both , the mutants lacking chitinase or T2SS components attach in a comparable yet weaker fashion to the fungal hyphae but are not capable of similar fusion events ( Figure 6I–J ) . Moreover , cell membrane staining to fungus has shown that the fungus does not engulf the bacterium by endocytosis . Overall , this strategy permits a traceless entry into the fungal cells , thus guaranteeing that host integrity is not affected . In summary , we identified a T2SS and a secreted chitinase as two molecular mechanisms involved in the attachment and infection process of an agricultural and medicinal relevant bacterial-fungal interaction . Secretion of chitinase and presumably further effector proteins translocated via a T2SS help to locally soften the fungal cell wall allowing bacterial entry and preventing the disintegration of fungal hyphae ( Figure 8 ) . Considering the growing number of reports about endobacteria in mycorrhiza and other fungi ( Bianciotto et al . , 2000; Bonfante and Anca , 2009; Kobayashi and Crouch , 2009; Hoffman and Arnold , 2010; Frey-Klett et al . , 2011 ) as well as the first indications about their functional implication in the bacterial-fungal and plant-fungal relationships ( Partida-Martinez and Hertweck , 2005; Sharma et al . , 2008 ) it is striking that there is such a lack of knowledge about the acquisition and the establishment of such associations . 10 . 7554/eLife . 03007 . 014Figure 8 . Model of processes involved in bacterial invasion . Chitinase as well as other effector proteins are secreted via a bacterial T2SS and induce a local dissolution of the fungal cell wall . This enables bacteria to enter and colonize the fungal cell and induce sporulation . DOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 014 As the fates of bacteria and fungi are ecologically intimately connected in soil one can envision that the endosymbiotic associations could be much more widespread in nature . Indeed , the genetic repertoire for chitinolytic enzymes and a T2SS can be found in other endobacteria , implying an involvement of these systems in other bacterial-fungal interactions . Bacteria from the order Burkholderiales are among the most frequently identified intracellular bacteria in fungal hyphae ( Hoffman and Arnold , 2010; Frey-Klett et al . , 2011 ) . Thus , our findings could present a model system for many other horizontally acquired symbionts and might help deepen the understanding of the common mechanisms involved in the interaction of proteobacteria with eukaryotic cells . Overall , this is the first report of the molecular basis of bacterial invasion of a fungus and the first visualization of the invasion process . We believe that the mechanisms employed are widespread and occur in the growing number of known bacterial-fungal endosymbioses as an alternative pathway to endocytosis ( vesicular uptake ) . Burkholderia rhizoxinica ( isolate B1 , HKI 0454 ) and the Rhizopus microsporus ( ATCC62417 ) harboring endobacteria as well as the symbiont free R . microsporus ( ATCC62417/S ) ( Partida-Martinez et al . , 2007c ) were used in this study . Pure cultures of B . rhizoxinica were grown in MGY medium ( M9 minimal medium supplemented with 1 . 25 g l−1 yeast extract and 10 g l−1 gycerol ) at 30°C or TSB M9 medium ( 10 g l−1 glycerol , 3 g l−1 yeast extract , 15 g l−1 tryptone soy broth , M9 salts ) respectively . Wild type and cured strains of R . microsporus were grown on Potato Dextrose Agar ( PDA ) at 30°C or in TSB respectively . All the strains used in this study are listed in Table 3 . 10 . 7554/eLife . 03007 . 018Table 3 . Bacterial and fungal strainsDOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 018StainsCharacteristicsReferencesBurkholderia rhizoxinica HKI-0454Wild type , isolated from Rhizopus microsporus ATCC62417Rhizopus microsporus ATCC62417Fungal host harboring bacterial endosymbionts , isolated from rice seedlings†B . rhizoxinica ΔgspD::KanrT2SS mutant B . rhizoxinica with deletion of gspDThis studyB . rhizoxinica ΔgspC::KanrT2SS mutant B . rhizoxinica with deletion of gspCThis studyB . rhizoxinica Δchit::KanrB . rhizoxinica with deletion of chitinase geneThis studyB . rhizoxinica Δcbp::KanrB . rhizoxinica with deletion of chitin-binding protein geneThis studyB . rhizoxinica Δchts::KanrB . rhizoxinica with deletion of chitosanase geneThis studyB . rhizoxinica/pHKT4B . rhizoxinica wt harboring a RFP expression vectorThis studyB . rhizoxinica ΔgspD::Kanr/pHKT2B . rhizoxinica gspD mutant harboring a GFP expression vectorThis studyB . rhizoxinica ΔgspC::Kanr/pHKT2B . rhizoxinica gspC mutant harboring a GFP expression vectorThis studyB . rhizoxinica Δchit::Kanr/pHKT2B . rhizoxinica chit mutant harboring a GFP expression vectorThis studyB . rhizoxinica Δcbp::Kanr/pHKT2B . rhizoxinica cbp mutant harboring a GFP expression vectorThis studyB . rhizoxinica Δchts::Kanr/pHKT2B . rhizoxinica chts mutant harboring a GFP expression vectorThis studyR . microsporus + B . rhizoxinica ΔgspD::Kanr/pHKT2Reinfected cured fungal host with gspD mutant harboring a GFP expressing vectorThis studyR . microsporus + B . rhizoxinica ΔgspC::Kanr/pHKT2Reinfected cured fungal host with gspC mutant harboring a GFP expressing vectorThis studyR . microsporus + B . rhizoxinica Δchit::Kanr/pHKT2Reinfected cured fungal host with chit mutant harboring a GFP expressing vectorThis studyR . microsporus + B . rhizoxinica Δcbp::Kanr/pHKT2Reinfected cured fungal host with cbp mutant harboring a GFP expressing vectorThis studyR . microsporus + B . rhizoxinica Δchts::Kanr/pHKT2Reinfected cured fungal host with chts mutant harboring a GFP expressing vectorThis studyEscherichia coli BL21 ( DE3 ) /pET28a-ChiE . coli with expression vector harboring chitinase geneThis studyPartida-Martinez and Hertweck ( 2005 ) Pathogenic fungus harbours endosymbiotic bacteria for toxin production . Nature 437:884–888 . †Ibaragi ( 1973 ) Studies on rice seedling blight . I . Growth injury caused by Rhizopus sp . under high temperature . Ann . Phytopathol . Soc . Jpn 39:141–144 . To address the involvement of both T2SS and chitinolytic proteins two genes of the type 2 secretion gene cluster as well as the gene annotated as chitinase were targeted by double crossover using a suicide-vector harboring a mutated phenylalanyl tRNA synthetase gene pheS as a counter-selectable marker as described previously with slight modifications ( Lackner et al . , 2011a ) . Flanking regions upstream and downstream of the selected gene were amplified using a proof-reading polymerase with primers containing 20 bp homologous to the flanking region and 20 additional bp targeting the 3′ and 5′ end of a kanamycin cassette . The same 20 bp have been used for primers amplifying a kanamycin cassette from pK19 as template . A triple overlapping PCR was performed with equimolar amounts of the flanking PCR fragments and a twofold amount of the kanamycin cassette PCR product using PhusionFlash High-Fidelity PCR Master Mix ( Thermo Fisher Scientific , Waltham , USA ) . The yielded fragments were subsequently transformed into pCR4Blunt-TOPO vector ( Invitrogen , Paisley , UK ) and after restriction digest ligated into pGL42a . Knockout constructs for cbp and chts were obtained following the previous described amplification method ( Lackner et al . , 2011a ) . The vectors pNM89 ( GspC ) , pNM91 ( GspD ) , pZU02 ( Chi ) , pGL47 ( Chts ) and pGL49 ( Cbp ) were introduced into competent cells of B . rhizoxinica by electroporation . Transformants were selected on standard nutrient agar supplemented with 50 µg ml−1 kanamycin . Colonies were inoculated in liquid MCGAVT medium ( Lackner et al . , 2011a ) for 3 to 4 days and subsequently spread on MCGAVT agar plates to obtain single colonies . This procedure was repeated several times and obtained clones were checked for correct integration of the knockout construct into the genome via PCR targeting an internal fragment of the respective gene ( int ) the pheS gene ( pheS ) as well as a region spanning the two recombination sites with ArmA and ArmB amplifying wt fragments and ArmC and Arm D mutant fragments , respectively . The complete list of primers used in this study is found in Table 4 . 10 . 7554/eLife . 03007 . 017Table 4 . Primers used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 03007 . 017NameOligo sequencePrimers used for generating KO mutantsTII_D_fl1_fw5′-GCTACGGATCCCTGCCAGGTATTGCCGTATT-3′TII_D_fl1_rv5′-GCTACAAGCTTCAATCAGCTTGTCGAATTGC-3′TII_D_fl2_rv5′-GCCCAGTAGCTGACATTCATCCCCGATCAATTATGCAAGCAG-3′TII_D_fl2_fw5′-TTCTTGACGAGTTCTTCTGATGGACTGGATGTCTGGATCA-3′TII_C_fl1_fw5′-GCTACGAATCTCAGATCTGTGCGAGGATTG-3′TII_C_fl1_rv5′-GCTACAAGCTTCAACTCGCCTTTACGTACCC-3′TII_C_fl2_rv25′-GCCCAGTAGCTGACATTCATCCGACGGCATGATGAGTTTGTG-3′TII_C_fl2_fw25′-TTCTTGACGAGTTCTTCTGAAGCAAGCTGGTCAGGAACAT-3′Kan_F5′-ATGATTGAACAAGATGGATTGC-3′Kan_R5′-GCCTTCTTGACGAGTTCTTCTGA-3′Chi_Fl1_F5′-GAACTAGTCTCGATCATGGGGGTATTTG-3′Chi_Fl1_R5′-GCCCAGTAGCTGACATTCATCCCAGGTGCTTTTTCATTGCTTC-3′Chi_Fl2_F5′-GCCTTCTTGACGAGTTCTTCTGACGTGACGTATCGTGCAAAGT-3′Chi_Fl2_R5′-ATCCCGGGACGCGGTCAAGTCGATGTAG-3′Kan_Chi_F5′-GAAGCAATGAAAAAGCACCTGGGATGAATGTCAGCTACTGGGC-3′Kan_Chi_R5′-ACTTTGCACGATACGTCACGTCAGAAGAACTCGTCAAGAAGGC-3′P1_chtos5′-GCTACGGGCCCGGCATCGGTGACTATCGTAAC-3′P2_chtos5′-GCTACTTAATTAAGCTAGCGTAGCACAGCCGATACCGTAAGC-3′P3_chtos5′-GCTACGCTAGCTTAATTAAGTAGCGCAATGGAGCAAGCTGATGG -3′P4_chtos5′-GCTACGCGGCCGCAACGTGCGCGACGATACGTTC-3′Kan_chts_F5′-GGATGAATGTCAGCTACTGGGC-3′Kan_chts_R5′-TCAGAAGAACTCGTCAAGAAGGC-3′P1_chtbdp5′-GCTACGCGGCCACGCCGAGATGATGTTG-3′P2_chtbdp5′-GCTACTTAATTAAGCTAGCGTAGCCGATCGTGCGTGAGTAAG-3′P3_chtbdp5′-GCTACGCTAGCTTAATTAAGTAGCAGCCAACCGACGTACCTACC-3′P4_chtbdp5′-GCTACGGGCCCAAGACGGCGGGCGTATTACC-3′Kan_cbp_F5′-GGATGAATGTCAGCTACTGGGC-3′Kan_cbp_R5′-TCAGAAGAACTCGTCAAGAAGGC-3′Primers used for RT-qPCR studiesTIISS_D_RT_F5′-GAGCAGCGATACCAACATCC-3′TIISS_D_RT_R5′-TTGAATGCGGAGACCGAAG-3′TIISS_C_RT_F5′-AGCGTCACTTACTGGGTCATC-3′TIISS_C_RT_R5′-CGAGCCGAACAGAGTTTGAG-3′Chi_RT_F5′-CGCTGGATACGGTCAACATC-3′Chi_RT_R5′-GCCTTGCACGTCATTCTT-3′CBP_RT_F5′-ACGACAGCGCATAATCCTTC-3′CBP_RT_R5′-GGGTGCATCGTAAATCAGGTChtos_RT_F5′-AGGTGGACTGACCCGTATTGChtos_RT_R5′-TTGCACGCTGTATTGGATGT-3′rpoB_RT_F5′-ATTTCCTTCACCAGCACGTT-3′rpoB_RT_R5′-TTCGGGGAAATGGAAGTGT-3′Primers used for control of generated mutantsArm_A_rv ( Kan ) 5′-AGTGACAACGTCGAGCACAG-3′Arm_B_fw ( Kan ) 5′-CGTTGGCTACCCGTGATATT-3′TIISSD_C_A_fw5′-TCACCTCACGTAGCAGATCG-3′TIISSD_C_rv5′-GCATCGACGAAATTCAAGGT-3′TIISSD_D_fw5′-GATAACCGGATCGTCAAGGA-3′TIISSD_D_B_rv5′-CCGGACAAGTCGTACTCGAT-3′TIISSD_Int_fw5′-GTCGAGGGACCAAAGTTTCA-3′TIISSD_Int_rv5′-GGCGTAGACAGGATGTTGGT-3′TIISSC_CA_fw5′-ACTCCAGCCCGCATACATAC-3′TIISSC_C_rv5′-ATTCAGCGCACGTAGATCGT-3′TIISSC_D_fw5′-GCGTCACTTACTGGGTCATC-3′TIISSC_DB_rv5′-AGGAAGTGCTGCGTGTAACC-3′Chi_Ctrl1_KO_F5′-GAACCATTCGCCTTCTTCAC-3′Chi_Ctrl1_WT_R5′-ATCGCTTTCAACAGGTGCTT-3′Chi_Ctrl2_WT_R5′-CCAGTTGTGGCAAATGATTG-3′Chi_Ctrl2_KO_R5′-ATTTCGGCTCTGACGTGACT-3′Chi_Int_fw5′-TGACCTCCATCGCCAAGTCG-3′Chi_Int_rv5′-CGGAACACCTGCGTGAATGC-3′Chtos_ext_for_15′-GAAGCGTGATGTGATTGAAG-3′Chtos_ext_rev_15′-AAGTCGCATCCAGACATTG-3′Chtos_int_for_15′-GACGCCAAGACGATCTACCA-3′Chtos_int_rev_15′-TTGGGCTTTGACCTTGCTAC-3′Cbp_ext_for_15′-ACTTTCTGAATACAGCTTGC-3′Cbp_ext_rev_15′-CAGTCATGATGCAATACGTG-3′Cbp_int_for_15′-GCGGTCTAGTCCCTGCTTAC-3′Cbp_int_rev_15′-GAGGCTATTGGTCGTCACCT-3′pBS_nspI_for_I5′-AGCTCACTCAAAGGCGGTAA-3′pBS_nspI_rev_I5′-TTTTTGTGATGCTCGTCAGG-3′ Low-temperature scanning electron microscopy of uncoated samples was performed as described previously ( Schubert et al . , 2007; Jennessen et al . , 2008 ) . Fungal mycelium from the cured R . microsporus ATCC62417/S was inoculated on NA agar plate and incubated overnight to start hyphal growth . On the edges of hyphal spreading 2 µl of B . rhizoxinica culture was spotted and the plate was again incubated for approximately 2 hr . From the interaction zone several parts were selected and excised with a surgical blade as small agar blocks , and transferred to a copper cup for snap-freezing in nitrogen slush . Agar blocks were glued to the copper ( KP-Cryoblock , Klinipath , Duiven , Netherlands ) . Samples were examined in a JEOL 5600LV scanning electron microscope ( JEOL , Tokyo , Japan ) equipped with an Oxford CT1500 Cryostation for cryo-electron microscopy ( cryoSEM ) . Electron micrographs were acquired from uncoated frozen samples , or after sputter-coating by means of a gold/palladium target for three times during 30 s . Micrographs of uncoated samples were taken at an acceleration voltage of 3 kV , and consisted out of 30 averaged fast scans , and at 5 kV in case of the coated sample . RNA was isolated using the RiboPure Bacteria or RiboPure Yeast Kit ( Ambion , Texas ) following the manufacturers' instruction either from pure bacterial culture or from cocultivation of B . rhizoxinica with sterile fungus . 40 ng of total RNA served as template for one-step RT qPCR using gene specific primers and Quanta sybr green Kit ( Quanta BioSciences , Gaithersburg , MD ) . Realtime PCR was performed on an Eppendorf realplex mastercyler ( Eppendorf , Hamburg , Germany ) in triplicate for each sample , and a control reaction without enzyme was included for each sample . The rpoB gene was used as an internal standard for calculation of expression levels and normalization . For cycling parameters we followed the manufacturers' protocol . Controls without template were included for each primer pair . Cycle threshold ( Ct ) values were calculated by the realplex software and used for quantification of expression levels via the 2−ΔΔCt method ( Livak and Schmittgen , 2001 ) . Homology search was carried out using the NCBI BlastP and the Burkholderia genome database . Domain prediction using Robetta ( Kim et al . , 2004 ) and Coils ( Lupas et al . , 1991 ) , visualization and modification was done in SwissPdbViewer and VMD . PROMALS3D was used for structure based alignment of chitinases ( Pei and Grishin , 2007 ) . The structural model of B . rhizoxinica chitinase was generated by threading the amino acid sequence to PDB database and building the model using Pymol . For phylogenetic analysis , protein sequences were aligned by the ClustalW algorithm implemented in the MEGA 3 . 1 software package ( Kumar et al . , 2004 ) . The obtained alignment blocks were used for tree-construction by the neighbor-joining method . 10 , 000 bootstrap replicates were run to estimate reliability of the inferred groups . Bacterial cells were grown in MGY medium without antibiotics . Fungus was cultivated in 8-well plates in TSB-medium . For reinfection/sporulation assay 6-well plates were used . Each well was filled with 5 ml TSB-medium and inoculated with a pellet of R . microsporus mycelium from a 48 plate well . 200 µl of bacterial culture was added and incubated at 30°C . After 4–7 days , sporulation of plates was examined by eye . To a volume of 250 ml supernatant of B . rhizoxinica wt 15 mg of acid swollen chitin were added and the mixture was stirred for 60 min at room temperature to allow for chitin binding . Subsequently it was centrifuged at 6000×g for 10 min to pellet chitin with bound proteins , the supernatant was stored ( S1 ) . The pellet was then washed twice with 0 . 9% NaCl and resuspended in 0 . 05 M NaCl to remove bound proteins . Chitin was once again centrifuged to give fraction C ( pelleted chitin ) and 0 . 1 M Tris–HCl pH 7 was added to the supernatant ( S2 ) . All three fractions were loaded on a SDS polyacrylamide gel and bands were identified MALDI-TOF . 250 ml of bacterial cultures and bacterial/fungal cocultures respectively were centrifuged at 8000×g at 4°C for 20 min and the obtained supernatant was sterile vacuum filtered ( 0 . 2 µM pore size ) and supplemented with 10 g l−1 of TCA for wt cultures and 50 g l−1 for mutants . The mixture was store at 4°C overnight to allow for protein precipitation . Proteins were pelleted by centrifugation at 12 , 000×g at 4°C for 30 min , the supernatant was removed and the pellet was rinsed twice in ice-cold acetone . The pellet was air-dried for 15 min at room temperature and subsequently resuspended in 300 μl 2D-lysis buffer ( 7 M urea , 2 M thiourea , 2% [wt/vol] CHAPS ( 3-[ ( 3-cholamidopropyl ) -dimethylammonio]-1-propanesulfonate ) , 1% [wt/vol] Zwittergent 3-10 , 30 mM Tris ) . To improve protein solubility the samples were sonicated for 5 min in an ultrasonic bat . After centrifugation at 14 , 000×g for 20 min at 4°C , the supernatant was collected . The protein concentration was determined according to the Bradford method . For the separation of proteins in the first dimension 11 cm IPG strips with a nonlinear pH range from both pH 3 to 11 ( GE Healthcare Bio-Sciences , Uppsala , Sweden ) which had been rehydrated overnight ( 7 M urea , 2 M thiourea , 2% [wt/vol] CHAPS , 1% [wt/vol] Zwittergent 3_10 , 0 . 002% [wt/vol] bromophenol blue , 0 . 5% [vol/vol] IPG buffer , 1 . 2% [vol/vol] De-Streak reagent [GE Healthcare Bio-Sciences , Uppsala , Sweden] ) were used as described ( Kniemeyer et al . , 2006 ) . Equal amounts of protein samples from B . rhizoxinica wt and mutant's pure culture as well as the respective cocultures were applied via anodic cup loading to IPG strips . Isoelectric focusing was conducted according to the following protocol: 4 hr at 300 V ( gradient ) , 3 hr at 600 V ( gradient ) , 4 hr at 1000 V ( gradient ) , 5 hr at 8000 V ( gradient ) and 48 , 000 V hr at 8000 V ( step ) . Subsequently strips were equilibrated for 10 min in 10 ml of equilibration buffer ( 6 M urea , 30% [vol/vol] glycerol , 2% [wt/vol] SDS ( sodium dodecyl sulfate ) , 75 mM Tris , 0 . 002% [wt/vol] bromophenol blue ) containing 1% ( wt/vol ) DTT and subsequently for 10 min in 10 ml of equilibration buffer containing 2 . 5% ( wt/vol ) iodoacetamide . Ettan DALT System ( GE Healthcare Bio-Sciences , Uppsala , Sweden ) was used to separate proteins in the second dimension . SDS polyacrylamide gels ( Mini-Protean TGX Precast Gels , AnyKD , BIORAD Hercules , CA ) were loaded with the strips and run for 50 min at 200 V . In order to identify the proteins by mass spectrometry ( MS ) , the gels were stained with colloidal Coomassie Brilliant Blue according to Kniemeyer et al . ( 2006 ) followed by manual excision of the spots . Protein spots were tryptically digested ( Shevchenko et al . , 1996 ) . Extracted peptides were measured and identified on an Ultraflex I and Ultraflextreme MALDI-TOF/TOF device using flexControl 3 . 3 for data collection and flexAnalysis 3 . 3 spectra analysis/peak list generation ( Bruker Daltonics , Bremen , Germany ) . Peptide mass fingerprint ( PMF ) and peptide fragmentation fingerprint ( PFF ) spectra were submitted to the MASCOT server ( MASCOT 2 . 3 , Matrix Science , London , UK ) , searching the B . rhizoxinica database . 25 ml of E . coli BL21 strains carrying the vector pET28a or pET28a/chitinase were cultured in TB ( 12 g l−1 tryptone , 24 g l−1 yeast extract , 4 ml l−1 glycerol ) buffered with 100 mM 2- ( N-morpholino ) ethanesulfonic acid ( MES ) at pH6 . The cells were induced at an OD600 of 0 . 6 with 1 mM IPTG , and grown overnight to OD600 of around 14 . For SDS polyacrylamide gel 500 µl of cultures were centrifuged for 5 min . Supernatant and pellet are separated . Cell pellets were dissolved in 500 µl TB-MES . 30 ml of the samples ( supernatant and pellet , respectively ) were suspended in 30 µl protein sample buffer . Samples were boiled for 15 min , and 20 μl run on 12% ( wt/vol ) SDS polyacrylamide gel . Gels were stained with Coomassie blue . Chitinase activity was detected using the aqueous solution of CM-Chitin-RBV as a substrate ( Loewe Biochemica GmbH , Sauerlach , Germany ) as described by ( Saborowski et al . , 1993 ) . The secretome of BL21 cells bearing pET28a and pET28a/chitinase were used to conduct the assay . Cells were harvested and supernatants were filter-sterilized , 250 µl of the secretome was mixed with 250 µl CM-Chitin substrate and buffered with 250 µl 0 . 1 M sodium acetate at pH6 . Triplicates of samples were incubated at 37°C for 15–90 min . An equal amount of 0 . 1 U Streptomyces griseus recombinant chitinase solution ( Sigma-Aldrich , St . Louis , MO ) was included as a positive control for chitinase activity . Each reaction was stopped by adding 250 µl of 0 . 1 N HCl and kept on ice for at least 5 min to ensure complete precipitation of the non-degraded substrate a low pH ( <3 ) . After centrifugation at 15 . 000×g for 10 min the absorbance of the supernatants was measured photometrically at 550 nm . Blanks without substrate or enzyme where run in parallel . To visualize the localization of the B . rhizoxinica mutant strains with respect to the fungal hyphae , the RFP encoding pHKT4 plasmid was transformed into the B . rhizoxinica wt and GFP encoding pHKT2 plasmid were transformed into B . rhizoxinica Δchit::Kanr , ΔgspC::Kanr and ΔgspD::Kanr mutant strains . Subsequently all resulting strains were co-cultured with sterile R . microsporus . After 3 to 4 days , a small piece of growing fungus from the co-culture as well as from the sterile fungus was examined . The endocytotic capacity of R . microsporus has been visualized by styryl dye FM4-64 . A piece of freshly growing R . microsporus mycelium has been co-incubated with growing B . rhizoxinica culture in 500 µl physiological saline . After 50 min 3 µM of FM4-64 and 5 µM of Syto9 has been added and incubated for 10 min . The live images have been taken in mounted slide to avoid drying . All images have been taken by using a Zeiss CLSM 710 confocal laser-scanning microscope ( Göttingen , Germany ) for fluorescence detection .	Many organisms live in what are known as symbiotic relationships , whereby two or more different species share a close and often long lasting association . Examples of symbiosis abound in nature , with well-studied examples including the sea anemone and the clownfish , corals and their photosynthetic algae , and nitrogen-fixing bacteria that live in association with legumes . In some cases , each member of the relationship benefits from the symbiosis . One such example is the rice seedling blight fungus Rhizopus microsporus and its bacterial symbiont Burkholderia rhizoxinica , which work together to produce toxins that kill rice plants . This frees up nutrients that nourish both the fungus and the bacteria . B . rhizoxinica lives inside the tissues of the fungus , but to do so the bacterial cells must first travel through the tough cell walls of the fungus . How these bacteria do this without also damaging the fungus was unknown . Moebius et al . discovered that B . rhizoxinica gains access into R . microsporus cells by using a series of proteins in the bacterium’s membrane called the type 2 secretion system , which transport proteins from the inside of the cell to the outside . On analyzing the proteins released by this system , Moebius et al . identified several enzymes that help the bacteria attach to the fungal cell wall and soften it so that the bacteria can penetrate into the cell . Only small amounts of enzyme are needed for the softening process , meaning that penetrating the cell wall is a relatively gentle process that causes no lasting damage to the fungus . Moebius et al . also captured images of the bacteria invading the fungal cells using a technique called cryo-electron microscopy , and provide the first known images of this type of infection in progress .	[ "Abstract", "Introduction", "Results", "Discussion", "Material", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2014	Active invasion of bacteria into living fungal cells
The Chromosome Region of Maintenance 1 ( CRM1 ) protein mediates nuclear export of hundreds of proteins through recognition of their nuclear export signals ( NESs ) , which are highly variable in sequence and structure . The plasticity of the CRM1-NES interaction is not well understood , as there are many NES sequences that seem incompatible with structures of the NES-bound CRM1 groove . Crystal structures of CRM1 bound to two different NESs with unusual sequences showed the NES peptides binding the CRM1 groove in the opposite orientation ( minus ) to that of previously studied NESs ( plus ) . Comparison of minus and plus NESs identified structural and sequence determinants for NES orientation . The binding of NESs to CRM1 in both orientations results in a large expansion in NES consensus patterns and therefore a corresponding expansion of potential NESs in the proteome . The exportin CRM1 ( Chromosome Region Maintenance 1 protein; also known as exportin 1 or XPO1 ) is the most prominent nuclear export receptor in the cell . CRM1 maintains the cellular localization of hundreds of diverse-functioning protein cargos , including many tumor suppressor , cell cycle proteins , and viral proteins ( Fornerod et al . , 1997; Fukuda et al . , 1997; Ossareh-Nazari et al . , 1997 ) . CRM1 is also a promising cancer drug target , and a small molecule inhibitor of CRM1 named Selinexor is currently in more than 40 clinical trials for a variety of cancers ( clinicaltrials . gov ) ( Lapalombella et al . , 2012; Etchin et al . , 2013; Sun et al . , 2013; Fung and Chook , 2014; Xu et al . , 2015 ) . CRM1 recognizes its protein cargos through 8–15 residue long nuclear export signals ( NESs ) in the proteins ( la Cour et al . , 2004; Kosugi et al . , 2008; Xu et al . , 2010 ) . NES sequences are highly diverse , and the peptides bind CRM1 with a large affinity range , with dissociation constants ( KDs ) ranging from low nanomolar to tens of micromolar ( Kutay and Guttinger , 2005 ) . Sequence , peptide-library , and bioinformatic analyses have found that NESs are best described by a set of six consensus sequences , which differ in the spacings between four key hydrophobic residues Φ1 , Φ2 , Φ3 , and Φ4 ( Figure 1A ) ( la Cour et al . , 2004; Kosugi et al . , 2008; Xu , Farmer et al . , 2012a ) . While sequence patterns are available to describe many NESs , there is limited structural information on diverse NESs and how they bind CRM1 . 10 . 7554/eLife . 10034 . 003Figure 1 . hRio2NES and CPEB4NES bind CRM1 in orientation opposite to the PKINES . ( A ) Six nuclear export signal ( NES ) consensus patterns ( Φ is Leu , Val , Ile , Phe or Met; X is any amino acid ) . ( B ) Structure of PKINES ( yellow cartoon ) bound to Chromosome Region of Maintenance ( CRM1 ) ( gray surface ) ( 3NBY ) on the right and PKINES was removed to show hydrophobic pockets P0–P4 in the CRM1 groove on the left panel . ( C ) Overall structure of the CRM1* ( gray ) -RanGppNHp ( orange ) -RanBP1 ( light purple ) -hRio2NES ( blue ) complex . ( D ) Structures of hRio2NES ( blue ) and CPEB4NES ( purple ) bound to the CRM1 groove ( gray surfaces ) . All NES peptides are in cartoon and their hydrophobic Φ residues shown as sticks . Their Φ residues and the corresponding P0–P4 CRM1 pockets that they bind are shown below . ( E ) Kick OMIT map meshes contoured at the 3 . 0σ level overlaid on the final , refined coordinates for hRio2NES and CPEB4NES . Kicked OMIT maps were generated by PHENIX by omitting the NES peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 00310 . 7554/eLife . 10034 . 004Figure 1—figure supplement 1 . Electron densities of the wild-type NES peptides . Stereo views of kicked OMIT map meshes contoured at the 3 . 0σ level , on the final , refined coordinates for ( A ) hRio2NES and ( B ) CPEB4NES as shown in sticks . Anomalous difference map for Se-met hRio2NESis calculated in PHENIX and contoured at the 3 . 0σ level in the leftmost panel in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 004 Structures are available for only three different NESs , from the cargos protein kinase A inhibitor ( PKI ) , Snurportin-1 ( SNUPN ) , and the HIV1-Rev protein , bound to CRM1 . The NESs bind in a hydrophobic groove , which is located on the outer/convex surface of the ring-shaped CRM1 ( Dong et al . , 2009; Monecke et al . , 2009; Güttler et al . , 2010 ) . The NESs use almost exclusively their side chains , especially their hydrophobic Φ side chains , to bind CRM1 . The NES-binding groove of CRM1 , which is wide at one end and narrow at the other end , consists of 5 hydrophobic pockets P0–P4 and is virtually identical in all CRM1-NES structures ( Figure 1B ) . NESs from PKI and SNUPN ( PKINES and SNUPNNES ) share a similar structure when bound to CRM1—an N-terminal 3-turn α-helix followed by a short C-terminal β-strand-like extension ( Figure 1B ) ( Dong et al . , 2009; Monecke et al . , 2009; Güttler et al . , 2010; Koyama et al . , 2014 ) . The NES helix binds the wide part of the CRM1 groove , while the β-strand binds the narrow end of the groove . The RevNES peptide binds the CRM1 groove in a different manner by adopting an entirely extended conformation ( Güttler et al . , 2010 ) . All three NES peptides bind in the same direction , with their N-termini at the wide part of the groove . The vastly different conformations of the extended RevNES compared to the helix-strand PKINES and SNUPNNES suggest that CRM1 may recognize divergent signal sequences in part by binding different peptide structures . The repertoire of conformations for CRM1-bound NESs remains unclear , but the asymmetric and seemingly structurally invariant NES-bound CRM1 groove presents physical constraints on structures of bound NESs . For example , the class 3 NES consensus of Φ1X ( 2 , 3 ) Φ2X ( 2 , 3 ) Φ3X2Φ4 with two intervening residues between Φ3 and Φ4 suggests a single long NES helix . The substitution of a narrow strand or extended chain at the C-terminus of an NES with a helix presents a steric problem as the thicker helix is unlikely to fit into the tapering CRM1 groove . In current NES databases , class 3 NES sequences are as prevalent as NESs of classes 1b , 1c , 1d , and 2 , but information of how they are able to bind CRM1 is missing ( Xu et al . , 2015 ) . We have developed a general strategy to crystallize CRM1 bound to NES peptides in order to study how diverse sequences , including the enigmatic class 3 NESs , bind the exportin . Crystal structures of two different class 3 NESs bound to CRM1 revealed a novel NES binding mode where polypeptide direction of the NES is reversed . We show that NES peptides can bind the CRM1 groove bidirectionally ( in both plus and minus directions ) , and biochemical and structural analyses identified determinants for one direction of binding vs the other . Bidirectional exportin-signal interactions suggest a significant expansion of the current NES consensus patterns that will enable new , previously unknown NESs to be identified . Crystallization of CRM1-Ran-NES peptide complexes has generally not been successful , possibly due to conformational flexibility and low affinities for the NESs ( Kutay and Guttinger , 2005 ) . Crystal structures of NESs bound to CRM1 were instead determined using the CRM1-Ran-SNUPN complexes , including ones where the SNUPNNES was replaced with the PKINES and RevNES ( Güttler et al . , 2010 ) . This strategy was limited because of severe mosaicity of the CRM1-Ran-SNUPN crystals ( Güttler et al . , 2010 ) . On the other hand , the ternary complex of Saccharomyces cerevisiae CRM1 ( ScCRM1 ) with RanGTP ( human or yeast RanGTP , Gsp1p ) and RanBP1 ( Yrb1p ) reliably yields crystals that diffract to high resolution and has been used to determine structures of several CRM1-inhibitor complexes ( Koyama and Matsuura , 2010; Lapalombella et al . , 2012; Etchin et al . , 2013; Sun et al . , 2013; Haines et al . , 2015 ) . We therefore used the CRM1-Ran-RanBP1 complex to determine structures of the exportin bound to the enigmatic class 3 NES peptides . RanBP1 binding normally stimulates NES release by closing the NES-binding groove of ScCRM1 ( Koyama and Matsuura , 2010 ) . We engineered CRM1 to shift the open-closed groove equilibrium toward the open state , in order for CRM1-Ran-RanBP1 to bind NESs . We started with a ScCRM1 construct ( residues 1–1058 , Δ377–413 , 537DLTVK541 to GLCEQ ) that is known to crystallize easily and has an NES groove that is virtually identical to that of human CRM1 ( Lapalombella et al . , 2012; Etchin et al . , 2013; Sun et al . , 2013; Haines et al . , 2015 ) . Koyama and Matsuura showed that mutation of the H9 loop of CRM1 , which packs against the back of a closed NES groove , stabilizes the open CRM1 groove ( Koyama and Matsuura , 2010 ) . Thus , we mutated the H9 loop ( Val441Asp ) to detach it from the back of the NES groove and to open the groove even when CRM1 is complexed with Ran and RanBP1 . The resulting CRM1* construct , with ScCRM1 residues 1–1058 , Δ377–413 , V441D and groove residues 537DLTVK541 mutated to GLCEQ ( to mimic human CRM1 , see methods ) , binds NES peptides in the presence of RanGTP and RanBP1 . We generated quaternary CRM1-RanGppNHp-RanBP1-NES complexes with two class 3 NES peptides , the hRio2NES ( 389RSFEMTEFNQALEEI403 ) and the CPEB4NES ( 379RTFDMHSLESSLIDI393; predicted Φ1–Φ4 positions are underlined ) and determined their structures to 2 . 3 Å and 2 . 1 Å resolution ( Figure 1C , D; crystallographic statistics in Table 1 ) . The crystals are isomorphous to previously crystallized inhibitor-bound CRM1-Ran-RanBP1 complexes , with one CRM1-RanGppNHp-RanBP1-NES complex in the asymmetric unit ( Sun et al . , 2013 ) . Residues modeled in the three proteins are CRM1* residues 1–440 and 460–1053 , Ran residues 9–216 , and RanBP1 residues 63–69 and 78–200 . hRio2NES residues 391–403 and CPEB4NES residues 379–393 were modeled in the respective structures . 10 . 7554/eLife . 10034 . 005Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 005ScXPO1-RanGppNHp-Yrb1p bound to NES of:Selenomethione-hRio2CPEB4Selenomethione-hRio2-RCPEB4-RPKI-Flip3Data collection Space groupP43212 Cell dimensions a , b , c ( Å ) 106 . 48 , 106 . 48 , 303 . 73105 . 96 , 105 . 96 , 304 . 00106 . 69 , 106 . 69 , 304 . 50106 . 48 , 106 . 48 , 303 . 73105 . 96 , 105 . 96 , 304 . 00 a , b , g ( ° ) 90 , 90 , 9090 , 90 , 9090 , 90 , 9090 , 90 , 9090 , 90 , 90 Resolution ( Å ) 50 . 00–2 . 28 ( 2 . 32–2 . 28 ) 50 . 00–2 . 10 ( 2 . 14–2 . 10 ) 50 . 00–2 . 28 ( 2 . 32–2 . 28 ) 50 . 00–2 . 94 ( 3 . 00–2 . 94 ) 50 . 00–2 . 55 ( 2 . 59–2 . 55 ) Rpim2 . 9 ( 37 . 7 ) 3 . 5 ( 43 . 4 ) 3 . 5 ( 38 . 6 ) 4 . 9 ( 40 . 6 ) 4 . 1 ( 46 . 5 ) I/sI24 . 3 ( 2 . 17 ) 19 . 5 ( 1 . 70 ) 22 . 5 ( 2 . 72 ) 13 . 3 ( 1 . 87 ) 19 . 0 ( 1 . 92 ) Completeness ( % ) 98 . 6 ( 99 . 8 ) 99 . 5 ( 100 ) 98 . 0 ( 99 . 2 ) 94 . 6 ( 96 . 0 ) 99 . 6 ( 100 ) Redundancy7 . 0 ( 5 . 9 ) 6 . 0 ( 6 . 1 ) 7 . 0 ( 7 . 0 ) 6 . 2 ( 5 . 7 ) 5 . 5 ( 5 . 5 ) Refinement Resolution ( Å ) 45 . 7–2 . 28 ( 2 . 32–2 . 28 ) 40 . 2–2 . 09 ( 2 . 12–2 . 09 ) 37 . 7–2 . 28 ( 2 . 31–2 . 28 ) 47 . 5–2 . 94 ( 3 . 02–2 . 94 ) 47 . 5–2 . 54 ( 2 . 60–2 . 54 ) No . reflections77 , 245 ( 2833 ) 98 , 659 ( 1793 ) 79 , 492 ( 3267 ) 34 , 265 ( 2013 ) 56862 ( 3361 ) Rwork/Rfree17 . 8 ( 25 . 8 ) /21 . 9 ( 27 . 3 ) 17 . 0 ( 23 . 8 ) /20 . 8 ( 27 . 0 ) 16 . 8 ( 24 . 7 ) /21 . 2 ( 27 . 6 ) 18 . 1 ( 25 . 2 ) /24 . 0 ( 31 . 3 ) 18 . 6 ( 25 . 0 ) /22 . 6 ( 30 . 6 ) No . atoms Protein10 , 85911 , 11410 , 82310 , 70810797 Ligand/ion6076595151 Water2716603588253 NES Peptide/Φ111/46122/43130/46112/43105/43 B-factors Protein42 . 039 . 343 . 953 . 846 . 5 Ligand/ion44 . 351 . 746 . 941 . 841 . 6 Water33 . 434 . 835 . 423 . 335 . 3 NES peptide/Φ80 . 5/77 . 377 . 6/70 . 467 . 5/61 . 781 . 2/80 . 598 . 6/96 . 0 R . m . s deviations Bond lengths ( Å ) 0 . 0030 . 0030 . 0060 . 0030 . 004 Bond angles ( ° ) 0 . 6170 . 6890 . 8350 . 5780 . 673PDB code5DHF5DIF5DI95DHA5DH9Highest resolution shell is shown in parenthesis . One crystal was used for each structure . Overall structures of the CRM1-Ran-RanBP1-NES complexes are highly similar to previously determined CRM1-Ran-RanBP1 structures ( all residue Cα rmsds 0 . 2–0 . 5 Å when compared to unliganded CRM1-Ran-RanBP1 ( PDB code: 3M1I , 4HB2 ) ( Koyama and Matsuura , 2010; Sun et al . , 2013 ) and to inhibitor-bound CRM1-Ran-RanBP1 complexes ( PDB code: 4HAT , 4HAU , 4HAV , 4GMX , 4GPT ) ( Lapalombella et al . , 2012; Etchin et al . , 2013; Sun et al . , 2013; Haines et al . , 2015 ) . Structures of the NES peptides were verified by kick-OMIT maps ( Praznikar et al . , 2009 ) generated without the peptide ( Figure 1E , stereo views in Figure 1—figure supplement 1 ) . Selenomethionine hRio2NES peptide was also generated and anomalous data were collected to confirm correct placement of its methionine , and unambiguously confirm the direction of the NES polypeptide chain ( Figure 1—figure supplement 1 ) . The CRM1-bound hRio2NES and CPEB4NES structures unexpectedly revealed that both NESs bound the CRM1 groove in opposite orientation ( termed the minus direction ) compared to previous NES structures ( PKINES , SNUPNNES , and RevNES bind in the plus direction ) ( Figure 1B , D ) . The NES groove of CRM1 is nearly invariant when bound to plus or minus NESs ( Cα rmsds 0 . 3–0 . 5 Å; all atom rmsds 1 . 0–1 . 3 Å , for CRM1 residues 521–605 in all available CRM1-NES structures [Dong et al . , 2009; Monecke et al . , 2009; Güttler et al . , 2010] ) . Although the polypeptide directions are reversed , local structures of the hRio2NES and CPEB4NES are similar to those of the PKINES and SNUPNNES . All four NES peptides are combinations of 3-turn α-helices and 2-residue β-strand-like extensions . Helices of the minus NESs are now at the C-termini of the peptides and their strands at the N-termini ( Figure 1D ) . Both plus and minus NES helices bind the same part of the CRM1 groove , with hydrophobic residues from one helix face occupying hydrophobic pockets P0–P3 of CRM1 ( Figure 1B , D ) . The structures show that the two minus NESs clearly match the consensus pattern Φ1XΦ2XXXΦ3XXΦ4XXΦ5 , which is the reverse of the class 1a pattern , Φ0XXΦ1XXΦ2XXXΦ3XΦ4 . The five hydrophobic residues of the hRio2NES and CPEB4NES , designated Φ1–Φ5 , bind the same P0–P4 pockets as the plus NESs , but in reverse order , with Φ1 in P4 and Φ5 in P0 ( Figure 1B , D ) . The narrow part of the CRM1 groove is still occupied by an extended strand motif , which is formed by Φ1 and Φ2 of the minus NESs as they occupy the CRM1 P4 and P3 pockets , respectively . The hRio2NES and CPEB4NES are in fact not class 3 NESs in the traditional sense , as the four previously designated Φ residues in hRio2NES ( 389RSFEMTEFNQALEEI403 ) and the CPEB4NES ( 379RTFDMHSLESSLIDI393; predicted Φ1–Φ4 positions are underlined ) do not occupy the P1–P4 hydrophobic pockets as predicted . The four hydrophobic residues that match the class 3 NES consensus , in fact form only a portion of an inverted class 1a pattern Φ2XXXΦ3XXΦ4XXΦ5 , with M393 of hRio2NES and M383 of CPEB4NES in the Φ2 positions . F391 of hRio2NES and F381 of CPEB4NES , which we had previously missed as consensus residues , are the Φ1 positions of the N-terminal ΦXΦ motif of an inverted class 1a pattern . Comparison of plus and minus NESs showed translational offsets of the helices along their axes ( Figure 2 ) . Cαs of minus NESs are shifted 1 . 3–3 . 5 Å from equivalent Cαs in the plus NESs , with the largest shifts observed for residues that occupy the P0 and P1 CRM1 pockets . In an α-helix , amino acid side chains are angled toward the N-terminus of the helix . Thus , since plus and minus helices progress from N- to C-terminus in opposite directions , side chains that emanate from the helices also project in opposite directions . The Φ0 , Φ1 , Φ2 , and Φ3 side chains of the plus NES helix project toward the wide end of the CRM1 groove near P0 to occupy the P0–P4 CRM1 pockets ( Figure 2 ) . In contrast , the equivalent Φ5 , Φ4 , Φ3 , and Φ2 residues of the minus NES helix project toward the narrow end of the CRM1 groove , thus necessitating a shift of the entire helix in the opposite direction to allow the hydrophobic side chains to reach the P0 , P1 , P2 , and P3 CRM1 pockets . 10 . 7554/eLife . 10034 . 006Figure 2 . Comparison of plus and minus NESs . Pairwise comparison of hRio2NES ( blue ) or CPEB4NES ( purple ) with PKINES ( yellow; 3NBY ) upon superposition of NES-bound CRM1 grooves . Hydrophobic NES residues ( Φs ) are shown as sticks and orientation of the CRM1 grooves is indicated by positions of the P0–P4 pockets . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 006 Because the entire minus NES helix shifts relative to a plus helix , it is not surprising that hydrophobic side chain preferences of the minus and plus helices are similar . We generated single amino acid mutants by replacing each of positions Φ3 , Φ4 , and Φ5 in hRio2NES with other hydrophobic residues , and tested binding of the mutants to CRM1 . Results of in vitro pull-down assays using immobilized GST-hRio2NES mutants , purified human CRM1 , and yeast RanGTP show that medium-sized hydrophobic side chains such as isoleucine and leucine are preferred at Φ4 and Φ5 for CRM1 interaction . Medium and larger hydrophobic side chains such as isoleucine , leucine , and methionine are preferred at Φ3 for binding CRM1 ( Figure 3A ) . These results are similar to ones previously shown in the mutagenesis study of the PKINES ( Güttler et al . , 2010 ) . 10 . 7554/eLife . 10034 . 007Figure 3 . Hydrophobic side chain preferences for hRio2NES binding to CRM1 . In vitro pull-down assay ( Coomassie-stained SDS/PAGE ) of purified human CRM1 binding to immobilized GST-hRio2NES mutants ( A ) Φ3 , Φ4 , or Φ5 or ( B ) Φ1 or Φ2 position mutated in the presence of excess ScRanGTP . Relative band intensities of triplicate experiments are plotted in histograms . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 007 The shift of the minus NES helix relative to the plus NES helix results in a corresponding translation of the preceding strand/loop segment that places the Φ1 and Φ2 side chains farther from the P3 and P4 pockets . In both the hRio2NES and CPEB4NES , large hydrophobic residues in the Φ1 ( phenylalanines ) and the Φ2 ( methionines ) positions within the extended segments enable a longer reach into the comparatively distal P3 and P4 CRM1 pockets . Mutagenesis of the Φ1 and Φ2 positions of hRio2NES and pull-down assays with CRM1 and Ran show that large hydrophobic side chains such as leucine , methionine , phenylalanine , and tryptophan are preferred in both positions ( Figure 3B ) . Smaller hydrophobic side chains like alanine , valine , and isoleucine in these positions are disfavored as the mutants do not bind CRM1 efficiently ( Figure 3B ) . The preference for large hydrophobic side chains is consistent with the need for side chains in the extended portions of the minus NESs to reach farther into their CRM1-binding sites . In contrast , the large aromatic residues phenylalanine and tryptophan are disfavored in equivalent Φ3 and Φ4 positions in the plus direction PKINES ( Güttler et al . , 2010 ) . Structural and mutagenesis data to compare plus and minus NESs suggest that placement of the strand-like ΦXΦ motif at the N-terminus of an NES generates a signal peptide that binds CRM1 in the minus direction , whereas a C-terminal ΦXΦ results in a plus direction NES . To first investigate whether features of the sequence such as spacings between hydrophobic residues and placement of the ΦXΦ motif are critical in determining directionality of NES binding , we reversed the sequence of the hRio2NES ( FEMTEFNQALEEI ) to generate hRio2NES-R ( IEELAQNFETMEF ) . We also reversed CPEB4NES ( FDMHSLESSLIDI ) to generate CPEB4NES-R ( IDILSSELSHMDF ) . Both reversed peptides match the class 1a NES pattern and were predicted to bind CRM1 like the PKINES , which is another class 1a NES . Binding affinities of NES peptides to CRM1 were measured in competition differential bleaching experiments using FITC-PKINES as a fluorescent probe , MBP-NESs as competitors and monitored with a microscale thermophoresis instrument ( Figure 4 , Figure 4—figure supplement 1 ) . The competition differential bleaching approach is explained in methods and representative titration data are shown in Figure 4—figure supplement 2 . Wild-type NESs MBP-hRio2NES and MBP-CPEB4NES bind CRM1 with KDs of 2200 nM [1600 , 2900] and 590 nM [400 , 840] , respectively ( Figure 4B ) . The ranges in brackets represent the 68 . 3% confidence intervals as calculated using F-statistics and error-surface projection method ( Bevington and Robinson , 1992 ) . When NES sequences are reversed , MBP-hRio2NES-R and MBP-CPEB4NES-R still bind CRM1 with similar affinities , KDs of 2400 nM [2100 , 2800] and 780 nM [610 , 980] , respectively ( Figure 4A , B ) . All of the NES peptides were also cloned into EYFP-NLS-NES fusions and tested for nuclear export activity in HeLa cells . They were all found to direct nuclear export in a Leptomycin B sensitive manner , suggesting that the reverse peptides function as active NESs in cells ( Figure 4C ) . 10 . 7554/eLife . 10034 . 008Figure 4 . hRio2NES , CPEB4NES , and their reverse counterparts bind CRM1 with similar binding affinities . ( A ) Sequences of NESs used . ( B ) Binding of FITC-PKINES and various MBP-NESs to CRM1 measured by differential bleaching , monitored by a microscale thermophoresis instrument . MBP-NESs compete with FITC-PKINES for CRM1 in competition titrations . Fitted binding curves are overlaid onto data points with error bars representing the mean and standard deviation of triplicate titrations . Dissociation constants ( KDs ) of the NESs are reported below the graphs with ranges in brackets representing the 68 . 3% confidence intervals . Binding of MBP-NPMmutANES ( a moderate CRM1 binder ) and MBP-SNUPNNES ( a weak binder ) is shown on the rightmost panel for reference . Experiments performed on separate days were fitted with a new triplicate set of direct bind titrations . ( C ) Leptomycin B ( LMB ) sensitive nuclear export activity of EYFP-NLS-NES fusions in HeLa cells . YFP ( pseudocolored in yellow ) , Hoechst ( pseudocolored in blue ) , and merged images were captured using spinning disk confocal microscope ( 60× ) . Images are maximum intensity projection of five confocal Z stacks spaced 0 . 3 µm apart . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 00810 . 7554/eLife . 10034 . 009Figure 4—figure supplement 1 . Binding affinities of shorter hRio2NES and CPEB4NES constructs . ( A ) Sequences of shorter hRio2NES and CPEB4NES constructs . ( B ) Binding of these NESs to CRM1 measured by differential bleaching . Comparison of hRio2NES constructs in Figure 4 and in this supplement shows that the addition of a lysine and glycine residue C-terminal of the NES helix weakens binding . In contrast , comparison of CPEB4NES constructs shows that addition of methionine and arginine C-terminal of the CPEB4NES helix improved binding . One possible explanation is addition of these residues may have affected helical propensity of the NESs , and hence affect binding affinity . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 00910 . 7554/eLife . 10034 . 010Figure 4—figure supplement 2 . Differential bleaching of fluorescence probe in a sigmoidal and binding-dependent manner . Fluorogram of normalized fluorescence signal collected from the direct titration of CRM1 into fluorescent-labeled PKINES peptide in presence of excess ScRanGTP is plotted on the top panel . Traces are colored in rainbow , indicating increasing concentrations of CRM1 . The shaded areas in blue and red represent the time frames where data collected represent pre-bleach and post-bleach fluorescence , respectively . Data points and the fitted curve are plotted in the panel below , with the data points colored according to their respective time-trace . Residuals are plotted on the bottom panel . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 010 Crystal structures of CRM1-bound hRio2NES-R and CPEB4NES-R peptides were solved at 2 . 3 Å and 2 . 9 Å resolution , respectively ( Figure 5 , Figure 5—figure supplement 1 and Table 1 ) . These structures show the peptides binding in the plus direction , that is , opposite that of their wild-type counterparts ( Figure 5A ) . The CRM1 grooves in the hRio2NES-R and CPEB4NES-R complexes are almost identical to those in the wild-type hRio2NES and CPEB4NES complexes ( Cα rmsds 0 . 2–0 . 3 Å ) . The N-terminal helices of hRio2NES-R and CPEB4NES-R that span Φ0–Φ3 bind CRM1 much like the helix of the plus direction PKINES . Their C-terminal strand-like ΦXΦ segments bind in the narrow part of the CRM1 groove , but are placed slightly outward toward solvent , perhaps to better accommodate the large Phe and Met side chains in the P3 and P4 CRM1 pockets ( Figure 5B ) . Pull-down assays with single amino acid mutants of hRio2NES-R reveal that smaller hydrophobic residues such as leucine in the Φ3 position and isoleucine , leucine , and valine in the Φ4 position are preferred for binding to CRM1 ( Figure 5C ) . The preference for smaller hydrophobic side chains in ΦXΦ segment can possibly be explained by the relief of steric constraints caused by the bulky phenylalanine in the native hRio2NES-R sequence . The structures of CRM1-bound hRio2NES-R and CPEB4NES-R peptides support the idea that the spacing between the hydrophobic residues is critical for determining the orientation the NES binds . When the sequence and the hydrophobic spacing pattern of an NES are reversed , the direction of the peptide binding CRM1 is also reversed . However , binding affinities of the NESs are similar regardless of binding orientation , consistent with the observation that hydrophobic interactions between the CRM1 groove and side chains in the Φ positions of hRio2NES and CPEB4NES , which likely govern CRM1-NES affinity , are preserved in hRio2NES-R and CPEB4NES-R . 10 . 7554/eLife . 10034 . 011Figure 5 . hRio2NES-R and CPEB4NES-R are plus NESs . ( A ) Structures of hRio2NES-R ( light blue ) and CPEB4NES-R ( light pink ) bound to CRM1 ( gray surfaces ) . ( B ) Pairwise comparisons of PKINES ( yellow ) , hRio2NES-R ( light blue ) , and hRio2NES ( blue ) when bound to CRM1 . ( C ) In vitro pull-down assay of purified human CRM1 binding to immobilized GST-hRio2NES-R mutants ( Φ3 or Φ4 mutated ) in the presence of excess ScRanGTP . Relative band intensities of triplicate experiments are plotted in histograms . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 01110 . 7554/eLife . 10034 . 012Figure 5—figure supplement 1 . Electron densities of the reverse NES peptides . Stereo views of kicked OMIT map meshes , on the final coordinates for ( A ) hRio2NES-R , ( B ) CPEB4NES-R and as shown in sticks . Anomalous difference map for Se-met hRio2NES-R is contoured at the 3 . 0σ level in the leftmost panel in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 012 To more rigorously test the idea that position of the ΦXΦ motif is critical for determining NES orientation , we flipped the C-terminal ΦXΦ ( LDI ) of PKINES ( SNELALKLAGLDI ) to the N-terminus of the peptide while preserving the sequence of the NES helix of wild-type PKINES ( Figure 6A ) . We named the new peptides PKINES-Flip and three variations were designed . PKINES-Flip1 has the inverted wild-type LDI at the N-terminus , giving sequence IDLNELALKLAGL . The two hydrophobic side chains in the N-terminal Φ-X-Φ were incrementally made larger to generate PKINES-Flip2 ( FDLNELALKLAGL ) and PKINES-Flip3 ( FDMNELALKLAGL ) mutants . PKINES-Flip1 does not interact with CRM1 in pull-down assays , while PKINES-Flip2 and PKINES-Flip3 show graded increases in CRM1 binding ( Figure 6B ) . We solved the structure of PKINES-Flip3 bound to CRM1 at 2 . 5 Å resolution and it indeed binds in the minus direction ( Figure 6C , Figure 6—figure supplement 1 and Table 1 ) . These results show that NES binding in the minus vs plus direction is determined by placement of the ΦXΦ pattern at the N- or C-terminal end of the NES peptide . Secondary to this positioning , hydrophobic side chains of the N-terminal ΦXΦ segment of a minus NES should be long enough to reach into binding pockets and pack with the CRM1 groove favorably . 10 . 7554/eLife . 10034 . 013Figure 6 . N-terminal ΦXΦ motif generates a minus orientation PKINES-Flip mutant . ( A ) Sequence alignment of PKINES and PKINES-Flip peptides with their hydrophobic residues of ΦXΦ motifs in red and the NES helix shown as a cylinder . ( B ) Pull-down assay of immobilized GST-PKINES mutants , purified CRM1 and RanGTP ( Coomassie-stained SDS/PAGE ) . ( C ) Structure of the PKINES-Flip3 peptide ( red , in cartoon with Φ residues in sticks ) bound to CRM1 ( gray surface ) with its sequence and CRM1 pockets for each Φ residue shown below . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 01310 . 7554/eLife . 10034 . 014Figure 6—figure supplement 1 . Electron densities of the PKINES-Flip3 NES peptide . Stereo views of kicked OMIT map meshes , on the final coordinates for PKINES-Flip3 . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 014 The discovery that CRM1 binds NESs in both the plus and minus directions almost doubles the number of possible NES consensus sequences . Of the six NES patterns in Figure 1A , class 1a , 1b , 1c , and 1d patterns are asymmetric , whereas class 2 and class 3 patterns are symmetric . Each of the asymmetric class 1a , 1b , 1c , and 1d patterns , which represent plus NESs , could be reversed to give class 1a-R , 1b-R , 1c-R , and 1d-R patterns that represent minus NESs ( Figure 7A ) . In principle , symmetric class 2 and 3 patterns can also bind CRM1 in both the plus and minus directions . For example , the class 2 RevNES binds CRM1 in the plus direction as an entirely extended chain , but it is also possible that hydrophobic side chains of another class 2 NES can be presented from a similar extended peptide in the minus direction . However , it remains to be determined whether any of the currently known class 2 and true class 3 peptides can indeed bind in the minus direction . Expansion of NES consensus by reversing class 1 NES consensus patterns to generate class 1-R patterns further suggests a corresponding increase of potential NESs in the proteome . 10 . 7554/eLife . 10034 . 015Figure 7 . Prevalence of putative minus NESs in the Dbase data set . ( A ) Consensus patterns for minus NESs in new NES classes 1a-R to 1d-R ( reverse of class 1 patterns ) . ( B ) The number of sequences in the 246 proteins in Dbase that match the class 1 ( + ) and class 1-R ( − ) consensus patterns . NES regions are defined according to original literature that experimentally identified CRM1 cargos and their NES regions . ( C ) The numbers of NES regions in Dbase divided into four categories according to the consensus matches they overlap with . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 015 We searched for sequences that match class 1-R ( minus ) patterns in the Dbase data set , which compiled 246 NES-containing CRM1 cargos from previously published literature ( Xu et al . , 2015 ) . Each CRM1 cargo contains multiple sequences that match NES consensus patterns but most of these sequences are not functional export signals . Dbase reports a total of 290 experimentally identified NES regions for the 246 CRM1 cargos in the database . Matches for both class 1 ( plus ) and class 1-R ( minus ) patterns appear to be similarly prevalent in the 246 CRM1 cargos ( 1849 minus vs 1950 plus matches ) ( Figure 7B ) . However , plus patterns seem to be somewhat enriched within NES regions ( 340 plus vs 230 minus matches; Chi-square test , p-value = 1 . 378e−05 ) ( Figure 7B ) . The bias for plus patterns in these previously reported NES regions may be a consequence of NES searches that were guided solely by the plus consensus patterns , since the minus patterns were unknown . The Dbase data set is further complicated by a lack of validation of direct CRM1-NES interactions . Only 60% of previously reported class 1 NESs that were tested recently were found to actually bind CRM1 ( Xu et al . , 2012a ) . Of the 290 NES regions in Dbase , 40% ( 116 ) contain sequences that match both class 1 ( plus ) and class 1-R ( minus ) patterns ( Figure 7C ) . 89 NES regions match class 1 pattern exclusively and 24 match class 1-R patterns exclusively ( Figure 7C ) , suggesting that there are still a significant population of putative minus NES even though the current NES annotation is biased and imperfect . We further investigated the 24 NES regions that contained only class 1-R matches , filtering out four because of overlap with the class 2 consensus , which was previously not considered in the analysis . The remaining 20 NES regions contain 22 sequences that match class 1-R patterns , which were tested for CRM1 binding in pull-down assays . Of the 22 sequences tested , one degraded and another aggregated during purification resulting in only 20 relevant NES sequences . 10 out of the 20 putative minus NESs , or 50% , bind CRM1 ( Figure 8 ) . This percentage is similar to the proportion of tested class 1 ( plus ) NESs that bind CRM1 ( Xu et al . , 2012a ) . These results suggest that there are a substantial number of functional NESs that likely bind CRM1 in the minus direction , even in an NES dataset like Dbase where many NESs were previously identified using the only plus NES consensus patterns . These possibly include cases where NES patterns have been mistakenly annotated or annotated as previously non-canonical patterns . This new expansion of NES consensus provides a means to identify previously unrecognizable NESs in previously identified and new CRM1 cargos . 10 . 7554/eLife . 10034 . 016Figure 8 . Putative minus NESs in the Dbase data set . ( A ) Summary of putative minus NESs ( in the Dbase data set that match class 1-R patterns exclusively ) tested for CRM1 binding . Nap1p ( * ) was previously shown to direct nuclear export in cells even though no CRM1 binding was observed ( Xu et al . , 2015 ) . ( B ) Putative minus NESs that bind CRM1 in pull-down assays with CRM1 and RanGTP . ( C ) Putative minus NESs that show no observable CRM1 binding . DOI: http://dx . doi . org/10 . 7554/eLife . 10034 . 016 The NES appears to be the only nuclear-targeting signal , and perhaps the only organelle-targeting signal , that has been shown thus far to bind its receptor in both polypeptide directions . This is in contrast to several modular-domain signaling systems , which are known to bind their linear motifs in both polypeptide chain directions ( Feng et al . , 1994; Lim et al . , 1994; Osawa et al . , 1999; Swanson et al . , 2004; Song et al . , 2005; Lorenz et al . , 2008; Ng et al . , 2008; Neufeld et al . , 2009 ) . Are protein systems that recognize linear motifs in opposite orientations unique or will most linear motifs bind their receptors in both orientations even though this phenomenon has not yet been observed for many ? Alternatively , do linear motifs that bind only in a single orientation do so because of spatial constraints inherent to their cellular functions ? Regularly spaced hydrophobic pockets in the CRM1-NES groove interact with similarly spaced NES side chains that often project from one face of amphipathic α-helices . Interestingly , several other linear motifs that bind in opposite orientations ( Paxillin LD motifs , HBP1/Mad1 Sin2-interaction domains , SH3-binding polyproline peptides , and various calmodulin targets [Osawa et al . , 1999; Swanson et al . , 2004; Lorenz et al . , 2008; Neufeld et al . , 2009] ) also present side chains from helices for recognition . Side chain to side chain distances within secondary structural elements of motifs are preserved regardless of polypeptide orientation , thus producing a feature that may be conducive for binding in opposite orientations . The required shift of the backbone to put the plus and minus side chains in the same position is more likely for linear motifs than for extensive interfaces between two folded proteins , since the latter are constrained by many additional contacts outside of the helix-binding groove . Thus , bidirectional recognition is probably more prevalent in recognition of linear helical motifs than in recognition of larger structured elements . Extended linear motifs such as phosphotyrosine peptides that bind SH2 domains also bind in opposite orientations ( Ng et al . , 2008 ) . Here , side chains , mainly the phosphotyrosine side chain , contribute the majority of binding energy , which can still be preserved when peptide orientation is reversed . Linear motifs that use mostly side chains for binding may be amenable to interactions in opposite orientations but those that make extensive contacts using their backbones may be limited to a particular orientation . For example , the IBB region of Importin-α , which is the nuclear localization signal or NLS that binds directly to Importin-β , is a long 28-residue helix that is preceded by a loop ( Cingolani et al . , 1999 ) . The IBB uses mostly charged and polar side chains to interact with Importin-β , and perhaps these side chain interactions could be preserved when polypeptide direction of the NLS peptide is flipped . Similarly , PY-NLS binding to Karyopherin-β2 ( also known as Transportin-1 ) involves mostly the NLS side chains , and we may observe these NLSs binding to Karyopherin-β2 in the opposite orientation in the future ( Lee et al . , 2006; Soniat et al . , 2013 ) . In contrast , the classical-NLS recognition by Importin-α and the Kap121-specific lysine-rich NLS ( also called the IK-NLS ) recognition by Kap121 involve extensive interactions with the NLS main chains and are therefore less likely to bind bidirectionally to their importins ( Conti et al . , 1998; Fontes et al . , 2000; Kobayashi et al . , 2015; Soniat and Chook , 2015 ) . Further studies will inform on orientation requirements for NLSs binding to their respective importins . We suggest that bidirectional recognition may , in fact , be widely present , but simply not widely observed . Components of modular-domain signaling and nuclear-targeting systems consist of mostly soluble proteins that bind linear motifs found within intrinsically disordered regions . These protein–peptide interaction systems are relatively free of spatial constraints compared to systems that bind organelle-targeting signals for delivery into membrane compartments . An example of the latter is the binding of ER signal sequences by the signal recognition particle SRP , which is likely constrained spatially by the nascent chain emerging from the ribosome and by subsequent delivery into the lumen of the translocon ( Janda et al . , 2010; Akopian et al . , 2013 ) . In principle , linear motifs that could bind in both orientations are sometimes constrained by other factors that limit them to only one . The CRM1-NES interaction is free from such spatial constraints as the entire exportin–cargo complex enters the nuclear pore complex for transport to the cytoplasm , thereby allowing some NESs to bind CRM1 in the plus orientation and others in the minus orientation . Finally , accurate prediction of NESs has been difficult because of the breadth and simultaneously , the insufficient coverage of the NES consensus . Many functional NES-containing regions of proteins contain multiple NES consensus matches and sometimes no NES match , suggesting that the set of NES consensus does not provide sufficient coverage for NES identification . Our study shows that structures of NESs bound to CRM1 can accurately define consensus patterns and sometimes identify new consensus patterns . Expansion of the NES consensus upon discovery of minus NESs leads to improved coverage of potential NESs , thus allowing identification of previously unrecognized NESs in known and new CRM1 cargos . However , the improved coverage afforded by the knowledge of bidirectional NES-binding is largely orthogonal to the problems in NES prediction that arise from false positive NESs ( Fu et al . , 2011; Xu et al . , 2012a ) . The majority of the NES patterns describe the ubiquitous 2-turn amphipathic helix , which are found in most helix-containing proteins , and many of these consensus-matching sequences are part of hydrophobic cores that are not accessible for CRM1 binding . In the development of NES predictors ( NESsential by the Horton Lab and LocNES by the Chook lab , Fu et al . , 2011; Xu et al . , 2015 ) , we found that prediction accuracy was improved by using both sequence and structural/biophysical properties ( such as disorder propensity and/or solvent accessibility ) as features for machine-learning methods . The latter features allowed consensus-matching sequences in the interior of folded domains to be flagged . A set of consensus sequences with high coverage rate such as the expanded set in Figure 7A is desirable when employed as a pre-filter in NES prediction as the machine-learning process that follows serves to reduce false positive matches . Future identification of minus NESs will also increase the size , diversity , and accuracy of experimental NES databases , which are the training/testing data sets for the development of our next generation NES predictors . In summary , we have found that NES peptides can bind the narrow CRM1-NES groove in two opposite orientations , which we now describe as the plus and minus orientations . Whether an NES binds CRM1 in the plus or minus orientation is determined by the location of its ΦXΦ strand motif . A C-terminal ΦXΦ motif that follows a helix dictates a plus NES , while an N-terminal ΦXΦ followed by a helix results in a minus NES . The five hydrophobic pockets in the CRM1-NES groove interact with hydrophobic side chains that are presented in many different ways on NES peptides , by different secondary structural elements and in both polypeptide chain directions , to enable specific recognition of diverse NES sequences . ScCRM1 ( 1–1058 , Δ377–413 , 537DLTVK541 to GLCEQ , V441D ) was cloned into the previously described pGEX-TEV vector ( Chook and Blobel , 1999 ) . As previously described in Sun et al . ( 2013 ) polypeptide segments that make up the ScCRM1 and human CRM1-NES grooves are 81% identical in sequence , with complete conservation in residues lining the groove that contact NESs and inhibitors ( Sun et al . , 2013 ) . In order to maximize similarity to the human CRM1-NES groove , we mutated the only stretch of ScCRM1 groove residues that has more than 2 non-conserved residues , 537DLTVK541 in the NES-binding groove of ScCRM1 to the human CRM1 sequence GLCEQ ( Sun et al . , 2013 ) . Yrb1p ( residues 62–201; or RanBP1 ) was cloned into pGEX-TEV and human Ran ( full-length ) was cloned into the pET-15b vector . Various NESs were cloned into the pMal-TEV vector . Sequences of NES peptides used for crystallization after TEV cleavage are hRio2NES: GGSY389RSFEMTEFNQALEEI403; hRio2NES-R: GGSYGKIEELAQNFETMEFSR; CPEB4NES: GGSY379RTFDMHSLESSLIDI393: CPEB4NES-R: GGSYRMIDILSSELSHMDFTR; PKINES-Flip3: GGSYRSFDMNELALKLAGLD . Sequences of NESs used for binding affinity measurements are listed in Figure 4 . All proteins were expressed separately in E . coli BL-21 ( DE3 ) by induction with 0 . 5 mM isopropyl β-D-1-thiogalactopyranoside for 10 hr at 25°C . GST-ScCRM1 and GST-RanBP1 cells were lyzed in buffer containing 40 mM HEPES ( pH 7 . 5 ) , 2 mM MgOAc , 200 mM NaCl , 10 mM dithiothreitol ( DTT ) and protease inhibitors , purified by affinity chromatography using glutathione Sepharose 4B beads ( GE Healthcare Life Sciences , PA ) , followed by cleavage with TEV protease and finally size-exclusion chromatography in GF buffer ( 20 mM HEPES pH 7 . 5 , 100 mM NaCl , 5 mM MgOAc , and 2 mM DTT ) . Cells expressing His-Ran were lyzed in buffer containing 50 mM HEPES ( pH 8 . 0 ) , 2 mM MgOAc , 200 mM NaCl , 10% ( vol/vol ) glycerol , 5 mM imidazole ( pH 7 . 8 ) , 2 mM DTT and protease inhibitors , purified by affinity chromatography with Ni-NTA Agarose ( Qiagen , Hilden , Germany ) and further purified by gel filtration chromatography in TB buffer ( 20 mM HEPES pH 7 . 5 , 110 mM KOAc , 2 mM MgOAc , 10% glycerol , and 2 mM DTT ) . Ran was loaded with non-hydrolyzable GTP analog GppNHp by nucleotide exchange . Cells expressing MBP-NESs were lyzed in buffer containing 50 mM HEPES pH 7 . 5 , 100 mM NaCl , 10% glycerol , 2 mM DTT and protease inhibitors , purified by affinity chromatography using amylose resin ( New England Biolabs , MA ) and ion exchange chromatography using ( HiTrap Q , GE Healthcare Life Sciences ) with a salt gradient from 50 mM to 1 M NaCl . Purified MBP-NES proteins were concentrated , cleaved with TEV protease and NES peptides were then isolated by gel filtration chromatography in GF buffer . To assemble the CRM1-Ran-RanBP1-NES complex , the RanGppNHp-RanBP1 heterodimer was first purified by gel filtration chromatography . ScCRM1* , Ran-RanBP1 and NES peptides were then assembled in 1:3:10 molar ratio and the quaternery complexes were purified by gel filtration chromatography in GF buffer . Purified ScCRM1-Ran-RanBP1-NES complexes were concentrated to ∼10 mg/ml and excess NES peptides were added to stabilize the complex during concentration . ScCRM1-Ran-RanBP1-NES complexes were crystallized in 17% ( wt/vol ) PEG3350 , 100 mM Bis-Tris ( pH 6 . 4 ) , 200 mM ammonium nitrate , and 10 mM Spermine HCl . Crystals were cryoprotected with the same crystallization condition supplemented with up to 23% PEG3350 and 12% glycerol and flash cooled in liquid nitrogen . X-ray diffraction data were collected at 0 . 9795 Å at the Advanced Photon Source 19ID beamline in the Structural Biology Center at Argonne National Laboratory . Data were indexed , integrated , and scaled using HKL-3000 ( Minor et al . , 2006 ) . All crystals in this study were isomorphous to crystals of previously solved inhibitor-bound and unliganded ScCRM1-Ran-RanBP1 complexes and has space group P43212 . Therefore , structures were determined by multiple rounds of refinement of unliganded complex ( 4HB2 ) against collected data using PHENIX ( Adams et al . , 2010; Afonine et al . , 2012 ) and manual modeling in Coot ( Emsley et al . , 2010 ) . X-ray/stereochemistry and X-ray/ADP weights were optimized in phenix . refine in final stages of refinement . Structure validation was guided by Molprobity suite in PHENIX ( Chen et al . , 2010 ) . Ramachandran plots of the five structures showed that 97 . 3–97 . 9% of residues are in favored regions and 0 . 0–0 . 1% are in disallowed regions . Structure figures were generated with PyMOL ( Schrodinger , 2010 ) . NESs in Figures 2 and 5 were compared by superimposing H12A helices of their respective CRM1s . Full-length human CRM1 ( HsCRM1 ) was purified in the same manner as ScCRM1 with buffers supplemented with 10% glycerol . ScRan ( Gsp1p ) was expressed using pET21d-GSP1 ( GSP1 residues 1–179 , Q71L ) ( gift from Dr . Takuya Yoshizawa ) and purified as described above for human Ran ( buffers in HEPES pH 7 . 4 instead of pH 8 . 0 ) . After affinity purification , ScRan was loaded with GTP ( incubated with molar excess of ethylenediaminetetraacetic acid ( EDTA ) for 30 min on ice followed by incubation with excess GTP and MgOAC for 30 min at room temperature ) and then purified by ion exchange chromatography ( HiTrap SP , GE Healthcare Life Sciences ) . NESs were cloned into the pGEX-TEV vector ( Chook and Blobel , 1999 ) , purified , and immobilized on glutathione Sepharose beads ( GE Healthcare Life Sciences ) in TB buffer described above containing 15% glycerol . 2 . 5 μM HsCRM1 and 7 . 5 μM ScRanGTP were added to ∼10 μg of immobilized GST-NESs in TB buffer in total volumes of 200 μl for 30 min at 4°C . Unbound proteins were washed extensively with TB buffer and bound proteins on the Sepharose beads were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS/PAGE ) and visualized with Coomassie Blue staining . All binding assays were performed in triplicates . To compare the relative intensities of CRM1 bands to yield an estimate of binding activities of various NES mutants , SDS/PAGE gels were dried and scanned with an Epson V300 scanner and the images analyzed with ImageJ software . CRM1 band intensities were corrected for differences in GST-NESs band intensities and normalized to wild-type control intensity in each set of mutations . Corrected relative CRM1 band intensities were plotted as histograms with standard errors with GraphPad Prism . To test putative minus NESs identified from the Dbase data set , 5 μM HsCRM1 with or without 15 μM ScRanGTP were used instead . Putative NESs that show no CRM1 binding were expressed in larger scale and purified by size-exclusion chromatography to assess their aggregation states . They were also subjected to intact mass determination by mass spectroscopy to ensure that the GST-NES proteins were not degraded . Cellular localization of EYFP2-NLS-NES fusion proteins overexpressed in HeLa cells was observed using procedures as previously described ( Xu et al . , 2015 ) . Expression constructs for EYFP2-NLS-NES fusion proteins were cloned similarly into pEYFP2-SV40NLS vectors . Live cell images were collected using a spinning disk confocal microscope system , Nikon-Andor ( Nikon , NY ) , and MetaMorph software . Image analysis was performed similarly with ImageJ . CRM1 dependence was demonstrated by the nuclear accumulation of EYFP fusion proteins after treatment with 2 nM Leptomycin B for 16 hr at 37°C . Experiments were performed in at least duplicates with over a total of 150 transfected cells . Differential bleaching was used as a parameter to monitor binding of MBP-NES proteins to CRM1 . In short , we observed that the fluorophore reporter attached to PKINES ( FITC ) underwent a reproducible time-dependent bleaching when exposed to excitation light ( Figure 4—figure supplement 2 ) . Furthermore , this phenomenon was concentration-dependent , that is , the bleaching was accelerated when the FITC-PKINES probe was exposed to increasing concentrations of CRM1 . However , this phenomenon was saturable at high concentrations of CRM1 , indicating that it was a function of CRM1 binding and not simply the presence of the protein that was causing the change . The differential bleaching can be counteracted by titrating of mixture of FITC-PKINES and CRM1 with a known competitor , MBP-PKINES , which competes directly with the fluorescent probe for the NES binding groove in CRM1 . A sigmoidal appearance of the binding and competition isotherms is observed when differential bleaching , quantified as the average fluorescence at a time after bleaching normalized by the averaged fluorescence just after the beginning of bleaching , is plotted vs titrant concentration in a semilog graph ( see Figure 4 ) . This illustrates that this bleaching behavior can be described as a two-state system where unbound probe and CRM1-bound probe bleach at different but specific rates , and that these quantities report on the populations of bound and unbound FITC-PKINES . A detailed description of the data-fitting procedures will be described in manuscript in preparation by C . A . B . For error reporting , we used F-statistics and error-surface projection method to calculate the 68 . 3% confidence intervals of the fitted data ( Bevington and Robinson , 1992 ) . While error reporting using the error surface projection method is relatively uncommon , the ranges more accurately represent the true confidence intervals given the observed noise in the performed set of experiments because they explicitly account for the ability of the fitting algorithm to compensate for fitting defects by modifying correlated parameters . Thus , they provide better evaluation of the fitted data than other , more commonly used methods ( e . g . , error estimations from the parametric variance-covariance matrix ) . All proteins used were subjected to an extra gel filtration step and dialysis overnight in TB buffer with 15% glycerol to remove possible aggregation and ensures buffer matching . The FITC-PKINES peptide ( FITC-SGNSNELALKLAGLDINKT ) was chemically synthesized by GenScript , NJ and dissolved in the TB buffer with 15% glycerol . For the direct titration , HsCRM1 was serially diluted from 40 μM to 1 . 2 nM and incubated with mixture of 120 μM ScRan-GTP and 40 nM FITC-PKINES in 1:1 vol to a total volume of 20 μl , and incubated for 1 hr in the dark at room temperature . For competition experiments , MBP-NESs were serially diluted from 100 μM to 3 nM in presence of 40 nM of FITC-PKINES and incubated with mixture of 300 nM HsCRM1 and 120 μM ScRan-GTP in 1:1 vol to a total of 20 μl , and incubated for 1 . 5 hr in the dark at room temperature . All reactions mixtures were supplemented with 0 . 05% Tween-20 . Following incubations , reactions were loaded into NanoTemper's ‘Standard’ treated capillaries and fluorescence signals were monitored by NanoTemper Monolith NT . 115 equipment ( NanoTemper Technologies , München , Germany ) with 60% LED power for 10 s . Titrations for parallel comparisons were performed in triplicates on the same day . Data collected were then analyzed with PALMIST ( manuscript in preparation ) and imported to GUSSI for generating figures ( Brautigam , 2015 ) . Protein sequences in the Dbase data set , a non-redundant compilation of CRM1 cargos from two of the most recent NES databases , ValidNESs ( Fu et al . , 2013 ) , and NESdb ( Xu , Grishin et al . , 2012b ) ( http://prodata . swmed . edu/LRNes ) , were used for analyses of plus and minus NESs . All protein sequences in the Dbase data set were first compiled along with their annotated NES regions into an in-house database implemented by MySQL ( version 5 . 5 . 43 ) on Linux ( Ubuntu 12 . 04 ) . NES regions are defined according to original reports in the published literature that identified the CRM1 cargos . PHP ( version 5 . 3 ) regular expression with look-ahead assertions was used to capture all sequences ( including overlapping sequences ) that match the eight different class 1 ( plus ) and class 1-R ( minus ) NES consensus patterns . Duplicate matches ( such as 10-mer sequences that simultaneously match both class 1a and 1d patterns , or match both 1a-R and 1d-R patterns ) were removed using Linux command line tools , and the resulting numbers of consensus matches ( see Figure 7 ) were used for the enrichment test of plus consensus patterns within NES regions by Chi-square test using R ( version 2 . 14 . 1 ) . The same MySQL database was used to search for putative minus NESs , and 24 NES regions that match the 1-R patterns exclusively were identified . Four of these NES regions were removed because of overlap with class 2 patterns , resulting in 20 NES regions ( containing 22 1-R consensus matches ) for examination of CRM1 interactions by pull-down binding assays . Structures and crystallographic data have been deposited at the PDB: 5DHF ( CRM1-hRio2NES complex ) , 5DIF ( CRM1-CPEB4NES complex ) , 5DI9 ( CRM1-hRio2NES-R complex ) , 5DHA ( CRM1-CPEB4NES-R complex ) , 5DH9 ( CRM1-PKINES-Flip3 complex ) .	Many organisms keep their DNA within a structure inside their cells called the nucleus . Two layers of membrane surround the nucleus and keep the DNA separate from the rest of the cell's contents . Yet , proteins and other molecules can move in and out of the nucleus by passing through small pores in this nuclear membrane . To travel through these pores , larger molecules such as proteins rely on the assistance of transport receptors , including one called CRM1 . This transport receptor helps to export hundreds of different proteins from the nucleus by recognizing a part of their structure called the ‘nuclear export signal’ . Earlier work has shown that three different nuclear export signals interact with CRM1 in a similar ways by binding to a groove on its outer surface . But , there are several different types of nuclear export signal , and many are predicted to have three-dimensional structures that would seem to prevent them from binding to CRM1 in this way . As yet , it remains unknown how these diverse signals interact with this important transporter receptor . Protein crystallization is a technique that is used to visualize a protein's three-dimensional structure . Fung et al . have now used this approach to investigate how a particular class of nuclear export signals ( called ‘class 3’ ) bind to CRM1 . First , a modified form of CRM1 was crystallized once it had bound to a small fragment of protein that contains a class 3 nuclear export signal . The protein's molecular structure was then revealed by performing X-ray diffraction on the crystals . The results show , unexpectedly , that two different nuclear export signals in class 3 bind to the groove of CRM1 in the opposite direction to that reported previously . Additional biochemical and structural experiments then identified a particular feature or motif in the nuclear export signals that determines which way round they bind to CRM1 . This discovery advances our understanding of how these signals work , which will allow us to more accurately identify new nuclear export signals from genome sequences . As more CRM1-binding nuclear export signals are discovered in the future , the experimental data sets used to train the computational programs that are currently used to locate these signals in genomic sequences will be diversified and improved .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Structural determinants of nuclear export signal orientation in binding to exportin CRM1
Medullary thymic epithelial cells ( mTECs ) play a critical role in central immune tolerance by mediating negative selection of autoreactive T cells through the collective expression of the peripheral self-antigen compartment , including tissue-specific antigens ( TSAs ) . Recent work has shown that gene-expression patterns within the mTEC compartment are heterogenous and include multiple differentiated cell states . To further define mTEC development and medullary epithelial lineage relationships , we combined lineage tracing and recovery from transient in vivo mTEC ablation with single-cell RNA-sequencing in Mus musculus . The combination of bioinformatic and experimental approaches revealed a non-stem transit-amplifying population of cycling mTECs that preceded Aire expression . We propose a branching model of mTEC development wherein a heterogeneous pool of transit-amplifying cells gives rise to Aire- and Ccl21a-expressing mTEC subsets . We further use experimental techniques to show that within the Aire-expressing developmental branch , TSA expression peaked as Aire expression decreased , implying Aire expression must be established before TSA expression can occur . Collectively , these data provide a roadmap of mTEC development and demonstrate the power of combinatorial approaches leveraging both in vivo models and high-dimensional datasets . The thymus is the primary site of T-cell development and is required for the establishment of central immune tolerance ( Klein et al . , 2014 ) . Within the thymus , medullary thymic epithelial cells ( mTECs ) play a key role in enforcing tolerance by expressing a wide array of tissue-specific self-antigens ( TSAs ) that serve to eliminate potentially autoreactive developing T cells ( Derbinski et al . , 2008; Lancaster et al . , 2019; Metzger and Anderson , 2011 ) . The autoimmune regulator ( Aire ) protein plays a central and nonredundant role in the expression of a subset of TSAs ( Anderson et al . , 2002 ) and is believed to function , in part , by modulating chromatin accessibility ( Koh et al . , 2018; Perniola , 2018 ) . The transcription factor Fezf2 has also been shown to promote the expression of TSAs within mTECs in an Aire-independent manner ( Takaba et al . , 2015 ) . Aire-expressing mTECs are characterized by high levels of MHC class II expression in the adult thymus and have an estimated half-life of 12–14 days ( Gray et al . , 2006 ) . These mTEC-MHC Class II high cells ( mTEC-hi ) are thought to arise from mTEC-MHC Class II low progenitors ( mTEC-lo ) ( Gray et al . , 2007 ) through inductive signals that include signaling through receptor activator of NF-κB ( RANK ) ( Rossi et al . , 2007 ) . Within mTEC-hi cells , single-cell studies have shown that expression of TSAs is highly heterogeneous , with each cell expressing only 1–3% of all TSAs ( Anderson and Jenkinson , 2015; Brennecke et al . , 2015; Meredith et al . , 2015 ) . This heterogeneity is likely due to complex regulatory events that occur before expression of Aire and Fezf2 . Considerable effort has been devoted to identifying the progenitors of Aire-expressing mTECs in the mTEC-lo compartment . It has been suggested that these progenitors express high levels of Ccl21a ( Onder et al . , 2015 ) , which encodes a chemokine ligand known to regulate the migration of positively selected thymocytes into the thymic medulla ( Kozai et al . , 2017 ) . However , disruption of these cells in mice without lymphotoxin beta receptor ( LTBR ) does not impact Aire-expressing mTECs , implying that Ccl21a-expressing cells may not be the precursors of Aire-expressing mTECs ( Lkhagvasuren et al . , 2013 ) . Additionally , thymic cells expressing Sca-1 were recently found to self-renew and generate both mTEC and cortical thymic epithelial cell ( cTEC ) lineages , suggesting these may be a bi-potent progenitor of early TEC development and further obfuscating previous work describing epithelial subset relationships ( Ucar et al . , 2014; Wong et al . , 2014 ) . Recently , single-cell RNA-sequencing ( scRNA-seq ) has been utilized to define the mTEC compartment during neonatal and adult stages of development ( Dhalla et al . , 2020; Kernfeld et al . , 2018; Miragaia et al . , 2018 ) . Using this approach , we and others identified a previously unappreciated and highly-differentiated mTEC subset that bore striking similarity to peripheral tuft cells found at mucosal barriers ( Bornstein et al . , 2018; Miller et al . , 2018 ) . Although scRNA-seq is a powerful tool for characterizing distinct cell populations , including minor and rare populations ( Hwang et al . , 2018 ) , assigning precursor–product relationships is difficult based on gene expression alone . In the current manuscript , we reasoned that combining scRNA-seq with in vivo tools would better resolve our understanding of the developmental relationships between mTECs . First , we used a bioinformatic approach on scRNA-seq of control thymus to characterize mTEC lineage relationships . Second , we combined scRNA-seq with a lineage tracing approach ( Kretzschmar and Watt , 2012 ) in which the developmental relationships of Aire-expressing mTECs and their progeny are readily discernible ( Metzger et al . , 2013 ) . Third , we combined scRNA-Seq with a transient ablation model in which differentiation of mTEC-hi cells is selectively blocked after treatment with anti-RANK-ligand ( RANKL ) antibody . Inhibition of RANK signaling depletes the Aire-expressing mTEC population , which then recovers over 10 weeks ( Khan et al . , 2014; Metzger et al . , 2013 ) . Using these approaches , we created a molecular roadmap of Aire-expressing mTEC development that details the kinetics of TSA expression in relation to Aire and identified a transit-amplifying population of mTECs that we propose is the immediate precursor of the Aire-expressing and Ccl21a-expressing populations . First , we utilized a bioinformatic approach to identify TEC populations and lineage relationships at the single-cell level that could be subsequently validated using lineage tracing and transient in vivo ablation ( Figure 1a ) . We performed single-cell transcriptome analysis of total TECs purified by flow cytometry ( EpCAM+ CD45- ) and subjected to scRNA-seq on the 10x Genomics platform ( Zheng et al . , 2017; Figure 1a and Figure 1—figure supplement 1a ) . A total of 2434 cells were included in the downstream analysis . Single cells were projected into a reduced-dimensional space using UMAP and were clustered based on the top 13 principle components , yielding six total clusters ( Adey , 2019; Becht et al . , 2018; Butler et al . , 2018; Figure 1b and Figure 1—figure supplement 1b ) . To identify clusters , we performed differential gene-expression analysis and searched for gene signatures previously associated with thymic epithelial subsets . As described in previous TEC single-cell studies , the ‘cTEC’ cluster was marked by high expression of Ackr4 and Prss16 ( Bornstein et al . , 2018; Kernfeld et al . , 2018; Miragaia et al . , 2018; Figure 1c and d , Figure 1—figure supplement 1c ) . The ‘Ccl21a-high’ cluster was marked by high expression of Ccl21a ( Lkhagvasuren et al . , 2013 ) and Krt5 ( Figure 1c and d , Figure 1—figure supplement 1c ) . The gene-expression pattern within the ‘Ccl21a-high’ cluster was consistent with populations that have previously been described as jTEC ( Miragaia et al . , 2018 ) , mTEC2 ( Kernfeld et al . , 2018 ) , and mTEC-I ( Bornstein et al . , 2018 ) . The ‘Tuft’ cluster had high expression of Trpm5 , Dclk1 , and a gene-expression pattern that closely agreed with previous single-cell descriptions of thymic tuft cells ( Bornstein et al . , 2018; Miller et al . , 2018; Figure 1c and d , Figure 1—figure supplement 1c ) . The remaining three clusters , dubbed ‘TAC-TEC’ , ‘Aire-positive’ , and ‘Late-Aire’ based on subsequent analysis , included cells that expressed Aire and Fezf2 and resembled populations previously described as mTEC3 ( Kernfeld et al . , 2018 ) , mTEC-II and mTEC-III ( Bornstein et al . , 2018 ) , and mTEhi and mTEClo ( Miragaia et al . , 2018; Figure 1c and d , Figure 1—figure supplement 1c ) . As further validation of our clustering , we restricted the list of differentially expressed genes to only those annotated as transcription factors and chromatin modifiers because these functional categories are known to be important in TEC biology ( Wang et al . , 2019; Figure 1—figure supplement 1d ) . We anticipated that differential expression analysis between clusters would be enriched for specific genes previously associated with mTEC function . Reassuringly , Aire ( Ramsey et al . , 2002 ) , Ptma ( Moretti et al . , 2015 ) , Hmgb1 ( Guha et al . , 2017 ) , Irf7 ( Otero et al . , 2013 ) , Cited2 ( Michell et al . , 2010 ) , Pax1 ( Romano et al . , 2013 ) , and Spib ( Akiyama et al . , 2014 ) were all found in the differentially expressed gene set and have known knockout phenotypes in mice affecting thymus development and function ( Figure 1—figure supplement 1d ) . Next , we used computational predictive algorithms to define the lineage relationships between the identified mTEC subpopulations . To accomplish this , we applied single-cell velocity ( Bergen et al . , 2019 ) , the stochastic version of transcriptional dynamics used in RNA velocity , which uses intron retention to predict the future transcriptional state of a single cell and infer lineage relationships between cell populations ( La Manno et al . , 2018 ) . Application of single-cell velocity to our data predicted that ‘TAC-TECs’ preceded the ‘Aire-positive’ population and that the ‘Aire-positive’ population preceded the ‘Late-Aire’ population ( Figure 1e ) . Further , single-cell velocity predicted a branching model of mTEC development in which ‘TAC-TECs’ give rise to both the ‘Aire-positive’ and ‘Ccl21a-high’ populations ( Figure 1e ) , rather than a precursor-product relationship between Ccl21a-expressing and Aire-expressing mTECs ( Onder et al . , 2015 ) . Additionally , this model also predicted a path to minor mTEC subsets , including the ‘Tuft’ cluster . Therefore , single-cell velocity analysis suggested that the ‘TAC-TEC’ cluster populates the mTEC subsets observed in this study . To further characterize the 'TAC-TEC’ cluster , we generated an unbiased list of all genes specifically upregulated in these cells ( Figure 2—figure supplement 1a ) . We noticed that this restricted list included a preponderance of chromatin-modifying factors , including Hmgb2 , H2afz , Hmgn2 , Hmgb1 , and Hmgn1 ( Figure 2—figure supplement 1a ) . Because many of these genes were also upregulated in other mTEC populations , we restricted this analysis to include only those genes uniquely upregulated in ‘TAC-TECs’ ( 35 total genes ) . It had been hypothesized that a transit-amplifying cell exists in the thymus ( Gray et al . , 2007 ) therefore , we compared the transcriptional signature of the ‘TAC-TECs’ to the gene signature of a published transit-amplifying population ( Basak et al . , 2018 ) and noticed a striking overlap . To test this overlap , we performed a hypergeometric test and determined that there was in fact significant enrichment of previously described transit-amplifying genes in the ‘TAC-TEC’ population ( p=1 . 44×10−34; Figure 2a and Figure 2—figure supplement 1b ) . Transit-amplifying cells are defined by rapid cycling ( Rangel-Huerta and Maldonado , 2017; Zhang and Hsu , 2017 ) , so we hypothesized that the ‘TAC-TEC’ cluster would contain cycling TECs . To test this we used cyclone ( Scialdone et al . , 2015 ) , an algorithm that determines cell-cycle state based on expression of cell-cycle genes , which revealed that the majority ( 104/123 ) of cells in the G2/M phase were located in the ‘TAC-TEC’ cluster ( Figure 2b ) . Notably , the mapping of cycling cells to the ‘TAC-TEC’ population provides support for the RNA velocity prediction that ‘TAC-TECs’ populate other major mTEC subsets ( Figure 1E ) . Expression of Hmgb2 specifically correlated with cells being classified as G2/M and many cycling cells also expressed Fezf2 ( Figure 2b ) . Many cycling ‘TAC-TECs’ expressed Aire and Ccl21a but they were often not co-expressed in the same cells , consistent with heterogeneity described in other transit-amplifying populations ( Zhang and Hsu , 2017; Figure 2b and Figure 2—figure supplement 1c ) . To determine if the cycling population could be detected at the protein level , we performed intracellular flow cytometry for Ki67 . For this analysis , MHCII and Aire gates were drawn at the lower boundry of the high-density contours to separate bona fide , mature MHCIIhi and Airehi populations from all others ( negative-intermediate ) . Intracellular staining revealed a Ki67bright TEC population that was MHCIIint and Aire−/low , implying that cell division immediately precedes high levels of Aire expression in a subset of cells ( Figure 2c ) . However , ‘TAC-TECs’ were not found to express previously described canonical stem cell markers ( Arnold et al . , 2011; Lyle et al . , 1999; Redvers et al . , 2006; Yu et al . , 2006 ) , underscoring their identity as a transit-amplifying population and distinct from stem-like progenitors ( Figure 2—figure supplement 1d ) . These data demonstrate the transcriptional signature of the ‘TAC-TEC’ population is enriched for transit-amplifying genes and this population is also actively cycling , both consistent with the classification of ‘TAC-TECs’ as transit-amplifying cells . A key function of mTECs is to express and display TSAs to developing T cells . While mTECs are capable of expressing the majority of the protein-coding genome , little is known about the kinetics of TSA expression in vivo and how specific regulatory factors impact TSA expression during mTEC maturation . Performing pseudotemporal analysis on single-cell RNA-seq data can be used to predict lineage relationships and the order of events that occurs during development ( Cheng et al . , 2018; Cohen et al . , 2018; Loo et al . , 2019 ) . To provide insight into the process of TSA expression , we examined gene expression across predicted developmental time using pseudotime to computationally order cells . To accomplish this , we used the Fantom database to compile a list of TSAs by identifying genes detected in five or fewer tissues ( Brennecke et al . , 2015; Forrest et al . , 2014 ) . We determined the average expression of all TSAs in each individual cell and plotted cells along pseudotime scores provided by Slingshot ( Street et al . , 2018; Figure 2d ) . Based on pseudotime predictions , TSA expression did not peak until well after the initiation of Aire expression and was maintained even after Aire expression decreased ( Figure 2d and e ) . To determine if the increase of average TSA expression late in pseudotime was because of changes in the number of TSAs expressed or the expression level of TSAs , we observed both the percent of TSAs expressed per cell and the average expression of only the expressed TSAs ( Figure 2—figure supplement 2a and b ) . The percent of TSAs expressed peaked late in pseudotime and followed the pattern described for all TSAs ( Figure 2—figure supplement 2a ) . In contrast , the average expression of only expressed TSAs did not show this pattern and instead increased gradually throughout pseudotime ( Figure 2—figure supplement 2b ) . These results demonstrate that the number of TSAs expressed peaks late in pseudotime rather than the expression level of TSAs and suggest Aire expression must be well-established before TSA expression can occur . To further dissect TSA expression patterns , we used previously described Aire- and Fezf2-dependent gene sets identified through characterization of the respective knockout animal models ( Sansom et al . , 2014; Takaba et al . , 2015 ) . Although Aire expression was detectable in dividing mTECs ( Figure 2b ) , the pattern of Aire-dependent gene expression ( a slow increase to peak levels followed by a decrease ) was very similar to that of all TSAs ( Figure 2e ) . Fezf2 expression peaked earlier in pseudotime than Aire expression and , in contrast to Aire-dependent gene expression , Fezf2-dependent genes were expressed more uniformly across pseudotime ( Figure 2e ) . To ensure the timing of gene expression over pseudotime wasn’t affected by TSAs , we repated the analysis after removing TSA genes from the ordering and observed similar timing of events across pseudotime ( Figure 2—figure supplement 2c ) . Together , these results provide a model for the kinetics of gene expression across development of Aire-positive mTECs and illustrate the differences in expression between Aire and Fezf2 . While single-cell transcriptomics can identify subpopulations , widely used tools for describing cellular dynamics rely upon computational assumptions and predictions . Thus , bioinformatics can predict lineage relationships between mTECs but cannot , in isolation , prove those relationships . To provide experimental support for the predicted lineage relationships , we first utilized an Aire lineage tracing system in which Aire-expressing cells are inducibly and indelibly labeled with the fluorescent reporter protein ZsGreen after tamoxifen treatment ( AireCreERT2;Rosa26CAG-stopflox-zsGreen ) ( Metzger et al . , 2013 ) . This allows for discrimination between cells that express Aire or have passed through an Aire-expressing state ( positive for ZsGreen ) and those that have never expressed Aire ( negative for ZsGreen ) ( Figure 3a ) . Because transient Aire expression occurs during embryogenesis ( Nishikawa et al . , 2010 ) , tamoxifen was used to allow strict temporal control of Cre recombinase activity as previously described ( Metzger et al . , 2013 ) . In this system , mice must be treated with tamoxifen to induce Cre translocation to the nucleus and initiate reporter expression . This event occurs only in cells expressing Aire and , therefore , Cre . However , once the Stop codon has been removed labeling is permanent and the reporter will continue to be expressed in cells after Aire expression ceases . Following a 10 day tamoxifen label , total TECs were purified by flow cytometry ( EpCAM+ CD45- ) and subjected to scRNA-seq on the 10x Genomics platform ( Zheng et al . , 2017; Figure 3a ) . A total of 1387 cells were included in the downstream analysis . Single cells were projected into a reduced-dimensional space using UMAP and were clustered based on the top 12 principle components ( Becht et al . , 2018; Butler et al . , 2018; Stuart et al . , 2019; Figure 3b and Figure 3—figure supplement 1a ) . First , we used the gene-expression patterns from our initial bioinformatic analysis from control mice to provide putative identities to the clusters observed in the lineage tracing animals ( Figure 3c ) . Notably , there was strong agreement between the clusters identified in control mice and those observed in the lineage tracing mice , indicating lineage trace samples were comparable to the control mice ( Figure 3—figure supplement 1 ) . Next , we examined the robustness of the computational lineage predictions ( Figure 2d ) by comparing the expression pattern of the ZsGreen lineage reporter to Aire . Specifically , in the ‘Aire-positive’ cluster , 92% of cells expressed high levels of both Aire and ZsGreen ( Figure 3c and e , Figure 3—figure supplement 2 , and Supplementary file 1 ) . In the ‘Late-Aire’ cluster , 46% expressed both Aire and ZsGreen , while 43% expressed only ZsGreen , confirming prior widespread Aire expression ( Figure 3c and e , Figure 3—figure supplement 2 , and Supplementary file 1 ) . The ‘Late-Aire’ cluster was also marked by a small number of cells ( 14% ) expressing high levels of Krt10 , which has previously been described as a maker of post-Aire , terminally differentiated cornifying mTECs ( Metzger et al . , 2013; Wang et al . , 2012; Figure 3c , Figure 3—figure supplement 2 , and Supplementary file 1 ) . In contrast , 52% of cells in the ‘TAC-TEC’ cluster expressed both Aire and ZsGreen but 18% of cells expressed only Aire ( Figure 3c and e , Figure 3—figure supplement 2 , and Supplementary file 1 ) . This was expected because ZsGreen reporter expression requires multiple downstream events after Aire expression , including CreERT2 translation , nuclear localization in the presence of tamoxifen , Cre-mediated Rosa-Stop-flox excision , and ZsGreen transcription and translation , which delay the kinetics of ZsGreen expression in comparison to Aire . This transcriptional lag provides a useful signature ( Aire+ , ZsGreen- ) that is indicative of recent Aire induction and was the pattern uniquely observed within the ‘TAC-TEC’ cluster ( Figure 3c and e , Figure 3—figure supplement 2 , and Supplementary file 1 ) . In contrast , co-expression of ZsGreen and Aire in the ‘Aire-positive’ cluster was nearly ubiquitous and consistent with established Aire expression ( Figure 3c and e , Figure 3—figure supplement 2 , and Supplementary file 1 ) . There were also 34% of cells that highly expressed ZsGreen in the ‘Ccl21a-high’ population but only 2% of these cells also expressed Aire ( Figure 3d , e , and Supplementary file 1 ) . To determine if ZsGreen was expressed in the ‘Ccl21a-high’ population at the protein level , we performed flow cytometry on Aire lineage tracing mice treated with tamoxifen as above and stained for intracellular CCL21 protein . Approximately half of cells that expressed CCL21 also expressed ZsGreen reporter but 25% of these double-positive cells were MHCIIlo , indicating that a subset of ‘Ccl21a-high’ cells are likely downstream of an Aire-expressing population ( Figure 3f and Figure 3—figure supplement 3 ) . The finding that Ccl21a cells are marked by ZsGreen is consistent with the pseudotime prediction that the ‘Ccl21a-high’ population is downstream of the ‘TAC-TEC’ population rather than the progenitor of the ‘Aire-positive’ population . Taken together , these data are consistent with the pseudotime predictions and provide further support for the ‘TAC-TEC’ cluster as the transit-amplifying population feeding the ‘Aire-positive’ and ‘Ccl21a-high’ compartments ( Figure 3c–f , Figure 3—figure supplements 2 and 3 , and Supplementary file 1 ) . Notably , we again observed heterogeneity in the ‘TAC-TEC’ cluster with some cells expressing Aire and other cells expressing Ccl21a , but rarely high levels of both ( Figure 3e ) . This again demonstrates that the ‘TAC-TEC’ population is a heterogeneous population with some cells more closely resembling the ‘Aire-positive’ population and other cells more closely resembling the ‘Ccl21a-high’ population . The combination of bioinformatics with lineage tracing enabled the identification of a transit-amplifying population that precedes Aire-expressing mTECs and provided a high-resolution map of Aire-adjacent TEC subsets for further analysis . To provide additional , independent support for the mTEC lineage relationships inferred by our analysis of unperturbed thymus , we employed a model of transient mTEC ablation in which RANK signaling was blocked in vivo using anti-RANKL antibodies ( Khan et al . , 2014; Metzger et al . , 2013 ) . We have previously characterized this system and its kinetics in detail ( Khan et al . , 2014 ) . Notably , by flow cytometry , there is a pronounced decrease in the absolute number of AIRE+ and MHCIIhi mTECs following antibody blockade , while MHCIIlo mTEC numbers are less impacted ( Khan et al . , 2014 ) . Structurally , corticomedullary architecture is maintained and KRT5 remains easily detectable by immunofluorescence but AIRE staining is transiently absent ( Khan et al . , 2014 ) . Because Aire-expressing mTECs are slowly repopulated following treatment with anti-RANKL , we used this model to permit the observation of developmental progression of adult mTECs during recovery from acute , transient ablation . Wild-type mice were treated with anti-RANKL blocking antibody and TECs were isolated for scRNA-seq using the 10x Genomics platform ( Zheng et al . , 2017 ) over a 10 week time course ( Figure 4a ) . A total of 8453 cells were included in the downstream analysis . After initial processing , the ablation samples , control samples , and Aire lineage tracing sample were combined by Seurat’s canonical correlation analysis , using the top 1000 most highly variable genes from each sample to account for batch effects ( Butler et al . , 2018 ) . These canonical correlations were used to perform dimensionality reduction and clustering on all combined samples ( Figure 4—figure supplement 1a ) . To determine how cell populations changed following treatment with anti-RANKL , we inferred population labels for clusters from the combined experiment based on the identity of the Aire lineage tracing cells contained in the respective clusters ( Figure 3c and Figure 4b ) . For example , if cluster 1 contained Aire lineage tracing cells classified as ‘Aire-positive’ , cluster 1 would be classified using this label . We found that gene-expression patterns between populations closely resembled those seen in the Aire lineage tracing experiment ( Figure 4—figure supplements 1 , 2 and 3 ) . As expected , anti-RANKL treatment decreased the proportion of all Aire-expressing and post-Aire-expressing populations throughout the time course of treatment and recovery ( Figure 4c and Figure 4—figure supplement 1b ) . The decrease in all mTECs , and specifically AIRE+ mTECs ( both in absolute count and relative frequency ) was also observed by intracellular flow cytometry ( Figure 4—figure supplement 1d ) . Although the Ccl21a population appeared to expand following treatment with anti-RANKL , intracellular flow cytometry revealed no significant difference in absolute counts of CCL21+AIRE- cells at week four in treated and control mice ( Figure 4d and Figure 4—figure supplement 2b ) . Thus , only the Aire-expressing developmental branch was depleted by treatment with anti-RANKL . If the ‘TAC-TEC’ population precedes the ‘Aire-positive’ and ‘Late-Aire’ populations in development as predicted by single-cell velocity and Aire-lineage tracing , recovery should follow this chronological order after ablation . To test this hypothesis , we overlaid the predicted pseudotemporal ordering of the Aire-expressing branch from all samples generated by Slingshot ( Street et al . , 2018 ) with the RANKL time course ( Figure 4e ) . Two weeks following treatment with anti-RANKL , the majority of remaining cells were located in the ‘Late-Aire’ cluster ( Figure 4f ) . These ‘Late-Aire’ cells likely represented the final wave of Aire-expressing cells immediately before RANKL blockade , consistent with estimates of mTEC half-life ( Gray et al . , 2006; Figure 4f ) . Four weeks after treatment , cells at the earliest point in pseudotime began to return , while the number of cells late in pseudotime decreased ( Figure 4f ) . Six weeks into recovery , the cell populations recovering continued to shift further along the pseudotime axis , and by week 10 , the cell distribution over pseudotime closely resembled that of untreated controls ( Figure 4f ) in agreement with immunostaining showing that anti-RANKL thymi at week 10 were indistinguishable from control thymi ( Khan et al . , 2014 ) . The close correlation of the in vivo recovery of the Aire-developmental branch and the ordering predicted by pseudotime provides strong validation of the pseudotime bioinformatic model . Finally , comparison of gene expression between the final timepoint and isotype control treated animals confirmed mTECs had fully recovered by week 10 ( Figure 4—figure supplement 2c ) . Thus , the pattern of mTEC recovery after transient anti-RANKL ablation provides in vivo support for the predictions made based on lineage tracing and bioinformatic approaches . Our data suggest the TAC-TEC population is a heterogeneous population consisting of a Ccl21a-expressing and Aire-expressing population . To further explore the heterogeneity of this population , we examined the cells that remain in the ‘TAC-TEC’ population following anti-RANKL antibody treatment . Although the ‘TAC-TEC’ population was partially depleted after treatment with anti-RANKL , a small number of ‘TAC-TECs’ remained . Intracellular flow cytometry revealed that there were fewer Ki67+ cycling cells ( p<0 . 0005; Figure 5a ) . Ki67+ cells were nearly entirely depleted in the AIRE-hi population ( p<0 . 0006 ) , a 20-fold reduction , but there was only a 3 . 5-fold reduction in the AIRE-lo population ( p<0 . 001 ) suggesting that the TAC-TEC population is not uniformly affected during anti-RANKL treatment ( Figure 5a ) . To better understand the ‘TAC-TECs’ that persisted during antibody treatment , we performed focused analysis on all ‘TAC-TECs’ collected in the study ( n = 511 cells; Figure 5b ) . The persisting ‘TAC-TEC’ population displayed the same proportion of cells in G2/M and showed similar expression patterns of chromatin-modifying factors ( Figure 5c and Figure 5—figure supplement 1a and b ) . Across the timecourse , the proportion of Ccl21a-expressing ‘TAC-TECs’ was increased at the expense of Aire-expressing ‘TAC-TECs’ at two and four weeks following treatment but frequencies were beginning to normalize by week 6 ( Figure 5d and Figure 5—figure supplement 1c ) . Although the expression of Ccl21a and Aire changed throughout the timecourse , the expression of Fezf2 was less significantly affected by treatment with anti-RANKL ( Figure 5d and Figure 5—figure supplement 1a ) . While the proportion of Ccl21a expressing cycling cells increased following treatment with anti-RANKL , the number of cells in the ‘TAC-TEC’ population that expressed only Ccl21a remained similar throughout the timecourse ( Figure 5—figure supplement 1d ) . The consistency observed in the Ccl21a-expressing cycling population following treatment with anti-RANKL agrees well with our finding that the absolute number of ‘Ccl21a-high’ mTECs did not change dramatically over the timecourse and shows that Aire-expressing and Ccl21a-expressing ‘TAC-TECs’ are not equally affected by anti-RANKL antibody treatment . To better understand population dynamics within the ‘TAC-TEC’ cluster , we repeated canonical correlation analysis between all samples using only these cells . Dimensionality reduction and clustering revealed three sub-populations ( Figure 5e ) . To understand the cells within each cluster , we performed differential expression analysis between the three clusters ( Figure 5—figure supplement 2 ) . Many of the genes differentially upregulated in cluster 1 , such as Calcb , Aire , H2-ab1 , were also expressed in the ‘Aire-positive’ population ( Figure 5—figure supplement 2 ) . The genes upregulated in cluster 0 , such as S100a14 , Calcb , Cdx1 , and Aire were also expressed in the ‘Aire-positive’ population ( Figure 5—figure supplement 2 ) . The genes upregulated in cluster 2 , including Ccl21a , Krt17 , Krt5 , and Id3 , were also most highly expressed in the ‘Ccl21a-high’ population ( Figure 5—figure supplement 2 ) . The frequency of cycling cells and levels of chromatin remodeling factors , including Hmgb2 , Hmgn2 , and H2afx , and expression levels of Fezf2 were similar between the three subsets ( Figure 5—figure supplement 1e and f ) . However , clusters 0 and 1 were found to have the highest levels of Aire and lowest levels of Ccl21a , whereas cluster 2 was found to have the lowest levels of Aire and highest levels of Ccl21a , suggesting an early stage of lineage commitment ( Figure 5f ) . Relative to other clusters within each treatment timepoint , the proportion of cells in cluster 2 increased at weeks 2 and 4 after treatment , consistent with analysis of Ccl21a- and Aire-expressing cells ( Figure 5d and g ) . By 6 weeks after treatment the abundance of Aire-expressing cluster 1 cells had largely normalized ( Figure 5g ) . Notably , despite changes in frequency , the transcriptomic profile of each ‘TAC-TEC’ cluster remained consistent after RANKL blockade ( Figure 5g ) . Taken together , while the ‘TAC-TEC’ cluster is defined by cycling cells and expression of chromatin remodeling factors , it can be further subsetted to reveal Aire or Ccl21a primed states . These results demonstrate that the sub-populations are present in all samples , indicating that Aire- or Ccl21a- primed cells are present in control thymi and treated thymi . Further , these results demonstrate that the Aire-primed cells are selectively ablated following treatment with anti-RANKL . Next , we assessed the dynamics of gene expression within the total Aire-positive population following anti-RANKL treatment ( Figure 6a ) . While the expression of housekeeping genes , including Gapdh , and the expression of Fezf2 did not change following anti-RANKL treatment , Aire expression was markedly reduced and with the lowest point at week four followed by recovery ( Figure 6b and Figure 6—figure supplements 1b and 2a ) . The recovery of Aire expression at week six was consistent with the pseudotime prediction that Aire expression peaks early in mTEC development ( Figure 2e ) . To explore the recovery of TSAs , we determined the average expression of genes in each of the gene lists described above ( Aire-dependent , Fezf2-dependent and total TSAs ) throughout recovery . Although the expression of all non-TSA genes ( all genes , excluding these three gene lists ) did not change across the timecourse , average expression of total TSAs and especially Aire-dependent genes decreased ( Figure 6c and Figure 6—figure supplement 2b ) . TSAs and Aire-dependent gene expression began to recover at week six but did not fully recover until week 10 ( Figure 6c ) . The pronounced delay between the recovery of Aire-expressing cells and Aire-dependent TSA expression was surprising . However , it supports the bioinformatic prediction that Aire expression must be well-established before the expression of TSAs . The expression of TSA and Aire-dependent genes could have decreased because the expression of each gene decreased , fewer total genes were expressed within each group , or fewer cells were present within the Aire-positive population . Therefore , we determined how many genes from each list were observed throughout the time course . Before making this calculation , we normalized the number of unique molecular identifiers ( UMIs ) in each sample to avoid the influence of gene dropout rates ( Figure 6—figure supplement 2c–f ) . The cumulative fraction of expressed total TSA and Aire-dependent genes decreased from over 75% in untreated controls to less than 20% following ablation , with the lowest fractional expression at week 4 ( Figure 6d and Figure 6—figure supplement 3b ) . Although Aire expression was mostly recovered by week 6 , Aire-dependent gene expression and TSA expression remained below 25% ( Figure 6d and Figure 6—figure supplement 3b ) . The average expression of only TSAs detected at week four following ablation did not change dramatically following treatment ( Figure 6—figure supplement 3c ) . The number of Fezf2-dependent genes also decreased from over 80% to 50% ( Figure 6d and Figure 6—figure supplement 3b ) . Based on this analysis , expression of TSAs and Aire-dependent genes decreased because fewer genes from each category are expressed at week four than in the controls . TSAs in mTECs were previously shown to be co-expressed as gene sets within individual cells in a somewhat ordered pattern ( Brennecke et al . , 2015 ) . To determine if co-expressed TSA gene sets were expressed at different times during mTEC development , we investigated the recovery of a subset of co-expressed gene sets defined by Brennecke et al . , 2015 . We determined the cumulative fraction of genes in each of the co-expression groups that are expressed throughout recovery following treatment with anti-RANKL ( Figure 6d ) . All co-expression gene groups had similar patterns to the Aire-dependent genes and all TSAs . For all co-expression groups , we detected the fewest genes at week four following treatment ( Figure 6d ) . Genes from the co-expression groups began to recover by week 6 , had fully recovered by week 10 , and all co-expression groups appeared to recover with similar kinetics ( Figure 6d ) . To determine if fewer genes were detected because the ‘Aire-positive’ population contained fewer cells following anti-RANKL treatment , we down-sampled cells from the week 10 control to the number of cells observed in each of the recovery samples . After down-sampling to equivalent cell numbers , we still observed fewer Aire-dependent and total TSA genes expressed after treatment ( Figure 6—figure supplement 3d ) . The total number of genes expressed per cell did not change over the course of recovery although the number of total TSA genes per cell , and especially Aire-dependent genes , did decrease ( Figure 6—figure supplement 3d ) . Overall , the gene-expression patterns observed during recovery of mTECs mirrored the gene-expression dynamics predicted by pseudotime . Plots were made using ggplot2 in R using our personal package that have been made publicly available . The full Snakemake pipeline , our personal analysis package , and scripts to recreate all figures are available on GitHub: https://github . com/kwells4/mtec_analysis ( Wells , 2020; copy archived at swh:1:rev:d3955fad0d73dc404a93fc2d81b84141c9c79efe ) with the companion R package available: https://github . com/kwells4/mtec . 10x . pipeline . AireCreERT2;Rosa26CAG-stopflox-zGgreen have been previously described ( Miller et al . , 2018 ) . C57BL/6 ( Jax #000664 ) and B6 . Cg-Gt ( ROSA ) 26Sortm6 ( CAG-ZsGreen1 ) Hze/J ( Jax #007906 ) mice were obtained from The Jackson Laboratory ( Madisen et al . , 2010 ) . Mice were maintained in the University of California San Francisco ( UCSF ) specific pathogen-free animal facility in accordance with the guidelines established by the Institutional Animal Care and Use Committee ( IACUC ) and Laboratory Animal Resource Center and all experimental procedures were approved by the Laboratory Animal Resource Center at UCSF . Age-matched female mice aged 4–14 weeks were used for all experiments unless otherwise specified in the text or figure legends . For Tamoxifen treatment of mice possessing conditional alleles , Tamoxifen ( Sigma-Aldrich ) was dissolved in corn oil ( Sigma-Aldrich ) and 2 mg doses were administered by oral gavage every other day for 10 days with flexible plastic feeding tubes ( Instech ) . Anti-RANKL antibody ( Clone IK22/5 , BioXCell ) or isotype control antibody ( clone 2A3 , BioXCell ) was administered to mice at a dose of 100 ug in PBS every other day via intraperitoneal ( i . p . ) injections for a total of 3 injections for all experiments . Mouse thymi were isolated , cleaned of fat and transferred to DMEM ( UCSF Cell Culture Facility ) containing 2% FBS ( Atlanta Biologics ) on ice . For each scRNA-seq sample , three thymi were pooled . Thymi were minced with a razor blade and tissue pieces were moved with a glass Pasteur pipette to 15 ml tubes and vortexed briefly in 10 ml of media . Fragments were allowed to settle before removing the media and replacing it with 4 ml of digestion media containing 2% FBS , 100 ug/ml DNase I ( Roche ) , and 100 ug/ml Liberase TM ( Sigma-Aldrich ) in DMEM . Tubes were moved to a 37°C water bath and fragments were triturated through a glass Pasteur pipette at 0 min and 6 min to mechanically aid digestion . At 12 min tubes were spun briefly to pellet undigested fragments and the supernatant was moved to 20 ml of 0 . 5% BSA ( Sigma-Aldrich ) , 2 mM EDTA ( TekNova ) , in PBS ( MACS buffer ) on ice to stop the enzymatic digestion . This was repeated twice for a total of three 12 min digestion cycles , or until there were no remaining tissue fragments . The single-cell suspension was then pelleted and washed once in MACS Buffer . Density-gradient centrifugation using a three-layer Percoll gradient ( GE Healthcare ) with specific gravities of 1 . 115 , 1 . 065 , and 1 . 0 was used to enrich for stromal cells . Cells isolated from the Percoll-light fraction , between the 1 . 065 and 1 . 0 layers , were then resuspended in 0 . 5% BSA ( Sigma-Aldrich ) , 2 mM EDTA ( TekNova ) ( FACS buffer ) and counted . Single-cell suspensions were prepared as described and incubated with Live/Dead Fixable Blue Dead Cell Stain ( ThermoFisher ) in PBS for 10 min at room temperature followed by blocking with anti-mouse CD16/CD32 ( 2 . 4G2 ) ( UCSF Hybridoma Core Facility ) and 5% normal rat serum for 10 min at room temperature . Cells were then washed in FACS buffer and stained for surface markers for 20 min at room temperature . For intracellular staining , cells were fixed and permeabilized using the eBioscience FoxP3 Transcription Factor Buffer Set ( ThermoFisher ) according to the manufacturer’s instructions . Cells were either sorted directly into DMEM ( ThermoFisher ) containing 10% FBS using a BD FACS Aria Fusion ( BD Biosciences ) or analyzed using a BD LSR II ( BD Biosciences ) housed within the UCSF Single Cell Analysis Center . Flow cytometry data was analyzed using BD FACSDiva v8 . 0 or FlowJo v10 . 5 . 3 software ( TreeStar Software ) . The following antibodies were used in this study: Ly51-PE ( 6C3 , BioLegend ) , CD11c-PE-Cy7 ( N418 , eBioscience ) , CD45-PerCP ( 30-F11 , BioLegend ) , EPCAM-APC-Cy7 ( G8 . 8 , BioLegend ) , Aire-e660 ( 5H12 , eBioscience ) , Ki67-PE ( eBioscience ) , CCL21 ( 59106 , R and D Systems ) , Goat anti-Rat IgG-A488 ( ThermoFisher ) . To prepare the cells for droplet-based sequencing , mTECs were flow-sorted into tubes containing 750 ul DMEM + 10%FBS ( 50 , 000 cells were sorted for the Aire trace experiment , 60 , 000 cells were sorted for the week two control , 37 , 000 cells were sorted for the week two experiment , 44 , 000 cells were sorted for the week four experiment , 32 , 000 cells were sorted for the week six experiment , 50 , 000 cells were sorted for the week 10 experiment , and 24 , 000 cells were sorted for the week 10 control ) . Cells were spun down at 300 g for 5 min and all but 50–100 ul of supernatant was removed to aim for a final concentration of 700–1000 cells/µl . An estimated 4000 single cells per sample were then subjected to 10x Genomics single-cell isolation and RNA-sequencing following the manufacturer’s recommendations . Illumina HiSeq 4000 ( Illuminia ) was used for deep sequencing . Sequences from scRNA-seq were processed using Cellranger v2 . 2 . 0 software ( Zheng et al . , 2017 ) . For the anti-RANKL experiment , sequences were processed using the cellranger mm10-1 . 2 . 0 genome and gtf file . For the Aire lineage tracing experiment , the sequence for the ZsGreen transcript was added to the fasta and the gtf file . A complete reference was made running cellranger mkref with the updated fasta and gtf files as arguments . For each sample , cell ranger mkfastq and cellranger count were run with the transcriptome argument pointing to the mm10 reference for the anti-RANKL experiment and the mm10 + ZsGreen reference for the lineage trace experiment . Raw data generated by Cellranger were then read into the Seurat ( Butler et al . , 2018 ) v2 . 3 . 4 R package with at least 200 genes per cell and at least 3 cells . Cells were further filtered based on the number of genes per cell ( between 200–7500 ) and the percent of mitochondrial reads per cell ( 0%–10% ) . The remaining cells and genes were used for downstream analysis . The data were normalized by using ‘LogNormalize’ method and data scaled with ‘scale . factor=1000’ from Seurat . For each sample , variable genes were found by using ‘FindVariableGenes’ with the following options mean . function = ExpMean , dispersion . function = LogVMR , x . low . cutoff = 0 . 0125 , x . high . cutoff = 3 , y . cutoff = 0 . 5 . The week 2 and week 10 control samples were initially analyzed as explained in the section ‘Initial analysis of scRNA-seq data’ . After initial processing , the two samples were merged using the MergeSeurat function . Variable genes were again found for the combined object as described above . The returned variable genes were used as the gene list given to the RunPCA function . Clusters were determined using the ‘FindClusters’ function with the following options reduction . type = ‘pca’ , dims_use = 1:13 , resolution = 0 . 6 , random . seed = 0 . Further dimensionality reduction was performed by using RunUMAP using the options reduction . type = ‘pca’ , dims_use = 1:13 . Markers of each cluster were found using the ‘FindMarkers’ command and highly similar clusters were merged . A small cluster of cells resembling t-cells based on gene expression were removed from further analysis . Clusters were classified based on similarity of marker genes to the populations in the Aire lineage tracing experiment . Differential gene expression was performed by running FindMarkers on all pair-wise clusters ( so that differentially expressed genes could be shared between clusters ) . Genes were called differentially expressed if the adjusted p-value was less than 0 . 05 and the log fold change was greater than 1 . Differentially expressed genes were filtered to include only transcription factors using a list of mouse transcription factors ( http://genome . gsc . riken . jp/TFdb/tf_list . html ) and chromatin-modifying factors ( H2afz , Hmgb1 , Hmgn1 , H2afx ) . These genes were manually searched for previous connections to mTEC development and Aire expression . Differentially expressed genes were then filtered to only include genes differentially upregulated in only the ‘TAC-TEC’ population , and not any other population . These genes were then compared to the list of genes differentially upregulated in cluster 5 and cluster 8 ( identified as transit-amplifying populations ) from Basak et al . , 2018 . To determine enrichment of these genes , we performed a hypergeometric test ( overlapping genes = 21 , genes in TAC-TEC = 35 , genes in transit-amplifing clusters = 173 , total genes = 20 , 309 ) . The same p-value was determined using a fisher’s exact test . Cell-cycle state was determined using cyclone from scran v1 . 10 . 1 which uses a list of known cell-cycle genes to determine each cell a score for G2/M and G1 phase . Pairs of genes that change between cell-cycle phases ( from Scialdone et al . , 2015 ) are provided to cyclone . Scores are based on which gene within each pair is more highly expressed in each cell . These scores are used to assign each cell to G2/M , G1 , or S phase . Single-cell velocity trajectory analysis was performed in python using scvelo v0 . 1 . 24 from the Theis lab . First , a loom file were generated using run10x from velocyto ( La Manno et al . , 2018 ) . The loom file was than loaded into python . Normalization and filtering was done using filter_and_normalize with the options min_counts = 20 , min_counts_u = 10 , n_top_genes = 3000 . Next the moments were calculated using 30 PCs and 30 neighbors . Finally , velocity and the velocity graph were calculated . UMAP coordinates and cluster identification from the Seurat object were then added to the RNA velocity object . The velocity was plotted on the UMAP coordinates and colored by the cluster id . The scvelo output was used to determine lineage relationships and common transit-amplifying populations . Plots of RNA velocity were made using the command pl . velocity_embedding_stream from scvelo . Pseudotime scores were determined using slingshot ( Street et al . , 2018 ) v1 . 1 . 0 with the ‘TAC-TEC’ cluster as the argument for start . clus . These pseudotime scores were used as the pseudotime values in Figure 2E . TSA genes were determined using the Fantom database and filtering out genes that were seen in five tissues or less ( Forrest et al . , 2014 ) . Genes were further filtered to only keep protein-coding genes as described previously ( Brennecke et al . , 2015 ) . Aire-dependent genes were defined based on the 2014 Sansom et al . list of Aire-dependent genes ( Sansom et al . , 2014 ) . Fezf2 dependent genes were identified by determining genes with a 2-fold enrichment in the control mouse vs . the Fezf2 knockout mouse ( Takaba et al . , 2015 ) . The percent of TSAs per cell was calculated by determining the number of TSAs expressed in a given cell and dividing by the total number of TSAs . TSA expression was averaged within each cell by dividing the total expression of TSAs by the total number of TSAs or the total number of expressed TSAs . Slingshot analysis was repeated after regenerating the dimensionally reduced object without any TSA genes included . For the Aire lineage tracing sample , the returned variable genes were used as the gene list given to the RunPCA function . Clusters were determined using the ‘FindClusters’ function with the following options reduction . type = ‘pca’ , dims_use = 1:12 , resolution = 0 . 6 , random . seed = 0 . FindClusters embeds cells in a k-nearest neighbor graph with edges drawn between cells with similar gene expression . These are these partitioned into highly inter connected communities ased on Euclidean distance in PCA space . Further dimensionality reduction was performed by using RunUMAP using the options reduction . type = ‘pca’ , dims_use = 1:12 . Markers of each cluster were found using the ‘FindMarkers’ command and highly similar clusters were merged . A small cluster of cells resembling t-cells based on gene expression were removed from further analysis . Clusters were classified based on common marker genes previously described for each mTEC population . Clusters highly expressing Aire were further characterized by their expression of ZsGreen . This analysis was repeated after cell-cycle regression to ensure conclusions were robust ( Data not shown ) . Double positive percents were found by using a cutoff of 0 for ZsGreen , Aire , Ackr4 , Trpm5 , and Fezf2 , a cutoff of 2 for Krt10 and a cutoff of 4 for Ccl21a . Cell-cycle state was again determined using cyclone from scran v1 . 10 . 1 . All results in the controls replicated observations made from the control mice ( Figure 3—figure supplement 1 ) . All samples were processed as described in the ‘Initial analysis of scRNA-seq data’ . The top 1000 most highly variable genes from each sample were merged , keeping only genes that were present in all samples were used for CCA analysis . The 7 Seurat objects and the variable genes found above were used to generate a new Seurat object with the RunMiltiCCA function , using num . ccs = 30 . The canonical correlation strength was calculated using num . dims = 1:30 and the samples were aligned using dims . align = 1:20 . Clusters were found using the ‘FindClusters’ function based on the aligned CCA with the following options reduction . type = ( ‘cca . aligned’ , resolution = 0 . 6 , dims . use = 1:20 , random . seed = 0 ) . Further dimensionality reduction was performed by using RunUMAP using the options reduction . type = ‘cca . aligned’ , dims_use = 1:20 , seed . use = 0 . The cluster identities were assigned based on the Aire trace experiment . The Aire trace cluster identities were added to the Seurat object containing all samples . For each cluster , the identity was assigned based on which Aire trace population it shared the most cells with ( Figure 4b ) . There was a clear majority of cells aligning with the Aire trace populations for all clusters . Population assignments were checked based on gene expression of marker genes such as Ackr4 , Ccl21a , Aire , Krt10 , and Trpm5 by plotting expression of marker genes on the UMAP dimensional reduction ( Figure 4—figure supplement 1 ) , and by plotting heatmaps of marker genes separated by population for each sample ( Figure 4—figure supplement 3 ) . To ensure that the cells mapping to the ‘Aire-positive’ cluster were correctly mapping to the cluster we looked at expression of the tope 30 genetic markers of the ‘Aire-positive’ cluster ( using FindMarkers in the wild-type samples ) ( Figure 6—figure supplement 1 ) . A small cluster of cells resembling t-cells based on gene expression were removed from further analysis . Cell-cycle state was again determined using cyclone from scran . Pseudotime scores were determined using slingshot ( Street et al . , 2018 ) v1 . 1 . 0 with the ‘TAC-TEC’ cluster as the argument for start . clus . These pseudotime scores were used as the pseudotime values in Figure 4E . The density of cells across psuedotime for each sample was plotted using ggplot2 . Cells in the ‘TAC-TEC’ population for each sample were subset from the Seurat of all samples using SubsetData with subset . raw = TRUE . Data from each seven new Seurat objects were normalized by using ‘LogNormalize’ method and data scaled with ‘scale . factor=1000’ from Seurat . For each sample , variable genes were found by using ‘FindVariableGenes’ with the following options mean . function = ExpMean , dispersion . function = LogVMR , x . low . cutoff = 0 . 0125 , x . high . cutoff = 3 , y . cutoff = 0 . 5 . The top 1000 most highly variable genes from each sample were merged , keeping only genes that were present in all samples were used for CCA analysis . The 7 Seurat objects and the variable genes found above were used to generate a new Seurat object with the RunMiltiCCA function , using num . ccs = 10 . The canonical correlation strength were calculated using num . dims = 1:10 and the samples were aligned using dims . align = 1:10 . Clusters were found using the ‘FindClusters’ function based on the aligned CCA with the following options reduction . type = ( ‘cca . aligned’ , resolution = 0 . 7 , dims . use = 1:7 , random . seed = 0 ) . Further dimensionality reduction was performed by using RunUMAP using the options reduction . type = ‘cca . aligned’ , dims_use = 1:10 , seed . use = 0 . Differential gene expression was performed by running FindAllMarkers . Genes were called differentially expressed if the adjusted p-value was less than 0 . 05 and the log fold change was greater than 0 . 5 . A Seurat object containing only cells from the ‘Aire-positive’ population was created using SubsetData . UMIs for each sample were down-sampled to the number of UMIs in the week four sample using DropletUtils v1 . 2 . 1 downsampleMatrix ( Griffiths et al . , 2018; Lun et al . , 2019 ) . Effectiveness of the down-sampling was determined by observing the number of dropouts of housekeeping genes before and after the down-sampling ( Figure 6—figure supplement 3d ) . The percent of dropouts was computed by dividing the number of cells with expression values of 0 for each housing keeping gene ( Chmp2a , Emc7 , Psmb4 , Vcp , and Gapdh ) by the total number of cells for each sample . This was calculated before and after down-sampling UMIs for each sample . Cumulative fraction of genes was determined for TSAs , Aire-dependent genes , Fezf2-dependent genes , and all other genes ( genes not listed as TSAs , Aire-dependent , or Fezf2-dependent ) . For each gene list , a gene was counted as ‘expressed’ if at least one cell in a sample had an expression value greater than 0 . The percent of the gene list expressed was determined by dividing the total number of expressed genes by the total number of genes in the list . The cumulative fraction of genes detected was also determined for the control samples before downsampling UMIs to determine the number of genes detected in these samples ( Figure 6—figure supplement 3A ) . To determine the number of genes expressed per cell from each list , a gene was again counted as ‘expressed’ if its expression was greater than 0 for the given cell . To determine the number of genes from each gene set expected for a specific number of cells from the 10 week control ‘Aire-positive’ population , cell_sampler from dropseqr ( https://rdrr . io/github/argschwind/dropseqr/ ) was used with ncells set to the number of cells in the ‘Aire-positive’ population in each sample . For each subsample of the control , the percent of observed genes from each gene set was calculated 1000 times . The percent values were used to create a distribution for each sample . The actual observed percent of the gene set was plotted as a red line and a p-value was determined using a z-score using the mean and standard deviation from the distribution of subsampled cells from the control ( Figure 6—figure supplement 3D ) . The list of Aire-dependent genes was originally determined by comparing Aire knockout to WT mTECs ( Sansom et al . , 2014 ) . Fezf2-dependent genes were determined by performing differential expression of a Fezf2 knockout to WT mTECs from a published dataset using a adjusted p-value cutoff of 0 . 05 and a log fold change value of 1 ( Takaba et al . , 2015 ) . TSA genes were found using the Fantom database and curated as previously described ( Brennecke et al . , 2015; Forrest et al . , 2014 ) . Genes were called a TSA if they were observed in five tissues or less . All other genes were determined by taking all protein-coding genes and subtracting out genes from the Aire-dependent , Fezf2-dependent , and TSA gene lists .	Specialized cells in the immune system known as T cells protect the body from infection by destroying disease-causing microbes , such as bacteria or viruses . T cells use proteins on their surface called receptors to stick to infectious microbes and remove them from the body . Some newly developed T-cells , however , contain receptors that recognize and bind to cells that belong in the body . If these faulty T cells are released , they can attack healthy tissues and cause an autoimmune disease . After a new T cell is developed , it gets carried to a gland in the chest known as the thymus . Cells in the thymus called mTECs screen T cells for receptors that may bind to the body’s tissues . mTECs do this by presenting T cells with proteins that are commonly found on the surface of healthy cells in the body . If a T cell recognizes any of these ‘tissue specific proteins’ , it is destroyed or given a new role in the body . Some faulty T cells , however , still manage to evade detection . One way to uncover why this might happen is to investigate how mTECs develop . Previous work showed that mTECs transition through various stages before reaching their final form . However , the order in which these events occur remained unclear . To gain a better understanding of these developmental steps , Wells , Miller et al . extracted mTECs from the thymus of mice and analyzed the genetic make-up of individual cells . This uncovered a missing link in mTEC development: a new type of cell that is the immediate predecessor of the final mTEC . These ‘predecessor’ cells were actively growing , highlighting that mTECs can be constantly generated in the body . By probing the genes that generate tissue-specific proteins in mTECs , Wells , Miller et al . revealed that these proteins were only produced for short periods and in the late stages of mTEC development . These findings contribute to our understanding of how mTECs develop to screen T cells . Mapping these developmental stages will make it easier to identify when faulty T cells are able to evade mTECs . This will lead to earlier detection of autoimmune diseases which could result in better treatments .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation", "genetics", "and", "genomics" ]	2020	Combined transient ablation and single-cell RNA-sequencing reveals the development of medullary thymic epithelial cells
Organogenesis depends on orchestrated interactions between individual cells and morphogenetically relevant cues at the tissue level . This is true for the heart , whose function critically relies on well-ordered communication between neighboring cells , which is established and fine-tuned during embryonic development . For an integrated understanding of the development of structure and function , we need to move from isolated snap-shot observations of either microscopic or macroscopic parameters to simultaneous and , ideally continuous , cell-to-organ scale imaging . We introduce cell-accurate three-dimensional Ca2+-mapping of all cells in the entire electro-mechanically uncoupled heart during the looping stage of live embryonic zebrafish , using high-speed light sheet microscopy and tailored image processing and analysis . We show how myocardial region-specific heterogeneity in cell function emerges during early development and how structural patterning goes hand-in-hand with functional maturation of the entire heart . Our method opens the way to systematic , scale-bridging , in vivo studies of vertebrate organogenesis by cell-accurate structure-function mapping across entire organs . Organogenesis builds on cell–cell interactions that shape tissue properties , and tissue-level cues that control maturation of cell structure and function . During cardiogenesis , region-specific heterogeneity in cellular activity patterns evolves as the heart undergoes large-scale morphological changes: The spontaneously active heart tube develops into the mature heart , in which pacemaker cells near the inflow site initiate the rhythmic excitation that spreads with differential velocities through distinct regions of the myocardium . This controlled cardiac activation gives rise to an orderly sequence of atrial and ventricular calcium release and contraction . An integrated understanding of cardiogenesis at the systems level requires simultaneous cell and organ scale imaging , under physiological conditions in vivo . Here , we present a high-speed light sheet microscopy and data analysis pipeline to measure fluorescent reporters of cardiomyocyte location and activity across the entire electro-mechanically uncoupled heart in living zebrafish embryos during the crucial looping period from 36 to 52 hr post fertilization ( hpf ) . By noninvasively reconstructing the maturation process of the myocardium in its entirety at cellular resolution , our approach offers an integrative perspective on tissue and cell levels simultaneously , which has previously required separate experimental setups and specimens . Our method opens a new way towards systematic assessment of the mutual interrelations between cell- and tissue properties during organogenesis . The zebrafish is an appealing vertebrate model system with a simple , yet functionally conserved heart . Light sheet microscopy has proven to be supremely suited for obtaining in vivo recordings of the intact embryonic zebrafish heart ( Chi et al . , 2008; Scherz et al . , 2008; Arnaout et al . , 2007; Trivedi et al . , 2015 ) . Whole cardiac cycles have been reconstructed in 4D ( 3D + time ) using post-acquisition synchronization of high-speed light sheet movies in a z-stack . The resulting effective temporal resolution of about 400 volumes per second ( Mickoleit et al . , 2014 ) is unmatched by other in vivo volumetric imaging techniques such as light sheet microscopy with electrically focus-tunable lenses or swept , confocally-aligned planar excitation ( Bouchard et al . , 2015; Fahrbach et al . , 2013; Hou et al . , 2014; Liebling et al . , 2005 ) . We built a light sheet microscope tailored for high-speed imaging of the heart in the living zebrafish embryo . By fine-tuning the magnification and restricting camera readout to the center area of the chip , we balanced the field of view and the spatial and temporal sampling to record cardiac activation in the entire heart with cellular precision ( Materials and methods ) . We investigated whether post-acquisition synchronization could be extended to visualizing calcium transients in cardiac myocytes across the entire heart of living embryonic zebrafish expressing the fluorescent calcium reporter GCaMP5G under the myl7 promoter ( Figure 1a , Figure 1—figure supplement 1 ) . The genetically expressed calcium reporter provides a specific , consistent and non-invasive readout of cardiomyocyte activity in vivo ( Figure 1b , Videos 1 and 2 ) . In a side-by-side comparison , the calcium signal had good and stable fluorescent yield at low excitation power , superior to genetically expressed voltage reporters . Importantly , the calcium signal faithfully reports presence and timing of cell activation ( Figure 1—figure supplement 2 ) ( Kralj et al . , 2011 ) . To prevent interference of tissue movement and deformation with observed signals , we decoupled electrical excitation and mechanical contraction by inhibiting the formation of the calcium-sensitive regulatory complex within sarcomeres , using a morpholino against tnnt2a ( Materials and methods ) . By mounting zebrafish embryos in low concentration agarose inside polymer tubes , we could position the embryos for precise optical investigation without anesthesia ( Figure 1—figure supplement 1a , b ) . To attribute calcium dynamics to individual cardiomyocytes , we also recorded a fluorescent nuclear marker ( myl7:H2A-mCherry ) . The high temporal ( 400 Hz ) and spatial sampling ( 0 . 5 µm pixel size ) was adequate for computing normalized average calcium transients throughout the cardiac cycle for each cell across the entire heart ( Figure 1—figure supplement 3 , Video 3 , Materials and methods ) . To map cellular activation timings onto a 3D structural representation of the heart , we identified every single cell’s activity ( Figure 1c’ ) and quantified dynamic characteristics of all cardiac myocytes across space and time . The distribution of calcium transient rise times ( from 10% to 90% of calcium transient peak amplitude ) revealed the emergence of distinct upstroke characteristics in different locations within the 3D network ( by 52 hpf ) . Rise times were shortest for atrial muscle cells , intermediate in the atrio-ventricular canal ( AVC ) , and longest for ventricular cardiomyoctes ( Figure 1c’’ ) , in keeping with higher vertebrates where atrial myocyte contraction is faster than that of ventricular cells ( Brandenburg et al . , 2016 ) . To get a more quantitative understanding of the 3D distribution of activity patterns , a canonical description of cell locations was needed . We traced and parameterized the myocardium’s centerline ( Figure 1—figure supplement 4a , Materials and methods ) , along which we assigned a unique position to each cell between inflow and outflow . We identified a positive correlation between rise time and position of cells along the midline , with a clear discontinuity at the AVC ( Figure 1—figure supplement 4b ) , illustrating emergence of chamber-specific patterns of individual cell activation properties across the heart . Next , we studied the spatial patterns of sequential cell activation , as a read-out for the speed of electrical conduction across the heart . While cardiac activation in the early linear heart tube is slow and near-uniform , the chambered heart shows areas of elevated conduction velocity ( Dehaan , 1961; de Jong et al . , 1992; Moorman et al . , 1998; Chi et al . , 2008; Panáková et al . , 2010 ) . Common imaging-based methods for determining cell conduction speed tend to deliver only a metric , or ‘biophysical speed’ of conduction ( distance over time ) . To reflect biological progression of activation between cells ( of potentially different or – in contracting tissue – dynamically changing size ) , we assessed local cell topology across the entire heart ( Figure 1d , Figure 1—figure supplement 5 , Materials and methods ) to also calculate the ‘biological speed’ ( number of activated cells over time ) of conduction ( Video 4 ) . Our analysis revealed that biophysical and biological conduction velocities show differences between and within anatomical regions of the heart ( Figure 1—figure supplement 6 ) . Either descriptor identifies particularly slow conduction between the most proximal atrial cells and between the cells of the AVC , while faster conduction is seen among working myocardial cells of the atrium and ventricle . In the atrium , there is a bias towards higher biophysical speeds , due to larger cell dimensions ( Figure 1d , Figure 1—figure supplement 4c ) . The activation sequence in the pacemaker region is of particular importance to heart physiology and function , yet individual cells that give rise to earliest activation were difficult to identify with previous methods ( Van Mierop , 1967; Arrenberg et al . , 2010; Christoffels et al . , 2010; Tessadori et al . , 2012 ) . We found that , at 52 hpf , less than 10 cells per heart serve as activation origins . They are located in the sinus venosus at the heart’s inflow side , which is a homologue of the primary cardiac pacemaker region in adult heart ( Poon and Brand , 2013 ) . Our data further show that the ring-like arrangement of pacemaker cells ( Figure 1—figure supplement 7 ) , together with a preferential orientation of myocardial cells in that region perpendicular to the inflow-outflow direction ( Auman et al . , 2007 ) ( Figure 1—figure supplement 8 ) , generates the initial planar ‘ring-like’ activation front that propagates evenly into the atrium . To visualize the 3D conduction pattern , the heart can be represented as a curved cylinder with varying diameter and s-shaped deformations in the coronal and in the sagittal plane . We used the position along the midline ( τ ) and the local Frenet-Serret frame as intrinsic coordinates ( ɸ , z ) ( Figure 1e and Video 5 , Materials and methods ) . Interestingly , by following the orientation of this reference system along the midline , we noticed torsion associated with the cardiac looping , which is most pronounced around the AVC ( Männer , 2000; Christoffels et al . , 2000; Harvey , 2002 ) . Establishing the intrinsic coordinates of the curved cylinder allows straightening and untwisting by implicitly removing the morphological torsion . Neglecting the actual distance to the midline and plotting the position along the midline ( τ ) against the angle ( ɸ ) , we obtained a 2D projection ( Figure 1e , f and Video 5 , Materials and methods ) , in which the two outer curvature regions of atrium and ventricle are located side-by-side . A clear asymmetry in conduction speeds between cells at the inner and the outer curvatures was apparent in this representation ( Figure 1f ) . Irrespective of these differences , however , the activation wave travelled smoothly in a ring-like fashion along the heart , as indicated by the isochronal lines in the cylindrical projection ( Figure 1—figure supplement 9 ) . In order to document how heterogeneity in cardiac function arises during cardiac looping , we extended our 3D optical mapping towards earlier developmental stages . During the crucial period between 36 and 52 hpf , ventricular cell number increased by about 45% , the initial heart tube developed into a two-chambered organ , the midline of the heart became increasingly curved and twisted ( Figure 2—figure supplement 1 ) , and the activation frequency increased – all signs of organ maturation . In spite of a net increase in cell numbers , the time required for activation to propagate from inflow to outflow decreased . A progressive crowding of calcium transient activation dynamics indicated maturation of cells , with two groups differentiating from the early more homogeneous pattern: working cardiomyocytes in atrium and ventricle ( Figure 2a–f ) . At 36 hpf , activation propagated evenly across the myocardium , compatible with peristalsis . Subsequently , calcium dynamics became increasingly structured . By 52 hpf , propagation of activation was fast across atrium and ventricle , while it remained slow in the AVC ( Figure 2—figure supplement 2a ) . With organ maturation during these 16 hr , cells in the ventricular part of the network showed longer calcium transient rise time ( Figure 2—figure supplement 2b ) but faster inter-cellular spread of activation ( Figure 2g–l , Figure 2—figure supplement 2c ) . Conduction also changed within chambers , such as along the outer curvature in the atrium . Increasing deformation and twisting of the cardiac tissue was associated with changes in cell shape ( cf . Figure 2d–f ) and in conduction ( Figure 2—figure supplement 3 ) . By noninvasively reconstructing the maturation process of the myocardium in its entirety at cellular resolution , our approach offers an integrative perspective on tissue and cell levels simultaneously . We show functional maturation in line with structural patterning of the heart muscle during development: starting from similar initial states , functionally distinct characteristics of calcium transients and conduction properties develop with a highly reproducible pattern relative to the cell locations ( cf . Figure 1—figure supplement 8 ) . Myocardial cells in the two chambers remodel and specialize into functional tissue of working atrial and ventricular cardiomyocytes , while cells in the pacemaker and AVC regions continue to resemble the earlier phenotype from the tubular stage . We demonstrate that myocardial activity can be recorded and analyzed with cellular detail across the entire embryonic heart . Future technological advancements can extend the scope of our approach: First , genetically expressed voltage reporters with improved dynamic range may provide a direct readout of myocardial electrical activity . Second , cameras with higher speed and sensitivity would enhance the recording frequency of rapid volume scanning , needed to explore aberrant myocardial activation during arrhythmias . Third , the integration of an algorithm capable of tracking cells in 3D during cardiac contractions would allow investigations in fully functional hearts . Fourth , the addition of optically gated actuators , such as light-activated ion channels or photo-pharmacological probes , would enable contact-free stimulation to probe the roles of individual cells or groups of cells in pacemaking , conduction , and arrhythmogenesis . Our work further highlights the value of the zebrafish as a vertebrate model system for in vivo cardiology , especially when combined with high-speed light sheet microscopy and suitable data analysis pipelines . It opens the way to systematic , scale-bridging , in vivo studies of organogenesis by facilitating cell-accurate measurements across entire organs . Zebrafish ( Danio rerio ) were kept at 28 . 5°C and handled according to established protocols ( Nusslein-Volhard and Dahm , 2002 ) and in accordance with EU directive 2011/63/EU as well as the German Animal Welfare Act . Transgenic zebrafish lines Tg ( myl7:GCaMP5G-Arch ( D95N ) ) ( Hou et al . , 2014 ) , Tg ( myl7:H2A-mCherry ) ( Schumacher et al . , 2013 ) and Tg ( myl7:lck-EGFP ) md71 were used . The lck sequence was PCR amplified from pN1-Lck-GCaMP3 ( Addgene , #26974 ) with In-Fusion primers 5’-GCAAAAGATCTGCCACCATGGGCTGTGGCTGC-3’ ( forward ) and 5’-GCAAAGGGCCCCGAGATCCTTATCGTCATCGT-3’ ( reverse ) designed with http://bioinfo . clontech . com/infusion/convertPcrPrimersInit . do and cloned into pEGFP-N1 ( Clontech , #6085–1 ) . PCR product generated from attB-flanked BP primers 5’-GGGGACAAGTTTGTACAAAAAAGCAGGCTGGATGGGCTGTGGCTGCAGCTCAAACC-3’ ( forward ) and 5’-GGGGACCACTTTGTACAAGAAAGCTGGGTCTTACTTGTACAGCTCGTCCATGCCGAG-3’ ( reverse ) was BP Clonase II cloned into Gateway pDONR221 ( ThermoFisher Scientific , #12536017 ) to generate the middle entry clone that was further assembled with p5E_myl7 , p3E_SV40polyA ( Tol2kit #302 ) , and pDEST . Cryst . YFP76 ( Mosimann et al . , 2015 ) into Tol2 transgene plasmid using MultiSite Gateway assembly . See supplementary files for detailed digital plasmid maps of these vectors . To generate Tg ( myl7:lck-EGFP ) md71 Tol2-mediated zebrafish transgenesis was performed by injecting 25 ng/ml transgene plasmid together with 25 ng/ml capped Tol2 transposase mRNA , followed by subsequent screening of positive F0 founders . During the 1 cell stage , embryos were injected with morpholinos against tnnt2a to uncouple electrical and mechanical activity ( Sehnert et al . , 2002 ) . Before imaging , fluorescent embryos were selected for absence of cardiac malformations and contractions , using an Olympus stereomicroscope equipped with an LED for transmitted light microscopy and a metal-halide light source and filter sets that match the excitation and emission spectra of GCaMP5G and mCherry for fluorescence excitation . Embryos were mounted in either 0 . 1 or 1 . 5% low gelling temperature agarose ( Sigma A9414 ) inside cleaned polymer tubes ( FEP tubing , inner/outer diameter 0 . 8/1 . 6 mm , BOLA S1815-04 ) . We built a light sheet microscope for in vivo cardiac imaging in zebrafish embryos , based on a previously published design ( Mickoleit et al . , 2014 ) . Imaging was performed in live zebrafish embryos between 36 and 52 hr post-fertilization ( hpf ) at a temperature of 24°C . Heart rate at this temperature is 2 Hz , about 0 . 5 Hz lower than at the temperature recommended for breeding of 28 . 5°C ( Baker et al . , 1997; Kopp et al . , 2005 ) . Embryos were kept in a custom imaging chamber filled with E3 fish medium and illuminated with a static light sheet generated from Coherent Sapphire LP lasers ( 488 and 561 nm ) using a cylindrical lens and a Zeiss 10x/0 . 2 air illumination objective . Laser power was kept at or below 2 mW in the field of view ( measured at the back aperture of the illumination objective ) to exclude thermal effects on heart rate ( an increase in heart rate was detected at laser powers of 5 mW and above ) . Fluorescence was collected and recorded using a Zeiss W Plan-Apochromat 20x/1 . 0 objective , a Zeiss 0 . 63x camera adapter , a Hamamatsu W-View image splitter and a Hamamatsu Flash 4 . 0 v2 sCMOS camera . Embryos were held in place by a Zeiss Lightsheet Z . 1 sample holder and oriented using motorized translation and rotation stages ( Physik Instrumente GmbH , Karlsruhe , Germany ) . For imaging of GCaMP5G , a z-stack of movies covering the entire heart was recorded by moving each embryo through the light sheet ( 488 nm excitation , band-pass 525/50 nm emission filter , 2 . 5 ms exposure time = 400 Hz , 600 frames = 1 . 5 s/movie , 1 µm z-steps ) . To ensure an efficient recording and the best possible synchronization ( see below ) , the embryo was rotated by about 30 degrees , such that both atrium and ventricle were visible in the imaging plane for the majority of the z-stack . For imaging of H2A-mCherry , a matching z-stack was recorded immediately afterwards ( 561 nm excitation , long-pass 565 nm emission filter , 20 ms exposure time = 50 Hz , 1 µm z-steps ) . Image acquisition was controlled by a custom program written in LabView ( National Instruments ) . Images were streamed onto a RAID-0 array of four 512 GB solid-state drives . Transgenic zebrafish embryos expressing myl7:H2A-mCherry and myl7:lck-eGFP in the myocardium were screened for absence of cardiac malformations and contractions using an Olympus stereomicroscope at 2 dpf . Selected embryos were mounted in 1 . 5% low-melting point agarose ( Sigma A9414 ) inside glass capillaries using plungers ( Brand 20 μl Transferpettor caps and piston rod ) . After a few minutes , mounted embryos were carefully transferred onto a custom 3D-printed sample holder with their heart facing up . Once positioned , they were fixed in place using drops of agarose at both ends of the agarose column . The sample holder was placed in a 52 mm plastic dish and covered with E3 fish medium . Confocal and two-photon microscopy was performed with an upright Zeiss LSM 780 NLO equipped with a Zeiss W Plan-Apochromat 20x/1 . 0 objective lens , a Coherent Chameleon multiphoton laser set to 920 nm , a Coherent HeNe 594 nm laser and gallium arsenide phosphide ( GaAsP ) detectors . The fluorescence reporter myl7:H2A-mCherry was recorded using single-photon excitation at 594 nm , a descanned GaAsP detector and a confocal pinhole set to one airy unit . Two-photon excitation at 920 nm and a non-descanned GaAsP detector were used for the fluorescence reporter myl7:lck-eGFP . The pixel size was 0 . 148 µm2 and z-stacks were recorded with a 1 µm step size . The total acquisition time of one z-stack was about 60 min .	The heart has a built-in pacemaker that sets the rhythm of the heartbeat . Pacemaker cells produce electrical signals that spread across the heart in a coordinated wave . As each cell receives its signal , ion channels open in its membrane . Calcium ions rush in from the spaces around the cells , triggering the release of more calcium ions from internal stores . The rise in calcium ion levels causes the heart muscle to contract . Standard techniques for studying how the activation process spreads across the heart typically involve removing the organ from the animal . One reason for this is that no microscopy technique had been able to provide the detail needed to observe the activity of individual cells across the whole heart during its activation cycle . Zebrafish embryos have a simple heart with two chambers that can be visually explored because the embryos are transparent . Their hearts are activated in a pattern that has been maintained throughout evolution with principal similarities in many different species . These properties make fish embryos well suited for the non-invasive examination of the heart . Weber , Scherf et al . have studied genetically engineered zebrafish embryos whose heart muscle cells contained a calcium-sensitive fluorophore , using a technique called light sheet microscopy . This method illuminates the heart with a thin sheet of laser light , which causes the fluorescent dye to glow in a way that indicates changes in the concentration of calcium ions in the cells . A fast and sensitive camera detects these signals and stacks of movies are recorded and synchronized , allowing cardiac activation to be mapped in three dimensions as it spreads across the heart . Applying this new technique revealed that different parts of the heart conduct activation signals at different speeds . These speeds finely match the anatomical features of the heart , yielding planar progression of the activation signal over the increasingly complex shape of the developing heart . Weber , Scherf et al . also showed that the heart only requires a handful of pacemaker cells to reliably set the heart’s rhythm . Future modifications to the technique of Weber , Scherf et al . could help us investigate how the heart works in even finer detail . For example , it might reveal how electrical activity , calcium handling , and contraction influence one another , and how they individually and collectively respond to drug treatments . This will help us understand how the normal heart rhythm develops , how it can be modified , and how the heart adapts to changes in its environment , including damage during cardiac disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "structural", "biology", "and", "molecular", "biophysics", "tools", "and", "resources" ]	2017	Cell-accurate optical mapping across the entire developing heart
Cortical dynein generates pulling forces via microtubule ( MT ) end capture-shrinkage and lateral MT sliding mechanisms . In Saccharomyces cerevisiae , the dynein attachment molecule Num1 interacts with endoplasmic reticulum ( ER ) and mitochondria to facilitate spindle positioning across the mother-bud neck , but direct evidence for how these cortical contacts regulate dynein-dependent pulling forces is lacking . We show that loss of Scs2/Scs22 , ER tethering proteins , resulted in defective Num1 distribution and loss of dynein-dependent MT sliding , the hallmark of dynein function . Cells lacking Scs2/Scs22 performed spindle positioning via MT end capture-shrinkage mechanism , requiring dynein anchorage to an ER- and mitochondria-independent population of Num1 , dynein motor activity , and CAP-Gly domain of dynactin Nip100/p150Glued subunit . Additionally , a CAAX-targeted Num1 rescued loss of lateral patches and MT sliding in the absence of Scs2/Scs22 . These results reveal distinct populations of Num1 and underline the importance of their spatial distribution as a critical factor for regulating dynein pulling force . Proper positioning of the mitotic spindle is essential for successful cell division and is crucial for a wide range of processes including creation of cellular diversity during development , maintenance of adult tissue homeostasis , and balancing self-renewal and differentiation in progenitor stem cells ( Galli and van den Heuvel , 2008; Gómez-López et al . , 2014; Morin and Bellaïche , 2011; Siller and Doe , 2009 ) . In various cell types , spindle positioning involves attachment of the minus end-directed MT motor cytoplasmic dynein to the cell cortex , where it exerts pulling force on astral MTs that emanate from the spindle poles ( di Pietro et al . , 2016; Kotak and Gönczy , 2013; McNally , 2013 ) . While proteins involved in anchoring dynein have been identified ( Ananthanarayanan , 2016; Couwenbergs et al . , 2007; Du and Macara , 2004; Heil-Chapdelaine et al . , 2000; Kotak et al . , 2012; Nguyen-Ngoc et al . , 2007; Saito et al . , 2006; Thankachan et al . , 2017 ) and the mechanism whereby dynein steps along the MT is becoming elucidated ( DeSantis et al . , 2017; DeWitt et al . , 2015; Grotjahn et al . , 2018; Nicholas et al . , 2015; Urnavicius et al . , 2018 ) , how pulling forces are precisely regulated to achieve the appropriate spindle displacement remains incompletely understood . The budding yeast Saccharomyces cerevisiae provides an important model for studying spindle position regulation [for review see ( Xiang , 2017 ) ] . During metaphase , the yeast spindle moves into the mother bud neck via dynein-dependent sliding of astral MT along the bud cortex ( Adames and Cooper , 2000; Moore et al . , 2009; Yeh et al . , 2000 ) . In the current model , dynein is recruited from the dynamic plus ends of astral MTs to cortical foci containing the attachment molecule Num1; once anchored , dynein uses its minus end-directed motor activity to walk along the MT lattice , generating pulling forces on astral MTs along the bud cortex , thereby moving the connected spindle into the bud neck ( Lee et al . , 2005 , 2003; Markus et al . , 2011; Sheeman et al . , 2003 ) . In contrast to the yeast model , studies in C . elegans embryos and mammalian cells show that cortically anchored dynein is able to mediate spindle movement by pulling on astral MTs in an apparent ‘end-on’ fashion ( Guild et al . , 2017; Gusnowski and Srayko , 2011; Kiyomitsu and Cheeseman , 2012; Nguyen-Ngoc et al . , 2007; Redemann et al . , 2010; Schmidt et al . , 2017 ) . Indeed , in vitro reconstitution studies using either bead-bound brain dynein or barrier-attached yeast dynein show that dynein can capture dynamic MT plus ends and generate pulling force on the captured MT ( Hendricks et al . , 2012; Laan et al . , 2012 ) . These experiments suggest that the particular geometry of the interaction between the barrier-attached dynein and the captured MT might promote MT shrinkage due to the barrier effect . Why ‘capture-shrinkage’ mechanism is not observed for Num1-based ‘cortical pulling’ has remained enigmatic . On the one hand , a classic study hinted that dynein pulls on the MT tips by inducing MT catastrophe at the cell cortex ( Carminati and Stearns , 1997 ) ; on the other hand , a recent work suggested that dynein destabilizes astral MT plus ends regardless of their cortex interaction and that this activity might not be used for generating force for spindle movement ( Estrem et al . , 2017 ) . Additionally , the MT-cortex interactions described by Carminati and Stearns . ( 1997 ) occurred before or after the nuclei moved into the neck , thus it is unknown whether they were mediated by the Num1-based mechanism that moves the spindle into the neck . Intriguingly , another study implicated cortical dynein in helping Bud6 ( a cortical MT capture protein ) and Bim1/EB1 ( a plus end tracking protein ) to couple shrinking MT plus ends to the cortex during an ‘early’ MT capture-shrinkage pathway mediated by the kinesin Kip3 ( a MT plus end depolymerase ) ( Ten Hoopen et al . , 2012 ) . This study , however , shows that Num1 is not required for the ‘early’ MT capture-shrinkage pathway , which functions to mediate movement of the spindle pole body ( SPB ) toward the incipient bud site . Together these data raise the question of whether dynein-mediated MT capture-shrinkage is downregulated during spindle movement into the bud neck . Recent work suggests that organelles may also have an important role in regulating dynein function in spindle positioning . For example , mitochondria appear to drive the assembly of a subset of cortical Num1 patches , which in turn serve to anchor the organelle itself as well as dynein to the cell cortex ( Kraft and Lackner , 2017 ) . Num1 also appears to associate with cortical ER through interaction with the conserved ER membrane VAP ( vesicle-associated membrane protein-associated protein ) , Scs2 ( Chao et al . , 2014; Lackner et al . , 2013 ) . In yeast , the VAP homologues Scs2 and Scs22 ( hereafter abbreviated as Scs2/22 ) have been implicated in the formation of ER-PM tethering sites at the cell cortex ( Loewen et al . , 2007; Manford et al . , 2012 ) and the ER diffusion barrier at the bud neck ( Chao et al . , 2014 ) . The latter is important for limiting Num1 to the mother cell until M phase , thereby regulating the timing of dynein attachment in the bud compartment . However , the distribution and appearance of Num1 patches associated with ER , mitochondria , and PM appear to be different ( Chao et al . , 2014; Heil-Chapdelaine et al . , 2000; Klecker et al . , 2013; Kraft and Lackner , 2017; Ping et al . , 2016; Tang et al . , 2009 ) , suggesting that dynein might be differentially regulated by different pools of Num1 . Additionally , despite the identification of the organelles involved in Num1 recruitment , the nature of the MT-cortex interactions and the associated nuclear movements affected by each organelle remain unclear . In this study , we set out to determine how changes in cortical Num1 localization alter dynein function , localization , and pulling mechanism in cells lacking the ER tether proteins Scs2/22 . Consistent with previous work ( Chao et al . , 2014 ) , we show that Num1 is concentrated in foci at polarized sites in scs2/22∆ cells , instead of being distributed throughout the cell cortex . We then show that the population of Num1 at the bud tip appears to be independent of mitochondria and is strikingly sufficient for dynein function in nuclear migration . We report direct observation of Num1- and dynein-dependent MT capture-shrinkage activity at the bud tip , explaining why nuclear migration across the bud neck can proceed as normal ( albeit with a decreased efficiency ) in the absence of classical dynein-mediated MT sliding along the bud cortex . The observed MT capture-shrinkage events require dynein anchoring at the bud tip and dynein motor activity , as well as MT tethering activity by the CAP-Gly domain of the Nip100/p150Glued subunit of dynactin , but not the MT plus end depolymerase activity of kinesin Kip3 or Kar3 . Remarkably , defects in MT sliding in scs2/22∆ are corrected by a CAAX-targeted Num1 , which restores lateral Num1 patches along the bud cortex and rescues the frequency of nuclear migration to WT level , highlighting a role for the ER-dependent population of cortical Num1 . Our results suggest that , in situations where cortical pulling forces drive cellular positioning processes , spatial distribution of dynein attachment molecule could potentially offer a mechanism to regulate dynein pulling force by influencing the relative activity of lateral versus end-on dynein contacts with MT at the cell cortex . In WT cells , Num1 forms dim and bright patches throughout the cell cortex ( Figure 1A; Video 1 , top ) ( Heil-Chapdelaine et al . , 2000; Tang et al . , 2009 ) . We found that cells lacking both cortical ER tethers Scs2 and Scs22 exhibited a dramatic loss of dim Num1 patches ( Figure 1A; Video 1 , bottom ) and a significant reduction in the number of bright Num1 patches ( Figure 1B ) . More than 70 . 0% of scs2/22∆ budded cells displayed ≤2 bright patches compared to only 6 . 0% in WT budded cells . The remaining Num1 patches in scs2/22∆ were observed as stationary foci at the polarized ends of the cell ( i . e . the distal bud tip and the mother cell apex; Figure 1A and C ) and as motile foci in the cytoplasm ( Figure 1—figure supplement 1A ) . Loss of Scs2 alone had a similar effect , whereas loss of Scs22 alone had no effect ( Figure 1—figure supplement 1B–D ) . However , loss of both proteins was worse than the loss of Scs2 alone ( Figure 1—figure supplement 1C; 2 . 03 ± 1 . 1 versus 2 . 8 ± 1 . 3 patches per cell for scs2/22∆ and scs2∆ , respectively ) , suggesting that Scs22 may have a redundant role when Scs2 is absent . Thus , we carried out all subsequent analysis in the scs2/22∆ double mutant background . We asked whether Num1 stability is affected in scs2/22∆ cells . Immunoblot analysis revealed that Num1-13myc levels in scs2/22∆ were similar to WT cells ( Figure 1D ) . Additionally , whole-cell intensity measurements showed that Num1-GFP levels were quantitatively the same as WT ( Figure 1—figure supplement 1E ) . However , the mean intensity of individual Num1-GFP patches was approximately 2–3 folds higher in scs2/22∆ compared to WT ( Figure 1—figure supplement 1F ) . Thus , loss of Scs2/22 affected Num1 distribution along the cell cortex but not Num1 stability . We next examined whether Num1 mobility is affected in scs2/22∆ cells . FRAP analysis showed that cortical Num1-GFP patches in scs2/22∆ exhibited no fluorescence recovery after photobleaching ( Figure 1—figure supplement 2 ) , indicating that Num1-GFP was stably associated with the cortex , a result similar to that in WT cells ( Chao et al . , 2014; Kraft and Lackner , 2017 ) . Additionally , although deletion of Scs2/22 resulted in a severe loss of cortical ER ( Figure 1—figure supplement 3A ) ( Loewen et al . , 2007; Manford et al . , 2012 ) , the timing for the accumulation of Num1 at the bud tip appeared to be unaffected compared to WT cells , as evident by imaging of single cells over time during bud growth ( Figure 1—figure supplement 4 ) . Conversely , no significant loss in cortical ER was observed in num1∆ cells ( Figure 1—figure supplement 3B ) . Importantly , no effect on Num1-GFP clustering at the bud tip was observed in scs2/22∆ when mitochondrial segregation into the bud was disrupted by the single mmr1∆ or double mmr1∆ gem1∆ mutation ( Figure 1—figure supplement 5 ) ( Frederick et al . , 2008 ) , which contradicts the model in which Num1 clustering in the bud depends on mitochondrial inheritance ( Kraft and Lackner , 2017 ) . Our data suggest that localization of Num1 to polarized bud tips does not require Scs2/22 and mitochondria . To assess whether the Num1 population distributed along the cell cortex was associated with ER , we analyzed sedimentation profiles of Num1-13myc in sucrose density gradients and colocalization of Num1-GFP with Scs2-mRuby2 . Sucrose gradient sedimentation analysis showed that a pool of Num1-13myc co-fractionated with ER in an Scs2/22-dependent manner ( Figure 1E ) . Colocalization analysis revealed that most Num1-GFP patches ( 155 out of 200; 77 . 5% ) exhibited intensities that were correlated with the signal intensities of Scs2-mRuby2 ( Figure 1F; 0 . 5 ≤ Pearson’s correlation coefficient ≤ 1 ) . However , a minority of Num1-GFP patches ( 45 out of 200; 22 . 5% ) did not co-localize with Scs2-mRuby2 ( Pearson’s correlation coefficient < 0 . 5 ) . These results , when combined with our analysis of Num1 localization in scs2/22∆ cells , implicate the existence of distinct populations of Num1 patches at the cell cortex . Next , we asked whether the observed change in Num1 localization in scs2/22∆ affects dynein targeting and function , as would be expected if Num1 functions as a cortical anchor for dynein . In WT cells , Dyn1-3GFP localizes to the SPB , astral MT plus ends , and to cortical foci where it has been offloaded from the MT plus ends ( Lee et al . , 2003; Sheeman et al . , 2003 ) . In scs2/22∆ cells , we observed that Dyn1-3GFP localized similarly to the SPB and astral MT plus ends ( Figure 2A ) but the levels of Dyn1-3GFP at the MT plus ends were significantly enhanced compared to WT cells ( Figure 2B ) , consistent with a reduced number of available offloading sites . In accord with the change in Num1 localization , cortical Dyn1-3GFP foci were found at the bud tip and mother apex of scs2/22∆ cells ( Figure 2—figure supplement 1A ) . However , the mean fluorescence intensity of individual cortical Dyn1-3GFP foci was enhanced in scs2/22∆ relative to WT ( 2 . 1 and 3 . 1-fold higher for cortical foci found in the bud and mother , respectively; Figure 2C ) . A similar enhancement was observed for Jnm1-3mCherry ( dynactin p50dynamitin subunit ) at the MT plus ends and cortex ( Figure 2D and Figure 2—figure supplement 1B ) . The difference in dynein targeting between scs2/22∆ and WT could not be attributed to changes in the expression level or the stability of dynein or dynactin ( which is required for dynein-offloading ) , as determined by immunoblotting ( Figure 2E ) . Furthermore , in scs2/22∆ cells , as reported for WT cells ( Markus et al . , 2011; Moore et al . , 2008 ) , plus end targeting of Dyn1-3GFP depended on Pac1/LIS1 ( Figure 2—figure supplement 1C ) , and cortical targeting of Dyn1-3GFP depended on dynactin ( Figure 2—figure supplement 1C ) , suggesting that regulation of dynein targeting remains intact even though dynein anchoring is limited to the polar ends of the cell . We first assessed dynein pathway function using a single-time point spindle orientation assay . Strikingly , scs2/22∆ strain had only 0 . 7% of cells with a misoriented anaphase spindle phenotype , quantitatively similar to that observed for WT ( 0 . 9%; Figure 2F ) , indicating that dynein pathway is functional . In contrast , scs2/22∆ strain expressing Num1L167E+L170E ( hereafter referred to as Num1LL ) , which harbors two point mutations that abolish the Num1-dynein interaction but does not interfere with the Num1 cluster formation ( Figure 2—figure supplement 1D and E ) ( Tang et al . , 2012 ) , exhibited a high level of misoriented anaphase spindle phenotype ( 42 . 6%; Figure 2F ) similar to that observed for a dyn1∆ or num1∆ strain ( 40 . 2 and 48 . 2% , respectively; Figure 2F ) , indicating that Num1-dynein interaction is required for proper spindle orientation in the scs2/22∆ background . The same results were obtained when nuclear segregation was assayed by DAPI staining ( Figure 2—figure supplement 1F ) . These data demonstrate that the remaining Num1 patches in scs2/22∆ , albeit few in number , appear to be sufficient for dynein pathway function . We further assessed dynein function by assaying for synthetic growth defects with kar9∆ and cin8∆ . Budding yeast lacking Kar9 or Cin8 requires the dynein pathway for normal growth ( Geiser et al . , 1997; Gerson-Gurwitz et al . , 2009; Miller and Rose , 1998 ) . Tetrad dissection analysis revealed that scs2/22∆ kar9∆ and scs2/22∆ cin8∆ triple mutant progeny formed viable colonies , exhibiting no growth defects when compared with scs2/22∆ double mutant ( Table 1 ) , consistent with the dynein pathway being functional in scs2/22∆ . Additionally , no synthetic effect on growth was observed for triple mutant of scs2/22∆ with dyn1∆ ( Table 1 ) . These genetic data further support the notion that the residual Num1 patches in scs2/22∆ cells are sufficient for dynein pathway function . Given the dramatic change in Num1 localization , we wondered how dynein would mediate spindle positioning in scs2/22∆ cells . We assessed dynein-dependent spindle movements by assaying for anaphase spindle re-alignment from a misoriented position , hereafter referred to as spindle correction ( Yeh et al . , 2000 ) . Kar9 was deleted to increase the frequency of spindle misalignment and to enhance dynein-dependent spindle movements ( Moore et al . , 2009; Yeh et al . , 2000 ) . The mechanism of spindle correction was scored based on time-lapse images of astral MT interaction with the bud cortex ( as detailed in Materials and methods ) . In kar9∆ cells , spindle correction was predominantly mediated by MT sliding along the bud cortex ( 86 . 7% , n = 30 events; Figure 3A and B; Video 2 , top ) , as previously reported ( Adames and Cooper , 2000; Yeh et al . , 2000 ) . In contrast , in scs2/22∆ kar9∆ cells , we observed that spindle correction was primarily mediated by capture-shrinkage of the astral MT plus end at the bud tip ( 77 . 8% , n = 63 events; Figure 3A and B ) . Two-color movies of mRuby2-Tub1 and Num1-GFP revealed that capture-shrinkage of the astral MT occurred upon ‘end-on’ interaction of the plus end with a Num1 patch at the bud tip ( Video 2 , bottom ) . Notably , the plus end stayed in contact with the Num1 patch while shrinking , pulling the minus-end-attached spindle into the bud , causing spindle correction . In separate experiments , we acquired movies with a larger number of optical sections confirming that the astral MT did not slide over the surface of the bud tip ( Video 3; displayed as XY and XZ frames ) . The same capture-shrinkage phenomenon was also observed when hydroxyurea ( HU ) -arrested preanaphase scs2/22∆ kar9∆ cells were examined . These data indicate that the change in the distribution of cortical Num1 in scs2/22∆ cells has apparently altered the mechanism of dynein-mediated spindle positioning . Kymograph analysis of MT capture-shrinkage events revealed that Dyn1-3GFP persisted at the shrinking MT plus end contacting the bud tip ( 15 out of 16 events; Figure 3C; Video 4 , top ) , supporting the idea that dynein is involved in generating the cortex-coupled pulling force during MT depolymerization at the Num1 site . Consistent with this notion , loss of Dyn1 abolished spindle correction in scs2/22∆ cells ( 0 out of 138 spindles were corrected; Figure 3B and Figure 3—figure supplement 1A ) . We considered the possibility that other MT plus end depolymerases might also be involved in force generation at the Num1 site . However , we found that the frequency of observing MT capture-shrinkage was unaffected in scs2/22∆ kar9∆ cells lacking Kip3 ( kinesin-8 ) or Kar3 ( kinesin-14 ) ( Figure 3D; Video 5 , top ) , two kinesin motors with known plus end depolymerase activity ( Gupta et al . , 2006; Sproul et al . , 2005 ) , indicating that these motors are not responsible for the capture-shrinkage phenomenon seen at the Num1 site . On the other hand , we found that disrupting dynein-anchoring using the num1LL allele abolished MT capture-shrinkage and prevented spindle correction ( 0 out of 91 spindles were corrected; Figure 3A and B; Video 4 , bottom ) . In num1LL scs2/22∆ cells , no capture-shrinkage events occurred despite the fact that astral MT plus ends with accumulated Dyn1-3GFP were seen sweeping along the bud tip ( Video 4 , bottom ) , indicating that cortical anchoring is required for dynein to generate the cortex-coupled pulling force at the Num1 site . These results contradict a previous study postulating that dynein does not need to attach to the cortex to destabilize MT ends ( Estrem et al . , 2017 ) . Moreover , we noted that the mean astral MT length in num1LL scs2/22∆ was not only longer than in scs2/22∆ kar9∆ , but also quantitatively the same as in dyn1∆ scs2/22∆ ( Figure 3—figure supplement 1B ) , which further supports the model in which dynein acts as a MT destabilizer at the cortical Num1 site . To investigate whether dynein’s motor activity is required for MT capture-shrinkage at the Num1 site , we made use of the established dyn1K2424A mutant , in which ATP binding was inhibited by a point mutation in the Walker A motif of the AAA3 domain ( Reck-Peterson and Vale , 2004 ) . We found that , while dynein targeting to the plus ends and cortex was not significantly affected ( Figure 3—figure supplement 1C ) , spindle correction was abolished by the dyn1K2424A mutation in scs2/22∆ cells ( 0 out of 99 spindles were corrected; Figure 3B and Figure 3—figure supplement 1A ) , indicating that motor activity is required for the production of cortex-coupled pulling forces . Notably , in dyn1K2424A scs2/22∆ cells , astral MT plus ends that grew into the bud appeared to remain stably attached upon reaching the bud tip ( Video 5 , bottom ) . To quantitate MT attachment at the bud tip , we tracked the position of the plus ends over time ( see Materials and methods ) . Compared with WT DYN1 , dyn1K2424A mutant increased the duration of attachment by nearly three folds ( ∆t = 111 . 2 ± 12 . 1 vs . 297 . 7 ± 31 . 3 s , n ≥ 25 for each; Figure 3E ) . Despite having a prolonged end-on interaction with the bud tip , the astral MTs were never observed to undergo shrinkage that led to a productive spindle movement . Conversely , we often observed the attached MTs to continue to grow and buckle while their plus ends stayed in contact with the bud tip ( 69 . 2% of dyn1K2424A scs2/22∆ cells compared to 3 . 8% of DYN1 scs2/22∆ kar9∆ cells exhibited buckling phenotype , n ≥ 52 cells for each ) . These observations suggest that dynein’s motor activity is needed to destabilize MT plus ends at the bud tip , possibly by enhancing catastrophes , as previously suggested by in vitro studies ( Laan et al . , 2012 ) . It is possible , however , that dynein’s motor activity is only needed to maintain a dynamic connection between the MT plus end and the cortex at the bud tip , as the work to pull the spindle may be performed entirely by the shrinking MTs themselves ( Grishchuk et al . , 2005; Kozlowski et al . , 2007 ) . Next , we examined how dynactin might be required for the observed MT capture-shrinkage events at the bud tip Num1 site . The vertebrate p150Glued subunit of dynactin contains a CAP-Gly domain and a basic region , both of which have been shown to bind MTs and enhance the processivity of dynein in vitro ( Ayloo et al . , 2014; Culver-Hanlon et al . , 2006; King and Schroer , 2000; Kobayashi et al . , 2006; Waterman-Storer et al . , 1995 ) . MT tethering by these domains might prevent dynein from dissociating from a shrinking MT end during capture-shrinkage events at the cortex ( Figure 4A ) . To test this , we excised codons 2–103 from the endogenous NIP100 gene , the budding yeast homologue of p150Glued , to remove the analogous CAP-Gly and basic region of the protein ( Figure 4B ) . To assess how capture-shrinkage was affected , we recorded time-lapse movies of spindle correction in kar9∆ background , as above . The number of scs2/22∆ kar9∆ cells , in which spindle correction occurred via MT capture-shrinkage at the bud tip , was dramatically decreased by the truncated Nip100 ( Figure 4C ) . The reduction could not be attributed to a defect in MT growth toward the cell cortex , as we often observed astral MTs grow into the bud , making frequent contacts with the cell cortex ( Video 6 ) . Also , the average length of astral MTs was quantitatively the same for scs2/22∆ kar9∆ cells expressing the truncated or full-length version of Nip100 ( Figure 4D ) , indicating that loss of the CAP-Gly domain did not affect the stability of astral MTs . Additionally , immunoblot analysis showed that the truncation did not affect the expression level of Nip100 ( Figure 4B ) , indicating that the observed reduction in capture-shrinkage events could not be attributed to an overall reduction in protein stability . To examine the contribution of the CAP-Gly domain more closely , we tracked the position of the astral MT plus ends in the time-lapse movies and quantitated their interaction with the bud tip . In CAP-Gly∆ scs2/22∆ kar9∆ cells , we observed the plus ends to interact with the bud tip for a significantly shorter duration compared with scs2/22∆ kar9∆ cells expressing the full-length Nip100 ( ∆t = 55 . 2 ± 7 . 2 vs . 111 . 2 ± 12 . 1 s , n ≥ 21 for each; Figure 4E and Figure 4—figure supplement 1A ) . In these abbreviated interactions , we could sometimes observe how an astral MT plus end , after making contact with the bud tip , underwent a brief capture-shrinkage event ( coupled with SPB movement ) that was suddenly aborted by its release from the cortex ( Video 6 , cell 2 ) . This suggests that the MT tethering activity of the CAP-Gly domain of Nip100 is needed for the persistence of dynein-dependent MT capture-shrinkage events that power spindle correction through the bud neck . As an alternative , we considered whether decreased spindle correction could be due to poor localization of dynein to the bud tip Num1 site , which we have shown above to be necessary for MT capture-shrinkage events . In CAP-Gly∆ scs2/22∆ cells , we observed that the localization of Dyn1-3GFP to the astral MT plus ends and the bud tip was unaffected ( Figure 4—figure supplement 1B ) . We also quantitated the fluorescence intensity of individual foci and found that the amount of dynein per cortical focus was quantitatively the same as that observed in scs2/22∆ cells expressing the full-length Nip100 ( Figure 4—figure supplement 1C ) . Thus , the decrease in MT capture-shrinkage caused by the CAP-Gly∆ mutation was not due to defective anchoring of dynein to the cortical contact point . Our results thus far suggest that dynein has two modes of cortical pulling mechanisms for controlling spindle movement into the bud cell compartment . To examine the relationship between these two modes , we quantitated the extent and consequence of losing dynein function in CAP-Gly∆ scs2/22∆ kar9∆ mutant , where both MT capture-shrinkage and sliding were presumably defective . The number of cells , in which the anaphase spindle was misaligned in the mother cell compartment , was significantly enhanced for CAP-Gly∆ scs2/22∆ kar9∆ mutant compared with scs2/22∆ kar9∆ ( 47 . 2 vs . 35 . 7% , n ≥ 300; Figure 4F ) . Additionally , CAP-Gly∆ mutation ( capture-shrinkage disrupting ) displayed severe synthetic viability defects with scs2/22∆ mutation ( sliding disrupting ) ( Figure 4G ) . These data provide strong evidence indicating that loss of both dynein-mediated pulling activities could result in additive consequences to spindle positioning and cell viability . Our data suggest that changes in Num1 localization affect dynein pulling mechanism but not dynein pathway function . We next tested whether distribution of Num1 along the bud cortex could dictate the mechanism of dynein-mediated spindle positioning . To investigate this , we asked whether spindle correction via MT sliding could be rescued in the scs2/22∆ mutant if lateral patches of Num1 were restored along the bud cortex . We attached a CAAX motif to Num1-GFP and assessed dynein-dependent astral MT interaction with the bud cortex using a spindle correction assay . Previous work showed that Num1-GFP-CAAX assembles functional cortical patches similar to those observed for Num1-GFP ( Tang et al . , 2009 ) . We found that , unlike Num1-GFP patches ( Figure 1A ) , Num1-GFP-CAAX patches were not affected by deletion of Scs2/22 and were distributed throughout the cell cortex ( Figure 5A ) . Cortical foci of Dyn1-3mCherry and Jnm1-3mCherry were observed colocalizing with lateral Num1-GFP-CAAX patches in scs2/22∆ cells ( Figure 5—figure supplement 1A and B ) . Interestingly , as lateral Num1 patches along the bud cortex were restored , we observed that dynein-dependent MT sliding became the primary mechanism for spindle correction in NUM1-GFP-CAAX scs2/22∆ kar9∆ cells: 80 . 0% ( 32 out of 40 ) of misaligned spindles were corrected by MT sliding mechanism compared with 22 . 2% ( 14 out of 63 ) when Num1-GFP was expressed in the same background ( Figure 5B versus Figure 3B; Video 7 ) . Consistent with the rescue of MT sliding mechanism , we found that the angles of interaction between the astral MT and the cortical surface for productive MT-cortex interactions ( i . e . those followed by spindle correction ) were significantly more oblique in scs2/22∆ kar9∆ cells expressing Num1-GFP-CAAX ( 54 . 0 ± 20° ) compared with those expressing Num1-GFP ( 83 . 7 ± 12 . 8° ) ( Figure 5C ) . In control kar9∆ background , the angles for productive MT-cortex interactions for Num1-GFP-CAAX ( 40 . 3 ± 14 . 2 ) was similar to those observed for Num1-GFP ( 54 . 0 ± 16° ) , in agreement with the idea that Num1-GFP-CAAX forms functional patches like Num1-GFP . Additionally , considering all MT-Num1 interactions in the bud , we found that sliding was correlated with MT interacting with a Num1 patch located within the proximal three quarters of the bud cortex , whereas end-on pulling was correlated with MT interacting with a Num1 patch located within the distal quarter of the bud ( Figure 5D ) . Thus , when combined with our results in Figure 3B , wherein spindle correction is primarily mediated by MT capture-shrinkage mechanism when Num1 is limited to the bud tip , the aforementioned observations indicate that the distribution of Num1 along the bud cortex could govern the mechanism of dynein-mediated spindle positioning . Moreover , the rescue of MT sliding by Num1-GFP-CAAX in the absence of Scs2/22 suggests that the primary role of Scs2/22 in the dynein pathway is to distribute Num1 along the cell cortex to facilitate dynein-dependent MT sliding . To examine the contribution of MT sliding to dynein pathway function in nuclear migration more closely , we quantitated spindle oscillation in HU-arrested cells , scoring for preanaphase spindle movements through the bud neck in a kar9∆ background . In kar9∆ cells , these movements coincided with lateral sliding of an astral MT along the cell cortex ( Moore et al . , 2009 ) . Lateral distribution of Num1 along the cortex might be necessary to promote efficient spindle movement across the bud neck . Compared with kar9∆ , scs2/22∆ kar9∆ mutant lacking lateral Num1 patches exhibited a significantly lower number of cells in which the preanaphase spindle moved from the mother cell compartment through the bud neck ( 22 . 5 vs . 11 . 2% , p<0 . 0001; Figure 5E , left ) . Moreover , in cells where the spindle was able to penetrate the bud neck , it moved for a significantly shorter distance ( Figure 5F , left ) . The observed differences could not be attributed to changes in astral MT dynamics in scs2/22∆ kar9∆ mutant ( Table 2 ) . However , we found that Num1-GFP-CAAX , which restored lateral Num1 patches and lateral MT sliding in scs2/22∆ kar9∆ ( Figure 5A and B ) , rescued the frequency of spindle movement across the bud neck to a level similar to that observed in kar9∆ ( 12 . 8 versus 16 . 7% , p=0 . 096; Figure 5E , right ) . Num1-GFP-CAAX also rescued the spindle penetration distance to a kar9∆ level ( Figure 5F , right ) , consistent with a role for lateral Num1 patches and MT sliding in increasing the efficiency of nuclear migration . These analyses uncovered a compromised dynein function in the scs2/22∆ cells , albeit without resulting in a spindle misorientation phenotype ( Figure 2F ) . We next investigated how Num1 is targeted to the bud tip in scs2/22∆ cells . Our results thus far indicate that Num1 redistributes to the bud tip when cortical ER tethering ( along the cell periphery ) is disrupted by deletion of Scs2/22 . Interestingly , a previous study using overexpressed epitope-tagged proteins showed that Num1 co-precipitated with the formin Bni1 ( Farkasovsky and Küntzel , 2001 ) , a polarisome component that nucleates actin cables in the bud ( Evangelista et al . , 2002; Sagot et al . , 2002 ) , suggesting that Bni1 and/or actin may play a role in Num1 targeting to the bud tip . We found that the percentage of scs2/22∆ cells exhibiting a Num1-GFP patch at the bud tip was significantly decreased upon deletion of Bni1 ( 72 . 3 to 33 . 3%; Figure 6A and B; Video 8 ) . The reduction in Num1 bud tip localization was accompanied by a striking spindle misalignment phenotype ( Figure 6C ) : 64 . 3% of bni1∆ scs2/22∆ cells exhibited a misaligned anaphase spindle compared with 2 . 5% of scs2/22∆ and 1 . 5% of WT cells . Notably , the levels of the spindle misalignment phenotype in bni1∆ scs2/22∆ cells were enhanced significantly ( by ~3 . 3 fold ) relative to those observed in bni1∆ single mutant ( Figure 6C ) , indicating a synergistic defect between bni1∆ and scs2/22∆ in anaphase spindle alignment . Furthermore , we found that depolymerization of F-actin using latrunculin A did not perturb Num1 localization at the bud tip ( Figure 6D ) , even though F-actin was completely disassembled , as judged by rhodamine-phalloidin staining ( Figure 6—figure supplement 1 ) . These data show that maintenance of Num1-GFP patches is independent of F-actin in scs2/22∆ cells , consistent with a previous study in WT cells ( Heil-Chapdelaine et al . , 2000 ) . Together , these results support that Bni1 itself , rather than its actin nucleation activity , is required for Num1 localization and function at the bud tip . Alternatively , Bni1 might be required early in the cell cycle to establish a binding site for Num1 attachment later in the cell cycle . Given the dual modes of dynein pulling mechanisms , we wondered whether they might function together to regulate spindle movement into the bud neck in WT cells . Interestingly , MT sliding movies from previous studies showed that productive sliding events in WT cells were often initiated along the lateral bud cortex and terminated when the plus end of the sliding MT encountered the bud tip ( see Video 1 in Lee et al . , 2003 ) . Additionally , Yeh et al . ( 2000 ) reported that astral MTs frequently undergo depolymerization at the bud tip after a dynein-dependent sliding event that pulled the anaphase spindle into the bud neck ( see Figure 7 in Yeh et al . , 2000 ) . To interrogate this further , we examined MT behavior during the end of MT sliding events in a kar9∆ , but an otherwise WT , background . In 16 out of 33 ( ~49% ) spindle correction events that began as sliding , we observed that the astral MT stopped sliding upon reaching the bud tip ( Figure 7A and Video 9 ) . Significantly , as shown in Videos 10 and 11 , colocalization with Num1-GFP indicated that stoppage of sliding occurred when the MT plus end encountered a Num1 patch or a cluster of Num1 at the bud tip . In all cases , at the end of sliding , the cortex-plastered astral MT could be seen changing into a straight conformation , with the plus end remaining attached to the bud tip . Because the straightening event did not push the spindle back toward the mother cell , the transition ( from a plastered/bent conformation to a straight conformation ) suggested that the captured plus end was undergoing depolymerization at the bud tip . We observed that , in majority of the cases ( 10 out of 16; 63% ) , the attached MT continued to shorten and shrink , causing the spindle to move closer to the bud tip , further aligning the spindle along the mother-bud axis . In the remaining cases ( 6 out of 16; 37% ) , straightening was followed by the immediate release of the astral MT from the bud tip . Intriguingly , in kar9∆ cells lacking the formin Bni1 , which is required for Num1 localization to the bud tip ( Figure 6A and B ) ( Farkasovsky and Küntzel , 2001 ) , we observed a significant increase in the frequency of finding astral MT sliding that went past the bud tip instead of stopping upon reaching the bud tip ( Figure 7B and C; Video 12 ) . Together these results suggest that dynein-dependent MT capture-shrinkage at the bud tip has a role in regulating spindle movement in WT cells . Our studies show how changes in cortical distribution of Num1 can dramatically alter dynein-dependent spindle pulling mechanism . When Num1 is restricted to the bud tip , as in the case for scs2/22∆ cells , dynein generates pulling forces predominantly via the MT capture-shrinkage mechanism . In this configuration , anchored dynein is geometrically limited to interact with the very end of the astral MT . Additionally , our data suggest that , through CAP-Gly domain of Nip100/p150Glued , dynactin may act as a cortical linker to maintain the connection between the shrinking MT end and the cortex , enabling transmission of MT depolymerization into pulling force generated by the dynein motor at the bud tip . However , when Num1 is distributed along the bud cortex , as in the case for WT and NUM1-GFP-CAAX scs2/22∆ kar9∆ cells , dynein generates pulling forces primarily via the MT sliding mechanism . In this configuration , anchored dynein can pull on the spindle by moving laterally along the MT lattice . This type of lateral cortical contact is facilitated by a larger surface for dynein-MT interaction , explaining why lateral Num1 patches are more efficient in promoting spindle movement across the bud neck ( Figure 5E and F ) . Interestingly , our studies also show how stoppage of MT sliding is coupled with MT capture and shrinkage at the bud tip , providing a mechanism by which dynein-dependent cortical pulling is spatially regulated in the bud . We show that the population of Num1 at the bud tip depends on Bni1 . We propose that MT capture-shrinkage at the bud tip functions as a brake for MT sliding ( Figure 7D ) . This function may be important to prevent oversliding of MT beyond the bud tip , thereby ensuring that the spindle is correctly positioned across the bud neck . Our results suggest that cortical ER has a facilitative role in the dynein pathway . In wild-type yeast cells , the cortical ER is consisted of a network of sheets and tubules that tightly associate with the plasma membrane ( Pichler et al . , 2001; Prinz et al . , 2000; West et al . , 2011 ) . The cortical ER and the PM make extensive contacts along the periphery of the cell , forming structures called ER-PM junctions ( Prinz , 2014; Stefan et al . , 2013 ) , which have been implicated in various cellular processes , including phosphoinositide signaling ( Stefan et al . , 2011 ) , sterol lipid transport ( Schulz et al . , 2009 ) , as well as maintenance of ER morphology and regulation of the unfolded protein response in the ER ( Manford et al . , 2012 ) . At least three families of integral ER proteins – Scs2/22 , Ist2 , and Tcb1/2/3 – function to tether the cortical ER to the PM ( Eisenberg-Bord et al . , 2016; Manford et al . , 2012 ) . Among them , Scs2/22 appear to be the most important ( Loewen et al . , 2007; Manford et al . , 2012 ) . They contain a single transmembrane domain and a cytoplasmic MSP ( major sperm protein ) domain that can bind directly to PI lipids ( Kagiwada and Hashimoto , 2007 ) or a FFAT ( diphenylalanine in an acidic tract ) motif found in lipid transfer proteins ( Loewen et al . , 2003 ) . Interestingly , a recent study shows that Num1 has a putative FFAT motif in its N-terminal region that targets Num1 to the cortical ER by binding to the MSP domain of Scs2 ( Chao et al . , 2014 ) . This raises the possibility that cortical ER-PM junctions may play a role in anchoring Num1 ( and therefore dynein ) for mediating movement of the nucleus into the bud neck . Our study here provides evidence to support a specific role in facilitating MT sliding . First , in WT cells , Num1 localizes to numerous dim patches throughout the cell cortex . In particular , dim patches are observed along the lateral cortex of medium and large buds ( Figure 1A ) , where dynein-dependent MT sliding is thought to occur during movement of the nucleus into the bud neck . We speculate that these dim patches represent Num1 molecules that are anchored at the ER-PM contact sites , since loss of Scs2 ( Figure 1—figure supplement 1B ) or Scs2/22 ( Figure 1A ) resulted in a severe loss of dim patches . Second , the number of MT sliding events that occur along the bud cortex is significantly diminished by loss of Scs2/22 ( Figure 5—figure supplement 2A ) . Although some sliding events remained , they appeared to be less effective in moving the spindle ( Figure 5—figure supplement 2B ) . It is possible that the remaining sliding events were mediated by Num1 that binds to the PM via its C-terminal PH domain ( Yu et al . , 2004 ) or to specific membrane on the mitochondrial surface via its N-terminal CC domain ( Ping et al . , 2016 ) . However , the reduction in the spindle penetration distance observed in the absence of Scs2/22 ( Figure 5F and Figure 5—figure supplement 2B ) is consistent with the idea that Num1 attachment to ER-PM junctions can provide a stronger resistive force , which may be needed to firmly anchor dynein for efficient pulling of the spindle from the cell cortex . Our study demonstrates for the first time that dynein pulling forces are spatially mediated and regulated by two differential populations of Num1 patches in the WT yeast buds . The first population , namely the Scs2/22-dependent lateral patches ( as discussed above ) , appears to initiate and facilitate MT sliding along the bud cortex . The second population , which localizes to the bud tips , terminates MT sliding by capturing MT plus end and inducing MT catastrophe . Consistent with the idea that dynein pulling activity is tightly regulated , previous work shows that the ER diffusion barrier also functions at the bud neck to confine Num1 to the mother compartment until M phase ( Chao et al . , 2014 ) . The diffusion barrier appears to be important for regulating the start of dynein pulling activity , by preventing premature localization of Num1 into the bud compartment . Our study now provides an additional level of control for modulating dynein pulling activity , which appears to be important for ending the spindle movement once it is started in the bud ( Figure 7D ) . Intriguingly , we did not observe premature accumulation of Num1 to the tips of small buds in scs2/22∆ cells ( Figure 1—figure supplement 4 ) , where the ER diffusion barrier is presumably disrupted , suggesting that additional components might be required to recruit Num1 to the bud tip in a timely manner . How Num1 switches the dynein motor from a side-on motor to an end-on motor remains an open question at this point . Our data rule out Scs2/22 as being required for motor activity , given that MT sliding can be rescued in the absence of Scs2/22 by a CAAX-targeted Num1 . Since the mechanism of membrane attachment for Num1 is likely to be different between the lateral cortex and the bud tip , it will be interesting for future investigations to examine whether cortical stiffness could play a role in regulating the switching of cortical dynein pulling mechanisms . All strains used in this study are listed in Supplementary file 1 and were derived from the genetic background of WT strains YWL36 and YWL37 ( Vorvis et al . , 2008 ) or the protease-deficient strain BJ5457 ( Jones , 1990 ) . Strains were generated by standard genetic crosses or by PCR product-mediated transformations ( Longtine et al . , 1998 ) . Diploids resulted from each cross were sporulated and tetrad dissected and the progeny were then examined by marker analysis . Transformations were performed using lithium acetate protocol ( Knop et al . , 1999 ) . Transformants were purified twice by streaking to single colonies on selective media plates . Proper deletion or insertion of fluorescent protein tagging cassette at the genomic locus was confirmed by diagnostic PCR and fluorescence microscopy . All fluorescent protein tagging was done at the chromosomal locus and imaging was performed using live cells unless stated otherwise . At least two independent transformants were chosen from each disruption or tagging procedure for subsequent experiments . To label MTs , strains were transformed with HindIII-digested HIS3p::mCherry-TUB1::LEU2 ( Zhu et al . , 2017 ) or BsaBI-digested HIS3p:mRuby2-TUB1+3’UTR::URA3 and HIS3p:mRuby2-TUB1+3’UTR::LEU2 ( Markus et al . , 2015 ) , or undigested GFP-TUB1::LEU2 ( Song and Lee , 2001 ) . Transformants were screened and selected by fluorescence microscopy . To label endoplasmic reticulum , we constructed a plasmid expressing EGFP-HDEL via Gibson assembly reaction ( Gibson et al . , 2009 ) for integration at the URA3 locus . Briefly , PCR products containing the TEF1 promoter ( TEF1p , 459 bp upstream of TEF1 start codon ) , ER-targeting signal sequence ( SS , 126 bp of the 5’ end of KAR2 ) , 3XGlyAla linker , and EGFP-HDEL sequence ( amplified from pFA6a-GFP ( S65T ) -TRP1 plasmid ( Longtine et al . , 1998 ) with HDEL sequence built into the reverse PCR primer ) were assembled together and cloned into BamHI and NotI digested pRS315 , a LEU2-containing vector ( Sikorski and Hieter , 1989 ) . Next , we subcloned TEF1p-SS-3xGlyAla-GFP-HDEL as a HindIII-SacI fragment into pRS306 , a URA3-containing vector , generating pRS306-TEF1p-SS-3xGlyAla-GFP-HDEL::URA3 . Strains were transformed with StuI-linearized pRS306-TEF1p-SS-3xGlyAla-GFP-HDEL::URA3 for integration into URA3 locus . Ura+ transformants were selected , colony purified , and screened by fluorescence microscopy . To generate in-frame deletion of the CAP-Gly and basic domain of Nip100 at the endogenous chromosomal locus , we used the two-step approach for constructing unmarked genomic mutagenesis ( Gray et al . , 2005 ) . Briefly , the URA3 marker was amplified from pRS306 with primers containing sequences flanking the targeted region of Nip100 ( amino acid 2–103 ) . We verified the substitution of the targeted sequence with URA3 by diagnostic PCR from the genomic DNA . The resulting strain was transformed with a second PCR product containing an in-frame fusion of the sequences flanking the targeted region ( 60 bp on one side of the URA3 insertion and 1209 bp on the other side ) , amplified from WT genomic DNA , along with a carrier plasmid containing the LEU2 marker ( pRS315 ) . Transformants were replica plated to 5-fluoroorotic acid ( 5-FOA ) plates to select for the removal of URA3 . Deletion of the targeted sequence was confirmed by diagnostic colony PCR and DNA sequencing . Confocal images and fluorescence recovery after photobleaching ( FRAP ) experiments were acquired using a 1 . 4 NA 60X oil immersion objective on a Nikon A1R confocal microscope equipped with a LU-NB laser launch system housed in the IALS Nikon Center of Excellence microscopy facility at UMass Amherst . Pinhole size was set to 0 . 7 airy unit . To FRAP , we bleached for 3 s using 488 nm laser at 5% laser power . After photobleaching , single focal plane images were acquired every 30 s at 0 . 3% laser power . Wide-field fluorescence images were acquired using either a 1 . 45 NA 100X objective on a Nikon 80i upright microscope equipped with piezo Z control ( Physik Instrumente ) and a cooled electron-multiplying charged-coupled device ( EMCCD ) camera ( Cascade II; Photometrics ) or a 1 . 49 NA 100X objective on a Nikon TiE inverted microscope system equipped with a laser launch ( 405/488/561/640 nm; LUN4; Nikon ) and a EMCCD camera ( iXon 888; Andor ) . Filter cube sets ( 31000 v2 , 49002 , 49008 , and TRF89901; Chroma ) were used for imaging DAPI , GFP and mRuby2/mCherry fluorescence . All three microscope systems were controlled by NIS-Elements software ( Nikon ) . Yeast strains were grown to mid-log phase in synthetic-defined media ( Sunrise Science Products , CA ) at 30°C and mounted on 1 . 7% agarose pad for imaging . All images were acquired at room temperature . For three-dimensional reconstruction of Num1-GFP localization , we acquired up to 25 optical confocal sections spaced 0 . 3 µm apart encompassing the entire thickness of cells . Image stacks were deconvolved where indicated using the 3D Deconvolution tool in NIS-Elements software . To minimize phototoxicity to cells and photobleaching during time-lapse imaging , we acquired frames at intervals as indicated in the videos and with up to seven optical sections spaced 0 . 5 µm apart . To quantify the number of cortical Num1 patches per cell , we used the analyze particle tool in ImageJ to determine the number of Num1-GFP foci from maximum intensity projection of confocal Z-stack images . Cortical patches were defined as foci having intensities above the threshold set by the average background intensity measured within the cytoplasmic area . To determine the spatial distribution of Num1 patches , we used the multipoint tool in ImageJ to determine the x-y coordinates of individual Num1-GFP patches relative to the bud neck from a maximum intensity projection of Z-stack images . The position of each patch was then normalized along the y-axis , with the distance from the bud neck to the bud tip as 0 to −1 , and the distance from the bud neck to the mother apex as 0 to 1 ( see diagram in Figure 1C ) . To measure the intensity of Num1 patches , we used the circle tool in ImageJ to encompass individual Num1-GFP foci from maximum intensity projection of Z-stack images . To subtract the background intensity from each measurement , we moved the circle tool from the patch to a nearby cytoplasmic area within the same cell . To quantify colocalization of Num1-GFP with Scs2-mRuby2 , we used the Colocalization tool in NIS-Elements software . Pearson’s correlation coefficients between Num1-GFP and Scs2-mRuby2 were calculated for individual Num1 patches in two-color deconvolved wide-field images of Num1-GFP and Scs2-mRuby2 . Time-lapse videos displaying XY and XZ views were generated using XYZ projection tool in ImageJ . Kymographs were generated using the MultipleKymograph plugin for ImageJ . We used the mTrack plugin for ImageJ to track the MT plus end in cells expressing mRuby2-Tub1 . The position of the plus end relative to the bud neck was tracked over time in movies acquired with 5 s intervals . The duration of plus end attachment at the bud tip ( ∆t ) was scored as the length of time between the plus end making contact with the bud tip and the time it depolymerizes away from the bud tip . For cold spindle misorientation assay , mid-log cultures expressing fluorescently-labeled tubulin were grown in YPD and then shifted to 16°C for 15 hr before imaging . For cold nuclear segregation assay , mid-log cultures were grown in YPD and then shifted to 16°C for 15 hr , fixed with 70% ethanol and stained with DAPI . For spindle oscillation assay , strains expressing mRuby2-Tub1 were grown to mid-log and arrested with 200 mM hydroxyurea for 1–1 . 5 hr before imaging , as described ( Moore et al . , 2009; Tang et al . , 2012 ) . The velocity of spindle movement was defined as ∆D/∆T , in which ∆D was the distance the spindle traveled in a continuous bud-directed movement , and ∆T was the time for the movement . For spindle correction assay , we scored for misoriented anaphase spindles that moved into the bud neck and became aligned along the mother-bud axis during a 10 min movie in the kar9∆ background . Spindle correction was scored as mediated by ‘sliding mechanism’ if the astral MT displayed lateral association with the bud cortex while the spindle moved into the bud neck during its realignment , and by ‘capture-shrinkage mechanism’ if the astral MT exhibited end-on interaction with the bud tip followed by depolymerization of the astral MT concomitant with spindle movement into the bud neck . MT growth rate , shortening rate , catastrophe frequency , and rescue frequency were measured as described ( Gupta et al . , 2006 ) . The angle of interaction between astral MT and the bud tip was measured using the angle tool in ImageJ . Rose histograms were plotted using MATLAB . For growth assays , strains were grown to mid-log phase in YPD media , then ten-fold serial dilutions were spotted on YPD plates and grown at 30°C for 2 days . To depolymerize F-actin , cells were grown to early log phase , collected by centrifugation , and resuspended in synthetic-defined medium containing 200 µM latrunculin A or 0 . 5% DMSO for 20 min before imaging . To verify loss of F-actin , cells were fixed and stained with rhodamine-phalloidin as previously described ( Waddle et al . , 1996 ) . To immunoblot for Num1-13myc , Dyn1-TAP , Jnm1-13myc , Nip100-13myc , and CAP-Gly∆−13myc , yeast strains were grown overnight in 5 ml of rich media ( YPD ) at 30°C . Cell pellets were resuspended in ice cold lysis buffer containing 20 mM Tris pH 7 . 5 , 150 mM NaCl , 1 mM EDTA , 1 . 5% Triton X-100 , supplemented with protease inhibitor cocktail tablet ( Millipore Sigma ) . Equal amount of cells were lysed by bead beating in round-bottom glass tubes for 6 × 30 s with 2 min interval between beatings . Following centrifugation ( at 21 , 130 g for 10 min at 4°C ) , the resulting supernatants were separated on 8% ( for Num1-13myc and Jnm1-13myc ) or 6% ( for Dyn1-TAP ) or 4–15% ( for Nip100-13myc and CAP-Gly∆−13myc ) SDS-PAGE and then electro-blotted to PVDF or nitrocellulose membrane in 25 mM Tris , 192 mM glycine , 0 . 05% SDS , and 20% methanol for 80 min . Membranes were probed with either mouse 9E10 anti-c-Myc antibody ( BioLegend ) at 1:250 or 1:500 or 1:1000 dilution , or rabbit IgG antibody ( GenScript ) at 1:5000 dilution . Goat HRP-conjugated anti-mouse ( BioLegend ) and anti-rabbit antibodies ( Jackson ImmunoResearch ) were used at 1:10 , 000 dilutions . Chemiluminescence signals were acquired and imaged using a ChemiDoc Imaging System ( Bio-Rad ) or a G:BOX Chemi HR16 ( Syngene ) equipped with a 16-bit CCD camera ( Sony ICX285AL; pixel size of 6 . 45 x 6 . 45 µm ) . Immunoblots were exposed for durations ranging from 3 s to 5 min without saturating the camera’s pixels . For sedimentation analysis of Num1-13myc , we poured 10 ml 20–60% sucrose step gradients and allowed them to equilibrate for 9 hr at 4°C before use . Each step of the gradient contained 2 ml of 20 , 30 , 40 , 50 , or 60% sucrose in sedimentation buffer ( 10 mM Tris pH 7 . 5 , 10 mM EDTA , and 50 mM NaCl ) . WT and scs2/22∆ strains expressing Num1-13myc were grown to mid-log phase in 60 ml of YPD media , collected by centrifugation , and resuspended in ice cold lysis buffer containing 20 mM Tris pH 7 . 5 , 1 mM EDTA , and 50 mM NaCl supplemented with protease inhibitor cocktail tablet . Cells were then lysed by glass bead beating for 6 times 30 s with 2 min intervals between beatings . Lysates were clarified at 500 g for 10 min at 4°C and 0 . 5 ml of the supernatant was loaded directly onto a 10 ml sucrose gradient prepared as above . Centrifugation was performed in a Beckman SW41 Ti rotor at 36 , 000 rpm for 17 . 5 hr at 4°C . Fractions of 0 . 5 ml were collected from the top of each gradient for analysis by Western blot using the mouse 9E10 anti-c-Myc antibody ( for Num1-13myc ) and the anti-Sac1 antibody ( a kind gift from Dr . Charles Barlowe for detection against the ER marker Sac1 ) . All statistical analyses were performed using GraphPad Prism software . A two-tailed Student’s t test or one-way ANOVA test was used to determine statistical significance where indicated . At least two independent experiments were performed for each analysis .	Cells must divide so that organisms can grow , repair damaged tissues or reproduce . Before dividing , a cell creates two identical copies of its genetic information – one for each daughter . A molecular machine known as the mitotic spindle then moves each set of genetic material to where it will be needed when the daughter cells form . For the process to work properly , however , a motor protein known as dynein must correctly position the spindle by pulling it into place from the outskirts of the cell . When a baker’s yeast cell divides , it first forms a ‘bump’ , which grows into a bud that will ultimately become another yeast . The spindle needs to be precisely placed at the midpoint between the original cell and the bud , so the genetic material can get into the future daughter cell . To do so , dynein travels to the bud , where a protein called Num1 helps it attach to the periphery and pull the filaments of the mitotic spindle ( known as microtubules ) to the correct position . Num1 also attaches to other cellular structures in the bud , including one known as the endoplasmic reticulum . It was unclear how this connection changes where dynein is located , and how it can pull on the spindle . To study this , Omer et al . labeled Num1 , dynein and microtubules with fluorescent markers so they could be followed in living baker’s yeast using time-lapse microscopy . Mutant yeast strains were also used to disrupt how these proteins associate , which helps to tease out their roles . The experiments show that there are several populations of Num1 in the bud . One associates with the endoplasmic reticulum , and it helps dynein grab the side of a microtubule and make it slide into the bud . The other does not attach to the reticulum , but instead is located at the very tip of the bud . There , it makes dynein capture the end of the microtubule; this destabilizes the filament , which starts to shorten . As the microtubule shrinks , the spindle is pulled closer to the bud’s tip , which aligns it in the right position . The yeast cells thus need Num1 in both locations to fine-tune the pulling activity of dynein , and the spindle’s final positioning . In the human body , not all divisions create two identical cells; for example , the daughters of stem cells can have different fates . This is due to a precise asymmetric division which dynein partly controls . The results by Omer et al . could help to unravel this mechanism .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2018	Cortical dynein pulling mechanism is regulated by differentially targeted attachment molecule Num1
Fabry disease ( FD ) is a life-threatening X-linked lysosomal storage disorder caused by α-galactosidase A ( α-GAL ) deficiency . Small fiber pathology and pain are major FD symptoms of unknown pathophysiology . α-GAL deficient mice ( GLA KO ) age-dependently accumulate globotriaosylceramide ( Gb3 ) in dorsal root ganglion ( DRG ) neurons paralleled by endoplasmic stress and apoptosis as contributors to skin denervation . Old GLA KO mice show increased TRPV1 protein in DRG neurons and heat hypersensitivity upon i . pl . capsaicin . In turn , GLA KO mice are protected from heat and mechanical hypersensitivity in neuropathic and inflammatory pain models based on reduced neuronal Ih and Nav1 . 7 currents . We show that in vitro α-GAL silencing increases intracellular Gb3 accumulation paralleled by loss of Nav1 . 7 currents , which is reversed by incubation with agalsidase-α and lucerastat . We provide first evidence of a direct Gb3 effect on neuronal integrity and ion channel function as potential mechanism underlying pain and small fiber pathology in FD . Thermal hyposensitivity and pain based on small nerve fiber pathology are cardinal neurological symptoms of the progressive and life-threatening X-linked Fabry disease ( FD ) ( Zarate and Hopkin , 2008 ) . While genetically caused reduction or loss of α-galactosidase A ( α-GAL ) with consecutive lysosomal accumulation of globotriaosylceramide ( Gb3 ) has long been considered the pathophysiological key event in FD ( Romeo and Migeon , 1970 ) , the mechanisms underlying sensory impairment , pain , and denervation remain unclear . Men more than women with FD report reduction in cold and warm sensation over time , which is reflected by increased thermal perception thresholds and paralleled by loss of intraepidermal innervation ( Üçeyler et al . , 2011 ) . Starting in early childhood , patients also report mainly episodic , acral , burning pain triggered by heat , fever , or physical activity ( Üçeyler et al . , 2014 ) . The finding of potential Gb3 deposits in dorsal root ganglion ( DRG ) neurons ( Gadoth and Sandbank , 1983; Lakomá et al . , 2016; Marshall et al . , 2010 ) and the presence of peripheral neuropathy mostly of the small fiber type ( Üçeyler et al . , 2011 ) supports the hypothesis of a pathophysiological mechanism involving Gb3 accumulation . Having investigated the largest and oldest cohort ( >24 months ) reported thus far for the α-GAL deficient ( GLA KO ) mouse model of FD , we previously described an age-dependent development of thermal hyposensitivity mirroring the clinical phenotype ( Üçeyler et al . , 2016 ) . Whether Gb3 accumulation might link neuronal pathology with sensory impairment , pain , and peripheral denervation remains to be determined . We hypothesized that neuronal Gb3 deposits interfere with ion channel expression and function , and neuronal integrity , contributing to the sensory phenotype in FD . We investigated GLA KO mice stratified for age using a comprehensive approach . Our data provide first combined molecular , histological , electrophysiological , and behavioral evidence for a direct and age-dependent influence of intracellular Gb3 deposits on neuronal integrity and ion channel function as a potential mechanism of progressive Fabry-associated sensory disturbance , pain , and skin denervation . First , we examined DRG neuron size by analysing neuronal area ( Figure 1A–D ) and found larger DRG neurons in young GLA KO compared to young WT mice ( p<0 . 01; Figure 1E ) . Neurons of old GLA KO mice were larger compared to old WT ( p<0 . 001 ) and young GLA KO mice ( p<0 . 001; Figure 1E ) . We also asked if Gb3 deposits are present and where they are located in DRG neurons of young and old GLA KO mice . We assessed semithin sections and found intraneuronal deposits in young and even more so in old GLA KO mice , while DRG neurons from wildtype ( WT ) mice displayed normal histology ( Figure 1F–I ) . We then applied antibodies against CD77 to detect Gb3 and saw marked immunoreaction in DRG of old GLA KO mice , which was not detectable in young mice and in WT littermates ( Figure 1J–M ) . Interestingly , Gb3 immunoreactivity was not restricted to neurons , but was also present extra-neurally ( Figure 1M , arrowheads ) . Applying confocal microscopy and co-immunoreaction with antibodies against β- ( ΙΙΙ ) -tubulin , we found that Gb3 is mainly located in the cytoplasm of DRG neurons of old GLA KO mice but also in the very proximal parts of sensory axons , in extra-neural connective tissue , and cellular membranes ( Video 1 ) . To investigate whether Gb3 accumulation in DRG neurons is associated with endoplasmic stress , we performed cellular binding immunoglobulin protein ( BiP ) expression analysis . BiP was homogeneously distributed in neurons of young GLA KO and WT mice ( data not shown ) and in old WT mice ( Figure 1N ) . In contrast , in neurons of old GLA KO mice , condensed BiP was located within and around the nucleus ( Figure 1O ) indicating enhanced endoplasmic stress . We then asked , whether increased neuronal Gb3 deposition and endoplasmic stress are associated with a reduction of peripheral innervation , a phenomenon reported for young GLA KO mice ( Lakomá et al . , 2014 ) and known in patients with FD ( Maag et al . , 2008; Üçeyler et al . , 2011 ) . We quantified intraepidermal nerve fiber density ( IENFD ) in skin obtained from mouse hind paws and found a marked reduction of cutaneous innervation in young and old GLA KO mice compared to their WT littermates ( Figure 2A–D ) , surpassing the physiological reduction of IENFD with aging ( p<0 . 001 each , Figure 2E ) . Furthermore , we assessed whether Gb3 accumulates not only in DRG , but also in axons of the sciatic nerve and in skin . We did not find any Gb3 depositions in the sciatic nerve ( Figure 2F–K ) or footpad skin ( Figure 2L–Q ) of old GLA KO and WT mice . To investigate the degree of apoptosis in DRG neurons in the course of Gb3 accumulation and potential endoplasmic stress , we performed a NucView 488 Caspase 3 Enzyme Substrate Assay . We quantified the percentage of caspase 3 positive neurons in cultured DRG neurons of old GLA KO and WT mice ( Figure 3A–D ) . DRG neuron cultures of old GLA KO mice in the naïve state displayed a higher percentage of caspase 3 positive neurons compared to old WT mice ( p<0 . 001 , Figure 3E ) indicating enhanced apoptosis . Additionally , positive control neurons of both genotypes incubated with 500 nM staurosporine for 16 hr showed a higher percentage of caspase 3 positive neurons compared to cultured DRG neurons in the naïve state ( p<0 . 05 each , Figure 3E ) . We further determined the percentage of neurons with neurite outgrowth . Cultured DRG neurons of old GLA KO mice showed less neurite outgrowth compared to neurons of WT mice ( p<0 . 001 , Figure 3F ) . Heat intolerance and heat induced pain are key symptoms reported by Fabry patients ( Üçeyler et al . , 2014 ) . We thus investigated transient receptor potential vanilloid 1 ( TRPV1 ) channel expression and function as the major neuronal ion channel that is primarily involved in heat perception and pain . While TRPV1 gene expression did not differ between genotypes and age-groups ( Figure 4A ) , we found an increased number of TRPV1 immunoreactive DRG neurons in young and old GLA KO mice compared to their WT littermates ( p<0 . 001 each , Figure 4B–F ) . We also analyzed the distribution of TRPV1 immunoreactivity across different neuronal sizes and quantified TRPV1 positive neuron diameters; neuron populations were stratified as small ( <25 µm in diameter ) and large ( ≥25 µm in diameter ) neurons ( Figure 4G ) ( Cesare and McNaughton , 1996; Hoheisel et al . , 1994; Lawson et al . , 1993 ) . TRPV1 immunoreactivity was mainly observed in small diameter neurons independent of genotype and age . Next , we investigated capsaicin induced TRPV1 current densities with patch-clamp analysis in five days old cultured DRG neurons . Neurons appeared enlarged and carried deposits in GLA KO mice , while were of normal shape in WT mice ( Figure 4G , H ) . We observed a tendency for higher current densities in young GLA KO mice ( exemplified current in Figure 4I ) , but the difference was not significant between genotypes ( Figure 4J ) . In contrast , cultured DRG neurons of old GLA KO and littermate WT mice did not respond to capsaicin at all . We investigated neurons obtained from different culture periods ( 24 hr , three , five , and eight days ) so that we do not miss time-dependent TRPV1 currents that might be present only at distinct time points in primary cell culture . TRPV1 currents were also not evoked by capsaicin using calcium-free bath solution to prevent tachyphylaxis . To test for a potential influence of genetic background , we patched DRG neurons of a 14 months old C57BL/6N male mouse , and again did not find capsaicin induced TRPV1 currents under any of the conditions detailed above . Since increased neuronal TRPV1 protein expression may be associated with heat hypersensitivity , we determined paw withdrawal latencies after intraplantar injection of capsaicin in old GLA KO mice at a dose that induced only mild and short lasting pain behavior in WT mice ( Carey et al . , 2017; Sakurada et al . , 1992 ) . Indeed , old GLA KO mice showed heat hypersensitivity compared to baseline 24 hr after capsaicin ( p<0 . 01 Figure 4L ) . We then studied potassium/sodium hyperpolarization-activated cyclic nucleotide-gated ion channels ( HCN ) and focused on HCN2 as a pacemaker current influencing neuronal action potential frequency and pain in several animal models ( Emery et al . , 2012 ) . There was no intergroup difference for HCN2 gene expression in DRG of GLA KO and WT mice ( Figure 5A ) , while HCN2 immunoreactivity increased with age in both genotypes ( p<0 . 05 , Figure 5B–F ) . In contrast , patch-clamp analysis of DRG neurons revealed that hyperpolarization-activated ( Ih ) current densities ( exemplified current in Figure 5G ) , which are carried by all four isoforms of HCN channels , were markedly reduced in old GLA KO mice compared to old WT mice ( p<0 . 001 each , Figure 5H ) , but did not differ between mice of young age-groups . Lacking a HCN2 specific blocker , further electrophysiological HCN channel subclassfication was not possible . Since HCN2 conditional knockout mice are protected from heat and mechanical hypersensitivity after peripheral nerve lesion ( Emery et al . , 2011 ) , we applied chronic constriction injury ( CCI ) at the right sciatic nerve of GLA KO and WT littermates . Indeed , heat hypersensitivity only developed in old WT mice ( p<0 . 01 ) lasting up to day 28 after surgery , while old GLA KO mice were spared ( p<0 . 01 , Figure 5I ) . Also , mechanical withdrawal thresholds remained at baseline level for the entire study period of 28 days in old GLA KO mice , while WT mice displayed hypersensitivity to mechanical stimuli up to day 28 after CCI ( p<0 . 01 , Figure 5J ) . Finally , we investigated neuronal voltage-gated sodium channel 1 . 7 ( Nav1 . 7 ) expression and function in GLA KO and WT littermates as a key contributor to neuropathic pain ( Yang et al . , 2018 ) . Nav1 . 7 gene expression in DRG was not different in both age-groups and genotypes ( Figure 6A ) . Since Nav1 . 7 immunostaining failed using five different Nav1 . 7 antibodies detailed above , we applied enzyme-linked-immuno-sorbent assay on DRG samples and found no difference in Nav1 . 7 protein expression in DRG between genotypes and age-groups ( Figure 6B ) . Patch-clamp analysis revealed that sodium current densities ( exemplified currents in Figure 6C ) were not different between young GLA KO and WT littermates , but were notably reduced in old GLA KO mice compared to old WT mice ( p<0 . 001 each , Figure 6D ) . We applied tetrodotoxin ( TTX ) to DRG neurons obtained from young GLA KO mice that had normal sodium currents with fast inactivating kinetics at baseline ( black trace in Figure 6E ) . These sodium currents were sensitive to TTX already at a concentration of 100 nM ( red trace in Figure 6E ) and recovered after washout with bath solution ( grey trace in Figure 6E ) , such that the observed sodium currents were identified as being predominantly produced by Nav1 . 7 , a channel which has been shown to contribute about 70% of the TTX sensitive current in small DRG neurons ( Vasylyev et al . , 2014 ) . We then investigated whether reduced neuronal Nav1 . 7 currents may be associated with protection from heat and mechanical hypersensitivity in an inflammatory pain model , as known for Nav1 . 7 conditional knockout mice ( Nassar et al . , 2004 ) . Indeed , intraplantar injection of complete Freund`s adjuvant ( CFA ) led to heat hypersensitivity in all mice groups except for old GLA KO mice ( p<0 . 001 , Figure 6F ) , in which heat withdrawal latencies did not change from baseline for the entire study period of seven days ( p<0 . 001 , Figure 6F ) . Similarly , all mice developed mechanical hypersensitivity starting one hour after CFA injection compared to baseline ( p<0 . 001 , Figure 6G ) , which was less pronounced in old GLA KO mice compared to old WT mice after CFA injection ( Figure 6G ) , and all mice remained mechanically hypersensitive until day seven after CFA injection . Finally , we investigated if cellular Gb3 accumulation interferes with Nav1 . 7 currents . For this , we silenced α-GAL in human embryonic kidney 293 cells ( HEK ) expressing Nav1 . 7 with small hairpin RNA ( shRNA ) directed against the human α-GAL transcript as an in vitro model . Few HEK cells transfected with control shRNA ( control HEK cells , Figure 7A–C ) showed mild Gb3 deposition , while the majority of HEK cells transfected with shRNA against α-GAL ( shRNA HEK cells , Figure 7D–F ) displayed a marked increase in Gb3 accumulation within only one week of transfection . These Gb3 deposits were reversible by incubation with 1 . 32 µg/ml agalsidase-α ( 1 mg/ml , Shire , Saint Helier , Jersey , UK ) and 250 µM lucerastat ( N-butyldeoxygalactonojirimycin , Biomol , cat# Cay19520-1 , Hamburg , Germany ) applied for 24 hr prior to patch-clamp recordings ( Figure 7G–L ) . Electrophysiological analysis of Nav1 . 7 currents in Gb3-positive HEK cells revealed a marked decrease of sodium currents after shRNA treatment compared to control HEK cells ( p<0 . 01 , Figure 7J , K ) , which recovered after agalsidase-α and lucerastat incubation ( agalsidase-α: p<0 . 05; lucerastat: p<0 . 01 , Figure 7N ) . Our data give first evidence for the involvement of neuronal Gb3 deposits in the pathophysiology of skin denervation and a direct and major role in sensory impairment , and pain of patients with FD . The exact mechanisms , however , remain to be elucidated , we show that neuronal Gb3 deposits result in an overall reduction of ion channel current densities and provide a HEK cell based in vitro model as a potent tool for further pathophysiological research and pharmaceutical testing of new Fabry-specific drugs . Gb3 influences neuronal function and integrity , thus , a sustained normalization of intracellular Gb3 load by drugs providing permanently low Gb3 levels without recurrent end-of-dose peaks is crucial which may be achieved with new pharmaceutical formulations . Our study also underscores the importance of investigating further neuronal ion channels like Nav and HCN isotypes and of studies in other organ systems , such as the heart and kidneys , to better understand the effect of Gb3 on for example cardiomyocytes in the generation of lethal arrhythmias . We believe that such approaches will open new avenues for mechanism-based diagnostics and treatment options for patients suffering from the life threatening FD . Our study was approved by the Bavarian State authorities ( Regierung von Unterfranken , # 54/12 ) . Animal use and care was in accordance with institutional guidelines . Mice were held in the animal facilities of the Department of Neurology , University of Würzburg , Germany . They were fed standard chow ( commercially prepared complete diet ) and had food and water access ad libitum . We used 95 GLA KO mice ( 45 male , 50 female ) of mixed genetic background ( C57BL6 and SVJ129 ) carrying a targeted disruption of the α-galactosidase A gene ( GLA ) as previously described ( Ohshima et al . , 1997 ) . Additionally , 96 WT littermate mice ( 45 male , 51 female ) were assessed . To ensure that our KO and WT mice have an identical genetic background , we first crossed GLA KO mice with C57BL6/N mice to generate heterozygous off-springs . These heterozygous mice were then cross-bred with each other to obtain homozygous female and male GLA KO and WT mice . In the further course of breeding , we mated these two homozygous lines only with genetically matching mice ( KO x KO , WT x WT ) of the respective strain . Thus , we generated two mouse strains ( homozygous GLA KO and WT ) with an identical genetic background . To maintain the purity of these strains , we performed genotype analysis on every single mouse born in our animal facility . For genotyping , we used the Kapa2G fast PCR Kit ( Kapa Biosystems , Wilmington , USA ) and the following primers: oIMR5947 , AGGTCCACAGCAAAGGATTG; oIMR5948 , GCAAGTTGCCCTCTGACTTC; oIMR7415 , GCCAGAGGCCACTTGTGTAG . Since FD shows age-dependent progression , we investigated young ( 3 months ) and old ( ≥12 months ) mice; old mice reached an age of up to 24 months . Animals were not stratified for sex , since in GLA KO mice α-GAL knockout results in a complete loss of enzyme activity in both sexes ( i . e . homozygous female mice , hemizygous male mice ) with similar pain behavior in contrast to human patients ( Lakomá et al . , 2014; Lakomá et al . , 2016; Rodrigues et al . , 2009; Üçeyler et al . , 2016 ) . Mice were sacrificed in deep isoflurane anesthesia ( CP-Pharma , Burgdorf , Germany ) and lumbar L3 and L5 DRG were dissected for quantitative real-time PCR . Tissue was obtained in the naïve state and was flash-frozen in liquid nitrogen for storage at −80°C before further processing . L4 DRG were collected for immunohistochemistry ( see below ) and were embedded in optimal cutting temperature medium ( TissueTek , Sakura Finetek , Staufen , Germany ) ; ganglia were stored at −80°C before further processing . For neuronal cell cultures , ten to twelve DRG pairs were dissected within 30 min after mice were sacrificed . Skin of footpads was dissected and incubated in 4% paraformaldehyde ( PFA , Merck Millipore , cat# 1 . 04005 , Billerica , Massachusetts , USA ) for three hours . After washing three times with phosphate buffer , skin samples were incubated in 10% sucrose at 4°C , were embedded in optimal cutting temperature medium , and stored at −80°C before further processing . Right L4 DRG of young and old GLA KO and WT mice were collected in 4% PFA ( Merck Millipore , cat# 1 . 04005; Billerica , Massachusetts , USA ) in 2% glutaraldehyde ( 25% stock solution , Serva , cat# 23115 , Heidelberg , Germany ) . Briefly , tissue was postfixed with 2% osmiumtetraoxid ( Chempur , cat# 006051 , Karlsruhe , Germany ) and dehydrated with an ascending aceton row ( Sigma-Aldrich , cat# 15364-56-4 , Taufkirchen , Germany ) . After embedding in plastic , 0 . 5 µm semithin sections were prepared using an ultramicrotome ( Leica EM UC7 , Leica Microsystems , Wetzlar , Germany ) and were stained with toluidine blue for light microscopy ( Axiophot two microscope , Zeiss , Oberkochen , Germany ) . Ten-µm DRG and sciatic cryosections were prepared with a cryostat ( Leica , Bensheim , Germany ) . We performed hematoxylin-eosin staining . Briefly , DRG cryosections were incubated in hematoxylin ( Sigma-Aldrich , cat# H3136 , Taufkirchen , Germany ) for 10 min and 25 s with 1% eosin ( Sigma-Aldrich , cat# 23251 , Taufkirchen , Germany ) . Afterwards , cryosections were dehydrated with an ascending ethanol row . To quantify cell size , neurons were surrounded using Fiji software ( ImageJ 1 . 50 g , Wayne Rasband , National Institute of Health , USA ) ( Schindelin et al . , 2012 ) and perimeter was calculated . For immunofluorescence , antibodies against TRPV1 ( goat , 1:500 , Santa Cruz , cat# SC-12498; Santa Cruz , California , USA ) , and HCN2 ( rabbit , 1:200 , Alomone Labs , cat# APC-030; Jerusalem , Israel ) were used . Five different Nav1 . 7 polyclonal antibodies were tested ( anti-rabbit , Alomone Labs: cat# ASC-008; anti- rabbit , cat# ASC-027; anti-guinea pig , cat# AGP-057 , Jerusalem , Israel; anti-mouse , Abcam , cat# ab85015 , Cambridge , UK; rabbit anti-Nav1 . 7: Y083 , generated from rat a . a . sequence 514–532 , Center for Neuroscience and Regeneration Research , Yale Medical School and Veterans Affairs Hospital , West Haven , Connecticut , USA ) . Additionally , antibodies against β- ( ΙΙΙ ) -tubulin ( chicken , 1:500 , Abcam , cat# ab41489 , Cambridge , UK ) , BiP ( rabbit , 1:5000 , Abcam , cat# ab21685 , Cambridge , UK ) and CD77 ( i . e . Gb3 , rat , 1:250 , Bio-Rad , cat# MCA579; Hercules , California , USA ) were used to document endoplasmic stress responses under pathophysiological conditions ( Lee , 2005 ) . We used goat anti-rabbit IgG , rabbit anti-goat IgG and goat anti-chicken IgG labelled with cyanine 3 . 18 fluorescent probe ( 1:50 , Amersham; Piscataway , New Jersey , USA ) and Alexa Fluor 488 anti-rat IgM ( 1:300; Jackson Laboratory; Bar Habor , Maine , USA ) as secondary antibodies . Negative controls were prepared by omitting the primary antibody . Photomicrographs were assessed manually ( Axiophot two microscope , Zeiss , Oberkochen , Germany ) using Spot Advanced Software ( Windows Version 5 . 2 , Diagnostic Instruments , Inc , Sterling Heights , USA ) . For quantification of ion channel positive cells , the total number of neurons per DRG sections ( three sections per mouse ) were counted with Fiji software ( ImageJ 1 . 50 g , Wayne Rasband , National Institute of Health , USA ) ( Schindelin et al . , 2012 ) and the percentage of immunoreactive neurons relative to the total number of neurons with a clear nucleus was calculated by an observer blinded to the genotype . Additionally , diameter of TRPV1 positive neurons were measured with Fiji software ( ImageJ 1 . 50 g , Wayne Rasband , National Institute of Health , USA ) ( Schindelin et al . , 2012 ) and neurons were categorized into small ( <25 µm ) and large ( >25 µm ) neurons . Forty-µm skin sections from footpads were prepared with a cryostat ( Leica , Bensheim , Germany ) . For immunofluorescence , antibodies against protein gene product-9 . 5 ( PGP9 . 5 , rabbit , 1:500 , UltraClone Limited , Isle of Wight , England ) were used . We applied goat anti-rabbit IgG labelled with cyanine 3 . 18 fluorescent probe ( 1:50 , Amersham; Piscataway , New Jersey , USA ) as secondary antibody . Intraepidermal nerve fibers were counted and the number of fibers per millimeter was calculated applying published counting rules ( Lauria et al . , 2005 ) . Confocal microscopy was performed on 10 µm cryosections of DRG obtained as described above . For immunofluorescence , antibodies against CD77 ( i . e . Gb3 , rat , 1:250 , Bio-Rad , cat# MCA579; Hercules , California , USA ) and β- ( ΙΙΙ ) -tubulin ( chicken , 1:500 , Abcam , cat# ab41489 , Cambridge , UK ) were used . We applied rabbit anti-rat IgM labelled with cyanine 3 . 18 fluorescent probe ( 1:50 , Amersham; Piscataway , New Jersey , USA ) and Alexa Fluor 488 coupled anti-chicken ( 1:300; Jackson Laboratory; Bar Habor , Maine , USA ) as secondary antibodies together with 4' , 6-diamidino-2-phenylindole ( 1:10 . 000; Sigma-Aldrich , cat# 28718-90-3 , Taufkirchen , Germany ) . Photomicrographs were acquired using an inverted IX81 microscope ( Olympus , Tokyo , Japan ) equipped with an Olympus FV1000 confocal laser scanning system , a FVD10 SPD spectral detector and diode lasers of 405 , 473 , 559 , and 635 nm . All images shown were acquired with an Olympus UPLSAPO60x ( oil , numerical aperture: 1 . 35 ) objective . For high-resolution confocal scanning , a pinhole setting representing one Airy disc was used . High-resolution confocal settings were chosen to meet an optimum resolution of at least three pixels per feature in x–y direction . In z-direction , 600 nm steps were used . 12-bit z-stack images were processed by maximum intensity projection and were adjusted in brightness and contrast . Images are shown as red-green-blue images . Image and video processing was performed with Fiji ( ImageJ 1 . 50 g , Wayne Rasband , National Institute of Health , USA ) ( Schindelin et al . , 2012 ) . Frozen DRG samples were processed using a Polytron PT 3100 homogenizer ( Kinematica , Luzern , Switzerland ) . Total RNA was isolated using TRIzol reagent ( Invitrogen , Carlsbad , CA , USA ) following the manufacturer’s instructions . Five hundred ng of RNA were then reverse transcribed with TaqMan Reverse Transcription Reagents ( Applied Biosystems , Darmstadt , Germany ) . Five µl of cDNA per sample were assessed with quantitative real-time PCR using TaqMan Universal Master Mix and the following target specific predesigned mouse TaqMan Gene Expression Assays ( Applied Biosystems , Darmstadt , Germany; Assay-IDs in brackets ) : TRPV1 ( Mm01246302_m1 ) , HCN2 ( Mm00468538_m1 ) , Nav1 . 7 ( Mm00450762_s1 ) . 18 s rRNA ( Hs99999901_s1 ) was used as an endogenous control . Quantitative real-time PCR reactions were performed in the 96-well GeneAmp PCR System 9700 cycler with the following cycler conditions: 2 min , 50°C; 10 min , 95°C; ( 15 s , 95°C; 1 min , 60°C ) x40 . Relative gene expression was calculated using the 2-ΔΔCt method . For protein analysis , ten to twelve DRG pairs per mouse were dissected ( see above ) and frozen at −80°C until further processing . To achieve sufficient tissue weight ( i . e . ≥300 mg ) , DRG of at least three mice were pooled on ice and were processed using a Polytron PT 3100 homogenizer ( Kinematica , Luzern , Switzerland ) in 500 µl phosphate buffered saline containing 20 µl protease inhibitor . The suspension was centrifuged 15 min at 1500 g and the supernatant was separated in aliquots à 200 µl . A mouse Nav1 . 7 enzyme-linked immunosorbent assay kit ( BlueGene , 0 , 1 ng/ml , cat# E03N0034 , Shanghai , China ) was used to determine Nav1 . 7 protein expression together with provided standards , following the manufacturer`s instructions and using undiluted samples . Mouse DRG neurons were dissected and cultivated in culture medium ( 100 ml TNB-100 , Biochrom , cat# F8023; Berlin , Germany , 25 mM glucose; 2 ml PenStrep , Life Technologies , cat# 15140–122; Carlsbad , CA , USA; 100 µl L-glutamine , Life Technologies , cat# 25030–032; Carlsbad , CA , USA; 2 ml protein-lipid-complex , Biochrom , cat# F8820; Berlin , Germany ) containing 25 ng/ml nerve growth factor ( 2 . 5S , Alomone Labs , cat# N-240; Jerusalem , Israel ) according to a previously published protocol ( Langeslag et al . , 2014 ) . DRG neurons of old GLA KO and WT mice , were dissected and cultured for 48 hr as described above . To analyze apoptosis , we performed a NucView 488 Caspase 3 Enzyme Substrate Assay ( Biotium , cat# 10403 , Fenton , California , USA ) according to the manufacturer`s protocol . As a positive control , cells of both genotypes were incubated with 500 nM staurosporine ( Abcam , cat# ab120056 , Cambridge , UK ) for 16 hr prior to performing the NucView 488 Caspase3 Enzyme Substrate Assay . For quantification of apoptosis , the percentage of caspase three positive neurons and the percentage of neurons with neurite outgrowth was determined . Whole-cell recordings were performed at room temperature three to eight days after isolation of DRG neurons and after axonal outgrowth . Bath solution consisted of 135 mM NaCl , 5 . 4 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 10 mM glucose , and 5 mM HEPES ( Eberhardt et al . , 2017; Hamill et al . , 1981 ) . Bath solution for HEK cells consisted of 140 mM NaCl , 3 mM KCl , 1 mM CaCl2 , 1 mM MgCl2 , and 10 mM HEPES . Patch pipettes were pulled from borosilicate glass capillaries ( Kimble Chase Life Science and Research Products , Meiningen , Germany ) and were heat-polished to reach an input resistance of 2 to 3 MΩ ( whole-cell ) . The pipette recording solution contained 140 mM KCl , 2 mM MgCl2 , 1 mM EGTA , 1 mM ATP , and 5 mM HEPES for DRG neuron analysis and 140 mM CsF , 1 mM EGTA , 10 mM NaCl , and 10 mM HEPES for HEK cell analysis . Currents were recorded with an EPC9 patch-clamp amplifier ( HEKA , Ludwigshafen , Germany ) with a sampling rate of 20 kHz . Stimulation and data acquisition were controlled by the PULSE/PULSEFIT software package ( HEKA , Lambrecht , Germany ) on a Macintosh computer , and data analysis was performed off-line with IGOR software ( WaveMetrics , Lake Oswego , Oregon , USA ) . To quantify TRPV1 currents , 500 nM capsaicin ( Merck Millipore , cat# 21127 , Billerica , Massachusetts , USA ) was used on DRG neurons . To investigate Ih currents , we used a series of depolarizing and hyperpolarizing step voltage pulses . To identify sodium channels , TTX ( Alomone Labs , cat# T-550; Jerusalem , Israel ) was applied to DRG neurons at a concentration of 100 nM and 1 µM using a standard perfusion system ( Solution Exchange System ALAVC3-8 , ALA Scientific Instruments , Farmingdale , New York , USA ) . Sodium currents were recorded continuously . For the quantification of TRPV1 and sodium currents , we performed measurements at maximum potential amplitudes; Ih currents were recorded at −120 mV . Current density was calculated by normalizing the measured potentials to cell size . DRG neurons with less than 25 pF capacity were considered as nociceptors . We investigated the effect of intraplantar injection of one µg capsaicin in 10 µl normal saline ( Merck Millipore , Billerica , Massachusetts , USA ) to the right hind paw of old GLA KO and WT mice under isoflurane narcosis . In a previous study a comparable dosage of intraplantar capsaicin led to short lasting ( <10 min ) pain behavior in mice ( Carey et al . , 2017; Sakurada et al . , 1992 ) . We determined heat withdrawal latencies one , six , and 24 hr after capsaicin injection in old GLA KO and WT mice . To model neuropathic pain , old mice of both genotypes received CCI of the right sciatic nerve ( Bennett and Xie , 1988; Sommer and Schäfers , 1998 ) . Briefly , mice were anesthetized with isoflurane and the right sciatic nerve was exposed . Three ligatures ( 7–0 prolene , Ethicon , Norderstedt , Germany ) with a distance of one mm each were loosely tied around the nerve proximal to its trifurcation until the ipsilateral hind paw flicked shortly . Behavioral tests were performed at baseline , three , seven , 14 , 21 , and 28 days after CCI . To induce inflammatory pain , mice of both genotypes and age-groups received an intraplantar injection of CFA ( Sigma-Aldrich , Taufkirchen , Germany ) . Ten µl CFA ( concentration: 20 pg/µl ) were applied intraplantarly to the right hind paw under isoflurane anesthesia . Behavioral tests were performed at baseline , one and 48 hr , and seven days after CFA injection . As a control , ten µl of normal saline 0 . 9% ( Braun , Melsungen , Germany ) were injected into the right hind paw of each control mouse . All behavioral tests were performed by the same experienced investigator blinded to the genotype and treatment groups . All animals were examined three times , each with a test interval of 1–2 days before interventions . Heat withdrawal latencies were determined using the Hargreaves method with a standard Ugo Basile algometer ( Comerio , Italy ) ( Hargreaves et al . , 1988 ) . Mice were placed on a glass surface within acrylic glass boxes and a radiant heat stimulus ( 25 IR ) was positioned under the plantar surface of the hind paw after 60 min of adaptation . The paw withdrawal latency was measured automatically . To avoid burn lesions , a stimulus cut-off time of 16 s was set . Each hind paw was tested three times ( at intervals of 5 min ) . Mechanical withdrawal thresholds were determined using the von-Frey test based on the up-and-down-method ( Chaplan et al . , 1994 ) . Mice within acrylic glass boxes were placed on a wire mesh . After adaption for 60 min , the plantar surface of the hind paw was touched with a von-Frey filament ( beginning at 0 . 69 g ) . Upon paw withdrawal the next thinner von-Frey filament was applied . If no paw withdrawal was observed , the next thicker von-Frey filament was used . Each hind paw was tested six times ( at intervals of 5 min ) . The 50% withdrawal threshold ( i . e . force of the von-Frey hair to which an animal reacts in 50% of the applications ) was calculated . HEK cells expressing Nav1 . 7 were prepared as described previously ( Cummins et al . , 1998 ) . Cells were cultured in Dulbeccos`s modified eagle medium ( DMEM ) /F12 ( Life Technologies , cat# 10565018; Carlsbad , California , USA ) containing 10% fetal calf serum , 1% PenStrep ( Life Technologies , cat# 15140–122; Carlsbad , California , USA ) , and 0 . 6 mg/ml Geneticin ( G418 , Life Technologies cat#10131035; Carlsbad , California , USA ) . To knock down the α-GAL gene expression , cells were transfected with small hairpin RNA from the MISSION shRNA Bacterial TRC2 library ( Sigma-Aldrich , Taufkirchen , Germany ) . TRC2 pLKO . 5-puro non-mammalian shRNA ( SHC202 ) was used as a control . TRC2-pLKO-puro vector containing shRNA sequence CCGGGATTCGCCAGCTAGCTAATTACTCGAGTAATTAGCTAGCTGGCGAATCTTTTTG ( Clone ID:NM_000169 . 2–458 s21c1 ) was amplified and transfected into HEK cells using lipofectamine 3000 transfection reagent ( Life Technologies , cat# L3000008 , Carlsbad , California , USA ) . Cells were transfected according to the manufacturer protocol in a six-well plate . Cells were co-transfected with shRNA plasmid and a plasmid expressing green fluorescent protein . HEK cells were incubated in DMEM/F12 medium containing transfection medium for three days ( 37°C , 5% CO2 ) . Transfection was repeated and cells were incubated for another three days . Cells transfected with shRNA and those with non-mammalian shRNA as a control were used for patch-clamp analysis and immunocytochemistry . We then treated transfected HEK cells with 1 . 32 µl ( 1 mg/ml ) agalsidase-α ( Shire , Saint Helier , UK ) and 250 µM lucerastat ( N-butyldeoxy-galactonojirimycin , Biomol , cat# Cay19520-1 , Hamburg , Germany ) to investigate , if functional ion channel alteration by Gb3 is reversible . Agalsidase-α is used as biweekly intravenous enzyme replacement therapy to treat patients with FD ( Eng et al . , 2001 ) . Lucerastat is an inhibitor of glucosylceramide synthase and provides a new therapeutic approach for Fabry disease patients ( Guérard et al . , 2018; Welford et al . , 2018 ) . Transfected HEK cells were incubated for 24 hr before patch-clamp analysis . To visualize Gb3 deposits in HEK cells , antibodies against CD77 ( i . e . Gb3 , rat , 1:250 , Bio-Rad , cat#; Hercules , California , USA ) were used . We applied Alexa Fluor 488 anti-rat IgM ( 1:300; Jackson Laboratory; Bar Habor , Maine , USA ) as secondary antibody together with 4' , 6-diamidino-2-phenylindole ( 1:10 . 000; Sigma-Aldrich , cat# 28718-90-3 , Taufkirchen , Germany ) . Photomicrographs were assessed manually ( Axiophot two microscope , Zeiss , Oberkochen , Germany ) using Spot Advanced Software ( Windows Version 5 . 2 , Diagnostic Instruments , Inc , Sterling Heights , USA ) . Statistical analysis and graph design were performed using SPSS software Version 23 ( IBM , Ehningen , Germany ) and GraphPad PRISM Version 5 . 03 ( GraphPad Software , Inc . , La Jolla , CA , USA ) . Data distribution was tested using the Kolmogorov-Smirnov test . The non-parametric Mann-Whitney U test for group comparisons was applied , since data were not normally distributed . Behavioral data were analyzed using a two-way ANOVA followed by Tukey’s post-hoc test after data transformation applying Johnson`s procedure . Data are expressed as line charts representing the mean and standard error of the mean . All other data are visualized as box plots representing the median value and the upper and lower 25% and 75% quartile and bar graphs representing the mean and standard error of the mean as appropriate . p-values<0 . 05 were considered significant .	Fabry disease is a life-threatening disorder that runs in families and affects many parts of the body . Symptoms begin in early childhood , often with episodes of burning pain in the hands and feet . As patients with Fabry disease grow older , sensory nerve fibers in their skin start to break down . As a result , affected individuals may often struggle to detect heat or cold against their skin . Mutations in a gene called alpha-galactosidase A cause Fabry disease . These mutations prevent the alpha-galactosidase A ( alpha-GAL ) enzyme from working properly . This enzyme breaks down fatty substances in the cells , in particular a molecule named globotriaosylceramide ( Gb3 ) . In patients with Fabry disease , Gb3 accumulates inside cells and is thought to cause pain , reduced temperature sensitivity , and loss of nerve fibers in the skin . But how it does this is still unclear . To find out more , Hofmann et al . studied mutant mice with a disrupted alpha-GAL gene , which consequently lack enzyme activity . Like patients , the mice accumulate Gb3 inside their sensory nerve cells as they age . This build-up of Gb3 damages the cells and reduces the function of ion channels ( passages for charged ions to enter and leave a cell ) in their membranes . This may contribute to the loss of nerve fibers and the reduced cold-warm sensitivity in Fabry patients . However , one particular ion channel is more abundant in elderly mutant mice than in normal animals . This channel , called TRPV1 , responds to high temperatures and also to capsaicin , the chemical that makes chilli peppers hot . Hofmann et al . propose that the accumulation Gb3 may be linked to the excessive activation of TRPV1 in the sensory nerve cells of patients with Fabry disease . This may in turn contribute to the heat-induced pain . By providing insights into the mechanisms underlying some of the symptoms of Fabry disease , these findings will assist researchers to develop new treatments . They will also be useful for clinicians who manage patients with the disorder . Further studies should investigate the exact cellular mechanisms linking Gb3 accumulation with changes in cellular activity .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2018	Characterization of small fiber pathology in a mouse model of Fabry disease
Meiotic chromosomes are highly compacted yet remain transcriptionally active . To understand how chromosome folding accommodates transcription , we investigated the assembly of the axial element , the proteinaceous structure that compacts meiotic chromosomes and promotes recombination and fertility . We found that the axial element proteins of budding yeast are flexibly anchored to chromatin by the ring-like cohesin complex . The ubiquitous presence of cohesin at sites of convergent transcription provides well-dispersed points for axis attachment and thus chromosome compaction . Axis protein enrichment at these sites directly correlates with the propensity for recombination initiation nearby . A separate modulating mechanism that requires the conserved axial-element component Hop1 biases axis protein binding towards small chromosomes . Importantly , axis anchoring by cohesin is adjustable and readily displaced in the direction of transcription by the transcriptional machinery . We propose that such robust but flexible tethering allows the axial element to promote recombination while easily adapting to changes in chromosome activity . Meiosis is a specialized developmental program in which a diploid cell undergoes two nuclear divisions to produce the haploid gametes required for sexual reproduction . A key event in meiosis is the programmed recombination of homologous chromosomes , which ensures proper chromosome segregation during the first meiotic division and also provides genetic variation in the offspring . Homologous recombination events initiate with programmed DNA double-strand breaks ( DSBs ) that are introduced by the topoisomerase-like enzyme Spo11 . Upon endonucleolytic removal of Spo11 from break ends , a subset of DSBs are repaired by crossing-over thereby physically linking homologous chromosome pairs for segregation during meiosis I . The genomic distribution of meiotic DSBs is decidedly non-random ( Gerton et al . , 2000; Blitzblau et al . , 2007; Buhler et al . , 2007; Pan et al . , 2011a ) , but the mechanisms driving this distribution , and thus the patterns of recombination , are not yet well defined . DSB formation and repair occur in the context of a specific chromosome architecture characterized by linear arrays of chromatin loops with the bases attached to a proteinaceous axis , known as the axial element ( Moens and Pearlman , 1988; Zickler and Kleckner , 1999 ) . In Saccharomyces cerevisiae most DSBs occur within the chromatin loops , whereas axis association sites are cold for DSB formation ( Blat et al . , 2002; Panizza et al . , 2011; Ito et al . , 2014 ) . The axial element in yeast contains the axis proteins Red1 and Hop1 ( Hollingsworth et al . , 1990; Smith and Roeder , 1997 ) , as well as the meiotic cohesin complex ( Klein et al . , 1999 ) . Red1 and Hop1 physically interact with each other ( de los Santos and Hollingsworth , 1999 ) , and Red1 helps recruit Hop1 to chromosomes ( Smith and Roeder , 1997 ) . Meiotic cohesin is not essential for chromosomal binding of Red1 and Hop1 , but Red1 and Hop1 distribution is highly abnormal in cohesin mutants ( Klein et al . , 1999; Panizza et al . , 2011 ) . Notably , this phenotype is not due to a loss of sister chromatid cohesion , as axis protein distribution is largely normal on unreplicated chromosomes ( Blitzblau et al . , 2012 ) . It is currently unclear whether cohesin directly interacts with Red1 and/or Hop1 in budding yeast . One of the first functions of Red1 and Hop1 is to correctly localize the essential DSB factors Mer2 , Rec114 and Mei4 to axis sites ( Panizza et al . , 2011 ) . Accordingly , DSB levels are strongly reduced in hop1Δ or red1Δ mutants ( Mao-Draayer et al . , 1996; Schwacha and Kleckner , 1997; Xu et al . , 1997 ) . Moreover , reduction in Red1 and Hop1 binding in the absence of cohesin correlates with reduced Spo11 binding and low DSB levels on several larger chromosomes ( Kugou et al . , 2009; Panizza et al . , 2011 ) . Conversely , increased binding of axis proteins on small chromosomes correlates with increased DSB levels and crossover recombination ( Kaback et al . , 1992; Blitzblau et al . , 2007; Pan et al . , 2011a; Panizza et al . , 2011 ) . The molecular determinants of axis attachment sites remain poorly defined . Red1 was reported to favor AT-rich segments on chromosome III at a resolution of 3 kb ( Blat et al . , 2002 ) , whereas Hop1 has preferential affinity for G-rich sequences and specific DNA topologies in vitro ( Kironmai et al . , 1998 ) . However , on meiotic chromosomes , Hop1 and Red1 binding coincides with cohesin binding sites and regular distribution of Hop1 requires cohesin ( Panizza et al . , 2011 ) , suggesting that cohesin is a major determinant of axis attachment sites . The yeast meiotic cohesin complex consists of Smc1 , Smc3 , Scc3 , and the kleisin Rec8 , which replaces the mitotic kleisin Scc1/Mcd1 ( Klein et al . , 1999; Watanabe and Nurse , 1999 ) . Despite the different subunit composition , the cohesin-associated chromosomal sites are highly conserved between mitosis and meiosis ( Blat and Kleckner , 1999; Glynn et al . , 2004 ) . A strong correlation between cohesin sites and regions of convergent transcription has been observed ( Lengronne et al . , 2004 ) , and changes in transcriptional activities can result in altered cohesin localization ( Lengronne et al . , 2004; Bausch et al . , 2007 ) . Because cohesin can encircle DNA from two sister-chromatids ( Ivanov and Nasmyth , 2005 , 2007 ) it is hypothesized to be pushed along genes ahead of the RNA polymerase II ( RNAPII ) complex , thus accumulating at convergent regions ( Ocampo-Hafalla and Uhlmann , 2011 ) . However , sliding has not been addressed in meiotic prophase , where cohesin may be stably incorporated in the axial element . Moreover , the mechanisms of meiotic axis protein loading and the functional link to cohesin remain to be elucidated . Here , we probed axis assembly using high-resolution chromatin binding studies of axis protein distribution and in vivo detection of physical associations between axis proteins . We found that axis proteins dynamically accumulate between convergent genes in a distinctive two-peak pattern , which we demonstrate to be a direct consequence of the underlying transcriptional activity . We show that cohesin is responsible for this behavior and that Rec8 interacts closely with both axis proteins via Red1 . In the absence of Rec8 , Hop1 becomes essential for chromosomal Red1 recruitment , defining a cohesin-independent mode of axis protein recruitment . Intriguingly , Hop1 but not Rec8 is required for the preferential loading of Red1 to short chromosomes , possibly by preventing excess Red1 accumulation on large chromosomes and at centromeres . Our findings demonstrate that the well-dispersed binding and chromosome size-biased enrichment of axis proteins are controlled by two independent recruitment modes and provide important insight into the mechanistic hierarchy of meiotic axis assembly . We used chromatin immunoprecipitation followed by deep sequencing ( ChIP-seq ) to determine the chromosomal distribution of two axis proteins , Red1 and Hop1 , as well as two cohesin subunits , Rec8 and Smc3 . To validate our data , we compared the resulting profiles with previous lower-resolution ChIP–chip data of Hop1 , Red1 , and Rec8 ( Blitzblau et al . , 2012 ) . A high correlation indicated that the identification of axis association sites was consistent between the two approaches ( Pearson's r = 0 . 80 ) . As observed previously , the genome-wide distribution of axis proteins and cohesin was highly correlated ( Figure 1—figure supplement 1A , B ) , with the exception of centromeric regions where cohesin was more highly enriched than axis proteins . Because of the high correspondence in localization , we chose the Red1 data set as the representative axis protein data set for subsequent analyses . We observed Red1 binding sites distributed across all chromosomes , with a notable bias toward the smallest chromosomes . Statistical analysis ( Zhang et al . , 2008 ) derived a total of 774 Red1 binding sites , in close agreement with previous reports , which identified 656 and 802 axis association sites , respectively ( Panizza et al . , 2011; Blitzblau et al . , 2012 ) . Binding was particularly strong on the three shortest chromosomes , which displayed a higher overall level of axis protein enrichment ( Figure 1—figure supplement 1C ) ( Panizza et al . , 2011 ) . Significantly , Rec8 exhibited no such bias , suggesting that this chromosome-specific patterning of axis proteins is not dictated by Rec8 ( see later sections ) . The enrichment of Red1 on small chromosomes mirrored a previously noted bias of average meiotic DSB levels , which are also highest on the three shortest chromosomes ( Blitzblau et al . , 2007; Pan et al . , 2011a; Thacker et al . , 2014 ) . Indeed , the numbers of axis sites on individual chromosomes were highly correlated with the number of Spo11 cleavage events ( Spo11-linked oligonucleotide counts; Figure 1—figure supplement 1D ) . Given that axis proteins are required for the majority of meiotic DSB formation ( Mao-Draayer et al . , 1996; Schwacha and Kleckner , 1997; Xu et al . , 1997; Kugou et al . , 2009 ) , their preferential enrichment on small chromosomes may contribute to the observed chromosome-size bias of meiotic DSB formation . Consistent with previous findings , DSB hotspots were largely excluded from sites of axis protein binding ( Blat et al . , 2002; Pan et al . , 2011a; Panizza et al . , 2011; Ito et al . , 2014 ) . Most axis sites detected by ChIP-seq were broad , ranging from 1 to 3 kb ( Figure 1—figure supplement 1E ) . Spo11 cleavage activity , as determined by sequencing Spo11-associated DSB ends ( Spo11 oligos ) , was strongly depleted in these sites ( Figure 1—figure supplement 1F ) . This pattern further corroborates the notion that DSBs occur preferentially within loop regions that are not bound by axis proteins . As is the case with DSB hotspots , motif discovery did not yield strong consensus sequences associated with Red1 binding . A repetitive sequence pattern consisting of varying numbers of a 3-bp GAN unit was found modestly enriched at Red1 peaks ( E-value = 6 . 6e-6 ) , with maximal enrichment at the summits of Red1 peaks ( Figure 1—figure supplement 1G , P-value = 5 . 0e-4 ) . However , this motif was not predictive of Red1 binding and was only present in 16 . 1% ( 125/774 ) of Red1 peaks , indicating that DNA sequence is at best a minor determinant of axis protein binding . To better understand the binding preference of axis proteins , we examined where Red1 binds relative to genes . We plotted the average Red1 binding signals with respect to all open reading frames ( ORFs ) . We found that Red1 enrichment was significantly biased toward the ends of ORFs , reaching a summit in the 3′ ends ( Figure 1A ) . Consistent with the high signal density on short chromosomes , we found an overall elevated signal along genes situated on short chromosomes . To further interpret the enrichment of axis proteins at 3′ ends of genes , we confined the analysis of binding peaks to 500 bp around Red1 summits . We found that 94 . 6% ( 732/774 ) of Red1 peaks were located at the 3′ ends of genes ( 250 bp on either side of the stop codon; Figure 1—figure supplement 1H ) . This bias contrasts with DSBs , which are predominantly detected at promoter regions ( Blitzblau et al . , 2007; Pan et al . , 2011a ) . Moreover , the Red1 signals were strongly enriched within the intergenic regions of convergent gene pairs as compared to tandem gene pairs ( Figure 1B and Figure 1—figure supplement 1H ) , a pattern previously described for cohesin-associated sites ( Filipski and Mucha , 2002; Lengronne et al . , 2004 ) . Closer inspection of convergent gene pairs in two independently generated data sets revealed that the Red1 binding signal typically consisted of two peaks at convergent 3′ ends , separated by a region of signal depletion ( Figure 1B ) . These data indicate that axis protein binding between convergent genes is subject to positional constraints . 10 . 7554/eLife . 07424 . 003Figure 1 . Red1 preferentially localizes to the 3′ ends of convergent genes . ( A ) Red1 distribution was plotted as an average across all genes on each of the 16 chromosomes as well as for the whole genome . The coding regions of genes were normalized to lengths of 1 kb ( 500 bp–1500 bp on x-axis ) . 500 bp upstream of the start codon and downstream of the stop codon were included in the plot . ( B ) Comparison of Red1 binding between convergent and tandem gene pairs . Average signals of Red1 binding were plotted within a 1 kb region from the midpoints of the intergenic regions . Signals were averaged at each bp among all the convergent ( red ) and tandem ( blue ) gene pairs across the genome . ( C ) qPCR analysis of Red1 binding at the YKL077W/YKL075C convergent gene pair before and after insertion of a URA3 transcription unit . Schematic shows the insertion of URA3 in two orientations between YKL077W and YKL075C . ChIP signals were normalized against input and an internal control; the control primer was chosen at the promoter region of YKL077W . Failure to detect a peak doublet upon URA3 insertion may be due to primer positions or because the URA3 transcript does not lead to overlapping transcription ( see Figure 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 00310 . 7554/eLife . 07424 . 004Figure 1—figure supplement 1 . Genome-wide localization of meiotic axis proteins and cohesin . ( A ) Distribution of axis proteins and cohesins as determined by ChIP-seq . Chromosome XI is shown as an example to show co-enrichment with the exception of the centromere ( indicated by black circle ) . Bottom panels are a zoom-in to show that axis proteins in general do not localize to DSB hotspots as measured by Thacker et al . ( 2014 ) . ( B ) Pairwise correlation between Red1 ChIP and cohesin ChIP signals . ( C ) Red1 and Rec8 ChIP signal per bp as a function of chromosome length . 25 kb to either side of the centromeres were excluded from this analysis to avoid biases caused by the strong centromere-proximal enrichment of Rec8 ( see A ) . The three shortest chromosomes are displayed as solid dots . ( D ) Numbers of Red1 peaks ( see ‘Materials and methods’ for peak calling ) were correlated ( Pearson's r ) with the numbers of DSB hotspots on each chromosome . ( E ) Distribution of widths for 774 Red1 peaks . ( F ) Spo11 oligos ( green ) were strongly depleted at the axis protein associated regions ( red ) . Average Spo11 oligo density is plotted as a function of distance from Red1 peak summits . Red1 signals were averaged among all the peaks . ( G ) Percentage of Red1 peaks ( top ) and DSB hotspots ( bottom ) in different positions relative to genes . Peaks were defined as a 500 bp-surrounding region of summits . 5′ regions of genes were defined as 500 bp upstream of the start codons , and 3′ regions of genes were defined as 250 bp on either side of the stop codons . Divergent ( Div . ) or convergent ( Conv . ) regions refer to the intergenic areas between two adjacent start or stop codons , respectively . ( H ) Putative motif at axis sites . Top: motif derived from Red1 ChIP-seq using MEME-ChIP ( Machanick and Bailey , 2011 ) . Bottom: probability of motif occurrence around Red1 peak summits . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 004 To examine whether convergent gene arrangement was required to recruit axis proteins , we inserted a URA3 gene into the intergenic region between YKL077W and YKL075C , which created a new convergent gene pair and a tandem gene pair ( Figure 1C ) . Strikingly , the enrichment at the 3′ end of the new tandem gene disappeared after the insertion , whereas the peak remained at the newly constructed convergent gene pair . To eliminate possible sequence bias , we inserted URA3 at the same locus but in the opposite direction . Again , Red1 binding was only detected between the convergent gene pair . These results indicate that convergent gene orientation is both necessary and sufficient for enrichment of axis proteins at this axis association site and demonstrate that DNA sequence is not a major determinant of axis protein binding . We further investigated the unexpected double peaks of axis protein association between convergent gene pairs . One important feature of the compact yeast genome is that convergent intergenic regions are usually very short , and convergent transcripts frequently overlap ( David et al . , 2006 ) . Accordingly , our analysis of published meiotic mRNA-seq data ( Brar et al . , 2012a ) revealed that 79% of convergent transcript pairs exhibited overlapping 3′ UTRs during the early stages of meiosis ( Figure 2A , B ) . After ranking the convergent gene pairs according to the extent of overlap of their 3′ UTRs , the presence and spacing of the two Red1 peaks appeared directly correlated with the amount of overlap . Gene pairs with an extensive overlap showed well-separated peaks , whereas gene pairs with little or no overlap exhibited a single peak at the midpoint of their intergenic regions ( Figure 2B ) . Red1 binding was maximally enriched ∼150 bp downstream of the 3′ ends of the respective transcripts , a behavior also observed for cohesin subunits ( Figure 2C ) . To test for a possible role of nucleosomes in axis protein localization , we also plotted nucleosome occupancy ( Pan et al . , 2011a ) . No overall correlation between nucleosome depletion and the gap between the axis protein binding signals was detected , although there was a decline of nucleosome density at a fraction of the convergent intergenic regions ( Figure 2B ) , indicating that nucleosomes do not position Red1 . Together , these data imply that the local transcriptional landscape shapes axis association patterns . 10 . 7554/eLife . 07424 . 005Figure 2 . Red1 accumulates next to transcript ends of convergent gene pairs . ( A ) Example of overlapping transcripts from a convergent gene pair . ORFs and transcripts of YKL077W and YKL075C were plotted with respect to their positions on chromosome XI . ( B ) Two-peak pattern of Red1 binding at convergent gene pairs correlates with the amount of transcript overlap . Convergent gene pairs were ranked by the extent of transcript overlap ( Brar et al . , 2012a ) . Red1 signals ( red ) were plotted within 1-kb regions centered at the midpoints between of transcript ends . Nucleosome occupancy signals ( blue ) ( Pan et al . , 2011a ) were also plotted as described above . The two blue curves represent the positions of transcript ends of each gene ( dark blue = forward direction , light blue = reverse direction; see schematic to the left ) . ( C ) Enrichment of cohesin subunits and axis proteins was averaged for all convergent transcript ends and plotted as a function of distance from the transcript ends . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 005 We investigated whether transcription levels of the underlying genes drive the distribution of Red1 . We first extracted gene pairs in which both genes were highly expressed or both genes were lowly expressed and tested for differences in Red1 signals . This analysis revealed higher and more confined Red1 occupancy between two highly transcribing convergent genes than between gene pairs with less transcriptional activity ( Figure 3A ) , indicating that ongoing transcription helps to focus axis proteins between convergent gene pairs . We also noted that the relative heights of the two Red1 peaks between a pair of convergent transcripts often differed substantially . Therefore , we sorted convergent gene pairs based on the relative transcriptional activity of the two genes . This analysis uncovered a strong reciprocal relationship between transcriptional strength and axis protein enrichment . In gene pairs with differential transcriptional activity , Red1 tends to accumulate downstream of the more highly expressed side . Accordingly , peaks of equal heights were present in gene pairs with similar transcriptional strengths . The reciprocal relationship was apparent both when average Red1 occupancy was calculated for six ranked quantiles and when the sorted gene pairs were displayed as a heatmap ( Figure 3B ) . This enrichment pattern suggests that axis proteins are not tightly bound to specific DNA sites but instead are focused between convergent genes driven by some aspect of their transcription . 10 . 7554/eLife . 07424 . 006Figure 3 . Transcriptional activity dictates Red1 binding patterns . ( A ) Convergent gene pairs were identified , in which both genes were strongly transcribed ( red ) or both genes were weakly transcribed ( blue ) and average Red1 signal was determined around the midpoints of the intergenic regions . The upper panel represents a cutoff of 50% ( both genes were among the top 50% highly transcribed or among bottom 50% lowly transcribed genes ) ; the lower panel represents a cutoff of 25% . ( B ) The Red1 binding signals are biased toward the more weakly transcribed gene at convergent gene pairs . Convergent gene pairs were ranked according to their differences in RPKM values ( ΔRPKM = RPKMforward − RPKMreverse ) , and Red1 binding signals were plotted ( right panel ) . The left panel shows the average Red1 binding signal in six quantiles of ranked ΔRPKM . Arrow sizes schematically represent the relative transcriptional activities of convergent gene pairs . ( C ) Gene expression of the GCY1/PFY1 convergent gene pair in response to different concentrations of copper was measured by RT-qPCR at the indicated time points . Copper was added to the sporulation medium at 2 hr , and samples were collected at 2 hr and 3 hr . Transcription levels of both convergent genes were examined in a wild-type strain and a strain harboring a pCUP1 promoter insertion upstream of GCY1 . Error bars: S . D . of three independent replicates . ( D ) ChIP-qPCR analysis of Red1 binding from the same experiment as in ( C ) . Schematic depicts relative position of primer pairs . Error bars: S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 00610 . 7554/eLife . 07424 . 007Figure 3—figure supplement 1 . Red1 position changes in response to transcription . ( A ) Gene expression of the GAL2/SRL2 convergent gene pair in response to different concentrations of copper was measured by RT-qPCR at the indicated time points . Copper was added to the sporulation medium at 2 hr , and samples were collected at 2 hr and 3 hr . Transcription levels of both convergent genes were examined in a wild-type strain and a strain harboring a pCUP1 promoter insertion upstream of GAL2 . Error bars: S . D . of three independent replicates . ( B ) ChIP-qPCR analysis of Red1 binding from the same experiment as in ( A ) . Schematic depicts relative positions of primer pairs . Error bars: S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 007 To further test this possibility , we examined whether changing the level of gene expression is sufficient to alter Red1 binding patterns at axis association sites . We utilized the copper-inducible CUP1 promoter ( pCUP1 ) to drive the expressions of two genes: GCY1 and GAL2 . Both genes exhibit low transcriptional activities during early stages of sporulation and are located in convergent gene pairs , in which the convergently transcribed neighbor is more highly expressed . Two concentrations of copper ( 5 μM and 20 μM ) were applied to induce different levels of expression during early meiosis . In wild-type control cells , the Red1 binding profile was skewed to the sides of the weakly transcribed GCY1 and GAL2 genes irrespective of copper concentrations ( Figure 3C , D shows GCY1; see Figure 3—figure supplement 1A , B for GAL2 ) . A similar profile was also observed for the pCUP1 strains in the absence of copper , although a shift of Red1 binding away from pCUP1 was apparent and correlated well with the leaky expression of this promoter ( compare Figure 3C , D; also Figure 3—figure supplement 1A , B ) . Addition of 5 μM Cu2+ induced a marked shift in the Red1 binding pattern toward the downstream gene . This shift was further exacerbated upon addition of 20 μM Cu2+ ( Figure 3D and Figure 3—figure supplement 1B ) , indicating that the transcription levels of convergent genes underlie the dynamic binding patterns of axis proteins . Such refocusing could either result from sliding motion or repeated recruitment to a chromatin modification or structure at the 3′ ends of transcription bubbles . Previous studies have shown that cohesin is able to change position in response to transcription ( Glynn et al . , 2004; Lengronne et al . , 2004; Bausch et al . , 2007 ) . Thus , association with cohesin may explain the transcriptional effects we observe for axis proteins . If cohesin plays a role in the transcription-dependent localization of meiotic axis proteins , these distribution patterns should no longer be apparent when the cohesin ring is disrupted . Our previous analysis had shown that in the absence of meiotic cohesin Rec8 , Hop1 still associates with chromosomes but displays large regions of depletion alternating with dense clusters of binding ( Panizza et al . , 2011 ) , a pattern that was also mirrored by Red1 ( Figure 4A ) . These large-scale changes in rec8Δ mutants were associated with markedly narrower Red1 peaks ( Figure 4—figure supplement 1A ) . Importantly , Red1 was no longer biased to the 3′ ends of genes ( Figure 4B ) and focusing of axis proteins at convergent regions was strongly reduced ( Figure 4C ) , implying that transcriptional focusing of axis proteins depends on cohesin . By contrast , the small-chromosome bias of axis protein binding persisted in rec8Δ mutants ( Figure 4D ) , indicating that Red1 is enriched on small chromosomes independently of Rec8 . 10 . 7554/eLife . 07424 . 008Figure 4 . Transcriptional focusing depends on cohesin . ( A ) Chromosomal localization of Red1 in rec8Δ mutants ( blue ) and WT ( gray ) on chromosomes XI and VI . Bottom panel shows zoom-in for a convergent gene pair on chromosome VI . ( B ) Red1 enrichment along genes in rec8Δ mutants and WT . ( C ) Average Red1 accumulation in convergent gene regions in rec8Δ mutants and WT . ( D ) Average Red1 ChIP signal per bp in rec8Δ mutants and WT as a function of chromosome length . ( E ) Alignment of chromosomal Red1 binding in rec8Δ mutants ( blue ) with regions exhibiting high coding density ( red ) . Red1 signals were plotted using a 5-kb smoothing window . Coding density was calculated using 0 ( intergenic ) or 1 ( coding ) at each nucleotide position and then smoothed using a 10-kb window and plotted as a heatmap . ( F ) Red1 signal per bp in rec8Δ mutants and WT as a function of gene lengths ( ORFs ) . Genes were clustered into 64 groups of 100 genes according to similar gene lengths . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 00810 . 7554/eLife . 07424 . 009Figure 4—figure supplement 1 . Red1 binding in rec8Δ mutants . ( A ) Distribution of widths of Red1 peaks in WT ( green ) and rec8Δ mutants ( blue ) . ( B ) Probability of GAN motif occurrence around Red1 peak summits in WT ( green ) and rec8Δ mutants ( blue ) . ( C ) Examples of correlation of Red1 binding and coding density along chromosomes III and X . Red1 signals were plotted unsmoothed . Coding density was calculated as in Figure 4E but using a smoothing window of 2 kb . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 009 We investigated the chromosomal landmarks that may direct the formation of axis protein clusters in cohesin mutants . Red1 enrichment over the GAN repeat motif was only modestly increased over background ( Figure 4—figure supplement 1B ) , indicating that this motif is not responsible for directing most Red1 binding in cohesin mutants . We also observed no apparent association with nucleotide bias , replication origins , tRNAs , or transposable elements . However , we did note a weak but consistent correlation ( Pearson's r = 0 . 37 ) between axis protein enrichment and the local coding density of the genome . Specifically , regions enriched for long ORFs or tightly clustered short ORFs exhibited increased axis protein association ( Figure 4E and Figure 4—figure supplement 1C ) . This effect appeared to be cumulative , as longer ORFs displayed higher Red1 binding per kb than shorter ORFs ( Pearson's r = 0 . 70; Figure 4F ) . Enrichment of Red1 on gene bodies was not apparent in wild-type cells ( Figure 4F ) , indicating that cohesin overrides this cohesin-independent axis recruitment activity . Thus , in wild-type cells , Red1 and Hop1 form axis attachment sites by recognizing a feature of cohesin-associated sites , but a cohesin-independent mechanism may contribute additional binding specificity . To further define the mechanisms of axis protein recruitment , we analyzed the mutual dependencies for binding to chromosomes . ChIP-seq analysis of N-terminally tagged V5-Red1 revealed that the majority of Red1 binding sites were unaltered for their positions in the absence of Hop1 ( Figure 5A ) , indicating that Hop1 is not essential for recruiting Red1 to axis association sites . We note , however , that Red1 binding appeared to increase at some positions ( see next section ) . Thus , Rec8 determines the positions of Red1 along chromosomes , while Hop1 seems to modulate its amount . Strikingly , Red1 DNA association is largely undetectable in hop1Δ rec8Δ double mutants ( Figure 5A ) , indicating that Hop1 is required for the cohesin-independent recruitment of Red1 . Hop1 , on the other hand , is undetectable on chromatin in absence of its partner Red1 even in the presence of Rec8 ( Figure 5B ) . Together , these data suggest that Red1 and Hop1 associate with DNA as a complex in the absence of cohesin . By contrast , Hop1 is not required for the recruitment of Red1 to sites of cohesin binding . Finally , with few exceptions , Rec8-HA signals were highly similar in wild-type or red1Δ mutants ( Figure 5C ) , establishing that cohesin recruits Red1 unilaterally . 10 . 7554/eLife . 07424 . 010Figure 5 . Chromosomal localization of Red1 requires Rec8 and Hop1 redundantly . ( A ) Red1 binding profiles in WT ( green ) , hop1Δ mutants ( blue ) , and rec8Δ hop1Δ mutants ( purple ) from ChIP-seq experiments ( NCIS normalized ) . A small ( Chromosome I ) and a medium sized chromosome ( Chromosome V ) are shown as examples to illustrate the reduced Red1 density on chromosome V in WT but not in hop1∆ mutants ( indicated schematically by the double-headed arrows ) . Arrowheads indicate increased Red1 binding confirmed by qPCR in ( E ) . ( B ) Genome-wide Hop1 ChIP-seq profiles in red1∆ mutants ( gold ) ( NCIS normalized ) . Chromosomes I and V are shown . ( C ) Genome-wide Rec8 ChIP-seq profiles in WT ( orange ) and red1∆ mutants ( red ) ( NCIS normalized ) . Chromosomes I and V are shown . ( D ) Red1 density/20 bp ( sum of Red1 signal after NCIS normalization divided by chromosomal length ) plotted against chromosomal length . The three smallest chromosomes exhibit highly increased binding in WT ( green dots ) , but this difference is largely abolished in hop1∆ mutants ( blue dots ) . Red1 density is reduced to noise level and no longer shows biased distribution in hop1∆ rec8∆ mutants ( purple dots ) . Horizontal lines indicate the mean of 16 chromosomes ( continuous lines ) , plus minus twofold standard deviation ( dashed lines ) . ( E ) White bars: ratio of V5-red1 ChIP-Seq signal ( hop1∆/HOP1 ) at six cohesin peaks that decrease ( a-f: chr3 219k , chr1 171k , chr1 195k , chr6 216k , chr1 95k , chr1 134k ) and six cohesin peaks that increase ( g-l: chr1 156k , chr5 274k , chr4 435k , chr4 712k , chr7 506k , chr16 576k ) upon HOP1 deletion ( data from the experiment shown in ( A ) ) . Yellow bars: ratio of V5-Red1 signals ( hop1∆/HOP1 ) obtained from ChIP-qPCR primer pairs located at the indicated positions . Gray bars: ratio of V5 signals ( untagged/V5-Red1 , HOP1 ) obtained at the corresponding positions . The gray line separates reduction ( below 1 ) from increase ( above 1 ) . Arrowheads indicate positions on chr1 and chr5 shown in ( A ) . DNA from each biological repeat was evaluated both at increasing and decreasing sites to exclude systematic bias in DNA preparation . Each single repeat confirmed the direction of change at each location . n indicates the number of repeats . ( F ) Average densities of V5-Red1 ChIP-seq signals over all 16 yeast centromeres , aligned at their midpoints . The averaged peaks are fourfold higher in hop1∆ mutants ( blue ) than in HOP1 cells ( green ) . The fold increase at centromeres exceeds the ∼twofold increase observed for the entire chromosomes ( see D ) . ( G ) Rec8-HA ( orange ) and Rec8-HA red1∆ ( red ) flank the centromere at a slightly wider distance ( around 345 bp ) than Red1 ( 270 bp ) . ( H ) V5-Red1 ( light green ) and Rec8-HA ( orange ) are shown as in ( G ) . The average transcription levels from 4 hr after meiotic induction ( Brar et al . , 2012a ) around centromeres ( forward—black , reverse—dark green ) reveal a transcription free pocket , in which Red1 resides . Meiotic cohesin peaks almost coincide with the ends of transcripts . Rec8-HA and V5-Red1 are shown on two different scales . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 01010 . 7554/eLife . 07424 . 011Figure 5—figure supplement 1 . Analysis of axis protein binding . ( A ) ChIP-qPCR analysis of axis proteins at selected positions in the indicated mutant backgrounds was performed with the precipitated DNA of the respective ChIP-seq experiment , before library amplification , for independent assessment of relative scaling of different profiles as shown in Figure 5 . Gray bars indicate signal independent of the respective tag ( V5-Red1 or Rec8-HA ) or antibody ( hop1∆ ) . X-axis labels indicate genomic positions of primer pairs . ( B ) Averaged densities of V5-Red1 ChIP-seq signals over all 16 yeast centromeres , aligned at their midpoints . Two rather low V5-Red1 peaks ( green ) flank the centromere midpoint at a distance of around 220 bp . rec8∆ mutants ( brown ) or rec8∆ hop1∆ double mutants ( purple ) show reduced peaks . ( C ) The meiotic axis remodeler Pch2 is not responsible for the loss of interaction of Hop1 ( or Red1 ) with Rec8 in red1∆ ( or hop1∆ ) mutants , respectively . Upper panels: detection of interaction at 4 hr in SPM using anti-H3K9me3 antibody . Lower panels: total Hop1-HA-H3 ( Red1-HA-H3 ) protein as determined by Western against the HA tag . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 011 Analysis of Red1 binding in hop1Δ mutants indicated that peak heights were affected in a non-uniform manner . Specifically , Red1 binding appeared to increase on large chromosomes and close to centromeres . In wild-type cells , we observed higher overall levels of Red1 on the 3 smallest chromosomes ( 2 . 5-fold over the rest , Figure 5D ) , correlating well with the higher enrichment of DSB machinery ( Panizza et al . , 2011 ) and increased DSB levels ( Blitzblau et al . , 2007; Pan et al . , 2011a ) on small chromosomes . The over-representation of Red1 was nearly abolished , falling to 1 . 25-fold , in the hop1Δ mutant ( Figure 5D ) . Thus , Hop1 modulates Red1 deposition in a chromosome-dependent manner . Relative scaling of ChIP signals between different experiments by NCIS ( Normalization of ChIP-Seq data with untagged control ) ( Liang and Keles , 2012 ) suggested that the reduced chromosome-size bias may be due to increased Red1 enrichment on the larger chromosomes in the hop1Δ mutant . To independently validate the ratio of Red1 occupancy between hop1∆ and wild-type strains , we used qChIP at 6 loci with increased and 6 loci with decreased Red1 enrichment . qChIP confirmed the direction of Red1 signal change predicted from ChIP-seq at all 12 genomic positions . Figure 5E shows the comparison between the ratios expected from ChIP-seq and those observed by qChIP . These results identify Hop1 as a chromosome-dependent regulator of Red1 deposition and suggest that it down-regulates accumulation of Red1 on the thirteen largest chromosomes . We also observed increased Red1 binding near centromeres in hop1Δ mutants . A doublet of small V5-Red1 peaks was found to flank the yeast centromeres at a stereotypic distance of about 270 bp from their centers ( Figure 5F ) . In rec8Δ and rec8Δ hop1Δ double mutants , these peaks were strongly decreased or missing ( Figure 5—figure supplement 1B ) . By contrast , in hop1Δ mutants , the centromere proximal peaks of Red1 were strongly enhanced ( Figure 5F ) , suggesting that Rec8 recruits Red1 next to the centromere against negative regulation by Hop1 , which may in turn possibly prevent unwanted centromere proximal recombination . The precise positioning of axis proteins at centromeres allowed us to probe the arrangement of cohesin and Red1 within axis association sites in more detail . Whereas the central 200 bp of the budding yeast point centromeres containing the cenH3 ( Cse4 ) hemi-nucleosome ( Henikoff and Furuyama , 2012; Henikoff et al . , 2014 ) were devoid of Rec8 , we detected two Rec8 peaks flanking the centromeres at roughly 345 bp from the center ( Figure 5G , H ) . Rec8 accumulation aligned with the ends of transcripts surrounding the largely transcript-free zone of the centromeres ( Figure 5H ) and was unchanged in red1Δ mutants . Red1 peaks are slightly closer to the centromeres ( ∼270 bp ) , raising the possibility that Red1 is sandwiched between the Cse4 hemi-nucleosome , which may act as a strong barrier , and cohesin , which in turn might be driven by the advancing RNAPII complex . Given that Rec8 is essential for Red1 recruitment along chromosome arms in hop1Δ mutants , we investigated whether the two proteins physically interact . Because interaction between Rec8 and Red1 could preferentially occur in insoluble chromatin , we first used M-track ( Zuzuarregui et al . , 2012 ) to test their interaction in vivo . M-track uses HKMT , a histone lysine methyltransferase and its acceptor sequence , a histone H3 fragment , each fused to one of the proteins suspected to interact , to create permanent methyl marks on the acceptor even upon transient interaction of the proteins of interest ( Figure 6A ) . When HKMT was fused to Rec8 , Red1-HA-3xH3 received a robust H3K9me3 signal ( Figure 6B ) . The signal existed , as long as Red1 was detectable . We also observed methylation of Hop1-HA-3xH3 by Rec8-HKMT ( Figure 6C ) . In agreement with our ChIP-seq data ( Figure 5A ) , methyl transfer to Hop1 , but not Hop1 protein stability , depended on the presence of Red1 ( Figure 6C ) , showing the specificity of the assay . Unexpectedly , Red1-HA-3xH3 depended on Hop1 for its interaction with Rec8 ( Figure 6C , right panels ) . This result was at odds with the observation that chromosomal recruitment of untagged Red1 did not require HOP1 ( Figure 5A ) . To investigate this discrepancy , we performed ChIP analysis of Red1 with or without the C-terminal HA-3xH3 tag . Only Red1 tagged at the C-terminus depended on Hop1 for chromosomal binding ( Figure 6D ) . Therefore , Hop1 contributes to the interaction between Red1 and Rec8 . Moreover , Red1 requires its intact C-terminus when interacting with Rec8 in the absence of Hop1 . 10 . 7554/eLife . 07424 . 012Figure 6 . Red1 and Hop1 interact with Rec8 . ( A ) Illustration of the physical proximity assay M-track . The writer ( Rec8 ) is tagged at on its C-terminus with a human derived histone lysine methyltransferase ( HKMT ) , which will transfer up to three methyl groups to lysine 9 of a small histone H3 fragment presented by the recipient ( Red1 ) depending on the lifespan of the interaction . Using an H3-K9me3 specific antibody , the interaction between writer and recipient can be visualized ( Zuzuarregui et al . , 2012 ) . ( B ) Stable proximity between Red1 and Rec8 . Upper panel: detection of interaction between Rec8 and Red1 in a meiotic time course using anti-H3-K9me3 antibody . Middle panel: total Red1-HA-H3 protein using anti-HA antibody . Lower panel: K9me3 antibody signal depends on the H3 tag . ( C ) Left panels: proximity between C-terminally tagged Hop1 and Rec8 depends on Red1 . Upper panel: detection of proximity between Rec8 and Hop1 at 4 hr in SPM using anti-H3-K9me3 antibody . Lower panel: total Hop1 protein as determined using anti-HA antibody . Right panels: proximity between C-terminally tagged Red1 and Rec8 depends on Hop1 . Upper panel: detection of proximity between Rec8 and Red1 at 4 hr in SPM using anti-H3-K9-me3 antibody . Middle panel: total Red1-HA-H3 protein as determined by Western against the HA tag . Lower panel: loading control . ( D ) When the C-terminus of Red1 is tagged , Hop1 becomes essential for the chromosomal recruitment of Red1 . qChIP ( using anti-HA antibody ) of Red1-HA2-H3 ( blue ) , Red1-HA2-H3 in the presence of Rec8-HKMT ( red ) , Red1-HA2-H3 in hop1Δ mutants ( green ) and untagged Red1 ( turquoise ) at 7 peak sites ( see Supplementary file 1B ) . The values for V5-Red1 in hop1Δ mutants ( purple ) were taken from the profiles shown in Figure 5A and put in proportion to Red1 WT ( see also qPCR for V5-Red1 in Figure 5—figure supplement 1A ) . ( E ) Left panel: at 4 hr in SPM Red1 was immuno-precipitated with anti-V5 antibody and co-precipitating Rec8 was detected with anti-HA antibody . Lanes from left to right are V5-RED1 hop1∆ , V5-RED1 HOP1 , RED1 HOP1 . Right panel: Western blot ( TCA ) of the cultures at the time of the IP . Lanes from left to right are V5-RED1 hop1∆ , V5-RED1 HOP1 , RED1 HOP1 , RED1 HOP1 . Rec8 was tagged with HA at the C-terminus except in the rightmost lane , where it was untagged . Swi6 was stained on the stripped blot as a loading control . ( F ) Same type of experiment as in ( E ) , but in two strains Scc1-HA was expressed from the REC8 promoter , while REC8 was deleted . Lanes from left to right are V5-RED1 pREC8-SCC1-HA rec8∆ , RED1 pREC8-SCC1-HA rec8∆ , V5-RED1 SCC1 REC8-HA . Right panel: Western blot ( TCA ) of the same cultures at the time of the IP . Swi6 served as loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 012 The robust signal produced by M-track suggested a stable physical interaction and was in line with previous mass spectrometric detection of Hop1 in a precipitate of Rec8 ( Katis et al . , 2010 ) , although Red1 was not detected in that experiment . To test for Red1-Rec8 interaction more directly , we precipitated V5-Red1 from sonicated and nuclease digested extracts using an anti-V5 antibody and tested the precipitate for the presence of Rec8-HA . Indeed , Rec8-HA co-precipitated and was dependent on the V5-tag of Red1 ( Figure 6E ) . Together with the ChIP-seq localization data this results suggests a model in which Rec8 recruits Red1 , which further recruits Hop1 . To investigate the importance of cohesin-mediated axis patterning for meiotic recombination initiation , we analyzed DSB formation in rec8Δ mutants . We used microarray-based measurement of DSB-associated single-stranded DNA ( ssDNA arrays ) to locate DSB hotspots in a dmc1Δ mutant background , which prevents DSB repair and allows cumulative DSB measurements ( Blitzblau et al . , 2007; Buhler et al . , 2007 ) . In the presence of cohesin , DSB hotspots were distributed across most of the genome . By contrast , DSB formation in rec8Δ mutants was largely restricted to small chromosomes and the vicinity of Red1 clusters ( Figure 7A , B ) . Thus , DSB formation is limited to the genomic neighborhoods that exhibited persistent Red1 and Hop1 binding in rec8Δ mutants . These findings are consistent with our previous observation that these regions selectively retain essential components of the meiotic DSB machinery ( Panizza et al . , 2011 ) . Strikingly , despite the loss of transcriptional focusing of axis association sites in the absence of cohesin ( Figure 4C ) and the overall altered patterns of Red1 distribution , DSB formation within Red1 clusters occurred at the same hotspots as in wild-type cells ( Figure 7B ) . Thus , hotspot designation does not depend on the precise location of axis binding sites . These data suggest that the general proximity of axis proteins is sufficient to activate DSB hotspots and that the primary role of Rec8 in controlling DSB formation is to broadly distribute axis proteins across meiotic chromosomes . 10 . 7554/eLife . 07424 . 013Figure 7 . Mitotic cohesin rescues transcriptional focusing but not genome-wide defects of rec8Δ mutants . ( A ) Relative DSB hotspot intensity of each chromosome in WT ( green ) and rec8Δ mutants ( blue ) as determined using ssDNA arrays . Chromosomes are sorted by increasing size . ( B ) DSB activity spatially correlates with Red1 enrichment along chromosomes . Distribution of Red1 enrichment ( top panels ) and DSB hotspots ( middle panels ) along chromosome XIII in WT ( green ) and rec8Δ mutants ( blue ) . Zoom-in in bottom panels shows that DSB hotspots location is unaltered at finer scales in rec8Δ mutants despite severely altered Red1 distribution . ( C ) Ectopic expression of mitotic cohesin does not rescue the defects in Red1 localization or DSB formation of rec8Δ mutants . Distribution of Red1 enrichment ( top panels ) and DSB hotspots ( bottom panels ) along chromosome XII in WT ( green ) and rec8Δ ( blue ) and rec8Δ pREC8-SCC1 ( orange ) . ( D ) Ectopic expression of mitotic cohesin partially rescues transcriptional focusing of Red1 in rec8Δ mutants . Average Red1 accumulation in convergent gene regions in WT ( green ) , rec8Δ ( blue ) , and rec8Δ pREC8-SCC1 ( orange ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 01310 . 7554/eLife . 07424 . 014Figure 7—figure supplement 1 . Ectopically expressed Scc1 localizes to the same sites as Rec8 in meiosis . Top panel: ChIP–chip analysis of Rec8-3HA in otherwise WT cells ( green ) . Bottom panel: ChIP–chip analysis of 3HA-Scc1 in rec8Δ , pREC8-3HA-SCC1 mutant cells ( purple ) . Signal along chromosome XII is depicted to allow comparison with the profiles shown in Figure 7C . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 014 We asked whether the need for assembling a DSB-competent meiotic chromosome axis could be one reason why cells express a meiosis-specific form of cohesin . For this purpose , we replaced the only meiosis-specific subunit of cohesin , Rec8 , with its mitotic counterpart Scc1/Mcd1 by deleting REC8 and expressing SCC1 from the REC8 promoter ( pREC8-SCC1 ) ( Toth et al . , 2000 ) . ChIP analysis revealed that Scc1-containing cohesin was targeted to the same chromosomal sites as Rec8-cohesin , indicating that Scc1 can fully substitute for Rec8 in recruiting cohesin to meiotic chromosomes ( Pearson's r = 0 . 93 , Figure 7—figure supplement 1 ) . However , Scc1-cohesin did not efficiently recruit Red1 ( Figure 7C ) . As a result , the DSB distribution of rec8Δ pREC8-SCC1 cells remained similar to rec8Δ mutants and unlike wild-type cells ( Figure 7C ) . Consistent with this observation , Scc1 failed to detectably precipitate with V5-Red1 when expressed instead of Rec8 from the REC8-promoter ( Figure 6F ) . We conclude that only meiotic cohesin can efficiently recruit meiotic axis proteins . Surprisingly , pREC8-SCC1 did restore transcriptional focusing to axis association sites ( Figure 7D ) . These data indicate that focusing of axis proteins and axis protein recruitment are separable activities of meiotic cohesin and suggest that Rec8 , but not Scc1 , provides the necessary Red1 contacts for evenly distributing axis proteins , and thus recombination events , along meiotic chromosomes . The establishment of a quasi-regular array of chromatin loops represents a considerable mechanistic challenge that could be solved by recruitment motifs or de novo patterning mechanisms . Our findings indicate that the dispersed genomic distribution of cohesin rings at convergent genes ( Glynn et al . , 2004; Kugou et al . , 2009 ) , and thus ultimately the transcriptional landscape , dictates the preferred attachment sites of the axial element . The enrichment of cohesin between convergent genes is not meiosis-specific ( Glynn et al . , 2004; Lengronne et al . , 2004 ) , but by replacing the Scc1 subunit of cohesin with a meiotic variant ( Rec8 ) , cohesin is converted into a landmark for axial element anchoring . Using oriented URA3 insertions , we demonstrate that such sequence-independent recruitment ensures that the patterning of axis attachment sites readily adapts to changes in gene arrangements . Moreover , this mechanism may inherently separate axis attachments from DSB hotspots , which occur in promoters and are inhibited by the immediate neighborhood of axis proteins ( Ito et al . , 2014 ) . We note that , similar to yeast , mitotic cohesin is enriched on or near transcribed genes in both flies and mammalian cells ( Wendt et al . , 2008; Gause et al . , 2010 ) , and cohesin complexes are required for forming the meiotic loop-axis architecture in mammals ( Novak et al . , 2008; Llano et al . , 2012; Hopkins et al . , 2014 ) . Thus , the use of the transcriptional landscape as a dynamic reference for meiotic axis assembly may well be conserved through evolution . Perhaps , linking both DSB hotspots and axis association sites to landmarks of active , RNAPII-dependent transcription ensures that meiotic recombination remains confined to euchromatic and thus non-repetitive regions of the genome , reducing the potential for non-allelic homologous recombination and genome instability ( Sasaki et al . , 2010 ) . Our data show that recruitment of axis proteins involves direct physical interaction with cohesin . The robust positive signal of an in vivo physical proximity assay between Rec8 and Red1 is corroborated by co-precipitation between Red1 and Rec8 . Consistent with this observation , Red1 is recruited exclusively to Rec8 binding sites on chromatin by Rec8 in the absence of Hop1 . In fact , among cohesin subunits , Rec8 plays a pivotal role , as its mitotic paralogue , Scc1 , does not interact notably with Red1 . In Schizosaccharomyces pombe , meiotic cohesin interacts with the homolog of ScRed1 ( Rec10 ) via Rec11 , the meiosis-specific S . pombe homolog of ScScc3 ( Sakuno and Watanabe , 2015 ) . Whether this mechanism of interaction also operates in S . cerevisiae is unknown , but as Scc1-cohesin does not recruit or stably interact with Red1 in S . cerevisiae , we suggest that Rec8 directly contributes to Red1 binding . These findings indicate that the meiotic cohesin ring forms a stable protein complex with Red1 and Hop1 to flexibly link the axial element with chromatin . How axis attachment sites congress to form the loop-axis array remains to be determined . Previous data showing that linear elements are lost in the absence of Red1 but remain detectable by electron microscopy ( EM ) in the absence of Hop1 ( Rockmill and Roeder , 1990; Klein et al . , 1999 ) , suggest that Red1 , but not Hop1 , is required to link axis attachment sites to form linear elements . Red1 has a well-known ability to oligomerize ( Woltering et al . , 2000 ) , which may support the congression of axis sites , although other proteins , including Hop1 , may contribute to the wild-type structure . Whether the linear organization emerges from simple nearest neighbor interactions of axis attachment sites along a contiguous piece of DNA remains to be shown . Another open question is whether all sites of cohesin binding ( with the exception of centromeres ) function as axis attachment sites on a given chromosome , and thus whether spacing of axis attachment sites can serve as a proxy for estimating loop sizes . Immunofluorescence analyses show Rec8 in continuous tracks , whereas Hop1 and Red1 frequently exhibit more focal staining ( e . g . , [Smith and Roeder , 1997; Jin et al . , 2009] and data not shown ) , which may indicate incomplete occupancy of potential axis attachment sites . On the other hand , EM analysis yields estimates of chromatin loop sizes of approximately 20 kb ( Moens and Pearlman , 1988 ) , which is only moderately larger than the average spacing of cohesin sites of 10 . 9 kb ( Glynn et al . , 2004 ) , indicating that loops generally do not span more than 2 or 3 potential axis association sites . Hop1 emerges from this work as an intriguing modulator of axis patterning . We show that Hop1 contributes to Red1 recruitment in three ways . First , Hop1 mediates Red1 binding to a number of chromosomal sites independently of Rec8 . Second , Hop1 contributes to the recruitment of Red1 by Rec8 , as revealed by the loss of Red1–Rec8 interaction in hop1∆ mutants when the C-terminus of Red1 is tagged . Third and most unexpectedly , Hop1 is a negative regulator of Red1 accumulation at selected chromosomal regions . It has been noted that DSB formation is dangerous and thus suppressed in the vicinity of centromeres ( Pan et al . , 2011a ) and that DSB density is higher on the three smallest chromosomes ( Blitzblau et al . , 2007; Pan et al . , 2011a ) , which also enjoy substantially higher meiotic recombination rates ( Kaback et al . , 1992; Kaback , 1996 ) . Hop1 strongly suppresses Red1 hyper-accumulation close to centromeres but also on medium and large chromosomes . In the presence of Hop1 , Red1 densities on chromosomes I , VI , and III are over twofold higher than on other chromosomes , but without Hop1 , Red1 densities for all chromosomes are nearly the same . These findings predict that preferential DSB formation on small chromosomes should be eliminated in a hop1∆ mutant . This is indeed the case based on Spo11 oligo mapping ( P Schlögelhofer , personal communication ) . Earlier results have shown that the overrepresentation of DSB machinery on small chromosomes is not dependent on Spo11 ( Panizza et al . , 2011 ) , allowing us to exclude models that involve a DSB feedback mechanism ( Thacker et al . , 2014 ) to explain the compensatory accumulations . Moreover , the preferential accumulation of Red1 is unaffected , if not exacerbated in rec8∆ mutants , consistent with the observation that Rec8 binding itself exhibits no chromosome-size bias . These data exclude a role of cohesin in creating this bias and support the conclusion that Hop1 is the key factor mediating compensatory DSB machine accumulation on small chromosomes . Our findings support the view that visibly condensed axial structures may not be prerequisite to efficient DSB formation . In particular , yeast rec8∆ mutants fail to assemble discernable linear chromosome axes by EM ( Klein et al . , 1999 ) but remain competent to induce DSBs in chromosomal domains that retain Red1/Hop1 binding . This correlation therefore likely reflects a quasi-local function of axis proteins in hotspot activation that is independent of linear axis formation but this remains to be shown . We previously demonstrated that axis proteins are required for the recruitment of several essential DSB proteins ( Panizza et al . , 2011 ) . Since axis proteins inhibit DSB formation within ∼800 bp of their binding sites ( Ito et al . , 2014 ) , the Red1/Hop1-dependent local concentration of DSB factors likely activates hotspots at a distance , involving higher-order chromatin interactions . This model explains why aberrant fine-scale positioning of Red1 in rec8∆ mutants still supported DSB formation at wild-type hotspots . The fact that the positions of DSB hotspots remain unchanged in these mutants strongly argues that axis proteins are not responsible for designating certain sequences as hotspots . This conclusion is consistent with growing evidence ( summarized in Borde and de Massy , 2013 ) that hotspot designation is primarily driven by epigenetic marks associated with gene promoters . Significantly , our findings indicate that anchoring of the axial element is highly flexible . We demonstrate that axis protein accumulation peaks ∼150 bp downstream of transcript ends and is actively driven by differential promoter activity , implying that axis proteins are not stably bound to DNA . For steric reasons , the observed double peaks likely reflect the average distribution of axis proteins within the assayed cell population , rather than the distribution along a single chromatid . Given that the cohesin ring was previously proposed to slide along chromatin in response to transcription ( Bausch et al . , 2007; Ocampo-Hafalla and Uhlmann , 2011 ) , these data indicate that cohesin sliding controls the sites of axis protein attachment . We note that this system allows DNA to be robustly anchored to the protein axis with enough flexibility to allow dynamic processes such as transcription to proceed undisturbed on the DNA ( Figure 8 ) . 10 . 7554/eLife . 07424 . 015Figure 8 . Model of the interface between axial elements and DNA . Using the topological protein–DNA interaction of cohesin , a robust linkage between a protein rod ( the axial element ) and a transcriptionally active chromosome can be established . The protein axis forms on top of cohesin , while the elongating RNAPII , likely unable to pass through the cohesin molecule , may pull the DNA through the complex to avoid a premature transcription block . Blue arrows indicate the direction of DNA flow , going in the opposite direction of transcription . Based on the range of transcriptional overlap , loop movements will in most cases not exceed 500 bp and are not drawn to scale . DOI: http://dx . doi . org/10 . 7554/eLife . 07424 . 015 In addition , the adjustable axis-chromatin interface offered by the cohesin ring may also be important during meiotic recombination . Although our data indicate that positional flexibility of axis attachment is dispensable for DSB formation , meiotic chromosome axis components have multiple roles in meiotic recombination , including chromosome pairing , checkpoint signaling and preventing non-productive recombination between sister chromatids ( Zickler and Kleckner , 1999; Humphryes and Hochwagen , 2014; Subramanian and Hochwagen , 2014 ) . Evidence for recruitment-independent roles of the cohesin ring comes from rec8Δ mutants , which are severely defective in maintaining homolog-directed repair bias at a hotspot on chromosome III ( Kim et al . , 2010 ) . This defect occurs despite normal levels of Hop1 and Red1 on this chromosome ( Brar et al . , 2009; Kugou et al . , 2009; Panizza et al . , 2011 ) ( this work ) . Intriguingly , expression of Scc1-cohesin , which we show restores transcriptional focusing for axis proteins , also partially rescues homolog-directed repair bias ( Kim et al . , 2010 ) . Given that targeted DNA repair requires substantial flexibility of DNA ends , these observations raise the possibility that the dynamic anchoring of axis proteins may also be an important prerequisite for controlled meiotic recombination . All strains used in this study are of the SK1 background . The genotypes are listed in Supplementary file 1A . To induce synchronous meiosis , strains were inoculated at OD600 = 0 . 3 in BYTA medium for 16 . 5 hr at 30°C . Then cultures were washed twice with water and resuspended into SPO medium at OD600 = 1 . 9 at 30°C ( details are described in Blitzblau et al . , 2012 ) . For the pCUP1 experiments , the meiotic culture was split into different flasks 2 hr after meiotic induction . CuSO4 ( stock of 50 mM ) was then added to induce expression . For experiments shown in Figures 5 , 6 and Figure 5—figure supplement 1 , strains were diluted 1:2000 from a saturated culture grown for 24 hr in YPD into SPS medium and incubated for 16 hr at 30°C . Then cultures were washed once with SPM medium and resuspended into SPM at OD660 = 1 . 1–1 . 2 at 30°C , shaking 200 rpm/min . ChIP was performed on samples collected at the 3 hr time point , unless specified otherwise . 25 ml of the meiotic culture was harvested and fixed for 30 min in 1% formaldehyde . The formaldehyde was quenched by addition of 125 mM glycine . Samples were processed as described in Blitzblau and Hochwagen ( 2013 ) . For ChIP-qPCR , ChIP samples were diluted 1:20 and input samples 1:2000 . 8 µl of each dilution were then used in a 20 µl reaction ( see RT-qPCR for details ) . The ChIP samples were quantified by qPCR using two-step normalization . First , the percentage of ChIP relative to input was calculated , followed by calculation of the fold enrichment against an internal control ( see Supplementary file 1B for primer sequences ) . For experiments shown in Figures 5 , 6 and Figure 5—figure supplement 1 , 50 ml of the meiotic culture was harvested at the 4 hr time point , fixed for 15 min in 1% formaldehyde and stopped with 131 mM glycine . For each of the biological replicates in Figure 5 , at least one increasing and one decreasing locus was included in the assay set to control against variable efficiency of ChIP between the two compared strains . qPCRs in Figure 6 and Figure 5—figure supplement 1 differed from previous ones in that the reaction volumes were 25 µl and qPCR at the non-axis locus ADP1 was used to standardize between different preparations . In Figure 5—figure supplement 1 , the second normalization factor ( ADP1 ) was plotted alongside , instead of being used for normalization . The genome-wide analysis of DSB positions using ssDNA enrichment was conducted as described ( Blitzblau and Hochwagen , 2011 ) . Library preparation was performed using Illumina TruSeq DNA Sample Prep Kits v1 but adapters were used at 1:20 dilution . Amplified ChIP DNA between 300 and 500 bp was gel purified on 1 . 8% agarose and sequentially quantified using the Qubit assay HS kit as well as an Agilent 2100 Bioanalyzer . Fragments were sequenced on an Illumina HiSeq 2500 instrument using a 51-bp single-end sequencing protocol . For experiments shown in Figure 5 and Figure 5—figure supplement 1B , library preparation was performed according to Illumina ChIP-seq DNA sample prep protocol with TruSeq adapters diluted 1:10 . Enzymes for fragment-end polishing and for adapter ligation were from New England Biolabs ( Ipswich , MA ) . Before and after library amplification , ChIP DNA between 200 and 500 bp was gel purified on 2% agarose and quantified using an Agilent 2100 Bioanalyzer . Fragments were sequenced on an Illumina Genome Analyzer using 36-bp single-end sequencing protocols . The raw data analyzed in this study are available from the NCBI Gene Expression Omnibus ( http://www . ncbi . nlm . nih . gov/geo/ ) , accession numbers GSE69232 and GSE70112 ( Sun et al . , 2015a , 2015b ) . Sequencing reads were mapped to the S288C ( sacCer2 ) genome using Bowtie unless specified otherwise ( reads were mapped to the SK1 SGRP genome for the transcriptional analyses [Brar et al . , 2012a] ) . Only perfect matches across all 51 bp were considered during mapping . Multiple alignments were not taken into account , which means each read only mapped to one location in the genome . Reads were also mapped to the SK1 genome with similar results . Reads were extended towards 3′ ends to a final length of 200 bp in MACS ( http://liulab . dfci . harvard . edu/MACS/ ) ( Zhang et al . , 2008 ) . For experiments shown in Figure 5 , sequencing reads were mapped to the S288C ( sacCer3 ) genome using NextGenMap , version 0 . 4 . 11 ( Sedlazeck et al . , 2013 ) ( reads were mapped to SK1 SGRP genome for the transcriptional analyses [Brar et al . , 2012a] ) , allowing for maximal 2 gaps or 3 mismatches in 36 bp . In case of reads mapping to N locations , the corresponding fraction 1/N was assigned to each location . Reads were extended towards 3′ ends to a final length of in average 139 nt after determining the local optimal extension size ( per 10 kb ) . For comparison between different experiments , profiles were normalized using NCIS ( Liang and Keles , 2012 ) , after which the untagged control was subtracted . NCIS uses windows of low and randomly distributed sequence counts within each experiment to define the background and to separate it from the signal . Profiles of tagged and untagged samples , as well as before and after background subtraction are deposited at GEO ( accession number GSE70112; Sun et al . , 2015b ) . To identify axis protein enriched regions ( peaks ) , the observed sequence read tag density along the chromosome was normalized against the measured background ( Input DNA ) . Median coverage of all base positions in the ChIP sample was calculated . The coverage at each position was then divided by the calculated median . The same division was also performed for the input sample , followed by subtraction of the input from the ChIP sample at each position . Peak identification was performed using MACS ( http://liulab . dfci . harvard . edu/MACS/ ) ( Zhang et al . , 2008 ) using a P-value of 1e-15 . Peak summits were defined as the local maximum of read coverage within Red1 peaks . Sequences of axis protein binding sites were extracted from the genome using Python scripts ( Supplementary file 2 ) . We used 500 bp upstream and downstream of the summits to perform motif discovery . All the peaks were ranked by their enrichment signals . An executable version of MDscan ( Liu et al . , 2002 ) was used with the following parameters: background model: S . cerevisiae intergenic; motif width: 6–15; number of top sequences to look for candidate motifs: 30; number of candidate motifs for scanning the rest sequences: 20; number of motifs to report: 5 . The online version of MEME-ChIP ( http://meme . nbcr . net/meme/cgi-bin/meme-chip . cgi ) ( Machanick and Bailey , 2011 ) was also used to perform motif discovery on sequencing data with the following parameters: motif site distribution: zero or one occurrence per sequence; count of motifs: 10 . M-track detects close proximity between proteins by transferring methyl groups via a methyltransferase ( HKMT ) fused to the bait to an acceptor moiety ( histone H3 peptide ) fused to the prey . M-track was performed as described by Zuzuarregui et al . ( 2012 ) . In brief , all fusions were generated by PCR-mediated C-terminal tagging at the endogenous chromosomal locus under the endogenous promoter ( genotypes provided in Supplementary file 1A ) . 2 × 10^8 cells were collected at the indicated time points . Protein extracts were prepared after trichloroacetic acid treatment as previously described ( Penkner et al . , 2005 ) and analyzed by Western blotting . Methylated proteins were detected using anti-H3K9me3 antibody ( gift from Egon Ogris , 1:100 ) and anti-HA antibody ( 12CA5 , 1:500 ) for methylation independent signal as loading control . A non-specific 12CA5 dependent band was used as an additional loading control . Anti-mouse horseradish peroxidase-conjugated antibody ( Pierce , 1:10 , 000 ) was used as the secondary antibody . Co-immunoprecipitation from meiotic culture was performed as described in de los Santos and Hollingsworth ( 1999 ) . In brief , 2 × 10^9 cells were collected at 4 hr in SPM and ruptured using a multibead shocker ( YASUI-KIKAI , Osaka ) at 2500 rpm , 15 cycles of 30 s on and 30 s off , at 4°C . Extracts were then sonicated using a Covaris S2 ( duty cycle: 15%; intensity: 10 . 0; cycles/burst: 500; total time: 240 s ) . After removing the cell debris by centrifugation , Benzonase ( 20 U/sample , Sigma ) as well as anti-SV5-coated magnetic beads ( 50 µl Pan mouse IgG , Dynabeads ) were added , followed by rotating incubation for 3 hr at 4°C . Rabbit anti-HA antibody ( Sigma , 1:1000 ) and anti-rabbit horseradish peroxidase-conjugated antibody ( Pierce , 1:10 , 000 ) were used for detection . 10 ml of meiotic culture was harvested at the specified time points and total RNA was extracted using glass beads , phenol chloroform and RNA buffer ( 300 mM NaCl , 10 mM Tris-HCl pH 7 . 5 , 1 mM EDTA , 0 . 2% SDS ) . After ethanol precipitation , 10 μg of RNA sample was digested using Turbo DNase ( Invitrogen ) at 37°C for 30 min followed by adding inactivation mix . We used 1 μg of RNA to generate cDNA , along with dNTP , RNaseOUT Recombinant Ribonuclease Inhibitor ( Invitrogen ) , oligo-dT primer and RevertAid Reverse Transcriptase ( Fermentas ) . Quantitative PCR was performed using the QuantiFast SYBR Green PCR Kit ( Qiagen ) . In a 20 µl of reaction , we used 10 µl 2× SYBR master mix , 1 µl of 20 µM primers , 8 µl of 1:10 diluted cDNA . Primers are listed in Supplementary file 1B . Each PCR reaction was performed in triplicate . Results of each sample were normalized against the expression level of an internal control ( GAL1 ) .	Chromosomes are long molecules of DNA that represent the genetic material of an organism . In most animal cells , chromosomes are found in pairs; with one inherited from the mother and the other from the father . Sex cells—egg cells and sperm—contain half the normal number of chromosomes , so that when they fuse , the resulting single-celled embryo inherits the full set . When sex cells are being produced , a ring made from a protein called cohesin encircles each pair of chromosomes and holds them together until they are ready to be separated . The paired chromosomes also swap sections of DNA via a process called recombination . Structures , referred to as axial elements , compact the chromosomes in each pair and bring them in close contact so that recombination can take place . In the sexually reproducing baker's yeast , axial elements contain three main proteins: cohesin , Hop1 , and Red1 , but it remains unclear how the entire structure is anchored to the underlying chromosomes . Furthermore , the genes encoded within the DNA of the compacted chromosomes remain active , but it is also not clear how this is possible . This is because the compacted structure would be expected to prevent the molecular machinery that expresses genes from accessing the DNA . Sun , Huang et al . have now studied this process in budding yeast cells by using a method called ChIP-seq to determine where cohesin and the Hop1 and Red1 proteins are found along the chromosomes . The experiments showed that cohesin , Hop1 , and Red1 are enriched in regions between two genes that run in the opposite directions to each other . Sun , Huang et al . also observed that cohesin recruits Red1 , which in turn , recruits Hop1 , and that all three proteins physically interact with one another . These findings imply that it is cohesin that anchors the axial elements to the underlying chromosomes . Further experiments showed that cohesin slides along chromosomes towards areas where genes are active . This suggests that cohesin provides a robust , but flexible , link between the axial elements and the chromosomes . This flexibility would enable recombination and gene expression to continue in compacted chromosomes . A loss of flexibility may be one of the reasons why mutations in cohesin components of the axial element cause infertility in men and condition called premature ovarian failure in women .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2015	Transcription dynamically patterns the meiotic chromosome-axis interface
Various pathologic conditions result in jaundice , a yellowing of the skin due to a buildup of bilirubin . Patients with jaundice commonly report experiencing an intense non-histaminergic itch . Despite this association , the pruritogenic capacity of bilirubin itself has not been described , and no bilirubin receptor has been identified . Here , we demonstrate that pathophysiologic levels of bilirubin excite peripheral itch sensory neurons and elicit pruritus through MRGPRs , a family of G-protein coupled receptors expressed in primary sensory neurons . Bilirubin binds and activates two MRGPRs , mouse MRGPRA1 and human MRGPRX4 . In two mouse models of pathologic hyperbilirubinemia , we show that genetic deletion of either Mrgpra1 or Blvra , the gene that encodes the bilirubin-producing enzyme biliverdin reductase , attenuates itch . Similarly , plasma isolated from hyperbilirubinemic patients evoked itch in wild-type animals but not Mrgpra1-/- animals . Removing bilirubin decreased the pruritogenic capacity of patient plasma . Based on these data , targeting MRGPRs is a promising strategy for alleviating jaundice-associated itch . Chronic pruritus , or itch , is a complex and often debilitating symptom that accompanies a range of cutaneous and non-cutaneous diseases ( Ständer et al . , 2007; Yosipovitch and Bernhard , 2013a ) . The most widely known pruritogen is histamine , which is secreted by mast cells in the skin and activates histamine receptors on nearby sensory neurons ( Bautista et al . , 2014; Ikoma et al . , 2006; LaMotte et al . , 2014; Yosipovitch and Bernhard , 2013b ) . While viable treatments exist for histamine-mediated itch , most non-histaminergic conditions are more difficult to treat because the mediators are often unknown ( Kremer et al . , 2011 ) . Jaundice , or yellowing of the skin , sclera , and mucosa due to abnormal accumulation of the yellow heme metabolite bilirubin , is commonly associated with chronic non-histaminergic pruritus . Jaundice often presents in patients with hepatobiliary disorders such as cholestasis , characterized by impaired bile flow . Physiologically , bilirubin is typically bound to albumin in serum and concentrates in the liver , where it is conjugated to glucuronic acid and subsequently excreted in bile . At physiologic and mildly elevated concentrations ( 0 . 2–2 . 7 mg/dL , 3 . 4–46 . 2 μM ) , bilirubin is benign . At highly elevated levels however , such as in cutaneous jaundice ( >5 mg/dL , >85 . 5 μM bilirubin ) , it is associated with pruritus , a correlation first noted by physicians as early as the second century B . C . E . ( Bassari and Koea , 2015 ) . Despite the long-standing association between jaundice and pruritus ( Talwalkar et al . , 2003 ) , bilirubin itself has not been described as a pruritogen . To determine whether bilirubin directly elicits pruritus , we subcutaneously injected bilirubin into the napes of mice . Pathophysiologic concentrations of bilirubin stimulated scratching in a dose-dependent manner at the site of injection ( Figure 1A ) . Pre-incubating bilirubin with excess human serum albumin , which binds bilirubin with high affinity ( Breaven et al . , 1973; Griffiths et al . , 1975; Jacobsen and Brodersen , 1983 ) , elicited fewer scratches ( Figure 1A ) . The behavioral profile of bilirubin-induced scratching mirrored that of two well-characterized pruritogens , histamine and chloroquine ( Figure 1B ) . Notably , histamine and chloroquine only elicit itch when injected into mice at millimolar concentrations despite having nanomolar affinity towards their receptors . In comparison , bilirubin elicited a similar degree of itch even when injected at lower concentrations than histamine or chloroquine ( Figure 1B ) . Since mice indiscriminately scratch at the nape if injected with substances that trigger either itch or pain , we also injected mice at the cheek . Unlike at the nape , painful sensations at the cheek evoke a distinct wiping behavior instead of scratching , whereas itchy sensations still elicit scratching ( Shimada and LaMotte , 2008 ) . Injecting bilirubin in the cheek prompted dose-dependent scratching just as it did at the nape ( Figure 1—figure supplement 1A ) . Bilirubin elicited neither wiping nor licking , indicating that it selectively triggers itch and not pain ( Figure 1—figure supplement 1B–C ) . We injected mice with metabolites structurally similar to bilirubin to determine the specificity of bilirubin’s pruritic activity ( Figure 1E ) . The two metabolites directly epistatic to bilirubin , hemin and biliverdin , did not induce scratching despite also being tetrapyrroles ( Figure 1D ) . While hemin , biliverdin , and bilirubin display only minor atomic and electronic differences between them , they vary substantially in their physiochemical properties and structures ( Figure 1E ) . To better understand these differences , we performed density functional theory ( DFT ) calculations ( Becke , 1993; Hohenberg and Kohn , 1964; Kohn and Sham , 1965; Stephens et al . , 1994 ) to determine the optimal geometry of each metabolite . Unlike in heme and biliverdin , bilirubin’s four pyrroles are extended and do not lie in the same plane ( Figure 1E ) . DFT calculations revealed that urobilinogen and stercobilin , two bacterial metabolites derived from bilirubin , adopt a similar extended conformation . Both urobilinogen and stercobilin were able to stimulate scratching behavior ( Figure 1D ) , indicating that bilirubin’s non-polar pyrroles may be important for its pruritic activity . Patients with jaundice-associated pruritus receive little benefit from antihistamines ( Bergasa , 2014 ) . Consistent with these clinical findings , the histamine receptor one blocker cetirizine ( 30 mg/kg , i . p . ) failed to alleviate scratching behavior in mice injected with bilirubin ( Figure 1—figure supplement 1D ) . Furthermore , bilirubin did not elicit a calcium response or induce appreciable histamine release from peritoneal mast cells ( Figure 1—figure supplement 1E–F ) . The Mas-related G-protein coupled receptor ( Mrgpr ) family of receptors is a major mediator of non-histaminergic pruritus ( Han et al . , 2013; Liu et al . , 2012; Liu et al . , 2009; Sikand et al . , 2011 ) . To test whether Mrgprs mediate bilirubin-induced pruritus , we injected mice lacking a cluster of 12 Mrgpr genes ( Mrgpr-clusterΔ−/− or Mrgpr-cluster KO ) with bilirubin ( Liu et al . , 2009 ) . Mrgpr-cluster KO animals scratched approximately 75% less than wild type ( WT ) mice , indicating that one or more of the 12 Mrgprs within the cluster mediates bilirubin-induced pruritus ( Figure 1C ) . To identify which Mrgpr is sensitive to bilirubin , we individually expressed each of the 12 Mrgprs deleted in the Mrgpr-cluster KO mouse in human embryonic kidney ( HEK ) 293 cells and monitored changes in intracellular calcium upon applying bilirubin . To ensure we would observe a calcium response following a true ligand-receptor interaction , we expressed the receptors in HEK293 cells stably expressing the G-protein alpha-subunit Gα15 , a Gα protein that couples GPCRs to intracellular calcium stores via phospholipase C ( PLC ) . Among the twelve cell lines expressing an Mrgpr , only MRGPRA1-expressing cells exhibited a calcium response to bilirubin ( EC50 of 145 . 9 µM ( Alemi et al . , 2013 ) ) ( Figure 2A , D ) . The same cells that responded to bilirubin also responded to FMRF , an MRGPRA1 agonist ( Dong et al . , 2001 ) . To ensure that bilirubin initiated signaling at MRGPRA1 and not downstream , we pre-treated MRGPRA1-expressing cells with inhibitors of GPCR signaling: the PLC inhibitor U73122 or the Gαq inhibitor YM-254890 . Both compounds abolished bilirubin-induced calcium responses ( Figure 2B–C ) . In addition to bilirubin , glucuronidated bilirubin is often upregulated in jaundice-associated itch . We assessed whether ditaurate bilirubin , a distinct but similar bilirubin derivative , could activate MRGPRA1 . Indeed , ditaurate bilirubin activated MRGPRA1-expressing cells ( Figure 2D ) . Hemin failed to activate MRGPRA1 ( Figure 2D ) , consistent with our earlier behavioral findings in which hemin did not evoke scratching . No other Mrgpr among the 12 that we screened responded to bilirubin ( Figure 2N , Figure 2—figure supplement 1 ) . The human MRGPRX family of receptors has functional similarities between species but have no obvious structural homologs in rodents ( Solinski et al . , 2014; Zylka et al . , 2003 ) . The mouse Mrgpra family is closest in sequence homology to the human MRGPRX family ( Dong et al . , 2001; Lembo et al . , 2002; Zhang et al . , 2005 ) . Of the four human MRGPRX receptors , only MRGPRX4-expressing cells responded to bilirubin ( EC50 of 61 . 9 µM ( Azimi et al . , 2017 ) ) ( Figure 2F , I ) . U73122 and YM-254890 inhibited bilirubin-induced calcium responses in MRGPRX4-expressing cells just as with MRGPRA1 ( Figure 2G–H ) . Conjugated bilirubin also activated MRGPRX4 , whereas hemin had no effect ( Figure 2I ) . To confirm that bilirubin directly binds the identified receptors , we assayed thermophoresis of each receptor in the presence and absence of bilirubin . Thermophoresis of a molecule is affected by physical parameters such as size , charge , and solvation . By extension , the thermophoresis of one molecule is altered when it interacts with another , and can therefore be used to measure interactions between molecules ( Duhr and Braun , 2006 ) . Using this approach , we determined that bilirubin bound MRGPRA1 with a KD of 92 . 9 ± 15 µM and MRGPRX4 with a KD of 54 . 4 ± 13 µM ( Figure 2E , J ) . Bilirubin exhibited little to no affinity for the closely related BAM8-22 receptor MRGPRC11 ( Figure 2O ) . Hemin , which did not activate MRGPRA1 or MRGPRX4 by calcium imaging ( Figure 2D , I ) , also did not bind MRGPRA1 or MRGPRX4 ( Figure 2E , J ) . Conjugated bilirubin bound both MRGPRA1 and MRGPRX4 , although with a lower affinity than unconjugated bilirubin ( Figure 2E , J ) . To make certain that bilirubin activates MRGPRA1 and MRGPRX4 upon binding , we measured exchange of guanosine diphosphate ( GDP ) for guanosine triphosphate ( GTP ) , one of the first events in GPCR signaling . Bilirubin increased GTP binding to MRGPRA1 and MRGPRX4 membrane complexes , but not to MRGPRC11 ( Figure 2K ) . To confirm that bilirubin activates MRGPRA1 in vivo to trigger itch , we generated an Mrgpra1 ( A1 KO ) knockout mouse line using CRISPR-Cas9 ( Jinek et al . , 2012 ) ( Figure 2—figure supplement 2 ) . A1 KO animals scratched significantly less than WT mice after exposure to either bilirubin or the established agonist FMRF , demonstrating that Mrgpra1 is functional in adult mice ( Figure 2L–M ) . The KD of bilirubin towards MRGPRA1 and MRGPRX4 suggests that bilirubin likely does not interact with these receptors in healthy individuals . Additional ligands with nanomolar affinities towards MRGPRA1 or MRGPRX4 may exist that modulate the receptors in normal physiology . We reasoned that if bilirubin triggers itch through MRGPRA1 and MRGPRX4 , bilirubin should activate these receptors in sensory itch neurons . Previous studies have demonstrated that both Mrgpra1 and MRGPRX4 are expressed in sensory neurons within the dorsal root ganglia ( DRG ) ( Dong et al . , 2001; Flegel et al . , 2015; Lembo et al . , 2002 ) . Mrgpra1 is expressed in a subset of adult DRG and trigeminal ganglia ( TG ) sensory neurons that innervate the skin and ramify in lamina I and II of the spinal cord ( Figure 3A–D ) . Bilirubin elicited robust action potentials in small-diameter ( <30 µm ) WT DRG sensory neurons at a proportion consistent with the percentage of sensory neurons that encode itch ( 5 of 50 ) . Bilirubin failed to elicit action potentials in A1 KO neurons ( 0 of 60 ) , suggesting bilirubin activates sensory neurons through MRGPRA1 ( Figure 3E ) . Bilirubin-sensitive neurons had an average somal diameter of 20 . 4 ± 1 . 3 µm , a diameter characteristic of itch sensory neurons ( Figure 3F ) . Applying bilirubin to neurons elicited calcium transients in approximately 5% of WT DRG neurons ( Figure 3G ) , whereas significantly fewer sensory neurons from either Mrgpr-cluster KO or A1 KO mice responded ( Figure 3H ) . We sought to determine whether expression of either MRGPRA1 or MRGPRX4 was sufficient to render neurons sensitive to bilirubin . To address this question , we infected Mrgpr-cluster KO DRGs with lentivirus carrying either Mrgpra1 , MRGPRX4 , or MRGPRX3 . Bilirubin activated 14% of Mrgpra1-and 32% of MRGPRX4-transduced Mrgpr-cluster KO DRGs ( Figure 3I–J ) . Mrgpr-cluster KO DRGs infected with the control gene MRGPRX3 did not respond to bilirubin . Bilirubin-responsive neurons partially overlapped with neurons that responded to 1 mM chloroquine , a ligand for MRGPRA3 that typifies itch sensory neurons ( Han et al . , 2013 ) ( Figure 4A–C ) . To validate that bilirubin activates MRGPRA3-positive itch neurons , we performed calcium imaging on DRG neurons isolated from Tg ( Mrgpra3-Cre ) ;lsl-tdTomato mice , which express the fluorescent protein tdTomato in Mrgpra3-expressing neurons . Bilirubin activated a substantial percentage of tdTomato-positive neurons ( Figure 4D ) . To confirm that bilirubin activates sensory neurons in vivo , we injected 5 μL of vehicle or bilirubin into paws of Tg ( Pirt-Cre ) ;lsl-GCaMP6s mice , which express the fluorescent calcium reporter GCaMP6s in DRG sensory neurons ( Kim et al . , 2014 ) . Bilirubin , but not vehicle , activated numerous DRG sensory neurons in the paws of GCaMP6s mice ( Figure 4E ) . Inhibiting transient receptor potential ( TRP ) and other Ca2+ channels with ruthenium red prevented bilirubin from activating sensory neurons ( Figure 4F–G ) ( Imamachi et al . , 2009; Liu et al . , 2009; Roberson et al . , 2013 ) . We wondered whether chronic elevation of bilirubin in vivo , like in cholestasis , stimulates Mrgpr-dependent itch . Bile is the primary means by which bilirubin is excreted , and patients with cholestasis exhibit elevated levels of bilirubin and other pruritogenic substances in their blood ( Alemi et al . , 2013 ) . To induce hyperbilirubinemia and model intrahepatic cholestasis , we administered α-napthyl isothiocyanate ( ANIT ) to mice ( Eliakim et al . , 1959 ) . We treated WT , Mrgpr-cluster KO , and A1 KO animals with 25 mg/kg ANIT for five days before assessing spontaneous itch ( Figure 5A ) . WT , Mrgpr-cluster KO , and A1 KO animals exhibited equivalently severe hepatocellular injury , judged by increases in plasma bilirubin , bile acids , alkaline phosphatase ( ALP ) , aspartate aminotransferase ( AST ) , alanine aminotransferase ( ALT ) , and gamma-glutamyl transferase ( GGT ) ( Figure 5D–E , Figure 5—figure supplement 1A–D ) . As expected , ANIT treatment significantly increased pruritus in all animals ( Figure 5B ) . However , Mrgpr-cluster KO and A1 KO mice scratched markedly less than WT mice ( Figure 5B ) , suggesting that MRGPRA1 mediates a component of hepatobiliary pruritus . In humans , bile acids , endogenous opioids , and LPA are often increased in cholestatic sera and have been shown to mediate pruritus ( Alemi et al . , 2013; Bergasa et al . , 1998; Bergasa et al . , 1992; Kremer et al . , 2010 ) . The serum of ANIT-treated animals exhibited elevated bilirubin and bile acids ( Figure 5D–E ) , whereas neither the endogenous opioid peptide met-enkephalin ( Thornton and Losowsky , 1989a; Thornton and Losowsky , 1989b ) nor the LPA-producing enzyme autotaxin were elevated ( Figure 5—figure supplement 1E–F ) . To assess whether other cholestatic pruritogens act at MRGPRs in mice , we injected WT and Mrgpr-cluster KO with deoxycholic acid ( a bile acid ) , opiates , and LPA . These other cholestatic pruritogens elicited equivalent degrees of scratching in WT and Mrgpr-cluster KO animals ( Figure 5—figure supplement 2A–D ) . Mrgprs are promiscuous receptors . It should be noted that there remain multiple bile acids , LPA molecules , and opiates which remain untested that may be agonists of Mrgprs . Based on this data , we hypothesized that Mrgpr-cluster KO and A1 KO mice scratched less with ANIT because MRGPRA1 mediates bilirubin-induced itch . To determine whether bilirubin is activating MRGPRA1 to stimulate itch in cholestasis , we induced cholestasis in a mouse that lacks the biosynthetic enzyme for bilirubin , biliverdin reductase ( BVR KO ) ( Kutty and Maines , 1981 ) ( Figure 1E , Figure 5—figure supplement 3A ) . BVR KO mice lack Blvra , the gene that encodes for bilirubin reductase . BVR KO mice do not have detectable levels of bilirubin in plasma ( Figure 5—figure supplement 3B–D ) . When treated with ANIT , BVR KO mice scratched significantly less than WT mice ( Figure 5C ) . Plasma levels of bile acids , ALP , AST , ALT , GGT , met-enkephaline , and autotaxin were indistinguishable between treated BVR KO animals and WT controls ( Figure 5—figure supplement 1A–D ) . Their diminished response to ANIT is not due to aberrant itch circuits , as BVR KO mice scratched normally when injected with either chloroquine or exogenous bilirubin ( Figure 5—figure supplement 3E–F ) . To confirm that the observed differences in cholestatic pruritus were not just specific to ANIT , we administered the hepatotoxin cyclosporin A to WT , A1 KO , and BVR KO mice ( Laupacis et al . , 1981 ) . We treated mice with either 50 mg/kg cyclosporin A or vehicle for eight days before assessing spontaneous itch ( Figure 5F ) . Cyclosporin A induced spontaneous itch in WT animals , whereas A1 KO and BVR KO mice again scratched significantly less than WT mice ( Figure 5G–H ) . Notably , we found that plasma bilirubin correlates poorly with cholestatic itch in patients and in cholestatic animals ( Figure 5I ) . We hypothesized that the levels of bilirubin in the skin would correlate better with itch than serum bilirubin largely because bilirubin likely binds and activates the sensory neurons in the skin . Unlike with serum , skin bilirubin appears to be a much stronger predictor of itch severity in mice ( Figure 5J ) . This is consistent with the anatomical distribution of itch sensory neurons and may explain why studies aimed at identifying plasma pruritogens that correlate with itch severity may have missed bilirubin . Secondarily , we find that plasma bilirubin does not correlate well with skin bilirubin , further suggesting that plasma bilirubin may be a poor predictor of itch severity and may not necessarily serve as a proxy for skin bilirubin ( Figure 5K ) . The amount of bilirubin in the skin is likely affected by several factors and equilibria , such as serum albumin . We assessed whether pharmacologically antagonizing MRGPRs could alleviate cholestatic itch . Recently , a 3-amino acid peptide , QWF , was identified as an MRGPRA1 antagonist ( Azimi et al . , 2016 ) . QWF abolished bilirubin-associated calcium signaling in MRGPRA1-expressing cells with an IC50 of 2 . 9 μM [1 , 5] ( Figure 6A ) . Mirroring its pharmacology in vitro , co-injecting 0 . 25 mg/kg QWF with bilirubin significantly alleviated pruritus associated with bilirubin ( Figure 6B ) . QWF specifically antagonized bilirubin , as it did not attenuate chloroquine-MRGPRA3 associated itch ( Figure 6C ) . Lastly , we evaluated whether the MRGPRA1 antagonist QWF could alleviate cholestatic pruritus . We dosed WT animals with ANIT as previously described , but intraperitoneally injected mice with either vehicle or 1 mg/kg QWF thirty minutes prior to behavioral analysis . Mice treated with QWF scratched significantly less than vehicle-treated animals ( Figure 6D ) . QWF treatment did not change plasma levels of total bilirubin , AST , ALT , or ALP , suggesting that QWF treatment did not alter the underlying liver pathology ( Figure 6E–H ) . Nasobiliary drainage is the most effective treatment for cholestatic pruritus ( Hegade et al . , 2016 ) . Based on this clinical observation , we predicted that plasma isolated from cholestatic animals would elicit pruritus ( Figure 7A ) . Indeed , plasma from WT animals with cholestasis elicited itch when injected into naïve WT animals ( Figure 7B ) . Cholestatic plasma isolated from BVR KO mice , which lacks bilirubin ( Figure 5—figure supplement 3B–D ) , elicited significantly fewer scratches than WT cholestatic plasma ( Figure 7B ) . The levels of ALP , AST , and ALT were indistinguishable between WT and BVR KO cholestatic plasma ( Figure 5—figure supplement 1A–D ) , presumably because ANIT induced similar hepatotoxicity in WT and BVR KO mice . Instead , BVR KO plasma likely results in less pruritus because it lacks bilirubin . We also isolated plasma from six patients suffering from various conditions that result in hyperbilirubinemia and six age- and sex-matched control patients ( Figure 7C–G ) . All six cholestatic patients’ plasma evoked itch in WT animals ( Figure 7D ) . When injected into A1 KO animals , each patient’s plasma elicited less itch ( Figure 7D ) . Compared to plasma from itchy patients , plasma from healthy donors with low levels of bilirubin elicited less itch in WT animals ( Figure 7C ) . Plasma from healthy donors injected into A1 KO animals elicited similar scratching behavior ( Figure 7C ) . To assess whether removing bilirubin from cholestatic plasma may be therapeutic , we depleted bilirubin both by selective oxidation with FeCl3 or an anti-bilirubin antibody and again evaluated its pruritic capacity . We verified depletion of bilirubin by HPLC ( Figure 7—figure supplement 1A–B ) . Injecting WT mice with plasma ( cholestatic patients 1 and 4 ) treated with FeCl3 or an anti-bilirubin antibody evoked less pruritus compared to untreated or control IgG-treated patient plasma ( Figure 7H–I ) . To date , bilirubin has largely been considered a neonatal neurotoxin or an inert biomarker in disease . Our results reveal that bilirubin itself is a pruritogen that evokes itch by binding and activating MRGPRs on sensory neurons and may be an overlooked source of some patients’ unrelenting itch . The KD of bilirubin towards MRGPRA1 and MRGPRX4 suggests that bilirubin likely only interacts and activates these receptors in individuals with markedly elevated bilirubin and not in healthy people . More narrowly in hepatobiliary diseases such as cholestasis , our data supports a model in which bilirubin is one of several pruritogens that contributes to itch . Specifically , we find that genetically removing either MRPGRA1 ( A1 KO ) or bilirubin ( BVR KO ) both strongly attenuated itch . However , these mutant mice still exhibit greater itch compared to untreated mice . This suggests that other pruritogens contribute to itch alongside bilirubin in hepatobiliary disease , consistent with the diverse and complex presentations of patients suffering from conditions such as chronic pruritus . Other responsible pruritogens could include bile acids , endogenous opioids , and LPA . While depleting bilirubin in jaundiced patients like in mice may be effective in reducing itch , not every patient who suffers from hepatobiliary pruritus is jaundiced . Accordingly , identifying and depleting other pruritogens may similarly reduce itch . Although our findings directly illustrate that bilirubin is pruritic , it is also clear that not every patient with jaundice experiences itch . For example , patients with genetic hyperbilirubinemias such as Dubin-Johnson syndrome , a disorder involving mutations in the bilirubin transporter ABCC2 , or Crigler-Najjar Type 1 , a disorder involving mutations in the bilirubin glucuronosyltransferase UGT1A1 , rarely complain of pruritus ( Levitt and Levitt , 2014; van der Veere et al . , 1996 ) . Moreover , neonates can have high levels of bilirubin in their skin but not itch . Bilirubin thus appears to exert selective pruritic activity in certain contexts , which we hypothesize may derive from its dynamic biophysical behavior and complex network of interactions . In isolated genetic hyperbilirubinemias , few – if any – other organic metabolites are elevated . In contrast , several other metabolites are elevated in addition to bilirubin in cholestasis , many of which may alter the equilibrium between bilirubin and albumin ( Alemi et al . , 2013; Jacobsen and Brodersen , 1983; Kalir et al . , 1990; Kozaki et al . , 1998 ) . Moreover , bilirubin’s affinity for albumin and other lipoproteins is disrupted by numerous agents in bile that are upregulated specifically in cholestasis . Bilirubin also exhibits distinct chemical behavior in cholestatic serum , and several groups have suggested that bile acids affect bilirubin’s solubility and conformation ( Ostrow and Celic , 1984; Rege et al . , 1988 ) . Accordingly , it is reasonable to speculate that bilirubin is more likely bound to albumin in isolated hyperbilirubinemias than in cholestasis , and is therefore less likely to enter the skin in conditions like Dubin-Johnson . Notably , a predictive metrics for cholestatic pruritus is the Mayo risk score , which considers both serum bilirubin and albumin levels ( Talwalkar et al . , 2003 ) . The Mayo risk score predicts that itch burden increases with increasing bilirubin and decreasing albumin levels; in such circumstances , bilirubin is less likely bound to albumin and is free to enter other tissues such as the skin . Crigler-Najjar patients may be even less likely to complain of itch than Dubin-Johnson patients because the standard treatments for Crigler-Najjar ( phenobarbital and light therapy ) may directly interfere with bilirubin’s pruritic activity . Specifically , light therapy induces photoisomerization and/or photolysis of bilirubin , which alter its structure and activity . Phenobarbital itself acts by broadly decreasing neural excitability , and may dampen itch circuitry alongside other central nervous system circuits . Notwithstanding these questions , our results suggest that blocking MRGPRX4 may offer relief to those suffering from jaundice and/or cholestatic-associated pruritus . All experiments were performed in accordance with protocols approved by the Animal Care and Use Committee at the Johns Hopkins University School of Medicine . Plasma from patients suffering from hyperbilirubinemia , specifically cholestasis , was isolated under a protocol approved by the Institutional Review Board at the Johns Hopkins University School of Medicine ( Study number: IRB00154650 ) . Control plasma was isolated from donors who did not exhibit kidney or liver disease , had no complaints of itch , and were free from any detectable viral infection ( HCV , HBV , HIV ) . Both cholestatic and control plasma were isolated under protocols approved by the Institutional Review Board at the Johns Hopkins University School of Medicine ( Cholestasis Study number: IRB00154650; Control study number: NA_00013177 , the Johns Hopkins Department of Dermatology Patient Database and Tissue Bank ) . Whole blood was collected into PAXgene tubes ( PreAnalytiX 761115 ) and centrifuged for 5 min at 300 g . Plasma was then collected , aliquoted , and stored at −20°C until experimentation . At time of plasma collection , a 5D itch questionnaire was administered . The following molecules were used: bilirubin IXα ( Frontier Scientific ) . α-naphthyl isothiocyanate ( ANIT , Sigma ) , biliverdin ( Sigma ) , chloroquine ( Sigma ) , compound 48/80 ( Sigma ) , cyclosporin A ( Sigma ) , hemin ( Sigma ) , human serum albumin ( HSA , Sigma ) , BAM8-22 ( Sigma ) , BOC-GLN-D- ( FORMYL ) TRP-PHE-BENZYLESTER ( QWF , Sigma ) , bilirubin ditaurate ( Lee Biosciences ) . cetirizine ( Tocris Biosciences ) , stercobilin ( Santa Cruz Biotechnology ) , urobilinogen ( Santa Cruz Biotechnology ) , FMRF peptide ( Sigma ) , cholera toxin ( Santa Cruz Biotechnology ) , U73122 ( Santa Cruz Biotechnology ) , YM-254890 ( Wako Chemicals ) , pertussis toxin ( Fisher Scientific ) , fibronectin ( Sigma ) , ruthenium red ( Sigma ) , Fluo 4-AM ( Molecular Probes ) , and Fura 2-AM ( Molecular Probes ) . Bilirubin is highly susceptible to oxidation and photolysis . Accordingly , bilirubin was freshly prepared just prior to each experiment in either DMSO or 0 . 1 M NaOH and then maintained in the dark . For calcium imaging analyses , bilirubin was diluted into calcium imaging buffer a few seconds before use . Final concentration of DMSO in all applicable tested solutions was <0 . 5% . ANIT and cyclosporin A were dissolved in olive oil and prepared freshly as needed . Urobilinogen and stercobilin were dissolved in phosphate buffered saline and adjusted to a pH of 7 . 4 before being stored at −20°C in 100 μl aliquots until needed . To maintain the integrity of bilirubin in human plasma samples , samples were stored at −80°C until use . Plasma stocks were maintained in the dark to minimize photolysis during injection or experimental manipulation . Plasma bilirubin was evaluated by HPLC as described above . All other drugs were prepared as 100 μl – 1 , 000 μl aliquots and stored at −20°C before thawing at 4°C . Freeze/thaw cycles were avoided whenever possible . To remove microprecipitates , we centrifuged our bilirubin solutions at 21 , 000 g for 20 min to ensure that bilirubin was in solution . Whenever physiologically and experimentally reasonable , we excluded divalent cations in our in vitro biophysical experiments . All applicable behavioral tests were performed and analysed with the experimenter blind to genotype . All mice used were 8–12 week old males and females ( 20 to 30 g ) that had either been generated on a C57BL/6J background or backcrossed to C57BL/6J mice for at least 10 generations . There were no significant differences in itch between male and female mice . All itch behavior experiments were performed between 8 a . m . and 12 p . m . On the day before the experiment , animals were placed in the test chamber for 30 min before being subjected to a series of three mock injections with 5 min break periods in between . On the day of the experiment , animals were first allowed to acclimatize to the test chamber for 10 min before injection . Pruritic compounds were subcutaneously injected into the nape of the neck ( 50 μL volume ) or cheek ( 10 μL volume ) , and scratching behavior was observed for 30 min . A bout of scratching was defined as a continuous scratching movement with either hindpaw directed at the area of the injection site . In the cheek injection model , a wipe was defined as a single forepaw stroking the site of the injection . Use of both forepaws on the face or cheek was considered as grooming behavior . Scratching behavior was quantified by counting the number of scratching bouts at 5 min intervals over the 30 min observation period . Wiping was quantified at 2 min intervals over a ten-minute observation period . For H1R block , 30 mg/kg of cetirizine HCl ( pH 7 . 4 ) was given intraperitoneally thirty minutes prior to injection of bilirubin . Licking behavior was quantified in seconds and identified as the licking of the toes or footpad of the hind paw site of injection that was neither preceded nor followed by licking of any other portion of the body . For QWF co-injection experiments , either 100 µM ( for bilirubin ) or 500 µM ( for chloroquine ) QWF was injected in the same volume as solubilized pruritogen . Mrgpr-clusterΔ-/- mice , Mrgpra1GFP mice , and MrgprdPLAP were generated as previously described ( Dong et al . , 2001; Liu et al . , 2009; Liu et al . , 2007 ) . Tg ( Mrgpra3-Cre ) mice were generated as previously described ( Han et al . , 2013 ) . Tg ( Pirt-Cre ) mice were generated as previously described via homologous recombination ( Kim et al . , 2014 ) . Rosa26-LoxP-STOP-LoxP ( lsl ) -GCaMP6s mice were purchased from Jackson Labs . Lsl-tdTomato mice ( Ai9 , 007909 ) were purchased from Jackson Labs . Mrgpra1-/- mice were generated using CRISPR-Cas9 on the C57BL/6 background using the following guide RNA sequence: TTCCCAGCAGCACCTGTGCAGGG . Blvra -/- mice were generated at Ozgene ( Australia ) on a C57BL/6J background . DFT calculations were performed with Spartan 16 and modelled with wxMacMolPlt . Geometry optimizations and single point energy calculations were carried out with DFT-Hartree Fock hybrid B3LYP theory with the 6-31G ( d ) basis set . Energies were calculated at ground state in the gas phase 298 K . All cultured cells were maintained in DMEM supplemented with 10% FBS and 1% Penicillin/Streptomycin at 37°C , 5% CO2 . All data included were generated from cells tested for mycoplasma and found to be negative . Cells were imaged in calcium imaging buffer ( CIB; 10 mM HEPES , 1 . 2 mM NaHCO3 , 130 mM NaCl , 3 mM KCl , 2 . 5 mM CaCl2 , 0 . 6 mM MgCl2 , 20 mM glucose , and 20 mM sucrose at pH 7 . 4 and 290–300 mOsm ) . To monitor changes in intracellular [Ca2+] ( [Ca2+]i ) , cells were loaded with either Fura 2-AM ( HEK293 cells ) or Fluo 4-AM ( DRG neurons and mast cells ) for 30 min in the dark at 37°C in CIB just prior to imaging . With Fura 2-AM , emission at 510 nm was monitored from excitation at both 340 nm and 380 nm . With Fluo 4-AM , emission at 520 nm was monitored from excitation at 488 nm . Cells were identified as responding if the intracellular [Ca2+] rose by either 50% compared to baseline or 50% compared to the [Ca2+]i change assayed during addition of 50 mM KCl ( neurons only ) . Damaged , detached , high-baseline , and motion-activated cells were excluded from analysis . Adult male mice 8–12 weeks of age were sacrificed through CO2 inhalation . A total of 25 mL of mast cell dissociation media ( MCDM; HBSS with 3% fetal bovine serum and 10 mM HEPES , pH 7 . 2 ) was chilled on ice before being used to make two sequential peritoneal lavages . Lavages were combined and spun at 200 g . The pellet was re-suspended in 2 mL MCDM , layered over 4 ml of an isotonic 70% Percoll suspension ( 2 . 8 ml Percoll , 320 ml 10% HBSS , 40 ml 1 M HEPES , 830 ml MCDM ) , and spun for 20 min at 500 g at 4°C . Mast cells were recovered in the pellet . Mast cells were re-suspended in DMEM with 10% fetal bovine serum ( FBS ) and 25 ng/mL recombinant mouse stem cell factor ( Sigma ) . Mast cells were purified as described and allowed to recover for 2 hr at 37°C . Cells were then seeded in 96-well plates coated with 20 mg/mL fibronectin at 300 cells/well . Plates were incubated at 37°C for 45 min before assay . For the assay , all compounds tested were diluted in CIB . Five minutes after compound addition , supernatant was aspirated and frozen at −80°C until histamine levels were determined with an HTRF histamine assay kit ( Cisbio Assays ) according to the manufacturer’s instructions . HEK293 stable cell lines expressing GFP-tagged MRGPRA3 , MRGPRC11 , MRGPRD , MRGPRX1 , and MRGPRX2 were generated in previously described reports ( Liu et al . , 2009; Liu et al . , 2012; McNeil et al . , 2015 ) . Briefly , plasmids containing the receptor of interest were transfected into HEK cells using Lipofectamine 3000 . After 3 days , cells were then selected using 0 . 5 mg/mL G418 . After 3 weeks , monoclonal colonies were established and each the highest expressing clones were identified . For this study , Mrgpra1 and MRGPRX4 were inserted into pEGFP-N1 and transfected into HEK293 cells . MRGPR-positive cells were selected using 0 . 5 mg/mL G418 for three weeks , after which GFP-positive cells were sorted by FACS and monoclonally expanded . Two lines expressing similar levels of MRGPRA1 and MRGPRX4 , as measured by GFP fluorescence , were selected for study . Binding isotherms for MRGPRA1 , MRGPRX4 , and MRGRPC11 towards various ligands were determined by microscale thermophoresis with the NanoTemper monolith NT . 115 instrument ( Duhr and Braun , 2006 ) . Ligands were pre-incubated with 10 µM of the GFP-tagged receptor of interest for 5 min at room temperature in binding buffer ( 20 mM Tris and 150 mM NaCl at pH 8 . 8 ) . Receptors were crudely purified as a membrane fraction from cells stably expressing the receptor ( Vasavda et al . , 2017 ) . Haem metabolites were freshly dissolved in 0 . 1 M NaOH in dim light and then diluted into assay buffer . Lyophilized BAM8-22 was dissolved in binding buffer . The pH of each ligand was evaluated prior to incubation with a receptor . Samples were loaded into NT . 115 Hydrophobic-Treated Capillaries from NanoTemper . Microscale thermophoretic experiments were executed using 20% LED power and 15% MST power . KDs were calculated using the law of mass action with data from three independent experiments . Binding between bilirubin and receptors was evaluated purely thermophoretically , whereas binding between BAM8-22 and MRGPRC11 was evaluated by T-Jump . Samples with dramatic deviations in initial fluorescence were excluded . MRGPR activation was determined by measuring binding of a radiolabelled and non-hydrolyzable form of GTP , [35S]guanosine-5’- ( γ-thio ) triphosphate ( [35S]GTPγS ) as previously described ( Vasavda et al . , 2017 ) . Briefly , 10 µg of crude membrane fractions were diluted into 175 µL assay buffer ( 50 mM HEPES , 5 mM MgCl2 , 100 mM NaCl , 1 mM EDTA , 0 . 1% Triton 80 ) supplemented with 10 µM GDP and incubated at room temperature for 5 min . Membranes were then incubated an additional 1 min in a final volume of 199 μL assay buffer supplemented with 50 μM bilirubin . Samples were then brought to 200 μL with the addition of 10 nM [35S]GTPγS . Samples were incubated for 2 hr at 4°C with gentle agitation . The experiment was terminated by rapid filtration onto GF/B filters and washed three times with wash buffer ( 50 mM Tris-HCl , 5 mM MgCl2 , and 50 mM NaCl at pH 7 . 4 ) . Filters were then immersed in scintillation cocktail and counted . Nonspecific binding was determined by competition with 10 μM unlabelled GTPγS . GTPγS binding assays were performed as two independent experiments , in triplicate . DRG neurons from all spinal levels were collected in cold DH10 media ( 90% Dulbecco’s modified Eagle’s medium ( DMEM ) /F-12 , 10% FBS , penicillin ( 100 U/mL ) , and streptomycin ( 100 μg/mL ) ) . DRGs were digested with a dispase ( 5 mg/ml ) /collagenase type I ( 1 mg/ml ) enzyme mixture at 37°C for 45 min . After trituration , cells were spun at 300 g and re-suspended in DH10 before being plated on glass coverslips coated with poly-D-lysine ( 0 . 5 mg/ml ) and laminin ( 10 μg/ml , Invitrogen ) . DRGs were cultured with DH10 supplemented with 50 ng/mL NGF at 37°C . Lentiviruses encoding various cDNA for MRGPRs were generated using psPAX2 and pMD2 . G . Virus was pelleted by centrifugation at 100 , 000 g for 4 hr , gently washed with twice with DH10 medium , and suspended in DH10 . One day after DRG isolation and culture , DRGs were infected with lentivirus 24 hr overnight . The following morning , medium was completely replaced with fresh DH10 supplemented with 50 ng/mL NGF . 24 hr after infection , cells were processed for calcium imaging . DRG neurons from 3 to 5 week old mice were collected as described . After culture for 1–3 days , DRG neurons were transferred into a chamber with extracellular solution containing ( in mM ) 144 NaCl , 2 . 5 KCl , 2 CaCl2 , 0 . 5 MgCl2 , 5 HEPES , and 10 glucose , adjusted to pH 7 . 4 with NaOH . Whole-cell current-clamp recordings were performed at ~23°C using borosilicate capillary glass electrodes ( Sutter Instrument ) with a tip resistance of 3–5 MΩ . Internal solution contained ( in mM ) 80 K-acetate , 30 KCl , 40 HEPES , and 1 CaCl2 , adjusted to pH 7 . 4 with potassium hydroxide ( KOH ) . Small-diameter neurons with diameter 15–25 μm were chosen for patch-clamp . Data were acquired using an Axopatch 700B Amplifier and Digidata 1322A Digitizer with pClamp9 . 2 software package ( Axon Instrument ) . Chloroquine ( CQ ) in 1 mM was added by perfusion for 20 s , and bilirubin freshly made in 50 µM was added by pipette . Solutions containing 50 mM KCl were applied at the end of each cellular recording . Only neurons that could fire action potentials after adding KCl were regarded as healthy and appropriate for inclusion in data analysis . Adult mice ( 5–6 weeks old ) were anesthetized by i . p . injection of chloral hydrate ( 20 μl/gram of 25 mg/ml solution ) and perfused with 30 ml 0 . 1 M phosphate buffered saline ( PBS ) ( pH 7 . 4 , 4°C ) followed with 30 ml of fixative ( 4% paraformaldehyde ( vol/vol ) , 4°C ) . Skin , trigeminal ganglia , dorsal root ganglia , and spinal cord were dissected from the perfused mice . Tissues were post-fixed in fixative at 4° for 2 hr . Tissues were cryoprotected in 20% sucrose ( wt/vol ) for up to 8 hr followed by 30% sucrose for 24 hr and then sectioned ( 25 μm width ) with a cryostat . The sections were dried at 37°C on slides for 1 hr and fixed with 4% paraformaldehyde at 21–23°C for 10 min . The slides were pre-incubated in blocking solution ( 10% normal goat serum ( v/v ) , 0 . 2% Triton X-100 ( v/v ) in PBS , pH 7 . 4 ) for 1 hr at 21–23°C , then incubated overnight at 4°C with primary antibodies . Secondary antibody incubation was performed at 21–23°C for 2 hr . For primary antibodies , we used rabbit antibody to CGRP ( T-4239 , Peninsula , 1:1 , 000 ) , rabbit antibody to GFP ( A-11122 , Molecular Probes , 1:1 , 000 ) , and Substance P ( rat monoclonal from Abcam , 1:250 dilution , M09205 ) . For secondary antibodies , we used goat antibody to rabbit ( A11011 , Alexa 568 conjugated; A11008 , Alexa 488 conjugated; Molecular Probes ) and Invitrogen 547 ( A-21247 ) for Substance P , all diluted 1:500 in blocking solution . To detect IB4 binding , sections were incubated with Griffonia simplicifolia isolectin GS-IB4 Alexa 568 from Invitrogen at 1:500 dilution ( I-21412 ) . Sections were washed three times with PBS and Fluoromount ( Southern Biotech ) was applied before cover slips were placed over section . Adult mice expressing Pirt-Cre and lox-stop-lox GCaMP6s were anesthetized by i . p . injection of chloral hydrate ( 20 μl/gram of 25 mg/ml solution ) . The back was shaved and disinfected with alcohol before application of ophthalmic ointment ( Lacrilube; Allergen Pharmaceuticals ) . A dorsal laminectomy was performed below the lumbar enlargement ( L5 ) targeting S1 . During the procedure , care was taken to keep dura intact . A 2 cm incision was made at the lumbar enlargement . 0 . 1 mL of 1% lidocaine was injected into paravertebral muscles before dissection to expose L3–L5 vertebrae . Using rongeurs , the surface aspect of the L4 DRG transverse process was removed and the underlying DRG exposed . Mice were laid abdomen-down on a custom-designed microscope stage and the spinal column was secured at two sites using clamps . Images were acquired using a laser-scanning confocal microscope ( Leica LSI microscope ) equipped with a 53 0 . 5 NA macro dry objective and fast EM-CCD camera . Live images were acquired at 8 to 10 frames in frame-scan mode per 7–8 s , at depths of 0 to 70 mm below the dura with the DRG in the focal plane . Images were taken 30 min after peripheral stimulation of DRG via injection of vehicle or bilirubin by Hamilton syringe ( 5 μl ) Throughout imaging , body temperature was maintained at 37 ± 0 . 5°C with a heating pad and rectal temperature monitoring . Anesthesia was maintained with 2% isoflourane and pure oxygen delivered through nosecone . Raw image stacks were collected , deconvoluted , and imported into ImageJ ( NIH ) . Optical planes from sequential time points were re-aligned and motion-corrected using the stackreg rigid-body cross-correlation-based image alignment plugin in ImageJ . Calcium signaling amplitudes were expressed as Ft /F0 as a function of time . F0 was defined as the average pixel intensity during the first two to six frames of each imaging experiment . All neurons that displayed a > 0 . 25 F0 change from baseline were selected for further analysis . In subsequent analysis , neurons that displayed a > 0 . 25 F0 change during either the baseline or the saline imaging periods were excluded from analysis . 1-naphthyl isothiocyanate ( ANIT; Sigma ) was solubilized in olive oil ( Sigma ) . Animals were dosed with 25 mg/kg ANIT per os daily for five days . On day five , animals were acclimatized for itch behavior . On day six , animals were placed in test chambers and videotaped for one hour . The number of scratching bouts , defined as a continuous scratching movement with either hindpaw , was counted and binned in five minute intervals during the one hour period . After itch behavior was assessed , animals were administered pentobarbital ( 50 mg/kg , i . p . ) . Blood was collected by cardiac puncture and placed into heparinized tubes ( BD Biosciences ) . After centrifugation , plasma was collected , aliquoted , and stored at −20°C until analysis . Bile acid levels were assessed by a fluorometric kit from Cell Biolabs . When applicable , mice were then proceeded for histology . For QWF antagonism of cholestatic itch , Day 5 ANIT-treated animals were dosed with either 1 mg/kg QWF dissolved in PBS or PBS vehicle i . v . via tail vein injection approximately 10 min before behavioral assessment of spontaneous itch . The dose was chosen based on previous studies ( Azimi et al . , 2017 ) as well as published pK data indicating stability in plasma ( t1/2 = 70 min ) . The skin of mice were exposed by applying a hair removal cream for 5 min , after which the skin was excised from the back and nape of mice and frozen at −80°C until processing . To extract skin bilirubin , skin was finely minced with a blade and then dounce homogenized at 4°C in 99% chloroform/1% glacial acetic acid ( v/v ) . The organic layer was separated from any remaining tissue by centrifugation at 4°C at 16 , 000 g for 10 min . The organic layer was subsequently washed in 1% glacial acetic acid , then 0 . 2 M NaHCO3 , and then H2O . The organic layer was evaporated with a speedvac until the pellet was dry , after which the pellet was resolubilized in buffer ( 50 mM Tris , 150 mM NaCl , 1% TritonX-100 , 5% glycerol at pH 7 . 4 ) . The final pellet was resuspended as thoroughly as possible , but was unfortunately relatively insoluble . The calculated yield was 1 . 83 ± 0 . 2463 ( SD ) % and accordingly factored in to normalize skin bilirubin . UnaG was expressed in pMAL-6P2-6xHIS in BL21 ( DE3 ) cells . Starter cultures were grown to saturation overnight in Luria Broth ( LB ) at 37°C . Starter cultures were then diluted 10-fold in LB and grown to an OD600 of 0 . 3 at 37°C , after which cultures were moved to 18°C . At OD600 of 0 . 6 , protein expression was induced by the addition of 400 μM isopropyl β-D-1-thiogalactopyranoside for 16 hr at 18°C . Cells were harvested by centrifugation , resuspended in 15 ml of resuspension buffer ( 50 mM HEPES , 300 mM NaCl , 0 . 5 mM TCEP , 10% glycerol , 1 mM PMSF , 2 . 34 μM leupeptin , 1 . 45 μM pepstatin at pH 7 . 4 ) per liter of culture , and flash frozen in liquid nitrogen for storage at −80°C . For purification , pellets were thawed on ice and sonicated to lyse . Lysate was clarified by centrifugation at 26 , 000 g at 4°C for 30 min and subsequently loaded onto an amylose column . Protein was eluted with 20 mM maltose in protease buffer ( 50 mM Tris , 150 mM NaCl , 0 . 5 mM TCEP , 10% glycerol , 0 . 01% TritonX-100 at pH 7 . 4 ) , after which the MBP was were removed by the addition of Prescission Protease ( GE Healthcare ) for 16 hr at 4°C . UnaG was further purified by a second nickel-affinity step to remove the cleaved tag and Prescission Protease . Protein was further purified on a gel-filtration column ( S-200 , GE Healthcare ) in 50 mM HEPES , 300 mM NaCl , 0 . 5 mM TCEP , 10% glycerol at pH 7 . 4 . The purity of peak fractions was assessed by SDS-PAGE . The purified protein was concentrated using a 10 kDa MWCO filter ( Amicon ) and flash frozen in gel filtration buffer supplemented with 30% glycerol for storage at −80°C . To measure plasma bilirubin , patient plasma was diluted 1:20 in HBSS containing purified UnaG . To measure skin bilirubin , UnaG was added directly After a 10 min incubation at 25°C , bilirubin was quantified by interpolating from a standard curve . Samples with blood and hemolysis were excluded from analysis . Skin bilirubin was ormalized to the dry weight of the skin . Plasma bilirubin was depleted either by selective oxidation by FeCl3 to biliverdin IXα/biliverdin XIIIα or by immunoprecipitation . FeCl3 was prepared as solution of 20% FeCl3 in 0 . 1N HCl/methanol . FeCl3 was fluxed with plasma at a final concentration of 1 . 5% FeCl3 at 37°C for 10 min . FeCl3 is a mild oxidant but exhibits a redox potential that the oxidation of bilirubin to biliverdin ( Dolphin , 1978 ) . Bilirubin was also immunopreciptated by incubating plasma with 5 μg of either normal rabbit IgG or anti-bilirubin antibody ( generated as previously described ( Doré et al . , 1999 ) ) coupled to protein A/G beads for 1 hr at 25°C . To quantify bilirubin depletion , bilirubin was extracted from samples with 100% methanol and subjected to HPLC and UV-visible spectroscopic analysis . Absorbance was adjusted to a baseline of 0 OD , and bilirubin was quantified by integrating the area under the chromatographic peak . Plasma bilirubin was detected by HPLC using an analytical LC-18 column , 25 cm ×4 . 6 mm ( Xterra , Waters Corporation ) . Bilirubin was eluted with gradients of mobile phases: 0 . 1 M ammonium acetate in 60% methanol/40% water ( v/v ) ( pH 5 . 2 ) ( Solvent A ) and 100% methanol ( Solvent B ) . Bilirubin was eluted as follows: 0 to 14 min: linear gradient from 100% A to 100% B; 14 to 19 min: linear gradient from 100% A to 100% B; 19–24 min: isocratic elution at 100% A . Bilirubin exhibited a retention time of approximately 14–15 min and was detected by measuring absorbance at 450 nm . The peak corresponding to plasma bilirubin was confirmed with the addition of 10 μM bilirubin to the sample as an internal standard . Group data were expressed as mean ±SEM unless otherwise noted . Two-tailed unpaired Student’s t-tests , Fisher’s exact test , and Chi-squared tests were used to determine significance in statistical comparisons , and differences were considered significant at p<0 . 05 . Statistical power analysis was used to justify sample size , and variance was determined to be similar among all treatment groups as determined by F test . No samples or animals subjected to successful procedures and/or treatments were excluded from analysis . All behavior experiments were designed in a blocked manner with consideration for both genotype and treatment .	Jaundice causes the skin to yellow as a result of a build-up of a pigment called bilirubin . Normally , bilirubin is made in the liver and removed from the body in digestive fluid called bile , but people with liver or gallbladder problems may end up with too much bilirubin that accumulates in their blood and skin . One side effect of jaundice is intense and uncontrollable itching . Researchers are not sure what causes this itching , and there are few treatments that help to relieve it . At the molecular level , itching sensations occur when compounds bind to particular receptors on the surface of nerve cells . One family of receptors that can trigger itch is called the Mas-related G-protein Coupled Receptor ( MRGPR ) . Could one of these receptors trigger jaundice-related itching ? Now , Meixiong , Vasavda et al . show that bilirubin binds to and activates MRGPRs to cause itch in mice . Whereas injecting bilirubin into normal mice causes them to scratch , mice that have been genetically engineered to lack MRGPRs do not itch when their own bilirubin levels rise , or when they are injected with bilirubin or with plasma from patients who experience jaundice-related itching . Furthermore , removing bilirubin from the plasma of patients before it was injected into normal mice reduced the amount of itching that the mice felt . Overall , the results reported by Meixiong , Vasavda et al . suggest that drugs that prevent bilirubin from attaching to MRGPRs might help to alleviate jaundice-related itching . However , researchers must first verify that bilirubin interacts with MRGPRs in people to cause itch . If bilirubin causes itch in people like in mice , scientists could then evaluate existing drugs or make new ones to prevent bilirubin from attaching to the MRGPRs .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	Identification of a bilirubin receptor that may mediate a component of cholestatic itch
Exercise has a wide range of systemic effects . In animal models , repeated exertion reduces malignant tumor progression , and clinically , exercise can improve outcome for cancer patients . The etiology of the effects of exercise on tumor progression are unclear , as are the cellular actors involved . We show here that in mice , exercise-induced reduction in tumor growth is dependent on CD8+ T cells , and that metabolites produced in skeletal muscle and excreted into plasma at high levels during exertion in both mice and humans enhance the effector profile of CD8+ T-cells . We found that activated murine CD8+ T cells alter their central carbon metabolism in response to exertion in vivo , and that immune cells from trained mice are more potent antitumor effector cells when transferred into tumor-bearing untrained animals . These data demonstrate that CD8+ T cells are metabolically altered by exercise in a manner that acts to improve their antitumoral efficacy . In humans , exercising cohorts have lower rates of cancer incidence ( Moore et al . , 2016 ) and better outcomes across a range of cancer diagnoses ( Cormie et al . , 2017; Friedenreich et al . , 2016 ) , proportionate to the degree and intensity of exercise . The mechanisms underlying these observations have remained elusive , although recent work has indicated a relationship between immune response and exercise-induced changes in malignant progression ( Pedersen et al . , 2016; Koelwyn et al . , 2017 ) . The metabolic demands of strenuous physical exertion generally induce significant changes in nutrient utilization , principally via central carbon metabolism ( Brooks , 1998 ) . These exercise-induced alterations in metabolism change the ratios of energy substrates utilized , and can shift intramuscular metabolite profiles . These shifts are reflected in systemic metabolite availability , which in turn modifies energy production throughout the body ( Henderson et al . , 2004; Lezi et al . , 2013; El Hayek et al . , 2019 ) . It is clear that cytotoxic T cells play a crucial role in controlling tumor growth . By recognizing mutation-derived neoantigens , T cells can identify and eliminate malignant cells in a process known as immunosurveillance ( Dunn et al . , 2002; Dunn et al . , 2004 ) . Escape from immune control is a critical step toward progressive malignant growth in many cancers , and tumors achieve this in a number of ways , amongst them the dampening of antitumor T cell responses ( Dunn et al . , 2002; Dunn et al . , 2004; Beatty and Gladney , 2015; Zappasodi et al . , 2018 ) . The activity of immune cells is tightly linked with their metabolism ( O'Neill et al . , 2016; Pearce and Pearce , 2013 ) . Many aspects of immune cell energetics are likely sensitive to the metabolic changes induced by exercise ( Henderson et al . , 2004 ) . Exercise is known to affect immune cell function , and an altered immune response has been suggested as a mechanism underlying effects of exercise on cancer risk and progression ( Christensen et al . , 2018 ) . In this study , we investigate the association between exercise , tumor growth , and CD8+ T-cell function . To address this , we undertook studies of exercise-induced changes in tumor progression , and asked what metabolites are released in response to exercise; as well as whether metabolites produced by exercise can alter cytotoxic T-cell function . We found that exercise itself can modify cytotoxic T-cell metabolism , and that exercise-induced effects on tumor growth are dependent on cytotoxic T-cell activity . To address the role of immunity on the effects of exercise in neoplasia , we first assessed how repeated voluntary exertion influenced tumor progression in mice , using a genetic model of mammary cancer induced by the MMTV-PyMT transgene on the FVB inbred strain background ( Figure 1—figure supplement 1A ) . FVB inbred mice are enthusiastic runners relative to most other inbred strains ( Avila et al . , 2017 ) , and the MMTV-PyMT model in many regard mimics the gradual progression of human breast cancer ( Lin et al . , 2003 ) . PyMT+ mice ran on average 6 km/day ( Figure 1—figure supplement 1B ) . Contrary to what was previously shown ( Goh et al . , 2013 ) , the running mice in our experiments showed no statistically significant differences in tumor growth between the groups ( Figure 1—figure supplement 1C- G ) . However , the infiltration of Granzyme B ( GZMB ) -positive cells was significantly higher in primary tumors of running mice , even though voluntary running had no effect on the frequency of CD3 , F4/80 , PCNA , or podocalyxin-expressing cells ( Figure 1—figure supplement 1H ) . This indicates that exercise in this tumor model modulates the infiltration of cytotoxic lymphocytes . Given that infiltration of cytotoxic T cells is linked to a favorable prognosis in many human neoplasms and in some animal tumor models ( Savas et al . , 2016 ) , we proceeded to explore the role of lymphoid cells , and in particular cytotoxic CD8+ T cells , on the process of exercise-induced reduction of tumor growth . As the progression of the transgenic MMTV-PyMT model is predominantly regulated by myeloid cells ( Lin et al . , 2003 ) , we carried out a further set of experiments , wherein immunocompetent FVB mice were given access to an exercise wheel 14 days prior to , and then following , subcutaneous injection with a murine mammary cancer line derived from the MMTV-PyMT model , the I3TC cell line ( Weiland et al . , 2012; Figure 1A ) . In this cell-line-driven model , allowing animals to exercise significantly reduced tumor growth and increased survival times ( Figure 1B ) relative to animals not given access to an exercise wheel . To determine whether immune cell populations were also affected by exercise in the subcutaneous I3TC tumor model , we carried out flow cytometric analyses of single-cell suspensions of inoculated tumors , spleens , and tumor-draining lymph nodes of running versus non-running animals . Here , only CD8+ T cells showed a significant increase in frequency across these three tissues in the running animals ( Figure 1E ) ; no significant changes were seen in the frequency of CD4+ T cells , NK cells ( Figure 1C and D ) , or intra-tumoral macrophages or neutrophils ( Figure 1—figure supplement 1I and J ) . To determine the role of increased cytotoxic T-cell populations in the reduction of tumor growth caused by exercise , we depleted CD8+ T-cells via weekly injections of anti-CD8 antibodies in both exercising and non-exercising mice during subcutaneous I3TC tumor growth . The CD8+ T-cell depletion reduced the CD8+ population of both the spleen ( Figure 1F ) and the tumor ( Figure 1G ) , and significantly reduced the beneficial effects of exercise on both tumor growth ( Figure 1H ) and long-term survival ( Figure 1I ) . This depletion study demonstrates a clear role for cytotoxic T-cells in the suppression of tumor growth by exercise in this model . Acute exercise induces changes in several known metabolic mediators of immune responses ( Henderson et al . , 2004 ) . To investigate how the systemic availability of metabolites can change in response to exercise , we undertook a metabolomic investigation of muscle and plasma in response to exercise . After habituation , an exhaustive endurance test was performed on wild-type mice on the FVB background , and following exercise , skeletal muscle and plasma were harvested immediately , snap frozen and analyzed by mass spectrometry ( GC-MS ) ( Figure 2—figure supplement 1A-C ) . Exercise affects a wide range of metabolic pathways; in particular , exercise reduced the frequency of glycolytic metabolites in skeletal muscle , for example , fructose-6-phosphate , glucose-6-phosphate and 3-phosphoglyceric acid ( 3 PG ) , while the frequency of several of the TCA metabolites , for example , citric acid , fumaric acid and malic acid , were significantly higher in muscle post-exercise ( Figure 2—figure supplement 1B ) . This is in line with mitochondrial metabolism being rate-limiting in high-intensity exertion ( Brooks , 1998 ) . Interestingly , the TCA metabolites citric acid , malic acid and alpha ketoglutaric acid ( aKG ) were all higher in plasma after exertion ( Figure 2—figure supplement 1C ) , suggesting release of these metabolites from muscle into plasma ( Figure 2—figure supplement 1D ) . Contribution of muscle citrate to plasma has previously been shown; for example , by femoral arterio-venous sampling during exercise in human subjects ( Nielsen and Thomsen , 1979; Schranner et al . , 2020 ) . To extend the information to include the lymphoid organs , an additional set of exhaustive endurance tests was performed on mice followed by immediate harvest of skeletal muscle , plasma , spleen , and muscle draining ( axillary ) and non-draining ( inguinal ) lymph nodes . In addition , human plasma was harvested before ( pre-exercise ) , directly after ( post-exercise ) and 1 hr ( 1 hr post ) after an intense endurance exercise session in sedentary individuals ( Figure 2—figure supplement 1E ) . These tissues were then analyzed by ultrahigh performance liquid chromatography/mass spectrometry ( Figure 2A ) . As well as altering the intramuscular ( Figure 2B ) and circulating levels of the TCA metabolites ( Figure 2C and D ) , acute exercise in these experiments also introduced shifts in the metabolic profiles of lymphoid organs , supporting the notion that acute exercise alters the metabolic environment of lymphoid organs ( Figure 2E–G ) . Notably , the number of significantly increased metabolites was markedly higher in the muscle draining lymph nodes when compared to non-draining lymph nodes ( Figure 2F and G , respectively ) . This indicates that changes seen in draining lymph nodes are likely attributable to muscle metabolite production . Interestingly , plasma , draining lymph nodes , and non-draining lymph nodes all showed higher levels of corticosterone after exercise ( Figure 2H ) . Changes in amino acid and fatty acid levels were also identified in multiple organs ( Figure 2—figure supplement 1H ) . However , the single most profound metabolic change induced by exertion is the transient increase in circulating lactate . Circulating lactate levels rise very rapidly during exercise , and can ( in response to high intensity exertion ) increase up to 100-fold in skeletal muscle ( Bonen et al . , 1998; Spriet et al . , 1987 ) , with a concomitant increase of more than 10-fold in plasma ( Goodwin et al . , 2007 ) . The rapid postmortem accumulation of systemic lactate in response to global oxygen deprivation made us unable to differentiate the lactate levels in murine organ samples ( Donaldson and Lamont , 2013; Keltanen et al . , 2015 ) . An increase in lactate as well as TCA cycle metabolites was however seen in human plasma post-exercise ( Figure 2D ) and when measuring lactate in blood from the tail vein in live animals directly after an acute treadmill exercise ( Figure 2—figure supplement 1G ) . In the human samples , these returned to close to resting levels 1 hr after exercise ( Figure 2—figure supplement 1F ) , indicating that the changes in central carbon availability are likely conserved between mice and humans . As shown above , exercise can reduce tumor growth in a CD8+ T-cell-dependent manner . Given that metabolism and T-cell differentiation are tightly linked ( O'Neill et al . , 2016; Pearce , 2010 ) , we next sought to determine whether any of the central carbon metabolites generated during exercise could be a determining factor in the action of CD8+ T cells in this process . To test this , we activated CD8+ T cells ex vivo for 3 days in the presence of increasing doses ( 25 μM-50mM ) of pH-neutralized lactic acid , pyruvic acid , citric acid , malic acid , succinic acid , fumaric acid , a-ketoglutaric acid , or sodium L-lactate . Reference human resting serum levels are provided in Figure 3—figure supplement 1A . Although the production of lactic acid is the immediate consequence of increased glycolytic flux during exercise , in healthy tissues , the increase in H+ is rapidly buffered in surrounding tissue and in plasma ( Péronnet and Aguilaniu , 2006 ) . For this reason , lactate produced during exercise does not significantly alter circulating plasma pH levels ( unlike the acidification seen in , for example , solid tumors ) ( Brand et al . , 2016 ) . We therefore , adjusted pH to 7 . 35 for all metabolites prior to adding them to the cell culture media . TCA metabolites can be transported across the plasma membrane with sodium carboxylate cotransporters ( Markovich , 2012 ) . Although none of the metabolites provided proliferative advantages after 3 days of activation , malate , succinate , and aKG were notably well tolerated by the T-cells at high doses ( Figure 3A , Figure 3—figure supplement 1B-C ) . Activation in the presence of increasing concentrations of pyruvate and citrate reduced CD8+ T cell expansion ( Figure 3A Figure 3—figure supplement 1B-C ) . Malate , succinate , fumarate , aKG , and lactate all induced a loss of CD62L at around 1 mM concentrations , ( Figure 3B , Figure 3—figure supplement 1D ) . Loss of CD62L surface expression is induced in response to TCR activation and instrumental for the migratory capacity of cells localized in secondary lymphoid organs . Based on the data from the metabolite screen , TCA metabolites appear to enhance the loss of CD62L in response to activation . Furthermore , sodium L-lactate induced a dose-dependent increase in inducible T-cell costimulator ( iCOS ) and Granzyme B ( GzmB ) expression at day 3 of culture ( Figure 3C–D , Figure 3—figure supplement 1E-F ) . We found no significant effect of metabolite exposure on T-cell CTLA-4 expression ( Figure 3—figure supplement 1G ) . The effect of sodium L-lactate could be detected starting at concentrations of approximately 6 mM . ICOS is expressed on activated T-cells and GzmB is the most highly expressed effector protein in activated mouse CD8+ T cells ( Hukelmann et al . , 2016 ) and is essential to granule-mediated apoptosis of target cells ( Heusel et al . , 1994 ) . In keeping with this , in co-culture with ovalbumin ( OVA ) expressing EL4 tumor cells , OVA-specific OT-I CD8+ T-cells activated in the presence of sodium L-lactate for 72 hr showed increased cytotoxicity against the tumor target cells ( Figure 3E ) . Because of our findings of altered T-cell function in response to exercise-induced metabolites , we next sought to investigate if there is evidence of an alteration in the CD8+ T-cell metabolome in vivo after intense exertion . To address if exercise can alter the central carbon metabolism of activated CD8+ T cells in vivo , we employed an OVA vaccination model and adoptive transfer of transgenic , OVA-specific OT-1 CD8+ T-cells where the timing of the T-cell activation can be controlled . In short , transgenic OT-I CD8+ CD45 . 1+ T-cells were administered to recipient animals , followed by vaccination with bone marrow derived macrophages ( BMDM ) presenting ovalbumin to activate the OT-1 T-cells . Two or 3 days after the vaccination , a bolus of [U-13C6]glucose was introduced to resting and exercising mice ( Figure 4A ) , and the central carbon metabolism of the OT-1 CD8+ T-cells isolated from the spleen was assayed . The [U-13C6]glucose was introduced after warm up on a treadmill for 10 min at low speed , so as to ensure maximal glucose uptake by the skeletal muscle at the time of injection . The spleens were harvested at 20 min post-exercise and at the equivalent time point in the resting animals . The data shows a diverging carbon metabolism of the exercised CD8+ T-cells at the level of m+three labeled pyruvate and m+two labeled aKG , indicating an altered enzymatic activity or contribution of extracellular labeled molecules ( Figure 4B and C ) , providing evidence that exercise can alter the metabolism of CD8+ T-cells in vivo . In order to determine the functional effects of exercise on the CD8+ T-cell population , adoptive transfer of naïve CD8+ T-cells from exercising OT-1 animals was carried out . These T-cells were transferred to C57Bl6 inbred strain recipient mice that had been inoculated with an OVA-expressing melanoma ( B16-F10-OVA ) ( Figure 5A and B ) . Following the transfer of T-cells to non-exercising mice , tumor growth was monitored for 40 days . Blood profiles on day 10 following transfer confirmed the expansion of the OT-1 population in the recipient mice , and also showed a significant increase in expression of iCOS in the cells transferred from exercising donors ( Figure 5C and D ) . The sedentary recipient animals that received T-cells from exercising donors showed an enhanced survival and reduced rate of tumor growth , when compared to sedentary animals receiving T-cells from sedentary donors ( Figure 5E–G ) . This indicates that there is a persistent and positive effect on the efficacy of anti-tumor CD8+ T-cells when the T-cells are derived from exercising animals . Given the above results that exercise-induced metabolites can alter the properties of immune cells , we wanted to investigate if the lactate-induced increases in CD8+ T cell differentiation markers and cytotoxic efficacy might extrapolate to affect tumor growth in vivo . Therefore , we performed daily infusions of sodium L-lactate into tumor-bearing animals at doses that result in plasma lactate levels similar to those seen during intensive exercise ( approximately 10–20 mM ) . As can be seen in Figure 6—figure supplement 1A , an intraperitoneal injection of a 2 g/kg dose of sodium L-lactate results in a 18 mM spike in serum lactate concentrations at 20 min post-injection . Following this dose , levels subside to 4 mM within 60 min; the expected time to reach baseline values from this magnitude of spike is approximately 180 min post-injection ( Lezi et al . , 2013 ) . This dose was chosen as an approximation of the levels and persistence of rises in plasma lactate that occur following intense short-term periods of exercise . Female FVB mice , given daily doses of 2 g/kg sodium L-lactate , showed a decrease in overall tumor growth after inoculations with the I3TC tumor cell line ( Figure 6A ) . Similar results were obtained with the colon adenocarcinoma MC38 cell line on C57BL/6 animals ( Figure 6C ) , with accompanying increases in tumor-bearing animal survival . Lower doses of lactate ( 0 . 5–1 g/kg ) did not significantly alter tumor growth ( Figure 6A ) , while a higher daily dose of Sodium L-lactate , 3 g/kg , caused a significant reduction in tumor growth , with approximately the same efficacy as the 2 g/kg dose ( Figure 6—figure supplement 1B ) . As shown in Figure 6B , there were significant changes in tumor infiltrating immune populations in animals treated with 2 g/kg of lactate , namely , increases in the frequency of total ( CD3+ ) T cells , including both CD4+ and CD8+ T cells . The frequency of tumor-infiltrating NK cells was reduced . Despite finding increased numbers of tumor infiltrating CD8+ T cells ( Figure 6B and Figure 6—figure supplement 1C ) , daily lactate administration did not change the percentage of GzmB+ cells or the GzmB , PD1 or CTLA intensity on CD8+ infiltrating cells ( Figure 6—figure supplement 1D-E ) . To determine whether the effect of lactate infusion is dependent on CD8+ T-cell-mediated effects on tumor growth , we depleted the cytotoxic T cell population via injection of an anti-CD8 antibody ( Figure 6D ) and observed that the net effect of lactate injection was reduced to statistical insignificance by the depletion of the CD8+ T cell population . This indicates that the effect of daily lactate injections on moderating tumor growth is mediated by CD8+ T cells . Here , we show that exercise requires cytotoxic T cells to affect tumor growth . Recent studies suggest that exercise reduces cancer recurrence and mortality , and the effect of exercise on tumorigenic progression has now been documented in a range of animal models ( Ruiz-Casado et al . , 2017 ) . Previously proposed underlying mechanisms for the anti-neoplastic effects of exercise include effects on weight control and endocrine hormone levels , as well as altered tumor vascularization ( Betof et al . , 2015 ) and immune function ( Gleeson et al . , 2011 ) . Exercise has been shown to reduce systemic inflammation ( Gleeson et al . , 2011 ) , cause increases in circulating numbers of immune cells ( Gustafson et al . , 2017 ) and enhance vaccination response ( Ledo et al . , 2020 ) . Systemic levels of both CD8+ T cell and NK cell levels increase transiently in response to acute exercise . There is evidence that these immune cells populations exhibit an effector phenotype ( Campbell et al . , 2009 ) and an increased peripheral tissue homing capacity ( Krüger and Mooren , 2007 ) , both important for anti-tumor activity . Exercise is a multimodal stimulus . Systemic signaling from activated skeletal muscle to immune cell function has been attributed previously to for example skeletal muscle-derived myokines such as IL-6 and Oncostatin-M ( OSM ) , and to catecholamine release ( Gleeson et al . , 2011 ) . In a recent study , Hojman et al . , showed that an increase in systemic levels of epinephrine during exercise together with skeletal muscle IL-6 could increase NK cell recruitment during tumorigenesis ( Pedersen et al . , 2016 ) . However , in the mouse models employed here we found limited infiltration and no effect of exercise on the CD3- NK1 . 1+ population , relative to our observation of an increased frequency of CD3+/CD8+ T cells . The results from depletion and adoptive transfer of the CD8+ population provide further support for the importance of cytotoxic T-cells in mediating the anti-neoplastic effects of exercise . Immune cells patrol all corners of the body and can reside in niches with highly differential metabolic profiles . Recent findings have provided evidence that these different environments can instruct effector functions of immune cells ( Buck et al . , 2017 ) . In the current study , we show how exercise modulates metabolic parameters significantly beyond local skeletal muscle . We found that lactate and TCA metabolites accumulated in both the active skeletal muscle and in plasma and secondary lymphoid organs after acute exertion . As trafficking immune cells circulate , they will often be exposed to high levels of accumulating metabolites , in both loci of disease and in metabolically active tissues . However , the principal shaping of immune function happens at the point of activation , primarily occurring in the lymphoid organs . In evaluating the role an altered metabolic environment plays in the process of modifying CD8+ T-cell function , we found that key metabolites , including lactate , had unexpected functions in the activation of CD8+ T cells , with multiple TCA metabolites stimulating the loss of CD62L . CD62L shedding from TCR activated CD8+ T-cells has been suggested to be mTOR and PI3K dependent ( Sinclair et al . , 2008 ) and aKG-mediated activation of mTORC1 has been reported in multiple cell lines ( DeBerardinis et al . , 2007; Durán et al . , 2013; Villar et al . , 2015 ) , as well as in activated CD8+ T-cells ( Suzuki et al . , 2018 ) , suggesting that the metabolite-induced loss of CD62L may be mediated through mTOR activation . In addition , we found that lactate induces an effector profile and an overall increase in cytotoxic activity . In contrast to our findings , within the local micro-environment , tumor-derived lactate has been shown to inhibit the antitumor functions of both T and NK cells due to intracellular acidification ( Brand et al . , 2016 ) . Likewise , lactate-associated acidity has been shown to impair CD8+ T cell motility ( Haas et al . , 2015 ) . Importantly , within poorly vascularized and fast-growing tumor masses , or in laboratory ex vivo cultures , lactic acid export from cells results in a significant acidosis . However , healthy tissue microenvironments are very efficiently buffered , and capable of maintaining a physiological pH range even when large amounts of lactate are produced . Our findings that circulating levels of TCA metabolites increase by 2–8 fold after exercise suggests that at peripheral sites , substantial amounts of these metabolites have been released into the lymphatic fluid draining the muscle tissue . Additionally , a recent paper using lymphatic cannulation showed that TCA metabolites are readily available in , for example , mesenteric lymph ( Zawieja et al . , 2019 ) . Thus , to mimic the activation of T cells in a lymphoid organ exposed to the physiologically buffered metabolites generated by exercising muscle , we activated CD8+ T cells in the presence of TCA metabolites and lactate in media adjusted to a physiological pH . The use of distally produced lactate as a fuel for oxidative respiration was described for inactive muscle already in 1970 ( Jorfeldt , 1970 ) and has recently been reported in both malignant and non-malignant cell types ( Faubert et al . , 2017; Hui et al . , 2017 ) , suggesting that lactate can be used as a primary fuel by CD8+ T cells , at least during early activation . In vivo administration of labeled glucose two and three days after vaccination , led to differential levels of labeled pyruvate and alpha-ketoglutarate in CD8+ T-cells of the spleen post-exercise . Dilution of 13C labeling between glycolysis and TCA cycle intermediates can imply oxidation of alternative fuels ( Hensley et al . , 2016 ) . The administered 13C-glucose was , as expected , used as a carbon source by CD8+ T cells . Although total glucose levels were similar between T-cells from exercising and non-exercising animals ( Figure 4—figure supplement 1A ) , the availability of labeled glucose appeared lower in exercised animals , reflected by lower total 13C-glucose counts ( Figure 4—figure supplement 1B ) . The reduced availability is most likely due to dilution of the 13C-glucose in the circulating blood glucose pool due to a contribution from hepatic glycogenolysis , stimulated by the increased metabolic demand during exercise . When normalized to the labeled glucose fraction , exercise altered the central carbon metabolism of activated CD8+T-cells in vivo at the level of m+3 labeled pyruvate and m+2 labeled aKG . This finding suggests an increased uptake of systemic metabolites by activated T-cells during exercise . When isolating the exercise effect on the CD8+ T-cell population by transferring cells from exercised animals to tumor bearing , non-exercising recipients , we show an enhanced efficiency in reducing tumor growth by the T-cells obtained from exercising individuals . These results demonstrate that in our models , the exercise effect is exerted by and inherent to the CD8+ T cell population . In the experimental setting where animals were given daily doses of high levels of sodium L-lactate , we found an increase in intratumoral CD8+ T-cells and a decrease in overall tumor growth . In addition , we found high levels of CD4+ cells and reduced NK-cell infiltration , supporting previous evidence that lactate can affect multiple cells of the immune system . These findings indicate that lactate infusion mimics some of the effects of exercise , but that exercise has additional , integrative , components beyond merely increased levels of lactate . In conclusion , we have shown that the antitumoral effects of exercise depend on CD8+ T-cells , and that intense exertion can alter the intrinsic metabolism and antitumoral effector function of cytotoxic T-cells . This indicates that , in the physiological setting , exercise-derived metabolites , whether systemically delivered or draining into an adjacent lymph node , could act to boost a nascent T cell response . This indicates that the adaptive immune system is a key component of exercise-induced suppression of tumorigenesis . All experiments and protocols were approved by the regional animal ethics committee of Northern Stockholm ( dnr N78/15 , N101/16 ) . Female wild type ( WT ) FVB-mice were bred with transgenic male mice carrying the PyMT oncogene under the MMTV promotor ( Lin et al . , 2003 ) on the FVB inbred strain background . The female offspring , being either PyMT-positive ( heterozygous ) or WT , were from an age of 4 weeks introduced to voluntary running conditions . For sprint and inoculation experiments , female wild type ( WT ) FVB-mice , 6–7 weeks of age were purchased from Janvier Labs ( France ) , allowed 7 days of acclimatization and housed two mice per cage in a temperature controlled room ( 20 ± 2°C ) with dark-light cycles of 12 hr and access to food and water ad libitum . For T cell purification , and adoptive transfer , wild type donor and recipient C57BL/6J animals were purchased from Janvier Labs . TCR-transgenic OT-I mice ( catalogue 003831 , The Jackson Laboratory ) were crossed with mice bearing the CD45 . 1 congenic marker ( catalogue 002014 , The Jackson Laboratory ) . The mice were habituated to the treadmill ( Columbus instruments OHIO Exer 3/6 ) for 4 to 5 days prior to the test . Habituation was performed by first placing the mice on a stationary band , with a 10° uphill slope , for a few minutes . The speed was then gradually increased for about 1 min , to a maximum of 6 m/min , which was then kept for 5 min . On the following days , the same procedure was performed , and time and speed were gradually increased in order to ensure that the mice were fully familiarized with running on the treadmill . During the test , the mice ran at 10° uphill . An initial 10 min warm-up ( 10 m·min− 1 ) was followed by gradual increases of 2 m·min− 1 every 2 min . Exhaustion was determined as the time when the mouse had withstood three mild motivations without attempts to continue running . Control mice were kept in the cage during the time of the test . Four-week-old PyMT+ females of the MMTV-PyMT strain were housed individually in a temperature controlled room ( 20 ± 2°C ) with 12 hr dark-light cycles and food and water ad libitum . The mice were allowed free access to a running wheel ( Med Associates Inc ENV-044 , St Albans , USA ) . Running patterns and distance were monitored wirelessly using appropriate software ( Med Associates Inc SOF-860 Wheel Manager and Med Associates Inc SOF-861 Wheel Analysis ) . A control group with identical but locked wheels was included in the study to control for environmental enrichment . Animal body weight was monitored weekly throughout the experiment and mammary gland tumors measured twice weekly using calipers . Tumor volumes ( V ) were approximated using the formula V = ( π/6 ) * l * w * h . Experiments were terminated when the animals reached 12 weeks of age . FVB WT animals were introduced to running or locked wheels and allowed to acclimate to running for 2 weeks prior to injections of 5 × 105 tumor cells of the I3TC cell line . The cells were trypsinized , washed once with PBS and suspended in 100 µL sterile PBS and injected subcutaneous ( s . c ) into running and non-running FVB WT mice . Animal body weight was monitored throughout the experiment . Tumors size was measured as previously described . Experiments were terminated 8 weeks after tumor cell injections . Upon termination of the experiments , tumor specimens were surgically removed and fixed in 4% paraformaldehyde ( Sigma , F8775 ) or dissociated for single cells suspension . For CD8 depletion , 200 µg of antibodies ( anti-mCD8α clone 53–6 . 72 , BioXCell or Rat IgG2a Isotype control , clone 2A3 , BioXCell ) was injected i . p at day 4 post-tumor cell injections and repeated weekly throughout the experiment . For T-cell adoptive transfer , 8 to 12 weeks old female C57BL/6J were inoculated sub-cutaneously with 5 × 105 B16-F10-OVA and conditioned 4 days later with peritoneal injection of 300 mg/kg cyclophosphamide ( Sigma , C0768 ) in PBS . At day 7 , 4 × 105 transgenic OT-I from running/non-running mice were injected peritoneally to the tumor bearing recipient mice . Tumor volume was measured every 2–3 days with electronic calipers until day 60 . At day 17 after OT-I cell transfer , peripheral blood was collected from tail vein , red blood cells lysed with water , and leukocyte cell suspension analyzed by flow cytometry . To evaluate the effect of lactate on tumor growth in vivo , FVB and C57/Bl6J WT animals were subject to daily intraperitoneal ( i . p ) Sodium L-Lactate ( Sigma , L7022 ) injections ( 0 . 5 , 1 , 2 , or 3 g/kg ) for 12 days prior to injections of 5 × 105 tumor cells of the I3TC or MC38 cell line . Tumor growth was monitored as previously described and daily injections of Sodium L-Lactate was continued throughout the experiment . To ascertain transient levels of sodium L-lactate in plasma induced by lactate injection , 2 g/kg of Sodium L-Lactate was administered i . p to FVB and C57Bl6 animals ( n = 1 ) , systemic levels of lactate were monitored through tail vein bleedings pre-injection , and at 5 , 10 , 20 , 40 , and 60 min post-injection ( AccutrendPlus ) . Eight healthy , non-smoking sedentary men aged 34–51 were recruited after filling in a questionnaire to assess their health status and performing a peak oxygen uptake test ( V˙O2 peak test ) to assess their maximal performance level . All subjects completed a single bout of acute endurance exercise which consisted of 30 min of cycling at 75% of their Wpeak . Food intake was controlled for by providing a standardized breakfast . Blood was sampled from V . mediana cubiti at three timepoints: immediately pre and 1 and 3 hr following acute endurance exercise using Li-Heparin Plasma separation tubes ( BD Vacutainer #367377 , BD Biosciences ) . Plasma was separated by centrifugation at 3000 g for 10 min and immediately frozen at −80°C . For the in vivo 13C glucose test , 8 to 12 weeks old female C57BL/6J mice were habituated to the treadmill . On day −1 , 2 × 106 transgenic OT-I-CD8+ T-cells were injected intra peritoneally . On day 0 , the mice were vaccinated using OVA-antigen presenting BMDMs , and on days 2 and 3 , the mice were split into running and resting mice ( n = 4 ) . The running mice were allowed to warm up on the treadmill , before 10 mg of [U-13C6] glucose was injected peritioneally and the running mice performed the 20 min incremental endurance test , as previously described . The mice were then allowed to recover for 20 min before cervical dislocation euthanasia and collection of organs . Spleens were harvested and placed in ice-cold PBS on ice and quadriceps muscle from one hind leg were dissected , frozen in liq N2 and stored at −80 °C until further processing . 2 × 106 CD45 . 1+ CD8+ splenocytes were isolated and frozen on a dry ice and ethanol slurry and stored at −80 °C until further processing . EL4 was a gift from Prof . H . Stauss ( UCL , London ) . B16-F10 and LLC were purchased from ATCC ( CRL-6475 and CRL-1642 , respectively ) . I3TC was originally derived from the FVB MMTV-PyMT breast cancer model ( Weiland et al . , 2012 ) . B16-F10 , LLC and cells were co-transfected with the transposon vector pT2 encoding OVA , eGFP and neomycin phosphotransferase and the vector encoding transposase SB11 . Three days later 400 mg/ml G418 ( Gibco , 10131035 ) was added to culture media to select for transgene-expressing cells . Successful integration was confirmed by analyzing eGFP fluorescence by flow cytometry . Limiting dilution was used to derive monoclonal OVA-expressing lines for each cell line . OVA presentation was confirmed by flow cytometry using a PE-labeled antibody against surface SIINFEKL bound to H-2Kb ( clone 25-D1 . 16 , BioLegend ) . Mouse femur and tibia were isolated from sacrificed mice . After sterilizing in ethanol and transferring to sterile PBS , muscle tissue was removed and tibia separated from femur . To isolate the bone marrow , bones were trimmed at both sides and flushed with 10 mL of sterile PBS to retrieve bone-marrow-derived cells . These were pelleted by 5 min centrifugation at 200 rcf and resuspended in 1 mL ACK lysis buffer for 2 min to lyse red blood cells . The reaction was stopped using 40 mL PBS and cells pelleted as before , and resuspended in BMDM media ( DMEM , 10% FBS , 1% PS , 10 ng GM-CSF , 10 ng M-CSF ) . After plating on 10 cm cell culture dishes , cells were cultured for 7 days; GM-CSF and M-CSF was replenished every 2 days . At day 7 , BMDM media was removed and replaced with RPMI ( with glutamine ) + 100 ng/mL LPS ( to activate BMDMs for antigen presentation by inducing MHC , CD80 , and CD86 expression ) , followed by 24 hr incubation . Next , BMDMs were lifted using 4 mL of Corning cell stripper ( Corning , Catalog #25–056 CI ) along with a cell lifter . After stopping the reaction in 8 mL RPMI media , cells were spun down , the pellet resuspended in pure RPMI , and counted . To this mixture , SIINFEKL ( an OVA fragment that can be presented by mouse MHC class I molecule H2kb ) was added to a final concentration of 100 ng/mL and cells incubated for 1 hr at 37 degrees Celsius , with shaking every 15 min to prevent attachment . two washes with PBS preceded resuspension in PBS at a concentration of 10106 cells/mL for injection . Spleens and lymph nodes were obtained from OT-1 transgenic mice ( Jackson laboratory #003831 ) and kept in PBS . Using a 40 µM strainer on a 50-mL tube , spleens and lymph nodes were filtered through the strainer using the flat end of a 1-mL syringe . This solution was spun at 200 rcf for 5 min to pellet cells . After discarding the supernatant , 450 uL of MACS buffer ( PBS , 2% FBS , 1 mM EDTA ) and 50 µL of CD8a Microbeads ( Miltenyi Biotec 130-117-044 ) were added per spleen and incubated at 4°C for 15 min . 20 mL of MACS buffer was then added to each tube and cells were pelleted as previously . The pellet was resuspended in 1–2 mL MACS buffer and transferred to a buffer-primed LS column in a magnetic holder with pre-separation filter . three column washes were carried out using 3 mL MACS buffer , and elution into a new tube ( containing 5 mL of MACS buffer ) using 5 mL of MACS buffer after removal from the magnetic holder , using the column plunger to force cells into the collection tube . 5 mL of MACS buffer was added to purified t-cells prior to counting . These cells were pelleted and resuspended in PBS at a concentration of 1106 cells/mL for injection . Tumor samples were dissected and minced with scissors , resuspended in 2 ml Digestion buffer composed of HBSS with 2 mg/ml Collagenase A ( 50-100-3278 , Fischer Scientific ) and 10 µg/ml DNase ( D5025 , Sigma-Aldrich ) , dissociated with a gentleMACS Dissociator ( Miltenyi Biotec , 130-093-235 ) and incubated 30 min at 37°C . 10 ml PBS with 10% FCS ( buffer A ) was added and the suspension was passed through a 100 µm cell strainer ( BD Biosciences ) . The resulting single-cell suspension was pelleted and resuspended in 1 ml ACK buffer ( 0 . 15 M NH4Cl , 10 mM KHC03 , 0 . 1 mM EDTA in distilled H2O ) and incubated on ice for 5 min . 10 ml of buffer A was added to stop the reaction . The suspension was passed through a 40-µm cell strainer ( BD Biosciences ) , pelleted and resuspended in 500 µL of cold PBS . Single-cell suspensions were subsequently subject to cell phenotyping by flow cytometry as described below . Spleens were harvested , mashed over a 40-µm cell strainer ( VWR , 10199–654 ) , and CD8+ T-cells purified by positive magnetic bead selection ( Miltenyi Biotec , 130-117-044 ) according to manufacturer’s instructions . Purified CD8+ T-cells were counted and cell diameter measured using a Moxi Z mini automated counter ( Orflo , MXZ001 ) . The cells were characterized using Flow cytometry described below or plated at 1 × 106 ( 24-well plate ) or 5 × 105 ( 48-well plate ) and activated with Dynabeads Mouse T-Activator CD3/CD28 ( Thermo Fisher , 11456D ) in a 1:1 beat to T-cell ratio and cultured in 2 ml ( 24WP ) or 1 ml ( 48WP ) RPMI 1640 ( Thermo Fisher , 21875 ) supplemented with 10% Fetal Bovine Serum , ( Thermo Fisher , 10270–106 ) , 50 µM 2-mercaptoethanol ( Thermo Fisher , 21985023 ) , 10 U/ml penicillin-streptomycin ( Thermo Fisher , 15140122 ) , and 10 U/ml recombinant human IL-2 ( Sigma , 11011456001 ) , and incubated at 37 °C for 3 days in a humidified CO2 incubator . The selected metabolites , citric acid ( Sigma-Aldrich , 251275–100G ) , malic acid ( Sigma-Aldrich , M1000-100G ) , succinic acid ( Sigma-Aldrich , S3674-100G ) , fumaric acid ( Sigma-Aldrich , 47910–100G ) , lactic acid ( Sigma-Aldrich , L1750-10G ) , pyruvic acid ( Sigma-Aldrich , 107360–25G ) , α-Ketoglutaric acid ( Sigma-Aldrich , K1750 ) , and oxaloacetic acid ( Sigma-Aldrich , O4126 ) were prepared in RPMI 1640 ( ThermoFisher , 21875 ) , supplemented with 50 ml Fetal Bovine Serum ( ThermoFisher , 10270106 ) , 5 . 5 ml Penicillin-Streptomycin ( ThermoFisher , 15140122 ) and 555 µl 2-ME ( ThermoFisher , 21985023 ) and pH adjusted to 7 . 2–7 . 4 , by adding sodium hydroxide ( wvr , pHenomenal ) . The metabolites were added to a 96-well plate round-bottom in a serial dilution prior to T-cell activation , as described below . Spleens were harvested from WT mice , mashed over a 40-µm cell strainer ( VWR , 10199–654 ) and CD8+ T cells were purified by magnetic bead selection ( Milentyie Biotec , 130-117-044 ) according to manufacturer’s instructions . Cells were counted ( BioRad , TC20 ) and resuspended in complete media . 1 × 105 cells placed in a 96-well plate with added metabolites were activated in a 1:1 bead to cell ratio with Dynabeads Mouse T-Activator CD3/CD28 ( ThermoFisher , 11456D ) and 10 U/ml recombinant human IL-2 ( Sigma , HIL2-RO ) . The treated cells were cultured in a humidified 5% CO2 incubator for 3 days . To assess cell proliferation of the activated CD8 T-cells in the presence of these metabolites , a resazurin reduction assay ( Sigma-Aldrich , R7017-1G ) was carried out on day 3 of the experiment . In short , 100 µl of the treated cells were transferred to a black 96-well plate , ( Corning , 3610 ) to which 20 µl of resazurin solution was added to a final concentration of 10 µg/ml and incubated for 3 hr . The relative fluorescence units ( RFU ) , proportional to the amount of metabolically active cells , was determined by measuring the fluorescence intensity of the media in the black 96-well plate at 530/25 nm excitation wavelength and 590/35 nm emission wavelength with a plate reader ( BioTek , Synergy HTX ) at three time points: 2 hr , 2 hr 30 min , and 3 hr of incubation . Cell proliferation curves of treated cells was determined by subtracting the RFU values from the background control ( media only ) and the change of RFU values over time was calculated and compared to the negative control ( media and cells ) . Single-cell suspensions were washed and stained with Fixable Near-IR Dead Cell Stain Kit ( Thermo Fisher , L10119 ) followed by staining of extracellular antigens with fluorochrome labeled antibodies . The Fixation/Permeabilization Solution Kit ( BD Biosciences , 554714 ) was used for exposing cytoplasmic antigens . Fluorochrome-labeled antibodies against mouse antigens was used in different combinations . CD44 ( clone IM7 ) , CD45 . 1 ( clone A20 ) , CD45 . 2 ( clone 104 ) , CD8 ( clone 53–6 . 7 ) , CTLA-4 ( clone UC10-4F10-11 ) , were purchased from BD Biosciences . CD62L ( clone MEL-14 ) , and ICOS ( clone C398 . 4A ) were purchased from BioLegend . Anti-mouse Granzyme B ( clone GB12 ) was purchased from Thermo Fisher . Cell counting was performed with CountBright Absolute Counting Beads ( Thermo Fisher , C36950 ) . Samples were processed in a FACSCanto II flow cytometer ( BD Biosciences ) . Data analysis was performed with FlowJo , version 8 . 8 . 7 ( Tree Star ) . Sodium L-Lactate ( Sigma , L7022 ) and Sodium Chloride ( Sigma , S5886 ) were prepared as 10x concentrated solutions in complete media . Compounds were added to purified OVA-specific OT-I CD8+ T-cells T-cells at the point of activation ( day 0 ) minutes before addition of CD3/CD28 Dynabeads . Sodium chloride or plain media was used as control . OVA-expressing GFP-positive EL4 ( EL4-GFP-OVA ) cells were mixed with their respective parent cell line ( EL4 ) in a 1:1 ratio in round-bottom 96-well plates . OVA-specific OT-I CD8+ T-cells were mixed with EL4 cells at effector to target ratios ranging from 20:1 to 1:50 . Cytotoxicity was assessed by flow cytometry 24 hr later . The ratio of GFP-positive events ( target ) to GFP-negative events in each test co-culture was divided by the ratio from cultures without addition of OT-I cells to calculate specific cytotoxicity . Mammary gland sections were deparaffinized with Tissue-Tek Tissue-Clear ( Sakura , Japan ) and rehydrated with graded ethanol . Antigen retrieval was performed either with Proteinase K ( for F4/80 ) or boiling the sections in a high pH antigen retrieval buffer ( Dako Target retrieval Solution , pH 9 , Dako , Denmark ) ( for Podocalyxin , CD3 , Granzyme B ) or low pH antigen retrieval buffer ( Dako Target retrieval Solution , pH 6 , Dako , Denmark ) ( for PCNA ) . Endogenous peroxidase activity was blocked using 3% H2O2 . Nonspecific protein interactions were blocked using 20% goat serum in PBS-T . Primary antibodies were used in mammary glands against CD3 ( #ab5690 , Abcam , UK ) to assess lymphocytic infiltration , F4/80 ( #MCA497 , Abd Serotec , Germany ) to assess macrophage density , podocalyxin ( #AF1556 , R and D Systems , Germany ) to assess capillary density , PCNA ( #M0879 , Dako , Denmark ) to identify proliferating cells , and Granzyme B ( #ab4059 , Abcam , UK ) to identify effector lymphocytes . Subsequently the sections were incubated with a suitable biotin-conjugated secondary IgG antibody . All sections were then incubated with an Avidin-Biotin peroxidase Complex ( ABC ) kit ( #PK6100 , Vector , USA ) and the peroxidase was stained with 3 , 3'-Diaminobenzidine ( DAB ) kit ( #SK4100 , Vector , USA ) . Nuclei were stained with hematoxylin and sections were mounted with Faramount Aqueous Mounting Medium , ( Dako , Denmark ) . Specific stainings were quantified using Image J software ( National Institute of Health , USA ) . Mammary glands were stained with H and E with a standard protocol and histologically scored for tumor stage in accordance with ( Lin et al . , 2003 ) , ranging from hyperplasia ( score 1 ) , adenoma/mammary intraepithelial neoplasia ( adenoma/MIN ) ( score 2 ) , early invasive carcinoma ( score 3 ) to late invasive carcinoma ( score 4 ) . The highest staging value of each specimen was used . For the metabolite analyses , blood and skeletal muscle ( quadriceps ) ( Metabolite analysis 1 ) or blood together with five different tissue types ( Metabolite analysis 2 ) were harvested from both exercising and control animals . Immediately after the exhaustion test , blood was collected from the mouse tail into a lithium heparin-covered microvette tube and kept at room temperature until further processing . Mice were subsequently euthanized by cervical dislocation prior to tissue collection . Biopsies of spleen , muscle , muscle-draining/non-draining lymph nodes and thymus were harvested and immediately frozen in liquid nitrogen . To assure that the right lymph nodes were collected , the muscle-draining lymph nodes had been identified in a previous experiment by injecting either the hind or the forelimb muscle of live animals with Patent Blue dye . The level of perfusion of local lymph nodes , under anesthesia , was then checked . The forelimb , muscle-draining lymph nodes showed to be the most easily accessible . Non-draining lymph nodes were harvested from the inguinal fat pad . Plasma was obtained by gradient centrifugation at 2000xg for 5 min at room temperature and stored at −80 . The initial metabolic profiling by GC-MS was performed at the Swedish Metabolomics Center in Umeå , Sweden . Information about reagents , solvents , standards , reference and tuning standards , and stable isotopes internal standards can be found as Supplementary information . Sample Preparation: Extraction was performed as previously described ( Jiye et al . , 2005 ) . For the plasma samples of extraction buffer ( 90/10 v/v methanol: water ) including internal standards were added to the plasma samples . The volume of plasma varied from 25 µL to 100 µL and the volume of extraction solution was adjusted accordingly keeping the volume ratio of plasma to extraction solution constant ( 10/90 v/v plasma: extraction solution ) . The samples were shaken at 30 Hz for 2 min in a mixer mill and proteins were precipitated for 2 hr at −20°C . The samples were centrifuged at +4°C , 14 , 000 rpm , for 10 min . 200 µL ( LC-MS ) and 50 µL ( GC-MS ) of the supernatant was transferred to microvials and solvents were evaporated to dryness . For muscle tissue samples ( all samples 20 mg + / - 2 mg with the individual weight of each sample noted ) , 1000 µl extraction buffer ( 90/10 v/v methanol: water ) including internal standards was added to the tissue samples . To all tissue samples two tungsten beads were added and samples were shaken at 30 Hz for 3 min in a mixer mill . The samples were centrifuged at +4°C , 14 , 000 rpm , for 10 min . 50 µL ( GC-MS ) of the supernatant was transferred to micro vials and solvents were evaporated to dryness . A small aliquot of the remaining supernatants were pooled and used to create quality control ( QC ) samples . The samples were analyzed in batches ( different sample types ) according to a randomized run order on GC-MS . Briefly , the second analysis was done on the Precision Metabolomics platform at Metabolon Inc ( Morrisville , NC , US ) samples were homogenized and subjected to methanol extraction then split into aliquots for analysis by ultrahigh performance liquid chromatography/mass spectrometry ( UHPLC/MS ) in the positive ( two methods ) and negative ( two methods ) mode . Metabolites were then identified by automated comparison of ion features to a reference library of chemical standards followed by visual inspection for quality control ( as previously described , Dehaven et al . , 2010 ) . For statistical analyses and data display , any missing values are assumed to be below the limits of detection; these values were imputed with the compound minimum ( minimum value imputation ) . Statistical tests were performed in ArrayStudio ( Omicsoft ) or ‘R’ to compare data between experimental groups; p<0 . 05 is considered significant and 0 . 05 < p 0 . 10 to be trending . An estimate of the false discovery rate ( Q-value ) is also calculated to take into account the multiple comparisons that normally occur in metabolomic-based studies , with q < 0 . 05 used as an indication of high confidence in a result . Mouse CD8+ T cells , purified from spleen of vaccinated animals , directly after an exhaustion test , was washed with PBS , and metabolic activity quenched by freezing samples in dry ice and ethanol , and stored at −80 °C . Metabolites were extracted by addition of 600 μl ice-cold 1:1 ( vol/vol ) methanol/water ( containing 1 nmol scyllo-Inositol as internal standard ) to the cell pellets , samples were transferred to a chilled microcentrifuge tube containing 300 μl chloroform and 600 μl methanol ( 1500 μl total , in 3:1:1 vol/vol methanol/water/chloroform ) . Samples were sonicated in a water bath for 8 min at 4°C , and centrifuged ( 13 , 000 rpm ) for 10 min at 4°C . The supernatant containing the extract was transferred to a new tube for evaporation in a speed-vacuum centrifuge , resuspended in 3:3:1 ( vol/vol/vol ) methanol/water/chloroform ( 350 μl total ) to phase separate polar metabolites ( upper aqueous phase ) from apolar metabolites ( lower organic phase ) , and centrifuged . The aqueous phase was transferred to a new tube for evaporation in a speed-vacuum centrifuge , washed with 60 μl methanol , dried again , and derivatized by methoximation ( 20 μl of 20 mg/ml methoxyamine in pyridine , RT overnight ) and trimethylsilylation ( 20 μl of N , O-bis ( trimetylsilyl ) trifluoroacetamide + 1% trimethylchlorosilane ) ( Sigma , 33148 ) for ≥1 hr . GC/MS analysis was performed using an Agilent 7890B-5977A system equipped with a 30 m + 10 m × 0 . 25 mm DB-5MS + DG column ( Agilent J and W ) connected to an MS operating in electron-impact ionization ( EI ) mode . One microliter was injected in splitless mode at 270°C , with a helium carrier gas . The GC oven temperature was held at 70°C for 2 min and subsequently increased to 295°C at 12 . 5°C/min , then to 320°C at 25°C/min ( held for 3 min ) . MassHunter Workstation ( B . 06 . 00 SP01 , Agilent Technologies ) was used for metabolite identification by comparison of retention times , mass spectra and responses of known amounts of authentic standards . Metabolite abundance and mass isotopologue distributions ( MID ) with correction of natural 13C abundance were determined by integrating the appropriate ion fragments using GAVIN . Statistical analyses were performed with Prism 7 version 7 . 0 ( GraphPad ) . Statistical tests and number of replicates are stated in figure legends . Principal component analysis ( PCA ) was carried out with normalized metabolite concentrations using the prcomp function in R version 3 . 6 . 1 . Fold changes and hierarchical clustering of metabolites was performed using the hclust function in R version 3 . 6 . 1 . All animal experiments were approved by the regional animal ethics Committee of Northern Stockholm , Sweden ( Dnr: N78/15 and N101/16 ) . In the human study , all subjects were informed about the study outline , gave written informed consent prior to inclusion and was familiarized with the experimental procedures . All procedures were performed in accordance with the Helsinki declaration and ethical standards of the institutional and national research committee Dnr: 2016/590–31 .	Exercise affects almost all tissues in the body , and scientists have found that being physically active can reduce the risk of several types of cancer as well as improving outcomes for cancer patients . However , it is still unknown how exercise exerts its protective effects . One of the hallmarks of cancer is the ability of cancer cells to evade detection by the immune system , which can in some cases stop the body from eliminating tumor cells . Rundqvist et al . used mice to investigate how exercise helps the immune system act against tumor progression . They found that when mice exercised , tumor growth was reduced , and this decrease in growth depended on the levels of a specific type of immune cell , the CD8+ T cell , circulating in the blood . Additionally , Rundqvist et al . found that CD8+ T cells were made more effective by molecules that muscles released into the blood during exercise . Isolating immune cells after intense exercise showed that these super-effective CD8+ T cells alter how they use molecules for energy production after exertion . Next , immune cells from mice that had exercised frequently were transferred into mice that had not exercised , where they were more effective against tumor cells than the immune cells from untrained mice . These results demonstrate that CD8+ T cells are altered by exercise to improve their effectiveness against tumors . The ability of T cells to identify and eliminate cancer cells is essential to avoid tumor growth , and is one of the foundations of current immune therapy treatments . Exercise could improve the outcome of these treatments by increasing the activation of the immune system , making tumor-fighting cells more effective .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation", "cancer", "biology" ]	2020	Cytotoxic T-cells mediate exercise-induced reductions in tumor growth
Humans and other animals base important decisions on estimates of number , and intraparietal cortex is thought to provide a crucial substrate of this ability . However , it remains debated whether an independent neuronal processing mechanism underlies this ‘number sense’ , or whether number is instead judged indirectly on the basis of other quantitative features . We performed high-resolution 7 Tesla fMRI while adult human volunteers attended either to the numerosity or an orthogonal dimension ( average item size ) of visual dot arrays . Along the dorsal visual stream , numerosity explained a significant amount of variance in activation patterns , above and beyond non-numerical dimensions . Its representation was selectively amplified and progressively enhanced across the hierarchy when task relevant . Our results reveal a sensory extraction mechanism yielding information on numerosity separable from other dimensions already at early visual stages and suggest that later regions along the dorsal stream are most important for explicit manipulation of numerical quantity . One largely debated theme in cognitive neuroscience is how the human brain developed the ability to perform mathematics . While mathematical skills certainly rely on the interplay of a wide range of cognitive functions ( De Smedt et al . , 2013; Fias , 2016; Iuculano and Menon , 2018 ) , an influential theory in the field proposes that a necessary prerequisite to develop such a sophisticated uniquely human ability resides in the ‘number sense’ ( Dehaene , 1997 ) . This is a phylogenetically ancient competence that enables humans and other animals to assess and mentally manipulate the approximate number of objects in sets . In humans the precision of the number sense ( or ‘numerical acuity’ , typically measured by visual number discrimination ) sharpens with age and with the acquisition of formal mathematical education ( Piazza et al . , 2013 ) , and correlates with arithmetical skills throughout the life-span ( Halberda et al . , 2008; Libertus et al . , 2011; Libertus et al . , 2013; Chen and Li , 2014; Anobile et al . , 2016a; Anobile et al . , 2018 ) . Deviations from the typical developmental trend of numerical acuity can be a symptom of developmental dyscalculia ( Piazza et al . , 2010 ) , a neurodevelopmental disorder that causes specific mathematical learning difficulties . The neural substrate subtending this sense of numerical quantity is thought to be shared across species and has been linked to a network of areas in the frontal and parietal cortices sensitive to changes in numerosity since very early in life ( Izard et al . , 2008; Hyde and Spelke , 2011; see for reviews: Cantlon , 2012; de Hevia et al . , 2017 ) . In these areas electrophysiological recordings in monkeys identified single neurons tuned to specific numerosities of visual arrays ( Nieder et al . , 2002; Nieder and Miller , 2004; Roitman et al . , 2007; Nieder , 2016 ) and fMRI studies in humans found activation in these areas to be modulated during quantity perception as well as during calculation ( for reviews see: Arsalidou and Taylor , 2011; Eger , 2016; Piazza and Eger , 2016 ) . While the first imaging studies in humans were limited by the low spatial resolution and univariate subtraction-based analyses , fMRI adaptation and multivariate pattern analysis methods provide higher sensitivity to finer-scale activity differences ( Kourtzi and Grill-Spector , 2005; Norman et al . , 2006; Tong and Pratte , 2012 ) . These methods allowed researchers to study the representation of individual numbers by recording the distance-dependent signal release from adaptation ( Piazza et al . , 2004 ) , or reading out patterns of number-related activity across multiple voxels of the frontal and parietal cortex ( Eger et al . , 2009 ) . Moreover , population-receptive field mapping ( pRF ) methods identified individual locations tuned to specific numerosities arranged in spatially organized maps in the parietal cortex ( Harvey et al . , 2013 ) . While these earlier findings mostly pointed at the key role of parietal and frontal areas in numerical representation , some recent studies found that it is possible to decode the number of items seen by the subjects from the fMRI activity patterns in early visual areas ( Bulthé et al . , 2014; Eger et al . , 2015; Bulthé et al . , 2015; DeWind et al . , 2019 , but see Castaldi et al . , 2016 ) . Moreover , spatially organized numerosity maps were recently claimed to extend to the occipital cortex ( Harvey and Dumoulin , 2017a ) and early ERP components compatible with generators in early visual areas responded to variations in the numerosity of visual arrays ( Park et al . , 2015; Fornaciai et al . , 2017; Fornaciai and Park , 2017 ) . Several properties characterizing numerosity perception , such as being ratio-dependent ( Weber’s law ) and being susceptible to adaptation , led some authors to suggest that number is a ‘primary’ visual property of the image that is directly perceived through specialized and dedicated mechanisms ( Burr and Ross , 2008; Ross , 2010; Anobile et al . , 2016b ) . However , in spite of dedicated efforts on modeling the extraction of numerosity from the visual image ( Dehaene and Changeux , 1993; Verguts and Fias , 2004; Dakin et al . , 2011; Stoianov and Zorzi , 2012; Morgan et al . , 2014 ) , the detailed neural processing mechanisms used by the brain to arrive at a representation of numerosity from the visual input remain little understood , and much less understood than the ones for other basic visual features such as orientation , colour , motion , etc . Numerosity is a notoriously difficult feature to study since changes in numerosity tend to be associated with changes in other quantitative features of the sets during natural viewing conditions ( e . g . , more items tend to occupy a larger area , or be spaced more densely ) , and it appears impossible to control for all of these associated quantities at the same time . For this reason , in spite of a large body of behavioral and neuroscientific work on this topic , it still remains debated whether the available evidence supports a sensory extraction mechanism directly sensitive to numerosity . Some have argued instead that numerosity might be judged indirectly by weighing a combination of other , non-numerical , quantitative features of the stimuli ( Gebuis and Reynvoet , 2012; Gebuis et al . , 2014; Leibovich et al . , 2016a ) . For example , numerosity can be mathematically defined as the product of density ( number of items per unit of area ) by field area; or by the total surface area divided by mean item size . Thus , decisions on numerical quantity could be taken merely indirectly , on the basis of representations of these non-numerical properties , without numerosity being encoded directly by perceptual systems . While this possibility is interesting , several behavioral findings argue against it: ( 1 ) the discrimination of numerosity and of one often correlated non-numerical feature ( item density ) follow different psychophysical laws ( Anobile et al . , 2016b ) , and ( 2 ) at least for relatively small numbers of not too densely spaced items , perceptual thresholds for numerosity discrimination are typically much smaller than the ones predicted from the thresholds for density and field area together ( Cicchini et al . , 2016 ) , making it unlikely that estimates of numerosity are based on the latter . For what concerns the neuronal level , a few recent studies have started to directly quantify the effects of non-numerical dimensions of non-symbolic numerical stimuli ( e . g . Park et al . , 2015; Fornaciai et al . , 2017; Harvey and Dumoulin , 2017b; Fornaciai and Park , 2018; DeWind et al . , 2019 ) . Those studies found that activity in earlier ( occipital ) or later ( parietal ) brain regions appeared to be linked to the numerical content of sets after taking into account effects of certain non-numerical dimensions . However , they mostly only considered the effect of one non-numerical variable at the time and compare it to that of number , without taking into account effects explained by all relevant non-numerical dimensions together . Thus , it still remains unclear to what extent activity evoked by non-symbolic numerical stimuli within early and later regions can be explained by a mechanism that encodes numerosity in itself , or by the ensemble of responses to the different non-numerical dimensions of the stimuli . Here , we implement a new approach to separate brain signals related to numerical and non-numerical quantities and test for a neuronal mechanism directly sensitive to the numerosity of visual sets along the dorsal visual stream hierarchy . We propose that the following signatures would advocate for the existence of such a mechanism: First , information on numerosity should be detectable in regional activity patterns after multiple important non-numerical quantities are simultaneously ( and not only individually ) taken into account . Second , and importantly , it should be possible to selectively amplify this numerical information depending on whether the numerical dimension of the stimuli is task relevant , similar to the attentional amplification that has been previously shown for other task-relevant primary features , such as orientation , contrast , color , direction etc . ( Jehee et al . , 2011; Ester et al . , 2016 ) . If a brain area encodes numerical information in a way that is separable from associated non-numerical dimensions , tasks involving selective attention to number should enhance the information about numerosity , without affecting the level of information on associated non-numerical dimensions . Thus , we propose that the presence of such independent attentional amplification is a key criterion in order to identify which brain areas explicitly encode information on numerosity . On the contrary , if activity patterns could be entirely accounted for by the combination of responses to multiple non-numerical dimensions of the stimuli , no information specifically related to number should be found in the patterns of activity once accounting for the other ( non-numerical ) dimensions simultaneously . Furthermore , if numerosity was not directly encoded but only indirectly inferred from percepts of non-numerical properties , attentional enhancement should not occur for signals related to numerosity , but if anything , only for other properties ( e . g . , density and field area ) that can jointly define it . To test these predictions , we created a novel stimulus space to disentangle the contribution of numerical and non-numerical dimensions to brain activity patterns , and designed a task where attention is selectively directed towards either of two orthogonal quantitative dimensions of the visual array ( number or item size ) . We exploited the enhanced sensitivity achieved by fMRI at ultra-high field ( 7 Tesla ) and specific multivariate pattern analyses to simultaneously model and separate the contributions of numerical and different non-numerical quantities to fine-scale activity patterns within multiple regions defined by a probabilistic atlas based on visual topography . Response accuracies for comparison of match stimuli were overall high and not significantly different across tasks ( 86% for the number task and 85% for the average size task , t ( 19 ) = 0 . 46 , p = 0 . 65 ) , suggesting that subjects attended to the correct stimulus dimension and the difficulty was on average successfully matched across tasks ( Figure 2A ) . We started the analysis of the functional imaging data by evaluating overall regional activation effects during both tasks . Surface-based random-effects group analysis identified similar bilateral activations in the occipito-parietal and frontal cortex during both tasks for sample stimuli against the implicit baseline where participants were just looking at the fixation point with blank screen and without performing any task ( Figure 2B and C , thresholded at p < 0 . 001 uncorrected ) . In both tasks the activity covered a wide occipito-parietal area starting from the superior occipital and transverse occipital sulci and extending throughout the intraparietal sulcus ( IPS ) up to the post-central sulcus . The frontal activity mainly covered the superior frontal gyrus . The direct contrast of sample stimulus-related activity during the number versus the size task revealed no area with significantly stronger activation for either of the two , despite the uncorrected significance threshold ( Figure 2D ) . Altogether , these results suggest that task difficulty was successfully matched and that under these conditions attending to different quantitative dimensions leads to equivalent overall activation of the brain regions involved in the task . Differences in overall activation level can therefore not confound the following more specific results on the within-dimension discriminability of quantitative features . As a critical test of whether the representations of numerical and non-numerical features of the stimuli could be dissociated across the dorsal visual stream , we performed Representational Similarity Analysis ( Kriegeskorte , 2008; Kriegeskorte and Kievit , 2013 ) which , unlike classification-based decoding , allows to assess the effect of multiple quantitative dimensions on activity patterns simultaneously . For each ROI and task , we obtained a neural representational dissimilarity matrix ( neural RDM , Figure 4A ) by computing the correlation distance between activation patterns for each possible pair of conditions . We then applied multiple regression analysis to test in how far the fMRI pattern dissimilarity structure could be explained by multiple predictor matrices reflecting the stimuli’s dissimilarity along several important quantitative dimensions: numerosity , average item size , total field area , total surface area and density ( Figure 4B ) . Of note , our design orthogonally manipulating numerosity , average item size and total field area ensured that numerosity was also partly decorrelated from density and total surface area ( as shown by the correlation values in the Predictor Correlation matrix , Figure 4B ) , yet not completely ( correlation between number and density predictors = 0 . 43; between number and total surface area predictors = 0 . 33 ) . Correlations between predictors in a multiple regression lead to a reduction of the unique variance attributable to each one of them , and to a greater variability of the estimated betas . An estimation of variance inflation factors ( VIF ) for each predictor in our case revealed that these remained reasonably low ( corresponding to 1 . 4874 , 1 . 1957 , 1 . 2048 , 1 . 3238 and 1 . 4591 for number , average item size , total field area , total surface area and density , respectively ) . By using a multiple regression approach we capitalize on the fact that the resulting beta weights reflect only the part of the variance that each one of these stimulus descriptors uniquely explained in the pattern of activity of a given ROI on top of all the others . Thus , by entering numerical and non-numerical dimensions together into a multiple regression , a significantly above zero beta for number would imply that the numerical information is contributing to the pattern of activity within a given ROI , over and above the contribution of the other non-numerical quantitative dimensions . Figure 5 displays the results of the estimated beta weights for various ROIs separately for the number ( Figure 5A ) and size tasks ( Figure 5B ) . Beta weights for the effect of number independent of the other dimensions ( black triangles ) were generally positive and progressively explained the activity patterns better when proceeding from lower to higher-level regions when task relevant . The evolution of the numerical information across the visual stream was attenuated during the size task , yet betas remained significantly above zero in all regions ( see p-values in Supplementary file 2 ) . Beta weights for the non-numerical dimensions ( other shapes in Figure 5 ) were pronounced predominantly in the earlier visual areas and , importantly , they appeared to be not clearly affected by task . To statistically test for differential modulation of the contribution of the different quantitative dimensions to activation patterns , beta weights were analyzed with repeated measure ANOVAs with ROI , task and dimension as factors . As for the classification analysis , we first focused on the three large regions corresponding to early , intermediate and higher-level areas and then further on individual ROIs from V1 up to IPS345 ( for results concerning the intraparietal sulcus excluding those regions defined by the atlas based on visual topography see Figure 3—figure supplement 2B and Supplementary file 4c ) . The significant triple interaction between ROI , task and dimension confirmed that the beta weights estimated for the different dimensions were differently affected by task across ROIs ( for the three large regions: F ( 4 . 23 , 80 . 40 ) = 3 . 32 , p = 0 . 01; for the individual regions: F ( 6 . 18 , 117 . 38 ) = 3 . 06 , p = 0 . 007 ) . To identify which dimension was maximally driving this effect , we quantified the changes in beta weights across ROIs and tasks for each dimension separately . Beta values for number were the only ones showing a significant interaction between ROI and task , when comparing the three large subdivisions across the visual stream ( F ( 1 . 35 , 25 . 62 ) = 5 . 97 , p = 0 . 015 ) . During the number task , betas for number were higher in intermediate and higher-level areas with respect to early visual areas ( although only the former comparison was significant , p = 0 . 04 , Cohen’s d = 0 . 70 ) . During the size task the betas for number were significantly lower ( significant difference across tasks in early: p = 0 . 007 , Cohen’s d = 0 . 78; intermediate: p = 0 . 000001 , Cohen’s d = 1 . 96; higher areas: p = 0 . 00001 , Cohen’s d = 1 . 43 ) and not different across regions . When focusing on the seven individual ROIs , the interaction between ROI and task was significant ( F ( 2 . 04 , 38 . 83 ) = 5 . 29 , p = 0 . 009 ) . Although post-hoc tests did not identify significant differences across ROIs , linear regression showed that the increase in beta weights for number across the dorsal visual stream was significant during the number task only ( F ( 1 , 5 ) = 14 . 23 , p = 0 . 01 , R2 = 0 . 74 ) , while during the size task betas for number were much more homogenous across ROIs ( F ( 1 , 5 ) =2 . 37 , p = 0 . 18 , R2 = 0 . 32 ) . Indeed the difference in beta weights between the number and size task was only minor or not significant in V1 and V2 , more pronounced in V3 , and highly significant from V3AB on ( difference across tasks: V1: p = 0 . 025 , Cohen’s d = 0 . 64; V2: p = 0 . 13 , Cohen’s d = 0 . 37; V3: p = 0 . 001 , Cohen’s d = 0 . 91; V3AB p = 0 . 000001 , Cohen’s d = 1 . 44; V7: p = 0 . 000008 , Cohen’s d = 1 . 54; IPS12: p = 0 . 000001 , Cohen’s d = 1 . 56; IPS345: p = 0 . 000112 , Cohen’s d = 1 . 28 ) . Different from number , beta weights estimated for the non-numerical dimensions were not modulated by task ( no significant interaction between ROIs and task , no significant main effect of task ) for any of the dimensions . Independent of the task , total field area best explained activity patterns in early visual areas , while its contribution was reduced when proceeding through intermediate to higher-level areas ( significant main effect of ROIs: F ( 1 . 23 , 23 . 29 ) =35 . 24 , p = 0 . 000002; significant differences in beta weights between primary and intermediate or higher-level ROIs: p = 0 . 000155 , Cohen’s d = 1 . 24 , p = 0 . 000008 , Cohen’s d = 1 . 80 , respectively ) . Beta values were highly significantly modulated also across the different individual ROIs ( main effect of ROIs: F ( 2 . 11 , 40 . 12 ) = 32 . 27 , p < 10−5 ) . Indeed , activity patterns in V1 , V2 and V3 were explained equally well by total field area and better than intermediate and higher regions , starting from V3AB on ( all p < 0 . 01 at least ) . Total surface area also most strongly modulated pattern dissimilarity in early visual areas . The significant main effect of ROI ( F ( 1 . 45 , 27 . 63 ) = 16 . 61 , p = 0 . 000078 ) and the following post-hoc tests showed that beta values for this dimension in the early visual areas were significantly higher than those estimated for the intermediate ( p = 0 . 000475 , Cohen’s d = 0 . 76 ) and higher-level ( p = 0 . 000943 , Cohen’s d = 1 . 30 ) ROIs , independent of the task . Beta weights for total surface area were comparable in V1 , V2 and V3 ( no significant difference across these ROIs ) and significantly higher than those of the others ROIs starting from V3AB/V7 on ( significant main effect of ROI: F ( 3 . 13 , 59 . 42 ) = 13 . 27 , p = 0 . 000001 , comparisons across regions: all p < 0 . 01 at least ) . Density modulated early visual areas during the number task and both earlier and higher-level areas during the size task . The main effect of ROI was significant ( F ( 1 . 41 , 26 . 72 ) = 4 . 05 , p = 0 . 04 ) , but additional post-hoc tests did not reveal any significant difference across the three large ROIs . Also at the level of individual regions the main effect of ROI was significant ( F ( 2 . 55 , 48 . 55 ) = 4 . 15 , p = 0 . 01 ) and the strongest difference across ROIs emerged when comparing the lowest beta weights estimated in V3AB with those obtained in V1 ( p = 0 . 003 , Cohen’s d = 1 . 13 ) and V7 , IPS12 and IPS345 ( p = 0 . 03 , Cohen’s d = 0 . 20 , p = 0 . 01 , Cohen’s d = 0 . 53 , p = 0 . 002 , Cohen’s d = 0 . 64 ) . Surprisingly , effects due to average item size could not be detected in any of the ROIs tested . In sum , while early visual areas contained independent information on multiple quantitative properties of which some explained more variance than numerosity , all regions were modulated to some extent by numerical distance over and above what was explainable by the non-numerical dimensions . Moreover , importantly , explicitly directing attention to number did enhance the representation of numerical information and did so selectively , without altering the representations of non-numerical quantities . Finally , although present starting from the earliest stages of visual analysis , the numerical information at this level was only to a minor extent modulated by task and the greatest contribution to explicit manipulation of numerical quantity was found in intermediate and higher-level regions . Overall , the model described in the previous sections , which takes into account multiple numerical and non-numerical dimensions simultaneously , seems to us the most appropriate way to address our main question , which concerns the contribution of numerosity to activity patterns above and beyond what can be explained by the non-numerical dimensions . However , since some correlations were present between the predictors for the different quantitative dimensions in our original model , we further explored in how far results might differ when modeling either only the orthogonal predictors ( i . e . number , average item size and total field area ) or only the non-numerical dimensions , some of which were correlated with number ( i . e . average item size , total field area , total surface area and density ) . Overall the results obtained with only the orthogonal predictors ( i . e . number , average item size and total field area; see Figure 5—figure supplement 1 ) are rather similar to the ones obtained with the full model . Beta weights for number were positive , all along the visual hierarchy , and were attenuated during the size task , yet remaining significantly above zero in all regions ( see p-values in Supplementary file 5 ) . Beta weights for average item size were almost never significant , while beta estimates for total field area were pronounced predominantly in the earlier visual areas . The most important difference with respect to the full model appears to be a somewhat higher contribution of the number predictor , especially pronounced in early visual regions . Results obtained from the model including only non-numerical dimensions were also similar to the full model ( see Figure 5—figure supplement 2 and p-values in Supplementary file 6 ) . Beta weights for average item size were around zero , those for total field area and total surface area were highest in early visual areas and then decreased with the visual hierarchy . Beta weights for density were positive both in early and in higher-level visual areas . The contributions of both total surface area and density show a tendency to be higher here compared to the full model , especially during the number task . However , in both of these analyses it remains ambiguous whether the effects observed for one given predictor are truly driven by that predictor , or potentially attributable to the unmodeled contribution of another one ( in the first case , the relatively enhanced effect of number in early visual regions might actually be due to density and TSA which are not included in the model , whereas in the second case , effects attributed to density and TSA could actually be due to the unmodeled contribution of number . Only the most complete model including both numerical and non-numerical dimensions allows us to dissolve this ambiguity . Our work exploited the enhanced spatial resolution provided by ultra-high field fMRI to reveal how the human brain represents multiple quantitative dimensions of non-symbolic numerical stimuli . Furthermore , we tested whether and at what cortical level the numerical information can be represented and specifically modulated by attention independently of non-numerical visual properties of the image . At the level of overall regional activity , attending to the numerosity or to the average size of the dots in the array recruited largely overlapping occipital and parietal areas , as also previously observed for perception and comparison of different types of quantities ( Pinel et al . , 2004; Dormal and Pesenti , 2009; Borghesani et al . , 2019 ) ; for a meta-analysis on other non-numerical representations see: Sokolowski et al . , 2017 ) . Two previous fMRI studies have investigated task-related differences in univariate regional activity elicited by numbers of items either in the subitizing ( Leibovich et al . , 2015 ) or in the estimation ( Leibovich et al . , 2016b ) range when participants attended to either the numerosity or to the total surface area of dot arrays while these dimensions varied congruently or incongruently with each other . During comparison of dot arrays within the subitizing range , the overall regional brain activity in the cingulate and right superior frontal gyrus was shown to be modulated by task order ( Leibovich et al . , 2015 ) . When testing larger numerosities , the activation in the right temporal parietal junction was modulated by the congruency between dimensions only when number , but not when total surface area , was the relevant dimension in a comparison task and several regions , including cingulate gyrus , anterior insula , superior frontal gyrus , orbitofrontal and inferior parietal cortex did either respond preferentially during the numerical or the non-numerical task ( Leibovich et al . , 2016b ) . While the fact that in the current experiment we do not observe significant differences between tasks at the univariate level might seem to contrast with the results by Leibovich et al . ( 2015 ) and Leibovich et al . ( 2016b ) , several methodological differences are worth noting: first , the non-numerical dimensions investigated are different , being average item size in our case and total surface area in theirs . Second , in the current experiment we made the changes in number and in average item size equally discriminable and we successfully matched task difficulty as demonstrated by the equal percentage of correct responses across tasks . The numerical and non-numerical dimensions were not matched for perceptual discriminability in the studies by Leibovich et al . ( 2015 ) and Leibovich et al . ( 2016b ) , leaving open the possibility that some of the differences in cortical activation may reflect the higher difficulty associated with the numerical as opposed to the non-numerical task . This possibility seems likely also because the experiments by Leibovich et al . ( 2015 ) and Leibovich et al . ( 2016b ) required participants to perform a comparison on every trial , thus the measured response not only reflects the perceptual process , but also the decision/comparison component . In the current experiment instead , our contrast is related to the sample trials , where participants were only required to perceive/memorize the value on the relevant stimulus dimension , without performing a comparison ( which was required only in the occasional match trials ) . Overall , having balanced task difficulty and excluded the contribution of the comparison process from the response , the current study found no differences between tasks at the univariate level . Only multivariate pattern analysis could detect differences in the way information along the different dimensions was encoded as a function of task in our study . Multivariate decoding analyses showed that the sample numerosity presented could be read out from brain activity significantly above chance all along the visual stream , however with important differences across regions . When explicitly attended , the numerical information could be read out with gradually higher accuracy following an occipital-parietal gradient , up to a maximum level in the parietal cortices . The effect of attention strongly affected the accuracy of the numerical discrimination in intermediate and higher regions while leaving the accuracy in the early visual areas unaffected . The successful read-out of information related to numerosity from parietal cortices in the current experiment contrasts with some previous studies where fMRI signals discriminative of numerosity could not be detected in the parietal regions ( DeWind et al . , 2019; Fornaciai and Park , 2018 ) . Differences in paradigms and sensitivity of the scanners used may account for this discrepancy . Most crucially , in those studies , participants were shown different numerosities and the task required detecting changes in the colour of the dots . Thus , participants’ attention may have not been directed to the numerosity of the visual arrays in that case , and the numerical information may have been reduced when focussing on the dots’ colour , similarly to what was observed for the size task in the current experiment . Although in the present study we could still read out numerical information even when it was irrelevant for the task , this signal may have remained undetected by less sensitive MRI scanners . Above chance decoding of numerosity during both tasks was observed here within both the medial IPS parts organized according to visual topography , as well as more lateral/inferior IPS parts outside the visual field maps , to a similar extent . Both regions also allowed to successfully decode the task carried out by the subject , with slightly higher accuracy in the most lateral/inferior IPS part with respect to the most medial IPS part . It has been recently proposed that there might be a sub-regional specialization along the intra-parietal sulcus , with the most medial/superior regions supporting processing of physical quantities and the lateral/inferior part supporting numerical operations , such as numerical comparison and calculation ( Harvey et al . , 2017; see also Castaldi et al . , 2019 ) . Our multivariate decoding results only marginally allow us to support distinct roles for these regions , however , addressing this precise question was not an explicit aim of the present study . It remains possible that the representation of numerical information in these different subparts supports different cognitive processes which cannot be differentiated here . For example , while the pattern of activity read out from the IPS field maps might underly the perception of numerosity , the numerical information reflected by the pattern of activity of the more lateral/task-responsive regions of IPS might provide input to internal manipulations of quantity during numerical operations , but further studies are necessary to explore this possibility . In our study the number presented could be decoded not only in parietal cortex , but already from the earliest stages of visual processing . However , since the multivariate classification analysis collapsed across the non-numerical dimensions of our stimulus set it is unclear whether the information underlying successful decoding was strictly numerical , especially in earlier regions . Some previous studies have dealt with the problem of correlations between numerical and non-numerical stimulus dimensions by controlling for non-numerical features one at the time and testing for fMRI adaptation effects , or replicability of decoding performance or layouts across conditions where individual non-numerical features where controlled for ( Piazza et al . , 2004; Eger et al . , 2009; Harvey et al . , 2013; Harvey and Dumoulin , 2017a; DeWind et al . , 2019 ) . When the effects of non-numerical dimensions were measured directly , this was done in some studies by computing the explained variance or classification performance for each feature in isolation and comparing it to the one for number ( Harvey and Dumoulin , 2017b; Cavdaroglu and Knops , 2019 ) , leaving unspecified the degree to which the simultaneous contribution of several non-numerical dimensions could account for the findings ( Gebuis et al . , 2014 ) . Some other previous studies have taken a different approach , by modeling jointly the effects of numerosity and two non-numerical dimensions ( termed ‘size in area’ and ‘spacing’ ) which were designed to be orthogonal to numerosity but do not necessarily constitute natural , perceptually relevant feature dimensions , but rather mathematically defined constructs ( DeWind et al . , 2015; Park et al . , 2015; DeWind et al . , 2019; Fornaciai and Park , 2018 ) . This design also allowed the authors to estimate , from the combined beta weights of numerosity and the mentioned two orthogonal dimensions , which feature represented by different directions in their stimulus space most accounted for the effects in a given ERP component or brain area . However , brain signals can reflect a combination of responses to multiple quantitative dimensions , and this approach does not permit to distinguish , for example , a modulation by numerosity from two independent modulations by field area and density . In our study , on the contrary , we separated the contributions of numerical and non-numerical stimulus dimensions by applying multiple regression to representational distance matrices which allowed us to test for the extent to which numerosity could explain the pattern of activity while taking into account simultaneously the variability explained by several important natural non-numerical features . Indeed , estimating significantly above zero beta values for number implies that information about numerosity is present in the pattern of activity over and above the contributions of all the other included non-numerical features . We found that information specific to number was detectable beyond the information related to the other dimensions , and that the numerical information was gradually enhanced when progressing along the visual stream when explicitly task relevant , and less strongly represented , although still detectable , when not task-relevant . Importantly , the level of information on other quantitative but non-numerical properties of the image , such as total field area , total surface area and density , although reliably detected especially in earlier brain regions , was not altered when explicitly attending to the numerical quantity . Given that numerical information becomes available in parallel with the other features and independently selectable by attention from early processing stages on , it appears that the human visual system has the capacity to detect signals related to numerosity and separate these from other quantitative dimensions , starting from very basic visual primitives . This supports the existence of a sensory processing mechanisms from which numerosity could be derived directly , rather than making numerical judgements merely indirectly on the basis of percepts of associated non-numerical quantities . Of course , the question of why numerical judgements are nevertheless often influenced by non-numerical quantities , and how these interactions might arise from the involved neuronal populations , remains an important one that deserves further study . Here we modeled the influence of several important natural non-numerical quantitative dimensions on brain activity , though our selection is necessarily non-exhaustive . We acknowledge the fact that we cannot formally rule out that a feature of a different type than those considered by us may have contributed to the effects observed here , and our conclusions hold to the extent of the non-numerical features tested . The enhancement of numerical information in activation patterns found here when number was the relevant stimulus dimension is extending a growing body of work on the neuronal correlates of feature-based attention . Neurophysiological studies have shown that attention to basic visual features either increases the gain or sharpens responses of neuronal populations preferentially responsive to these features in different visual areas ( e . g . Treue and Martínez Trujillo , 1999; McAdams and Maunsell , 2000; Reynolds et al . , 2000; Martinez-Trujillo and Treue , 2004; David et al . , 2008 ) , see also: Carrasco ( 2011 ) for a review ) . Correspondingly , fMRI decoding studies have found that directing attention to one feature dimension such as orientation , motion direction or color or to particular values within one given dimension improves the read-out of these features from brain activity in early sensory regions ( Kamitani and Tong , 2005; Kamitani and Tong , 2006; Serences and Boynton , 2007; Jehee et al . , 2011 ) but in some cases also in higher-level areas ( Liu et al . , 2011; Ester et al . , 2016 ) . According to one influential account , higher-level fronto-parietal areas such as the lateral intraparietal area ( LIP ) implement spatial ‘priority maps’ in which the level of activity at individual locations depends jointly on the different features of objects at these locations as well as on top-down factors such as their task relevance , associated reward , etc ( Itti and Koch , 2001; Thompson and Bichot , 2005; Gottlieb , 2007; Sapountzis et al . , 2018 ) . Independent of spatial priority , LIP neurons have also been found to represent higher-level factors such as learned category membership and other non-spatial information ( Freedman and Assad , 2009 ) and to flexibly switch between encoding of different visual features , such as color or motion , depending on the task ( Toth and Assad , 2002; Ibos and Freedman , 2014 ) . The idea of a role for intraparietal areas as mere ‘priority maps’ or reflecting entirely flexible encoding of information on task-relevant features ( without intrinsic selectivity ) can insufficiently account for our results , since it would predict an equivalent amplification of the representation of average size when this is the attended feature instead of number . This is not what we observed . Our results are thus more compatible with an enhancement of the responses of neuronal populations with intrinsic selectivity to the feature numerosity in these areas ( comparable to the one observed for other features in lower-level visual regions ) . While the existence of individual neurons tuned to different numbers of items in intraparietal cortex is well established ( Nieder and Miller , 2004; Roitman et al . , 2007 ) , the only electrophysiological study that recorded from neurons in the ventral intraparietal ( VIP ) cortex in macaque monkeys under changing task conditions ( Viswanathan and Nieder , 2015 ) found that neurons encoded numerosity to the same extent , regardless of whether the task required to attend to the number or the color of the items . This differs from our results which show a clear attentional amplification of numerosity information . Given that the human IPS 1–5 investigated in the current work is usually considered to be the equivalent of the macaque LIP/VIP complex ( Kastner et al . , 2017 ) , the difference between results may be due to a difference across species , but differences in paradigms and in the nature of the signal recorded in the two studies make it difficult to directly relate the two findings . For example , monkeys were trained initially with the color match to sample task , then re-trained to respond to number , thus implying comparisons across an extended time period and different context , whereas our participants switched between the two tasks within the same scanning session . In addition , it is possible that the color task with a single color per stimulus and a small number of highly distinguishable alternatives placed lower demands on attentional load compared to our average size task , therefore leaving number processing unaltered . Nevertheless , as a common denominator both studies agree on pointing to some degree of spontaneous encoding of numerosity in intraparietal areas under conditions of attention to an orthogonal stimulus dimension . The gradual enhancement of numerosity information observed by us in the number task when progressing along the dorsal visual stream is compatible with a multi-stage process of the extraction of numerosity where attention may operate at multiple levels over which attentional enhancements accumulate . If numerosity information can be retrieved from multiple levels of the cortical hierarchy , this does not need to imply that this feature is encoded by individual neurons at all these levels , but it may be detectable by multivariate methods even if it existed only in distributed form across the population of neurons . As one speculative interpretation , the numerical information read out from early visual areas could reflect a location map ( Dehaene and Changeux , 1993 ) , or the process of object segmentation where different individual items start to be separately represented , but this representation may not yet be in a form that is most easily read out for numerical discrimination . Higher areas may progressively transform and concentrate the initially distributed information onto individual neurons , which most likely constitute the base on which we operate when comparing numbers . This interpretation is in line with a recent study showing that although different numerosities could be discriminated based on the pattern of activity in early visual areas and parietal cortex , the behavioral precision of numerical discrimination was correlated with the decoding accuracy only in the latter region ( Lasne et al . , 2019 ) . While earlier behavioral research suggested that the precision of numerical representation is predictive of formal arithmetic and gets refined with development and mathematical learning ( Halberda and Feigenson , 2008; Nys et al . , 2013; Piazza , 2010; Piazza et al . , 2010; Piazza et al . , 2013 ) , other recent evidence has led to a slightly different view: what might be changing with development and mathematical competence could be the ability to focus on numerical information while filtering out non-numerical dimensions during a numerical comparison task ( Castaldi et al . , 2018; Piazza et al . , 2018; Starr et al . , 2017; Wilkey et al . , 2018 ) . The influence of non-numerical dimensions on numerical judgments decreases over normotypical development and both dyscalculic children ( Bugden and Ansari , 2016; Piazza et al . , 2018; Szucs et al . , 2013; Wilkey et al . , 2018 ) and adults ( Castaldi et al . , 2018 ) seem to be disproportionately affected by non-numerical dimensions during numerical comparison tasks . In light of these behavioral findings , it would be interesting to see whether in dyscalculic subjects , numerical information at the neuronal level is less precisely encoded overall or merely less accessible to attentional selection . It is possible that the capacity to selectively focus on the numerical information and to enhance it already from early levels of visual analysis on , as shown in the current study , is learned or emerging over development , and future studies should directly test this hypothesis . A surprising result of the current experiment is that we could not find information about average item size in the pattern of activity in any of the regions examined , even though this feature’s perceptual discriminability was equated with the one of numerosity . This suggests that the neural mechanisms supporting average size representation may differ from those engaged during single object size analysis which has been shown to overlap partly with numerosity maps in parietal regions ( Harvey et al . , 2015 ) . Mechanisms for average size perception , and in general for ensemble statistics are still unclear . It has been previously suggested that average item size perception , like density perception , may rely on texture processing mechanisms rather than individual item identification ( Im and Halberda , 2013 ) . Various regions along the ventral visual stream have been implicated in texture perception . In particular , adaptation studies have identified recovery of fMRI signal in the medial part of the posterior collateral sulcus that was selective for texture as opposed to color or shape of 3D irregular objects ( Cavina-Pratesi et al . , 2010 ) and the parahippocampal place area ( PPA ) showed equal release from adaptation for object ensemble and surface textures , suggesting that ensembles and textures are processed similarly ( Cant and Xu , 2012 ) . It is possible that average size is also represented in the ventral stream which was not covered here , and future studies should focus on these regions to try to detect a representation of average size . What we observed , however , was that beta weights for density obtained from RSA regression became significant in the parietal regions during the size task , suggesting that texture processing mechanisms may be automatically activated during the average size task . This interpretation , however , has to remain speculative and future studies should investigate neural mechanisms relating texture , density and average size processing . In conclusion , with this study using high-resolution , high-field fMRI we provide direct neuroscientific evidence for a sensory processing mechanism capable of disentangling signals related to visual numerosity from the ones related to associated non-numerical quantities from early stages of cortical processing on , which can then be independently and progressively amplified across the dorsal visual stream when numerical information is explicitly task-relevant . An important goal for the future will be to better understand what are the processing steps and transformations occurring at the different levels of the cortical hierarchy that allow the human brain to isolate numerical information , for example by comparing fMRI data against computational models simulating the visual extraction of numerosity . In addition , it will be important to understand how neuronal representations of numerosity are shaped developmentally and at which cortical levels they can be perturbed to given rise to impaired behavior . Twenty healthy adults with normal or corrected vision ( 10 males and 10 females , mean age 24 years ) participated in the study . The study was approved by the regional ethics committee ( CPP Ile de France VII , Hôpital de Bicêtre , No . 15–007 ) and all participants gave written informed consent . Sample size , although not specifically estimated prior to the study , was equal or larger than the one typically used in experiments in the field ( see for examples: Eger et al . , 2009; Eger et al . , 2015; Cavdaroglu et al . , 2015; Castaldi et al . , 2016; Borghesani et al . , 2019; Cavdaroglu and Knops , 2019; DeWind et al . , 2019; Fornaciai and Park , 2018 ) . Functional images were acquired on a SIEMENS MAGNETOM 7T scanner with head gradient insert ( Gmax 80mT/m and slew rate 333 T/m/s ) and adapted 32-channel head coil ( Nova Medical , Wilmington , MA , USA ) as T2*-weighted fat-saturation echo-planar image ( EPI ) volumes with 1 . 3 mm isotropic voxels using a multi-band sequence ( Moeller et al . , 2010 ) ( https://www . cmrr . umn . edu/multiband/ , multi-band [MB] = 2 , GRAPPA acceleration with [IPAT] = 2 , partial Fourier [PF] = 7/8 , matrix = 120×150 , repetition time [TR] = 2 s , echo time [TE] = 22 ms , echo spacing [ES] = 0 . 71 ms , flip angle [FA] = 68° , bandwidth [BW] = 1588 Hz/px , phase-encode direction left >>right ) . Calibration preparation was done using Gradient Recalled Echo ( GRE ) data . Sixty oblique slices covering the occipital , parietal and partially the frontal cortex were obtained in ascending interleaved order . Before the experimental runs two single volumes were acquired with the parameters listed above but with opposite phase encode direction to be used for distortion correction in the later analysis ( see Image Processing and Data Analysis ) . T1-weighted anatomical images were acquired at 0 . 8 mm isotropic resolution using an MP2RAGE sequence ( GRAPPA acceleration with [IPAT] = 3 , partial Fourier [PF] = 6/8 , matrix = 281×300 , repetition time [TR] = 6 s , echo time [TE] = 2 . 92 ms , time of inversion [TI] 1/2 = 800/2700 ms , flip angle [FA] 1/2 = 4°/5° , bandwidth [BW] = 240 Hz/px , ) . During scanning participants wore a radiofrequency absorbent jacket ( Accusorb MRI , MWT Materials Inc , Passaic , NJ , USA ) to minimize so-called ‘third-arm’ or ‘shoulder’ artifacts due to regions where the head gradient is unable to unambiguously spatially encode the image ( Wald et al . , 2005 ) . Head movement was minimized by padding and tape . Visual stimuli were back-projected onto a translucent screen at the end of the scanner bore and viewed through a mirror attached to the head coil . Participants held two response buttons in their left and right hands . During fMRI scanning participants were centrally presented with heterogeneous arrays of dots , half black , and half white , on a mid-gray background to ensure that total luminance was not a cue for number , a strategy used in many previous studies ( Anobile et al . , 2012; Anobile et al . , 2014; Anobile et al . , 2016c; Anobile et al . , 2016a; Anobile et al . , 2018; Cicchini et al . , 2016; Dakin et al . , 2011; Fornaciai et al . , 2016; Morgan et al . , 2014; Ross , 2010 ) . The generated sets of dots were orthogonally varied in number , average item size and total field area for a total of 18 conditions: six , ten or seventeen dots were presented with either small , medium or large average item area ( 0 . 04 , 0 . 07 , 0 . 12 visual squares degree ) and designed to fall within a small or large total field area ( defined by a virtual circle of either about 5 or 7 . 5 visual degree diameter ) . This implies that higher numbers were associated with higher total surface areas ( total surface area for number six , ten and seventeen respectively corresponded to: 0 . 25 , 0 . 42 , 0 . 71 vd2 for the smallest average item size , to 0 . 42 , 0 . 71 and 1 . 19 vd2 for the medium average item size and to 0 . 72 , 1 . 19 and 2 . 03 vd2 for the largest average item size , correlation between numerosity and total surface area: rho = 0 . 68 , p = 0 . 002 , see Figure 1—figure supplement 1A ) and higher density ( density for numerosity six , ten and seventeen respectively corresponded to: 0 . 30 , 0 . 50 and 0 . 85 dots/vd2 for the small total field area and to: 0 . 14 , 0 . 23 , 0 . 39 dots/vd2 for the large total field area , correlation between numerosity and density: rho = 0 . 71 , p = 0 . 0008 , see Figure 1—figure supplement 1B ) . Despite the significant correlations between numerical and non-numerical dimensions , some pairs of stimuli had equal or similar values of total surface area and density across numerosities and sizes ( for example the array of ten dots with the smallest average item size had total surface area equal to 0 . 42 vd2 which corresponded to the total surface area of the array of six dots with medium average item size , see Figure 1—figure supplement 1 for a more comprehensive visualization of the full stimuli set ) . Convex hull was not explicitly controlled and a-posteriori calculation showed that it was correlated with number ( average convex hull for numerosity six , ten and seventeen respectively corresponded to: 9 , 14 and 18 vd2 for the small total field area and to: 20 , 31 , 41 vd2 for the large total field area , correlation between numerosity and convex hull: rho = 0 . 57 , p = 0 . 01 ) and total field area ( correlation between total field area and convex hull: rho = 0 . 78 , p = 0 . 0001 ) . Numbers and average item sizes were chosen to be perceptually equally discriminable based on a previous behavioral study ( Castaldi et al . , 2018 ) . Total field areas were chosen so that arrays of dots could be sufficiently sparse ( ~1 dot/vd2 ) to target the ‘number regime’ ( Anobile et al . , 2014; Anobile et al . , 2015 ) . Within each run participants performed two tasks in different blocks , as indicated by the written task instructions provided at the beginning of each block . Instructions were shown for 2 s and specified whether participants had to attend either to the number of dots ( number task ) or to the average item size of the dots ( size task ) in the array . Six seconds after the instruction a delayed comparison task started with brief presentation ( 500 ms ) of a sample dot array stimulus . At each trial participants attended to the cued dimension of the sample stimulus and held this information in memory until the following trial was presented , knowing that a comparison response with the following trial may be required . After a variable ISI of 3 . 5–5 . 5 s , a second dot array was presented . If the color of the fixation point remained unchanged ( green ) , no comparison was required and participants only had to update their memory with the new sample stimulus . If instead the fixation point changed color ( turning to red 1 s before the stimulus presentation ) participants had to compare the current stimulus ( match stimulus ) with the one held in memory and decide whether the current stimulus was larger or smaller ( on the attended dimension ) than the previous one . Response was provided by button press and after 5 . 5 s the next sample stimulus was presented and the whole procedure started again . Match stimuli were designed to be ~2 JNDs larger or smaller than the previously presented sample stimulus on the attended dimension , based on each participant’s Weber fraction as measured in a behavioral test prior to the fMRI scanning , while the unattended dimension was the same as the previous sample stimulus . Twenty trials were presented in each block: one trial for each one of the 18 sample stimulus conditions ( 3 numerosity x 3 sizes x two total field areas ) and two match trials . The hands assigned to either the ‘smaller’ or ‘larger’ response were inverted in the middle of the scanning session , that is after the third run , and counterbalanced across subjects . Within the scanning session participants performed six runs of ~7 min and 44 s . Each run included four blocks where the two tasks alternated . The type of task with which the run started was balanced across runs and participants . To measure their numerical and average size acuity , participants performed a behavioral test prior to the fMRI scanning . In different sessions participants were shown two consecutive centrally presented arrays of dots and were required to perform a discrimination task on the attended dimension ( either numerosity or average item size ) by pressing the left or the right arrow ( to choose the first or the second stimulus respectively ) . The set of stimuli used included arrays of 5 , 7 , 9 , 11 , 15 and 20 dots ( ratios 0 . 5 , 0 . 7 , 0 . 9 , 1 . 1 , 1 . 5 and 2 with respect to the reference of 10 dots ) that could be displayed with the average dot areas of 0 . 05 , 0 . 06 , 0 . 08 , 0 . 11 , 0 . 15 and 0 . 2 visual square degrees ( ratios 0 . 5 , 0 . 6 , 0 . 8 , 1 . 1 , 1 . 5 and 2 with respect to the reference of 0 . 1 visual square degrees ) . Dots were randomly drawn within two possible virtual circles of ~5 . 8 and 7 . 6 visual degrees diameter . Reference and test stimuli could appear either as first or as second stimulus . After task instructions and twelve practice trials , participants performed three sessions of one task and three sessions of the other , with counterbalanced order across subjects . For each task participants performed a total of 432 comparisons ( 6 numerosities x six average item sizes x two total field areas x two presentation order x three sessions ) . To quantify participants’ precision in number and size judgments , we computed the JND for each task . The percentage of test trials with ‘greater than reference’ responses was plotted against the log-transformed difference between test and reference and fitted with a cumulative Gaussian function using Psignifit toolbox ( Schütt et al . , 2016 ) . The difference between the 50% and the 75% points yielded the JND . Stimuli and paradigms were generated and presented under Matlab 9 . 0 using PsychToolbox routines ( Brainard , 1997 ) . EPI images were motion-corrected and co-registered to the first single band reference image using statistical parametric mapping software ( SPM12 , https://www . fil . ion . ucl . ac . uk/spm/software/spm12/ ) . The single-band reference images of the two initial volumes acquired with opposite phase encode directions served to estimate a set of field coefficients using topup in FSL ( https://fsl . fmrib . ox . ac . uk/fsl/fslwiki/FSL ) , which was subsequently used to apply distortion correction ( apply_topup ) to all EPI images . Cortical surface reconstruction and boundary based registration of single band reference images to each subject’s cortical surface , as well as a minimal amount of surface constrained smoothing ( FWHM = 1 . 5 mm ) for noise reduction were performed in Freesurfer ( https://surfer . nmr . mgh . harvard . edu/ ) . The preprocessed EPI images ( in subjects’ native space ) were entered into a general linear model separately modeling the effects of the 36 sample conditions ( 3 numerosities x three average item sizes x two total field areas x two tasks , within each run the two repetitions for each condition were pooled together ) , the match stimulus separately for left and right hand and the written instructions at the beginning of the block as stick functions ( using the default of 0 duration for events ) convolved with the standard hemodynamic response function . The six motion parameters were included in the GLM as covariate of no interest . An AR ( 1 ) model was used to account for serial auto-correlation and low-frequency signal drifts were removed by a high-pass filter with a cutoff of 192 s . In each subject we contrasted the activation elicited by: all the sample stimuli during the number tasks against the implicit baseline ( contrast name: ‘Judge Number > Baseline’ ) ; all the sample stimuli during the size tasks against the implicit baseline ( contrast name: ‘Judge Size >Baseline’ ) ; all the sample stimuli during the number tasks against all the sample stimuli during the size tasks ( contrast name: ‘Judge Number > Judge Size ) . After creating the contrasts in each single subject’s volume space , the contrast images were projected onto the surface with Freesurfer , aligned to fsaverage and smoothed with a 3 mm fwhm Gaussian kernel . The second-level group analysis was then performed in the surface space . The beta estimates for the sample stimulus conditions from the first-level analysis ( one beta estimate per run and condition ) were entered into pattern recognition analysis . In each subject we defined anatomical regions of interest ( ROIs ) derived from a surface based probabilistic atlas ( Wang et al . , 2015 ) where regions are defined based on retinotopy . ROIs for V1 to IPS5 were created on the Freesurfer surface and projected back into each subject’s volume space . For each ROI we merged the left and right hemisphere . ROIs were further merged into three large ROIs corresponding to early ( V1 to V3 ) , intermediate ( V3A , V3B and V7 , also known as IPS0 ) and higher-level ( IPS one to IPS5 ) areas . In addition we focused the analysis on individual regions: V1 , V2 , V3 , V3AB ( merging V3A and V3B ) , V7 , IPS12 ( merging IPS 1 and 2 ) , IPS345 ( merging IPS 3 , 4 and 5 ) . Finally , for supplementary analyses , we defined a region along the intraparietal sulcus excluding the field map representation IPS 0–5 . This region was defined by excluding V7 ( also called IPS0 ) and IPS1-5 from the intraparietal and transverse parietal sulci ROI as defined by the Destrieux et al . ( 2010 ) . Within each one of these bilateral regions we selected on a subject-by-subject basis an equal number of 800 voxels that responded most strongly to the orthogonal contrast ‘all sample stimuli > baseline’ for pattern recognition analysis . To evaluate the degree of spatial consistency of the selected voxels across subjects we created an overlap map with Freesurfer ( Figure 3B ) : single subjects’ ROIs were aligned to fsaverage and the number of subjects for which a given location was included in their specific ROI was represented by a heat map ( with yellow color meaning that a given location was selected in all subjects ) . Pattern classification analysis was performed in scikit-learn ( Pedregosa et al . , 2011 ) using beta estimates after subtracting the voxel-wise mean across conditions by applying linear support vector machines ( SVM ) with regularization parameter C = 1 . Classification analysis was performed leaving patterns of one run out at each loop of the 6-fold cross-validation cycle . This implies that classifiers were trained on five betas per condition and tested with the left-out beta images ( one per condition ) . The classification accuracies obtained for each cycle were then averaged together . Pairwise classification was performed for all pairs of numerosities collapsing across the size and total field area dimensions , but keeping patterns separated by task . Classification accuracy was then averaged across all pairs of numerosities for each task . A one-sample t-test against the theoretical chance level of 50% was performed to evaluate significance of discrimination . Repeated measures ANOVAs where then performed on classification accuracies with ROI and task as factors . For the results described in the Supplementary Material , equivalent analyses were performed on the decoding accuracies when the classifier was trained and tested to discriminate between tasks . For representational similarity analysis ( Kriegeskorte , 2008; Kriegeskorte and Kievit , 2013 ) the GLM was performed concatenating the runs and obtaining one single beta per condition , task and subject . Comparable to the procedure of the pattern classification analysis , voxel-wise scaling was applied by subtracting the mean across conditions . Neural representational dissimilarity matrices ( neural RDMs ) for each task and ROI were created by computing the correlation distance ( 1 – the Pearson correlation across voxels ) between activity patterns associated with all possible pairs of conditions using CoSMoMVPA Toolbox ( Oosterhof et al . , 2016 ) . The neural RDMs were then entered in a multiple regression with five predictors corresponding to matrices encoding the distance between all pairs of conditions on a logarithmic scale for the different quantitative dimensions defining the dot arrays: number , average item size , total field area , total surface area and density . To explore potential effects of correlations between predictors , equivalent supplementary analyses included only orthogonal dimensions ( i . e . number , average item size and total field area ) , or all non-numerical dimensions except number ( i . e . average item size , total field area , total surface area and density ) . In the multiple regression analysis all distance matrices were z-transformed before estimating the regression coefficients . The obtained beta weights for each dimension and ROI were tested with one-sample t-tests against zero across subjects . The effects of ROI , dimension and task were analyzed with repeated measures ANOVAs .	Numbers and the ability to count and calculate are an essential part of human culture . They are part of everyday life , featuring in calendars , computers or the weekly shop , but also in some of humanity’s biggest achievements: without them the pyramids or space travel would not exist . A precursor of sophisticated mathematical skill could reside in a simpler mental ability: the capacity to assess numerical quantities at a glance . This ‘number sense’ appears in humans in early childhood and it is also present in other animals , but it is still poorly understood . Brain imaging techniques have identified the parts of the brain that are active when perceiving numbers or making calculations . As techniques have advanced , it has become possible to resolve fine differences in brain activity that occur when people switch their attention between different visual tasks . But how exactly does the human brain process visual information to make sense of numbers ? One theory suggests that humans use visual cues , such as the size of a group of objects or how densely packed objects are , to estimate numbers . On the other hand , it is also possible that humans can sense number directly , without reference to other properties of the group being observed . Castaldi et al . presented twenty adult volunteers with groups of dots and asked them to focus either on the number of dots or on the size of the dots during a brain scan . This approach allowed the separation of brain signals specific to number from signals corresponding to other visual cues , such as size or density of the group . The experiment revealed that brain activity changed depending on the number of dots displayed . The signal related to number became stronger when people focused on the number of dots , while signals related to other properties of the group remained unchanged . Moreover , brain signals for number were observed at the very early stages of visual processing , in the parts of the brain that receive input from the eyes first . These results suggest that the human visual system perceives number directly , and not by processing information about the size or density of a group of objects . This finding provides insights into how human brains encode numbers , which could be important to understand disorders where number sense can be impaired leading to difficulties learning math and operating with numbers .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	Attentional amplification of neural codes for number independent of other quantities along the dorsal visual stream
The Mullerian ducts are the anlagen of the female reproductive tract , which regress in the male fetus in response to MIS . This process is driven by subluminal mesenchymal cells expressing Misr2 , which trigger the regression of the adjacent Mullerian ductal epithelium . In females , these Misr2+ cells are retained , yet their contribution to the development of the uterus remains unknown . Here , we report that subluminal Misr2+ cells persist postnatally in the uterus of rodents , but recede by week 37 of gestation in humans . Using single-cell RNA sequencing , we demonstrate that ectopic postnatal MIS administration inhibits these cells and prevents the formation of endometrial stroma in rodents , suggesting a progenitor function . Exposure to MIS during the first six days of life , by inhibiting specification of the stroma , dysregulates paracrine signals necessary for uterine development , eventually resulting in apoptosis of the Misr2+ cells , uterine hypoplasia , and complete infertility in the adult female . In mammals , both sexes initially develop Mullerian ducts , consisting of a single layer of epithelial cells surrounded by undifferentiated mesenchymal cells . Mullerian Inhibiting Substance Receptor ( Misr2/Amhr2 ) expression is first detected in the Mullerian mesenchyme at around E13 . 5 both in males and in females ( Arango et al . , 2008 ) . In male mice , secretion of MIS ( also known as Anti-Mullerian Hormone or AMH ) by the developing testes causes regression of the Mullerian ducts during embryonic days 14 . 5–15 . 5 ( Jost , 1947; Josso et al . , 1976 ) . Regression of the male Mullerian ductal epithelium is mediated through non-cell autonomous paracrine signals emanating from the underlying Misr2+ mesenchymal cells in response to MIS , and is thought to be dependent on Wnt signaling and beta-catenin activity in the epithelium ( Mullen and Behringer , 2014 ) . The Mis/Misr2 ( Amh/Amhr2 ) pathway is highly specific to this process , since either ligand or receptor knockout mice present with identical phenotypes of persistent Mullerian duct syndrome ( PMDS ) , a rare form of male pseudohermaphroditism in mice ( Behringer et al . , 1994; Mishina et al . , 1996; Mishina et al . , 1999 ) and humans ( Imbeaud et al . , 1994 ) . It is thought that the narrow developmental window when Mullerian duct regression occurs is the only period when these Misr2+ mesenchymal cells are able to respond to MIS , beyond which further differentiation of the duct renders it insensitive to this inhibitory signal ( Josso et al . , 1976 ) . Indeed , transgenic mice with constitutive overexpression of MIS , driven by a metallothionein-1 promoter , during the critical period of Mullerian duct formation displayed Mullerian agenesis and quickly lost all germ cells in the ovary after birth ( Behringer et al . , 1990 ) . In females , which do not express MIS during embryonic development , the fate of these subluminal Misr2+ mesenchymal cells remains unclear . In adult mice , expression of Misr2 is restricted to the myometrium , suggesting a common origin for both cell types ( Arango et al . , 2005; Arango et al . , 2008 ) ; however , the lack of an inducible Misr2 reporter mouse has precluded a precise lineage tracing . Many cell types of the urogenital ridge primordium including the coelomic epithelium , mesonephric mesenchyme , and Wolffian duct epithelium are thought to contribute to the formation and development of the Mullerian duct as it further differentiates into the oviduct , uterus , cervix , and upper vagina ( Fujino et al . , 2009; Mullen and Behringer , 2014 ) . In rodents , most of the differentiation of the uterine layers occurs postnatally; at birth , the uterus consists of a single layer of luminal epithelium surrounded by undifferentiated mesenchyme which later develops into myometrium and endometrial stromal cells , while the epithelium subsequently gains the ability to form endometrial glands though invagination at postnatal day ( PND ) 6–9 ( Branham et al . , 1985; Brody and Cunha , 1989 ) . The timing of development of the endometrial stroma , and its contribution to the coordination of development of the myometrium and endometrial glands in early postnatal development are poorly understood . Mutations of genes which orchestrate early Mullerian mesenchyme development can have drastic consequences on female fertility and lead to Mullerian aplasia or uterine hypoplasia as observed in the Mayer-Rokitansky-Kuster-Hauser syndrome ( Patnaik et al . , 2015 ) . Therefore , the pathways involved in early postnatal specification of the uterine compartments are critical to our understanding of Mullerian development and uterine factor infertility . Here , we characterize the persistence of subluminal Misr2+mesenchymal cells beyond the embryonic period of sexual differentiation , document their retained sensitivity to MIS neonatally , and characterize their critical role in the development of the endometrial stroma . To determine the fate of the Mullerian subluminal mesenchyme in the developing uterus , we sought to identify specific markers whose expression in that cell type perdured postnatally . Although the embryonic development of the Misr2+ Mullerian mesenchyme has been extensively studied ( Jamin et al . , 2002; Arango et al . , 2008; Kobayashi et al . , 2011 ) , its early postnatal fate has not . Using lineage tracing in a Misr2-CRE/TdTomato reporter transgenic cross in C57BL/6 mice , we first confirmed that embryonic urogenital Misr2+ intermediate mesoderm gives rise to both the endometrial and the myometrial layers of the uterus , but not its epithelium ( Figure 1—figure supplement 1A ) . Because Misr2-CRE is not inducible , any Misr2 expression during early development will result in permanent expression of the TdTomato reporter ( Figure 1—figure supplement 1A ) Therefore , to track further the Misr2+ cells in Mullerian mesenchyme , we conducted careful spatiotemporal studies using Misr2 RNA in situ hybridization ( RNAish ) from the embryonic period ( E14-15 ) into postnatal life ( Figure 1A ) . As expected , expression of Misr2 is restricted to the mesenchyme surrounding the Mullerian duct in both male and female urogenital ridges during embryonic development ( E17-19 ) ( Figure 1A ) . Postnatally , Misr2 expression becomes increasingly restricted to a thin band of subluminal mesenchyme , while being excluded from the epithelium and developing myometrium ( Figure 1A , PND 0 , PND 2 ) ( Figure 1A , Figure 1—source data 1 ) . Following differentiation of the functional layers of the uterus around PND 6 ( Brody and Cunha , 1989 ) , Misr2 expression commences to be detectable in the myometrium consistent with previous findings ( Arango et al . , 2008 ) ( Figure 1A ) . To evaluate the effect of MIS on the development of the uterus , we chose to use rats which have larger litters and display a similar spatiotemporal pattern of Misr2 expression ( Figure 1—figure supplement 1D–E , Figure 1—source data 1 ) . Furthermore , in rats , Misr2 expression is gradually attenuated in the PND1-6 period , both at the proximal ( cervix ) and at the distal ( oviduct ) ends of the developing uterine horns ( Figure 1—figure supplement 1D ) coinciding with the timing of expansion of the endometrial stroma , and the rise in secretion of MIS from the developing ovaries , which was measured in the rat serum by ELISA ( Figure 1—figure supplement 1B and C ) . To evaluate the sensitivity of Misr2+ mesenchymal cells to MIS during this postnatal period , we treated rat pups with adeno-associated viral vectors ( AAV9 ) ( Pépin et al . , 2015; Kano et al . , 2017 ) delivering MIS at PND 1 ( Figure 1B and Figure 2A ) . Administration of a single dose of AAV9-MIS ( 5E10 particles/pup ) on PND 1 led to a robust induction ( 2 . 8 ± 0 . 7 µg/ml ) of circulating exogenous MIS as measured by ELISA on PND6 ( Figure 2B ) . Postnatal exposure to MIS led to an alteration in the appearance of the uterus by PND 6 , and severe uterine hypoplasia by PND 20 , suggestive of a failure of the uterus to develop beyond its perinatal state ( Figure 2C ) . Histomorphological analyses of transverse uterine sections revealed smaller uteri with underdeveloped endometrial stromal layers and smaller lumina in MIS-treated rats compared to their sibling controls at all time points analyzed ( PND 6 , 10 , and 20 ) ( Figure 2C and D , Figure 2—source data 1 ) . Treatment with MIS prevented the gradual decrease of Misr2+ cells , and inhibited the expansion of the endometrial stroma normally observed by PND 6 ( Figure 2E ) . For consistency and clarity , we will refer to these persistent Misr2+ subluminal mesenchymal cells in the MIS-treated animal as ‘inhibited progenitors’ , although this function remains presumptive . In response to MIS , these inhibited putative progenitors expressed high levels of Smad6 , a negative regulator of TGFβ signaling , and a previously reported canonical downstream target of MIS in the Mullerian mesenchyme ( Figure 2E ) , suggesting MIS signaling was both operational and autonomous to these Misr2+ cells ( Clarke et al . , 2001; Mullen et al . , 2018 ) . Interestingly , the endometrial stromal hypoplasia resulting from MIS exposure precluded later development of endometrial glands , as confirmed by absence of Foxa2 marker expression in immunofluorescence and qPCR ( Figure 2—figure supplement 1A and B ) . However , the development of the circular myometrial layers was unaffected by exposure to MIS , as shown in the MIS-treated transverse uterine sections by immunofluorescence , and qPCR analysis of smooth muscle markers such as smooth muscle actin ( Acta2 ) and transgelin ( Tagln ) , which remain similar at various time points between control and MIS exposed uteri ( Figure 2—figure supplement 1C and D ) . Surprisingly , the inhibited progenitor cells expressed Acta2 ectopically ( shown in red ) as well as Vimentin ( shown in green ) in response to MIS on PND 6 ( Figure 2F ) , suggesting a metaplastic effect of MIS . We next focused on differential gene expression in response to MIS treatment using our dataset of 6811 control and 2990 MIS-treated cells . In our cell atlas , the control and treated uterine cell datasets were of similar quality , and had comparable distributions of unique molecular identifiers ( UMI ) and gene numbers ( Figure 3D , Figure 3—figure supplement 1B , Figure 3A ) . While both the epithelial ( CTL 9 . 63% , MIS 8 . 06% ) and the myometrial ( CTL 11 . 98% , MIS 16 . 72% ) ( Figure 3—figure supplement 1B–C , Figure 3—figure supplement 4 ) cell types were re-presented in similar proportions in both datasets , ‘inner’ ( CTL 18 . 68% , MIS 3 . 28% ) and ‘outer’ ( CTL 15 . 64% , MIS 0 . 27% ) endometrial stroma clusters ( Figure 3—figure supplement 1B–C ) were almost entirely composed of control cells unlike the ‘inhibited progenitor’ cluster , which was unique to the MIS treatment ( Figure 3B–D , Figure 3—figure supplement 1B–C , Figure 4A , Figure 4—figure supplement 1 ) . Misr2 expression was enriched only in the inhibited progenitor cluster ( Figure 3C–D , Figure 4A ) as previously observed by RNAish ( Figure 2E ) . MIS stimulation caused the sub-luminal inhibited progenitor cells to persist longer and induced overexpression of genes related to the bone morphogenic protein and transforming growth factor beta ( BMP/TGFβ ) signaling pathways ( Smad6 , Bambi ) , epigenetic markers ( Hdac4 ) , and , strikingly , to express ectopically smooth muscle ( Thbs2 ) and epithelial ( Msx2 ) markers , reflecting the multipotent characteristics of the early Mullerian progenitors ( Figure 3E–G , Figure 4A , Figure 4—figure supplement 1A–D , Table 1 ) . QPCR analysis of these ectopically expressed genes revealed that the large differences observed at PND 6 diminished with time , and became less pronounced by PND 15 ( Figure 3G , Figure 4A , Figure 4—figure supplement 1; Figure 4—figure supplement 2; Figure 4—figure supplement 1B , Figure 3—source data 3 ) . Cleaved caspase-3 histological staining of the MIS-treated uteri confirmed that the inhibited progenitor cells underwent apoptosis at approximately 9 days after treatment , whereas normally developing endometrial stromal cells were negative for the apoptotic marker ( Figure 4—figure supplement 1E ) . QPCR and RNAish validation of differentially expressed genes in the stroma confirmed that both Wfikkn2 ( ‘outer stroma’ ) , Bmp7 ( ‘inner stroma’ ) failed to be induced over time in the MIS treated group ( Figure 3G ) , along with other markers such as Cpxm2 and Enpp2 ( Figure 4—figure supplement 3A–D ) , consistent with the hypothesis that MIS prevented the subluminal progenitor cells from undergoing stromal specification and amplification; instead they eventually underwent apoptosis ( Figure 4—figure supplement 1E ) . Our results indicate that neonatal exposure to MIS induces changes in gene expression in the epithelial cell cluster as visible by the shifted treatment population in the t-SNE plot ( Figure 3D , Figure 3—figure supplement 1B , Figure 4—figure supplement 4A ) . Even though expression of the basic luminal epithelial cell markers were not significantly different between control and MIS-treated cells ( Ecad , Cd24 , and Klf5 ) ( Figure 4—figure supplement 4 ( A-E ) , differential gene expression analysis of the epithelial cluster based on treatment ( CTL and MIS ) revealed gene candidates whose expressionwas significatly changed by MIS treatment ( Figure 4—figure supplement 4F–I ) despite the lack of Misr2 expression in epithelial cells ( Figure 1A , E ) . We confirmed that Id3 was downregulated in the MIS-treated epithelial cells ( Figure 4—figure supplement 4H ) . Another intriguing gene expression pattern in response to MIS treatment was the downregulation of the epithelial cell marker Msx2 at PND6 , which coincides with its ectopic expression in the inhibited putative progenitor ( Figure 3F–G ( Msx2 ) , Figure 4A , Figure 4—figure supplement 4F ) . These results are consistent with MIS mediating the indirect repression of epithelial Msx2 and Id3 through paracrine signals emanating from the Misr2+ mesenchymal cells , or the lack of normal endometrial stromal signals , which in turn may prevent subsequent endometrial gland formation at later time points ( Figure 4—figure supplement 4I , Figure 2—figure supplement 1A–D ) . Finally , to survey the paracrine signaling between the inner stroma and epithelium and catalog how it may be disrupted by MIS treatment ( inhibited putative progenitor and epithelium ) , we performed a comprehensive ligand/receptor analysis using the CellPhoneDB algorithm ( Vento-Tormo et al . , 2018 ) ( Figure 4B- Figure 4—figure supplement 4 , Figure 4—source data 1 ) . Briefly , significantly expressed ligand/receptor pairs were systematically cataloged between each functional cell types ( Figure 4B ) , revealing important developmental pathways dysregulated in MIS-treated uteri , such as Wnt and Igf2 signaling ( Figure 4B , Figure 4—figure supplement 4 , Figure 4—source data 1 ) . The timing of expression of Misr2 in the subluminal mesenchyme led us to hypothesize that limiting exposure to MIS during only the first 6 days of development could be sufficient to explain the observed long-term uterine hypoplasia . To test this hypothesis , we treated rat pups with recombinant MIS protein ( rhMIS ) during 6 days intervals ( 3 mg/kg ) starting from days 1 , 6 , or 11 ( Figure 5A ) . Consistent with this hypothesis , only when rats were treated during PND1-6 period were the uteri smaller than controls , at PND6 , 20 , 45 , and even up to 8 months ( Figure 5B , C , G ) . In contrast , when rhMIS was administered starting from PND6 or PND11 , uterine hypoplasia was muted or absent , suggesting a narrow window of susceptibility with long-lasting consequences ( Figure 5B ) . Therefore , only perinatal ( PND1-6 ) administration of rhMIS completely phenocopied the continuous exposure phenotypes observed with AAV9-MIS , including lower percentage of endometrial stromal cells , smaller luminal ducts , and absence of glandular development ( Figure 5B , Figure 5—figure supplement 1A ) as confirmed by downregulation of Foxa2 expression and absence of glandular ducts ( Figure 5D–E ) . Strikingly , this 6-day postnatal treatment was also sufficient to cause complete infertility later in adulthood ( n = 3 per group , p<0 . 05 ) ( Figure 5F ) . Analysis of the control and the 6-day rhMIS-treated uteri at a later time point ( 8 months ) confirmed that the profound endometrial stromal hypoplasia persists in the adult ( Figure 5G–H ) . In contrast , the ovaries fully recover from the short MIS inhibition of folliculogenesis ( PND1-6 ) and display normal ovarian sizes and follicular composition at this timepoint ( Figure 5H ) . The impaired uterine development and infertility is unlikely to be secondary to ovarian suppression since folliculogenesis only starts reaching early pre-antral stages by PND6 , and MIS treatment appears to have little effect on steroid hormones ( E2 and P4 ) during that time ( Figure 5—figure supplement 1B ) . Furthermore , as previously described ( Kano et al . , 2017 ) , MIS inhibition of folliculogenesis by rhMIS is reversible , and while the MIS-treated PND1-6 ovaries showed an initial delay in folliculogenesis at early timepoints ( PND 20 ) , it was resolved by PND 45 confirming no lasting impact on ovarian function ( Figure 5—figure supplement 1C–D ) . To confirm that the effect of MIS on the Misr2+ putative stromal progenitor was intrinsic to the uterus ( and not the ovary ) , we treated gonadectomized rat pups with control or AAV9-MIS on PND2 ( Figure 5—figure supplement 2A , B ) , which resulted in the same uterine hypoplasia phenotype by PND 10 ( Figure 1C ) . Finally , to confirm that the signaling in the ‘inhibited progenitor’ was dependent on the canonical MIS receptor , we treated Misr2-deficient female mice ( Misr2-/- ) with AAV9-MIS , which failed to recapitulate the uterine hypoplasia phenotype ( Figure 5—figure supplement 2B , C ) . Together , these results revealed a cell-autonomous effect of MIS in the subluminal mesenchyme , intrinsic to the uterus , and dependent on Misr2 , in which progenitors normally specified to form endometrial stromal layers at PND1-6 are inhibited by MIS , leading to long-term infertility . We sought to determine whether uterine subluminal mesenchymal cells expressing MISR2 are also present in the human female fetus preceding production of the ligand ( MIS ) by the ovary . We used paraffin-embedded archival tissue of human female reproductive tract from fetuses ranging from 22 weeks ( wk ) to 37 weeks of gestation . Using the adjacent fetal ovary tissue as a positive control for MISR2 RNAish , we analyzed the uterus at different developmental stages , revealing a spatiotemporal pattern of expression strikingly similar to that of the rodents ( Figure 6A–C ) . MISR2+ cells were present in the same location in the fetal uterus , directly adjacent to the lumen at 22 weeks of gestation and receded at later timepoints ( 24wk , 37wk ) , coinciding with the production of MIS by the human ovary , which is thought to begin at 24 weeks of gestation ( Kuiri-Hänninen et al . , 2011 ) . To determine if candidate genes suspected to cause Mullerian aplasia or hypoplasia in humans ( Nik-Zainal et al . , 2011 ) may be present in the developing rat uterus , we analyzed their expression in our cell atlas , revealing the enrichment of several candidates within the ‘inhibited progenitor’ cluster ( Figure 6D ) . Single-cell RNA sequencing of the PND6 uterus revealed that the subluminal Mullerian duct mesenchyme contains a previously uncharacterized cell type ( Figure 7 ) that plays a crucial role in the specification of the endometrial stroma during neonatal uterine development . We hypothesize that these Misr2+ cells represent stromal progenitors , which normally gives rise to the inner and outer endometrial stromal layer around PND6 in mice and rats ( Figure 1B ) . Surprisingly , these Misr2+ putative stromal progenitors retain sensitivity to inhibition by MIS postnatally , and can be reprogrammed to undergo apoptosis instead of developing into the endometrial stromal layers if exposed to MIS ( Figure 7 ) . Although postnatal MIS exposure is no longer able to induce regression of the uterine luminal epithelium , its normal function in male fetuses ( Jost , 1947 ) , it does irreversibly block its ability to form endometrial epithelial glands . We speculate that this retained sensitivity to MIS in the female may be a vestigial pathway of the male , which is normally silenced in females prior to the emergence of secretion of MIS by the ovary . However , it is unlikely that MIS itself is an important developmental trigger regulating endometrial stroma development , since the uterus develops normally in both Mis and Misr2 knockout mice ( Behringer et al . , 1994 ) ( Mishina et al . , 1999 ) ( Mishina et al . , 1999 ) . The postnatal response of the Misr2+ putative progenitors to MIS provides some unique insights into the developmental pathways elicited during fetal Mullerian duct regression . The nascent fetal Mullerian duct is mesoepithelial in origin , being derived from the invagination of the coelomic epithelium , and begins further differentiation into epithelium proper coincidentally with the timing of regression . Others have suggested that this epithelial differentiation may subsequently restrict the ability of the ductal cells to undergo the epithelial to mesenchymal transition characteristic of ductal regression ( Allard et al . , 2000 ) . This raises the possibility that the Misr2+ mesenchyme in the neonatal female may be responding to MIS similarly to male urogenital mesenchyme , but that the neonatal epithelium is unable to regress in response to those signals . Supporting this interpretation of recapitulated mesenchymal regression is the ectopic expression of many of the same genes and pathways previously identified in the regressing male fetal Mullerian duct ( Bambi , Smad6 , Wif1 , etc . ) ( Mullen et al . , 2018 ) . Even though gain of function MIS or MISR2 mutations have not been reported in women with Mullerian anomalies , the uterine hypoplasia observed in the present study is suggestive of Mayer-Rokitansky-Küster-Hauser syndrome ( MRKH ) , also known as Mullerian aplasia which affects 1 in 4500 women ( Nik-Zainal et al . , 2011 ) . Genes expressed in the Misr2+ putative progenitor cells in response to MIS treatment likely represent either pathways of Mullerian duct regression or of uterine endometrial stroma progenitor development . Therefore , we speculate that the markers described herein may represent candidate genes underlying developmental disorders of the Mullerian duct . Efforts to identify causative genes within regions of copy number variation in patients affected with Mullerian aplasia have turned up candidate genes such as KHDRBS2 , and GFRA1 ( Nik-Zainal et al . , 2011 ) , which we see uniquely expressed in the ‘inhibited progenitor’ cluster ( Figure 6D ) . Similarly , both Gata3 which causes hypoparathyroidism , sensorineural deafness , renal anomaly ( HDR ) syndrome with uterine hypoplasia ( Van Esch et al . , 2000 ) and Fgfr2 , which causes disorders of sexual dimorphism in males ( Bagheri-Fam et al . , 2015; Barseghyan et al . , 2018 ) , and decidualization defects in females ( Filant et al . , 2014 ) , are highly expressed in the inhibited MISR2+ progenitors ( Figure 6D ) . The importance of the endometrial stroma in the formation of endometrial glands and fertility has been demonstrated in multiple mouse models such as Wnt4 mutant mice , and neonatal diethylsilbesterol ( DES ) treatments , both of which carry phenotypes of endometrial glandular dysplasia also observed in our rodents with postnatal exposure to MIS ( Herbst et al . , 1980; Medlock et al . , 1988; Hayashi et al . , 2011; Prunskaite-Hyyryläinen et al . , 2016 ) . These mouse models display endometrial stromal hypoplasia , which precludes glandular development as a result of the disruption of Wnt pathways and the communication between stromal and epithelial compartments . Our single-cell transcriptomic analysis identified distinct inner and outer endometrial stromal layers with molecular signatures that might presage specialized stromal functions , such as regulation of the adjacent epithelium or myometrium . A comprehensive characterization of such paracrine signals across the cell types of our control and MIS-treated uterine atlases using the CellPhoneDB algorithm ( Vento-Tormo et al . , 2018 ) revealed the immense complexity of those cellular interactions , and their dysregulation by MIS ( e . g . Wnt signaling , Figure 4B ) . MIS treatment is likely especially disruptive to the interaction between the inner endometrial stroma and adjacent luminal epithelium , as seen by the absence of expression of Bmp7 , and coincidental downregulation of Msx2 in those cell types , respectively , which have been previously implicated in the coordination of endometrial gland formation in the uterus ( Phippard et al . , 1996; Kodama et al . , 2010; Yin et al . , 2015 ) and decidualization defects ( Daikoku et al . , 2011; Monsivais et al . , 2017 ) . The nature of the paracrine signals emanating from the inner stroma regulating epithelial development , and their dysregulation during MIS treatment , including the ephrin , notch , Igf , Tnf , Mdk/Ptn , Tyro3 , and Fn1 pathways ( Figure 3 , Figure 3—figure supplement 4 ) , should be systematically investigated in future studies . Furthermore , the involvement of premature exposure of the developing Mullerian ducts to MIS in disorders of sexual differentiation of the female is not fully appreciated ( Chen et al . , 2014; Van Batavia and Kolon , 2016 ) . Moreover , the recent identification of an Misr2+ subepithelial adult endometrial stem cell suggests that some neonatal endometrial stromal sprogenitor may persist into adulthood , where they could play a role in endometrial homeostasis and repair ( Yin et al . , 2019 ) . Interestingly , we have recently reported an increased incidence of preterm birth in PCOS patients with high circulating MIS ( Hsu et al . , 2018 ) raising the possibility of a causative link between high MIS exposure and uterine dysfunction . Conversely , the complete infertility resulting from a short treatment with MIS could have useful applications in the veterinary settings where it may be used as a permanent contraceptive that does not affect ovarian function ( Hay et al . , 2018 ) . Finally , given our findings that MIS is a potent inhibitor of MISR2+ putative endometrial stromal progenitors during uterine development , and that this cell type is likely also active in fetal human uteri , it would be of interest to explore possible clinical applications of MIS , or related pathways , in the context of Mullerian development pathologies , uterine infertilities , and endometrial stromal cancers , should these pathways become reactivated in those tumors ( Chiang and Oliva , 2013 ) . This study was performed in accordance with experimental protocols 2009N000033 and 2014N000275 approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee . Strains of Sprague–Dawley ( purchased from Envingo ) and Friend leukemia virus B ( FVB ) ( purchased from Charles River Laboratories ) were used for rat and mouse experiments , respectively . Misr2/Amhr2-Cre knock-in mice were purchased from the Mutant Mouse Regional Resource Centers ( MMRRC ) ( strain B6;129S7-Amhr2tm3 ( cre ) Bhr/Mmnc , backcrossed with C57BL/6J ) ( Jamin et al . , 2002 ) . Tail genotyping of the Misr2-cre knock-in and WT mice were done with REDExtract-N-Amp Tissue PCR Kit ( Sigma , #SLBT8193 ) with the previously described sets of primers ( Jamin et al . , 2002 ) . The adeno-associated virus serotype 9 ( AAV9 ) gene therapy vector was used for sustained delivery of a higher concentration of human MIS analog ( LR-MIS ) as described ( Pépin et al . , 2015 ) . To test the effect of continuous MIS exposure on uteri , rats or mice were injected subcutaneously with AAV9-recombinant human LR or RF MIS ( AAV9-MIS ) on postnatal day 1 , and their uteri were harvested at different time points for histological analysis ( 5E10 particles/pup ) . Blood was collected by cardiac puncture at endpoint , and centrifuged at 900 × g for 10 min at room temperature from control and AAV9-MIS-treated female rat pups on PND3 , 5 , 6 , 10 , 15 , and 30 ( n > 2 ) . To validate the RNA scope markers on tissue sections , mice were injected with AAV9-MIS on PND 1 ( 1E10 particles/pup ) , and sacrificed on PND 6 ( n = 3 both for the control and the treated ) . For each time point , the uteri were cut radially in halves . One half was fixed in formalin for histological analysis and immunohistochemistry , the other half was flash-frozen for RNA isolation and qPCR analyses . In mice , the paraffin-embedded fixed tissue was sectioned for RNAish analysis , while in rats it was used for histomorphological analyses . To test the window of sensitivity to exogenous MIS during uterus development , rats were injected subcutaneously with a human recombinant MIS ( LR-MIS ) protein ( Pépin et al . , 2013 ) daily ( 3 mg/kg/day ) for 6 days starting from PND1-6 , PND6-11 , or PND11-17 ( Figure 5A ) . Uteri of rat pups injected from PND1-6 were harvested and fixed on 20 , and 45 ( for PND20 , n = 2 for control , n = 3 for treated , for PND45 n = 2 both for the control and the treated ) . Uteri of the rat pups injected from PND6-11 , or PND11-17 , were harvested and fixed on PND45 ( n > 2 for all ) ( Figure 5A ) . To rule out the involvement of ovarian hormones as a contributor to the MIS-induced uterine hypoplasia , rat pups were gonadectomized two days after birth ( n = 4 ) , prior to receiving MIS treatment ( Figure 5—figure supplement 2A–B ) . For gonadectomy , the rat pups were anesthetized using isofluorane . Bilateral longitudinal incisions were made in the mid dorsal line through the skin and musculature one-third the distance between the base of the tail and the neck , directly over the position of the ovary . The ovaries together with the oviducts were extruded through the incision via the ovarian fat pad . The ovarian vasculature was ligated between the oviduct and uterine horn and the ovaries and oviducts were resected . The muscle layer was closed with silk sutures , and the outer skin was closed with a single metal clip . The rats were then kept warm until they had completely recovered from the anesthesia . Analgesia was provided for 3 days with Carprofen PO . Two of the rat pups were treated with AAV9-MIS , while the controls were treated with empty vector ( 5E10 particles , subcutaneously , n = 2 both for control and treated ) 6 hr after the surgery . The pups were then euthanized on PND10 , and their uteri were fixed for histomorphological analysis ( Figure 5—figure supplement 2A–B ) . To verify that the MIS Receptor 2 ( Misr2 ) is the mediator of the uterine phenotype caused by exogenous MIS , Misr2cre/+ and Misr2cre/cre transgenic mice ( Jamin et al . , 2002 ) were treated with empty vector or AAV9-MIS ( 5E10 particles , subcutaneously ) on day 1 and their uteri were analyzed on PND20 ( Figure 5—figure supplement 2C–D ) ( n = 2 for Misr2cre/+ , n = 1 for Misr2cre/cre ) . Misr2cre/cre males had retained Mullerian ducts confirming the loss-of-function of the Misr2 ( Jamin et al . , 2002 ) . The Beckman AMH ELISA ( Beckman , #A73818 ) , which can detect both endogenous murine MIS and exogenous human MIS secreted by the AAV9-MIS infected muscles was used to measure the serum MIS levels . To detect the murine endogenous MIS levels during the developmental time span of rat females , serum from control PND1 , 4 , 6 , and 20 rats , as well as AAV9-MIS-treated rats on PND six were measured by ELISA ( n = 3 for PND4 , 6 and 20 , and n = 2 for PND 1 , n = 3 for AAV9-MIS treated rat ) . Murine Estradiol and Testosterone serum levels were measured with specific ELISAs at the Ligand Assay and Analysis Core of the Center for Research in Reproduction at University of Virginia School of Medicine under a cooperative agreement ( The core is supported by the Eunice Kennedy Shriver NICHD/NIH ( NCTRI ) , Grant P50-HD28934 ) . ( n = 3 for PND3 and 30 control and treated , and PND 5 control; n = 2 for PND five treated; PND6 , 10 , 15 control and treated animals ) . Sprague–Dawley rats were injected subcutaneously with 3 mg/kg/day of recombinant MIS ( LR-MIS ) or 20 µl saline ( vehicle control ) from PND1-6 . One MIS-treated and one sibling control female were caged with one experienced breeder male ( n = 3 cages ) at 6 weeks of age . The male was separated from the cage after pregnancy was identified and returned after the pups were weaned . The total number of pups and litters from each female was monitored for a period of 4 months . Dissected uteri and ovaries were fixed in 4% ( wt/vol ) paraformaldehyde at 4°C ( for histology and immunofluorescence ) or in 10% neutral buffered formalin at room temperature overnight ( for RNAish ) . Tissues were embedded in paraffin blocks in an automated tissue processor ( Leica #TP1020 ) . 5 μm transverse uterine sections from the middle of the uterine horn ( i . e ‘b’ in Figure 1—figure supplement 1D ) were used for hematoxylin and eosin ( H&E ) staining , immunofluorescence ( IF ) , and RNAish using the RNA scope ( ACD bio ) system . Archival human fetal tissue sections were provided by the Massachusetts General Hospital , Gynecological Pathology Department through an IRB approved protocol ( IRB 2007P001918 ) . Tissue sections were rehydrated for IF in an alcohol series after deparaffinization in xylene . Antigen retrieval was performed by parboiling in 10 mM sodium citrate ( pH 6 . 0 ) , cooling at room temperature , blocking in 3% bovine serum albumin ( BSA ) in Tris-buffered solution ( TBS ) for 1 hr , followed by three washes ( 10 min each ) in TBS and the sections incubated in primary antibody overnight at 4°C . For double-labeling , the slides were blocked after washes , and then incubated with a second primary antibody overnight at 4°C . The sections were then incubated in fluorescently conjugated secondary antibodies ( Alexa Fluor 555-conjugated donkey anti-rabbit IgG antibody , # A31572; Alexa Fluor 488-conjugated donkey anti-rabbit IgG , #A21206 ) for one hour at room temperature and cover-slipped with vectashield mounting medium with DAPI ( Vector Laboratories # NC9265087 ) . For immunohistochemistry ( IHC ) , Dako EnVision + System horseradish peroxidase ( HRP ) Labeled Polymer Anti-Rabbit was used as the secondary antibody ( #K4002 ) , and the HRP signal was detected using the DAKO detection system ( Dako , #K5007 ) . Antibody dilutions for IF and IHC were as follows: Smooth muscle alpha action ( SMA ) ( 1:300 , abcam , #5694 ) , Vimentin ( 1:300 , abcam , #32547 ) , Foxa2 ( 1:500 , LifeSpan Biosciences , #138006 , 1:500 ) , cleaved caspase-3 ( 1:50 , cell signaling , #9661S ) , E-cadherin ( Cdh1 ) ( 1:200 , Invitrogen #13–1900 ) . RNAish was performed with the manual RNAscope 2 . 5 HD Reagent Kit ( RED ) ( ACD Bio , # 322350 ) following the manufacturer’s instructions as previously described ( Wang et al . , 2012 ) . The tissue sections were hybridized with pre-designed or custom-designed probes spanning mRNAs of the target genes ( see Table S1 for accession number , target region , and catalog number of each gene ) in the HybEZ hybridization oven ( ACD Bio ) for 2 hr at 40°C , following deparaffinization in xylene , dehydration , peroxidase blocking , and heat-induced epitope retrieval by the target retrieval and protease plus reagents ( ACD bio , #322330 ) . The slides were then processed for standard signal amplification steps , and a red chromogen development was performed using the RNAscope 2 . 5 HD ( Red ) detection Kit ( ACD Bio , #322360 ) . The slides were then counterstained in 50% hematoxylin ( Dako , #S2302 ) for 2 min , air-dryed and coverslipped with EcoMount . Middle sections ( ‘b’ in the scheme of Figure 1—figure supplement 1D ) of control and MIS-treated rat uteri were stained with H and Es for histomorphological analyses , which were conducted at different developmental time points ) ( Figure 2—source data 1 , Figure 5—source data 1 ) . Luminal duct height , area of the whole uterus , and area of the endometrium were calculated from the transverse sections using the image J software . For Figure 2 , n = 3 for PND3 control and treated; n = 2 for PND6 , 20 control and treated , n = 1 for PND10 control and treated samples . For Figure 4 , n = 2 for the PND20 control; PND45 control and the treated; n = 3 for the treated PND20 sample ( Figure 2—source data 1 ) . Human fetal tissue sections were imaged by the Keyence BZ-X800 microscope at 20x resolution , and the RNAish stains were auto-quantified by BZ-X800 analysis software . Approximately 60 images were obtained per stained section , and stitched together for the top panel of Figure 6a . 5 random 20x images were selected per time point for analysis , and areas of red RNA scope dots ( Misr2 transcripts amplified by RNA-scope ) were detected and labeled based on hue ( see red dots on the bottom section of Figure 4a , labeled as ‘enhanced Misr2’ ) . Total cell area was calculated by setting the masked area as the hematoxylin-stained region . RNAish dots/total area were auto-calculated by the same settings in 5 random 20X sections of human fetal tissues from 20 , 22 , and 37 weeks of gestation , n = 1 for each time point . Total RNA was extracted from the uteri of control and AAV9-MIS treated rats at different time points ( Figure 3—source data 3 ) using the Qiagen RNA extraction kit . For all the samples , cDNA was synthesized from 500 ng total uterine RNA using SuperScript III First-Strand Synthesis System for RT-PCR according to manufacturer’s instructions using random hexamers ( Invitrogen , # 18080–051 ) . The primers were designed to span the exon-exon junctions of the target genes ( see Figure 3—source data 3 for complete list of primers ) to avoid genomic DNA contamination . Expression levels relative to 18S ( for Acta2 ) and Gapdh ( for all the other genes analyzed ) were calculated by using cycle threshold ( Ct ) values logarithmically transformed using the 2−ΔCt function and the average value of the relative expression levels were normalized to PND three control set; and fold changes were calculated relative to PND three control time point with three technical replicates per sample . For Acta2 , Tgln , and Foxa2 expression levels were normalized to PND six control set . Sample sizes and p values for each gene and time point are listed in Figure 3—source data 3 . Newborn rats ( PND1 ) were injected with 5E10 particles of AAV9-MIS ( N = 3 ) or AAV9 empty particle controls ( N = 3 ) . On PND6 , rat pups were sacrificed and the uterine tissue was microdissected , taking care to exclude the oviduct , cervix , and ureter . Both uterine horns from each animals ( N = 3 per group ) were combined and placed in 5 mL of dissociation medium ( 82 mM Na2SO4 , 30 mM K2SO4 , 10 mM Glucose , 10 mM HEPES , and 5 mM MgCL2 - 6H2O , pH 7 . 4 ) containing 15 mg of Protease 23 ( Worthington ) , 100 U Papain with 5 mM L-Cysteine and 2 . 5 mM EDTA ( Worthington ) , and 1333 U of DNase 1 ( Worthington ) prewarmed to 34C . The samples were placed on a rocker at 34C for 15 min . The medium was then removed with a pipette and replaced with 5 mL chilled ( 4C ) stop medium ( dissociation medium containing 0 . 025% BSA ) supplemented with 0 . 5 mg trypsin Inhibitor and 0 . 5 mg ovomucoid protease Inhibitor . Samples were triturated 10 times with a 5 mL pipette , then 10 times with a 1000 µl micropipette , and finally filtered through a prewetted 100 µm filter . Filtered samples were spun at 1000 g for 10 min , and the cell pellet was resuspended in 1 ml of stop solution with 267 U of DNase . This step was repeated twice . Ten microliters was removed from each sample and combined with trypan blue to assess viability and concentration of cells . The cell mixture was spun a final time and resuspended in stop solution containing 20% Optiprep ( Sigma ) to a concentration of 150 , 000 cells/mL for inDrop sorting . Single-cell RNA sequencing ( inDrop ) inDrop microfluidic sorting was performed as previously described ( Hrvatin et al . , 2018 ) generating two libraries of approximately 5000 cells from each combined ( N = 3 rats ) cell suspensions ( control and MIS ) . Transcripts were processed as previously described and samples were sequenced on a NextSeq 500 ( Ilumina ) in a single combined lane ( Hrvatin et al . , 2018 ) . The analysis of the demultiplexed data was performed using the Seurat package in ‘R’ ( Butler et al . , 2018 ) . Filtering parameters were set to remove cells with fewer than 200 genes , and those with more than 3000 genes and/or 10 , 000 UMIs from the dataset . Dataset included 6811 control cells , and 2990 MIS treated cells which were jointly normalized . The Pearson correlation coefficient of UMI and nGene across the dataset was 0 . 97 . Principal component analysis was performed using 20 dimensions and FindClusters parameters were set at k . param = 30 , k . scale = 25 , and prune . SNN = 0 . 0667 with a resolution of 0 . 6 using 5346 variable genes . Data set and R codes are presented in Figure 3—source data 1 . For differential expression analysis of the myometrium and the epithelial cell clusters , data sets are presented in Figure 3—source data 2 and Figure 4—source data 2 . CellPhoneDB predicts signaling between cell clusters through analyzing co-expression of known Human receptors and secreted proteins . We downloaded ortholog information for Rat ( Rnor_6 . 0 ) and Human ( GRCh38 . p12 ) from ENSEMBL ( Release 95 ) . For non one-to-one orthologs , we assigned the maximum of gene expression values for all Rat to Human mappings . CellPhoneDB was subsequently run using default parameters ( Vento-Tormo et al . , 2018 ) . The resulting data set is presented in Figure 4—source data 1 . For serum MIS ELISA measurements of the control rats , two-way ANOVA analysis was used . For the serum ELISA measurements of the control and the treated rats on PND6 , the histomorphology , and the staining analyses , unpaired Student’s t test was used to compare the control and the treated samples using the Prism software ( Graphpad version 8 . 0 ) . p values are presented in Figure 2—source data 1 , and Figure 3—source data 3 , and Figure 5—source data 1 .	In the womb , mammals possess all of the preliminary sexual structures necessary to become either male or female . This includes the Mullerian duct , which develops into the Fallopian tubes , uterus , cervix , and vagina in female fetuses . In male fetuses , the testis secretes a hormone called Mullerian inhibiting substance ( MIS ) . This triggers the activity of a small group of cells , known as Misr2+ cells , that cause the Mullerian duct to degenerate , preventing males from developing female sexual organs . It was not clear what happens to Misr2+ cells in female fetuses or if they affect how the uterus develops . Saatcioglu et al . now show that in newborn female mice and rats , a type of Misr2+ cell that sits within a thin inner layer of the developing uterus still responds to MIS . At this time , the uterus is in a critical early period of development . Treating the mice and rats with MIS protein during their first six days of life eventually caused the Misr2+ cells to die . The treatment also prevented a layer of connective tissue , known as the endometrial stroma , from forming in the uterus . As a result , the mice and rats were infertile and had severely underdeveloped uteri . While the Misr2+ cells are present in newborn rats and mice , Saatcioglu et al . found that they disappeared before birth in humans . However , the overall results suggest that Misr2+ cells act as progenitor cells that develop into the cells of the endometrial stroma . Future work could investigate the roles these cells play in causing uterine developmental disorders and infertility disorders . Furthermore , the finding that MIS inhibits the Misr2+ cells could help researchers to develop treatments for uterine cancer and other conditions where the cells of the uterus grow and divide too much .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2019	Single-cell sequencing of neonatal uterus reveals an Misr2+ endometrial progenitor indispensable for fertility
T-box riboswitches are cis-regulatory RNA elements that regulate the expression of proteins involved in amino acid biosynthesis and transport by binding to specific tRNAs and sensing their aminoacylation state . While the T-box modular structural elements that recognize different parts of a tRNA have been identified , the kinetic trajectory describing how these interactions are established temporally remains unclear . Using smFRET , we demonstrate that tRNA binds to the riboswitch in two steps , first anticodon recognition followed by the sensing of the 3’ NCCA end , with the second step accompanied by a T-box riboswitch conformational change . Studies on site-specific mutants highlight that specific T-box structural elements drive the two-step binding process in a modular fashion . Our results set up a kinetic framework describing tRNA binding by T-box riboswitches , and suggest such binding mechanism is kinetically beneficial for efficient , co-transcriptional recognition of the cognate tRNA ligand . Riboswitches are cis-regulatory RNA elements that recognize and respond to defined external signals to affect transcription or translation of downstream messenger RNAs ( mRNAs ) ( Breaker , 2012; Serganov and Nudler , 2013; Sherwood and Henkin , 2016 ) . Riboswitches generally consist of two domains: a sensory or aptamer domain and a regulatory domain or expression platform . The expression platform can adopt different conformations in response to ligand binding to the aptamer , and in this way control gene expression outcome ( Breaker , 2012 ) . The aptamer of each riboswitch class contains conserved sequence motifs and unique secondary or tertiary structural elements that help distinguish and bind specific ligands ( McCown et al . , 2017 ) . Bacterial T-box riboswitches represent a unique class of riboswitches that do not bind small molecule ligands , instead they recognize and bind tRNA molecules and sense directly their aminoacylation state ( Zhang and Ferré-D'Amaré , 2015 ) . T-box riboswitches serve as excellent paradigms to understand RNA-RNA interactions and RNA-based regulation . T-box riboswitches are found in Gram-positive bacteria and are usually located in the region upstream of mRNA sequences encoding aminoacyl tRNA synthetases and proteins involved in amino acid biosynthesis and transport and hence participate directly in amino acid homeostasis ( Zhang and Ferré-D'Amaré , 2015 ) . In general , the aptamer domain of all T-box riboswitches contains a long stem , Stem I , responsible for specific tRNA binding ( Rollins et al . , 1997 ) . The expression platform can adopt either a terminator or anti-terminator conformation , depending on whether the bound tRNA is charged or uncharged ( Henkin , 2014; Zhang and Ferré-D'Amaré , 2015 ) . In most T-box riboswitches , binding of a charged tRNA to the T-box leads to rho independent transcription termination whereas an uncharged tRNA stabilizes the anti-terminator conformation and leads to transcription read-through ( Henkin , 2014; Zhang and Ferré-D'Amaré , 2015 ) . Whereas Stem I and the anti-terminator domain are highly conserved among T-box riboswitches , the region connecting them can vary . The Bacillus subtilis glyQS T-box riboswitch , involved in glycine regulation , represents one of the simplest T-box riboswitches ( Grundy et al . , 2002b ) in which only a short linker and a small stem , Stem III , connect Stem I and the anti-terminator domain ( Figure 1 ) . Recognition of tRNA by a T-box riboswitch involves three main structural elements of the tRNA: the anticodon region , the ‘elbow’ region formed by the conserved T- and D-loops , and the 3’ NCCA sequence ( Figure 1 ) . The anticodon and elbow regions of the tRNA interact with Stem I directly . Stem I contains several phylogenetically conserved structural motifs ( Rollins et al . , 1997 ) , including a K-turn motif , a specifier loop , a distal bulge , and an apical loop ( Rollins et al . , 1997 ) ( Figure 1 ) . Bioinformatics and structural analyses have collectively revealed the interactions between Stem I and the tRNA ( Grigg et al . , 2013; Lehmann et al . , 2013; Zhang and Ferré-D'Amaré , 2013 ) . Specifically , the co-crystal structures of Stem I/tRNA complexes show that Stem I flexes to follow closely the tRNA anticodon stem and interacts directly with the anticodon loop and the elbow through its proximal and distal ends , respectively ( Zhang and Ferré-D'Amaré , 2013 ) . The distal bulge and the apical loop fold into a compact structural module of interdigitated T-loops ( Chan et al . , 2013; Krasilnikov and Mondragón , 2003 ) , which interact directly with conserved unstacked nucleobases at the tRNA elbow ( Grigg et al . , 2013; Zhang and Ferré-D'Amaré , 2013 ) . In addition , the structures revealed that Stem I turns sharply around two hinge regions using a conserved dinucleotide bulge and the K-turn motif ( Grigg and Ke , 2013; Zhang and Ferré-D'Amaré , 2013 ) . Sensing of the aminoacylation state involves direct binding of the tRNA 3’ end to a highly conserved bulge in the T-box , the t-box sequence ( Grundy et al . , 2002a ) ( Figure 1 ) . A free NCCA end can base pair with the t-box sequence , enabling the anti-terminator conformation , whereas a charged NCCA end prevents the formation of the NCCA/t-box interactions , leading to the more stable terminator conformation ( Henkin , 2014; Zhang and Ferré-D'Amaré , 2015 ) . Importantly , discrimination between the charged and uncharged tRNA does not require any additional proteins , such as EF-Tu ( Suddala et al . , 2018; Zhang and Ferré-D'Amaré , 2014 ) , and is driven solely by RNA/RNA interactions . While Small Angle X-ray Scattering ( SAXS ) -derived models of the entire B . subtilis glyQS T-box riboswitch in complex with tRNA are available ( Chetnani and Mondragón , 2017; Fang et al . , 2017 ) , atomic-level structural details on the interactions between tRNA and the anti-terminator region are still lacking . In addition , there is a dearth of information on the kinetics of the binding process . Whereas it is clear that tRNA recognition involves several specific interactions , their binding temporal sequence remains elusive . In addition , it is unclear whether sensing of the 3’ end of the tRNA involves any additional conformational changes in the T-box . Here , by introducing donor-acceptor fluorophore pairs at several locations in the tRNA and the T-box riboswitch , and using single-molecule fluorescence resonance energy transfer ( smFRET ) , we demonstrate the temporal order of events in the trajectory of tRNA binding . Our results demonstrate that tRNA binds to the riboswitch in two steps , with its anticodon being recognized first , followed by NCCA binding accompanied with an inward motion of the 3’ region of the T-box riboswitch , including Stem III and the anti-terminator stem , relative to Stem I . In addition , by introducing mutations at different locations of the T-box , we further show that the two-step binding kinetics is regulated by the modular structural elements in the T-box riboswitch . To observe directly the binding of tRNA to the T-box , we placed the donor dye ( Cy3 ) on the 3’ end of a T-box fragment ( T-box182 ) , and the acceptor dye ( Cy5 ) on the 5’ end of the tRNAGly , where the subscript ‘182’ denotes the length of the T-box construct ( Figure 1A ) . In vitro transcribed and labeled T-box and tRNA were purified and refolded according to published procedures ( Chetnani and Mondragón , 2017; Zhang and Ferré-D'Amaré , 2013 ) ( Figure 1—figure supplement 1 ) . Labeling of the tRNA at the 5’ end had a modest effect on the binding affinity ( Figure 1—figure supplement 2 ) . T-box182 spans Stem I , the linker sequence , Stem III and the anti-terminator , but does not contain the terminator sequence , thereby preventing the transition to the terminator conformation . A short RNA extension sequence was added to the 5’ end of the T-box for surface immobilization ( Figure 2A , Supplementary file 1 ) . Single-molecule fluorescence images were recorded under equilibrium condition in the presence of 30 nM tRNAGly-Cy5 . Binding of tRNAGly-Cy5 results in a major distribution of FRET values around 0 . 7 , with 79 ± 4% of the traces showing a stable signal at 0 . 7 and 9 ± 5% traces sampling from 0 . 7 to 0 . 4 ( Figure 2B , C ) . The SAXS model ( Chetnani and Mondragón , 2017 ) predicts a distance between the labeling positions at the 3’ end of the T-box182 and the 5’ end of the tRNA to be around 52 Å . ( Figure 1B ) . Based on a Förster distance of 54–60 Å ( Ha et al . , 2002; Hohng et al . , 2004 ) , our measured FRET value is within the range of estimated FRET values ( 0 . 56–0 . 70 ) . Therefore , we assign the 0 . 7 FRET state to be the fully bound state of the tRNAGly by the T-box . In order to assign the 0 . 4 FRET value to specific tRNA binding states , tRNATyr-Cy5 and tRNAΔNCCA-Cy5 ( ‘ΔNCCA’ denotes a tRNAGly with a deleted 3’ NCCA sequence ) were flowed in the flow-chamber with pre-immobilized T-box182-Cy3 ( 3’ ) ( 3’ denotes that the label was added at the 3’ end ) . We did not observe any binding of tRNATyr-Cy5 ( Figure 2—figure supplement 1 ) , confirming that recognition of the anticodon by the specifier region is required for tRNA binding . In the presence of tRNAΔNCCA-Cy5 , we observed a fluctuating signal between 0 . 4 and 0 FRET ( Figure 2B , D ) , with a mean lifetime of the 0 . 4 FRET state of 3 . 6 ± 0 . 6 s and a mean waiting time before binding of 31 . 3 ± 5 . 3 s ( Figure 2—figure supplement 2B ) . Taken together with the results from the tRNAGly , tRNAΔNCCA and tRNATyr binding experiments , we assign the 0 . 4 FRET state to a partially bound state where only the anticodon interactions have been established . To further confirm the assignment of the FRET states , we generated T-box149 , where the anti-terminator sequence is truncated ( Figure 1A , Supplementary file 1 ) . Based on the structure model from the SAXS data ( Chetnani and Mondragón , 2017 ) we predicted that a Cy3 dye placed either at the end of Stem III ( T-box149 ) or at the end of the anti-terminator stem ( T-box182 ) are localized in close proximity in three dimensions , further confirmed by the distance measurement using smFRET ( Figure 2—figure supplement 3 ) . Therefore , we expect that if tRNAGly-Cy5 can reach the same fully bound state in T-box149 as in T-box182 , a high FRET state centered at 0 . 7 would be observed . However , using T-box149-Cy3 ( 3’ ) in combination with tRNAGly-Cy5 , we again observed transient binding of tRNAGly with a FRET value centered at ~0 . 4 with the same mean lifetime as observed with the T-box182-Cy3 ( 3’ ) and tRNAΔNCCA-Cy5 combination ( Figure 2B–D , Figure 2—figure supplement 2C ) . Therefore these two complexes ( T-box182 + tRNAΔNCCA and T-box149 + tRNAGly ) represent the same binding state of the tRNA , that is the state where binding of the anticodon to the specifier region has been established , but is unstable without the further interactions between the NCCA and the t-box region . Collectively , our results suggest a two-step binding model involving the separate establishment of the interactions with the anticodon and the NCCA . The fact that tRNATyr , which has a mismatched anticodon , but contains an intact NCCA 3’ end , does not show any binding activity suggests that interactions with the anticodon precede the interactions with the NCCA end of the tRNA and are necessary for the establishment of the NCCA contacts . Without the interaction between the NCCA and the t-box sequence the binding of tRNAGly is not stable . From the binding kinetics of tRNAΔNCCA , we estimated the association rate constant ( k1 ) and the disassociation rate constant ( k-1 ) for the first binding step to be ( 5 . 0 ± 1 . 6 ) x 105 M−1s−1 and 0 . 28 ± 0 . 05 s−1 , respectively ( Figure 5; Figure 6E ) . We classified smFRET traces for T-box182-Cy3 ( 3’ ) in complex with tRNAGly-Cy5 into three types ( Figure 2B ) : ( I ) traces stably sampling the 0 . 7 state ( 79 ± 4% of total traces ) , ( II ) traces transiently transitioning from the 0 . 7 state to the 0 . 4 state ( 9 ± 5% ) , and ( III ) traces only sampling the 0 . 4 state without reaching 0 . 7 state ( 12 ± 5% ) . The low percentage of Type III traces indicates that once the anticodon is recognized , the commitment to the next binding step , NCCA interactions , is high . The majority of the traces showed that the tRNAGly remained mostly in the fully bound state ( Type I ) until the fluorophore photobleached , with the actual lifetime limited by photobleaching ( τ0 . 7 > 24 s , where τ0 . 7 denotes the lifetime of the 0 . 7 FRET state ) ( Figure 2—figure supplement 2A ) . The observation that tRNAGly is able to transit from the fully bound state back to the partially bound state ( Type II ) suggests that the NCCA/t-box interaction can break occasionally ( Figure 2B ) . We estimated the lifetime of the transiently sampled partially bound state to be 0 . 35 ± 0 . 09 s in the presence of full-length tRNAGly ( Figure 2—figure supplement 2A ) , ~10 fold shorter than the partially bound state in the presence of tRNAΔNCCA . While the majority of the T-box molecules were already bound to tRNAGly before starting data acquisition , we could detect that some molecules show real-time binding during imaging acquisition . We observed only a few traces briefly sampling the 0 . 4 FRET state from the zero FRET ( unbound ) state before reaching the 0 . 7 FRET state , while most traces directly sampled the 0 . 7 FRET state without a detectable 0 . 4 FRET , likely due to our imaging time resolution ( 100 ms per frame ) . We post-synchronized the FRET traces at the transition point from the zero FRET state to the first sampled 0 . 4 FRET state , and plotted them in a time-evolved FRET histogram . From the time-evolved FRET histogram ( Figure 2F ) , we estimated roughly that the upper limit of the lifetime spent at the 0 . 4 FRET state is ~100 ms , very rapidly followed by establishment of NCCA/t-box interactions . In contrast , tRNAΔNCCA could not pass the 0 . 4 FRET state . To capture better real-time binding , we performed a flow experiment , where tRNAGly-Cy5 was flowed into a chamber with immobilized T-box182-Cy3 ( 3’ ) during imaging acquisition . The corresponding post-synchronized time-evolved FRET histogram again shows a fast transition into the fully bound state ( Figure 2—figure supplement 4 ) . In addition , the association rate constant of tRNAGly in the real-time flow experiment is ( 7 . 5 ± 0 . 7 ) x 105 M−1s−1 , consistent with the k1 of tRNAΔNCCA and confirming that the NCCA end of the tRNA does not participate in the first binding step . From the real-time binding kinetics of tRNAGly to T-box182 , we estimated a transition rate constant from the partially bound state to the fully bound state ( k2 ) to be on the order of 10 s−1 ( Figure 2F , Figure 2—figure supplement 4 ) . On the other hand , as transitions back to the partially bound state from the fully bound state were only observed in ~10% traces , we interpreted this to mean that the reverse transition rate constant ( k-2 ) is very small , and the second binding step in the wild-type ( WT ) T-box with uncharged tRNAGly is close to irreversible ( Figure 5; Figure 6E and see Discussion ) . We next investigated whether tRNA binding requires any conformational changes in the T-box itself . Using doubly labeled T-box182 , with Cy3 at the 3’ end and Cy5 at the 5’ hybridization extension , we observed a high FRET state ( centered at ~0 . 75 ) in the absence of tRNA ( Figure 3—figure supplement 1 ) . Based on the structural model ( Chetnani and Mondragón , 2017 ) , we estimated the distance between the 5’ and 3’ ends of the T-box182 to be ~36 Å ( Figure 1B ) . Our measured FRET value is slightly less than the predicted FRET value ( ~0 . 90 ) , likely due to the engineered 5’ extension sequence used to immobilize the T-box . No noticeable change was detected upon incubation with unlabeled tRNAGly ( Figure 3—figure supplement 1 ) , indicating that the 3’ portion ( Stem III plus the anti-terminator stem ) does not move away from the 5’ portion ( Stem I ) . Given that the measured FRET efficiency of 0 . 75 is already located beyond the FRET sensitive region , it is unlikely that any inward motion of the 3’ portion relative to the 5’ could be detected . To overcome this limitation , we added extensions at both the 3’ and 5’ ends ( Figure 3A , Supplementary file 1 ) . ITC experiments suggest that addition of a 5’ and/or a 3’ extension sequences to the T-box does not affect tRNA binding ( Figure 1—figure supplement 2 ) . With this intra-T-box FRET scheme , we observed a FRET shift from ~0 . 5 to~0 . 65 when tRNAGly was added ( Figure 3B ) , indicating that the 3’ half of the T-box moves closer to the 5’ half , potentially with the T-box becoming more compact due to the presence of the cognate tRNAGly . Adding non-cognate tRNAPhe or tRNAΔNCCA gave similar FRET values as the T-box alone ( Figure 3B ) , suggesting that the conformational change is associated with binding of both the anticodon and NCCA , not with anticodon recognition alone ( Figure 5 ) . Using the above two FRET pairs , we observed that the 3’ portion of the T-box moves towards the base of Stem I as well as the NCCA end of the tRNA during the second binding step . To ascertain whether the NCCA end of the tRNA also moves relative to the base of Stem I , we measured FRET between a Cy3 placed at the 5’ end of the T-box ( T-box182-Cy3 ( 5’ ) ) and tRNAGly-Cy5 ( Figure 4A ) . Using this FRET pair , binding of both tRNAGly and tRNAΔNCCA generated a similar FRET value centered at ~0 . 35 ( Figure 4B , C ) . However , the FRET traces behaved differently for these two tRNA molecules . For tRNAΔNCCA , the signal fluctuated between zero and 0 . 35 ( Figure 4D ) , with a lifetime of the 0 . 35 FRET state of 4 . 5 ± 1 . 0 s , reminiscent of the 0 . 4 FRET state using the tRNA/T-box182-Cy3 ( 3’ ) FRET pair ( Figure 4—figure supplement 1 ) . For tRNAGly , the signal was more stably centered at 0 . 35 ( Figure 4B ) . Since the tRNA-Cy5/T-box182–Cy3 ( 5’ ) FRET pair cannot distinguish the partially bound from the fully bound state , we fit the lifetime with a double-exponential decay . The fast dissociation fraction has a lifetime of 3 . 9 ± 0 . 7 s ( 46 ± 21% of population ) , consistent with the lifetime for the partially bound state , and the low dissociation fraction has a lifetime of 15 . 7 ± 0 . 8 s ( 54 ± 21% ) , representing the stable fully bound state ( Figure 4—figure supplement 1 ) . Overall , the measurements with the tRNA-Cy5/T-box–Cy3 ( 5’ ) FRET pair further validate the two-step binding model and reveal that the NCCA end of the uncharged tRNA maintains its relative position to the base of Stem I during the second binding step . The interdigitated T-loops structure formed by the interactions between the distal bulge and the apical loop at the distal end of Stem I has been shown to be important for tRNA binding ( Grigg et al . , 2013; Lehmann et al . , 2013; Zhang and Ferré-D'Amaré , 2013 ) . Specifically , C56 of the T-box stacks on a nucleobase in the D-loop of the tRNA , and a point mutation of C56 to U has been shown to reduce the tRNA binding affinity by ~40 fold ( Zhang and Ferré-D'Amaré , 2013 ) . We introduced the same mutation in the T-box182 backbone ( T-boxC56U ) ( Figure 6A , Supplementary file 1 ) . The smFRET trajectories for tRNAGly binding to T-boxC56U are overall similar to the trajectories for WT T-box182 , with a majority of traces ( 73 ± 6% of total traces ) showing stable binding at 0 . 7 FRET state , and 13 ± 4% of the traces showing transitions back to the 0 . 4 FRET state ( Figure 6C , D ) . τ0 . 7 was estimated to be at least ~23 s ( limited by the photobleaching of the fluorophore ) ( Figure 6E ) . Post-synchronized time-evolved histogram on the subset of traces that demonstrated real-time binding shows fast transition to the fully bound state ( Figure 6—figure supplement 1 ) . Comparison of tRNAGly binding to T-boxC56U and T-box182 suggest that the C56U mutation does not affect the second binding step . To investigate whether the mutation at the T-loop region affects the first binding step , we analyzed the binding and dissociation of tRNAΔNCCA-Cy5 to T-boxC56U-Cy3 ( 3’ ) . We found that the k1 of tRNA binding to T-boxC56U was roughly 20-fold slower compared to tRNA binding to T-box182 , and the dissociation was roughly 2 . 5-fold faster compared to T-box182 ( Figure 6E ) , leading to a ~ 50 fold higher dissociation constant for the first binding step . Our results suggest that the T-loop region of the T-box is critical during the first binding step , potentially aiding in anticodon recognition , but does not contribute significantly to the second binding step . The functional role of Stem III is unclear . It has been speculated that Stem III might serve as a transcription stalling site to allow co-transcriptional folding and regulation of the T-box riboswitch ( Grundy and Henkin , 2004; Zhang and Landick , 2016 ) . In addition , a SAXS data-derived model suggested coaxial stacking of Stem III and the anti-terminator stem , leading to a plausible role of Stem III in stabilizing the anti-terminator conformation in the presence of uncharged tRNAGly ( Chetnani and Mondragón , 2017 ) . To investigate the latter hypothesis , we generated a T-box mutant ( T-boxSIII-Δ4bp ) , in which four base pairs in Stem III are deleted to significantly shorten its length ( Figure 6A , Supplementary file 1 ) . smFRET studies using T-boxSIII-Δ4bp-Cy3 ( 3’ ) with tRNAΔNCCA-Cy5 and tRNAGly-Cy5 revealed insignificant difference in overall kinetics in the first and second step bindings ( Figure 6C–E , Figure 6—figure supplement 2 ) . Noticeably , the τ0 . 7 was around 50% shorter than that for the T-box182 ( Figure 6E ) , indicating that Stem III may contribute to the stabilization of the fully bound state , potentially through coaxial stacking with the anti-terminator stem , but the effect is minor . We next investigated the role of the K-turn in regulating tRNA binding kinetics . We disrupted the K-turn ( T-boxΔKT ) by changing the six bulged nucleotides to three nucleotides ( UCA ) to replace the K-turn with a three base pair stem ( Figure 6A , Supplementary file 1 ) . In contrast to binding of tRNAGly to T-box182 , binding to T-boxΔKT results in three FRET states centered on ~0 . 2 , 0 . 4 , and 0 . 7 . ( Figure 6B ) . While the exact boundary of each FRET state is difficult to determine accurately from the FRET histogram ( Figure 6C ) , a transition density plot ( TDP ) clearly reveals interconversion between the 0 . 2 , 0 . 4 , and 0 . 7 states ( Figure 6D ) , with transitions between the 0 . 2 and 0 . 4 FRET states , and between the 0 . 4 and 0 . 7 FRET states more populated . Binding of tRNAΔNCCA-Cy5 to T-boxΔKT , on the other hand , leads to the loss of population of the 0 . 7 state; however , both the 0 . 2 and 0 . 4 FRET states and fluctuations between these two states are frequently sampled ( Figure 6—figure supplement 3A , B ) . Comparing the tRNAGly and tRNAΔNCCA binding , we speculate that both the 0 . 2 and 0 . 4 FRET states observed in the case of the T-boxΔKT represent the partially bound state in which only the anticodon and elbow are recognized . In contrast to T-box182 , T-boxΔKT , with the K-turn replaced by an extension of Stem I , could potentially favor a relaxed conformation of Stem I , as observed in the NMR structure of an isolated K-turn and specifier loop domain ( Wang and Nikonowicz , 2011 ) , generating a lower FRET value centered at 0 . 2 . However , the sampling of the 0 . 4 FRET state in the T-boxΔKT construct suggests that the interactions between the specifier and anticodon of the tRNA may transiently force open the extended base pair region and bend the T-box to adopt a similar conformation to the one observed in T-box182 . The lifetimes of the 0 . 4 state before transition back to the 0 . 2 state and before transition forward to the 0 . 7 state are 0 . 30 ± 0 . 03 s and 0 . 13 ± 0 . 05 s , respectively ( Figure 6—figure supplement 3C ) , indicating that this forced bent state is energetically unfavorable . However , this 0 . 4 FRET state is very likely to be required for the NCCA/t-box interaction to occur , as in the presence of tRNAGly the transitions from the 0 . 2 FRET to the 0 . 7 FRET state often pass through the 0 . 4 FRET state ( Figure 6B ) . Furthermore , we observed a small region in the TDP corresponding to direct transitions between the 0 . 2 and 0 . 7 states . Given the very short lifetime of the 0 . 4 state , which is close to the time resolution of our experiments , the 0 . 2 to 0 . 7 state transition is likely to represent populations whose 0 . 4 FRET state lifetime is even shorter than the time resolution of the experiment . The 3-step kinetic scheme for tRNA binding to T-boxΔKT is presented in Figure 6—figure supplement 3C . The association rate constant k1 ( ( 5 . 4 ± 2 . 9 ) x104 M−1s−1 ) , estimated from binding of tRNAΔNCCA to the T-boxΔKT , is ~9 fold smaller than binding to T-box182 , suggesting that disruption of the K-turn affects anticodon recognition . Considering both the 0 . 2 and 0 . 4 FRET state as the partially bound state in T-boxΔKT , the apparent dissociation rate from the partially bound state k-1_app of tRNAΔNCCA to T-boxΔKT is 0 . 21 ± 0 . 06 s−1 , similar to that for T-box182 , suggesting that disruption of the K-turn does not affect the stability of the partially bound state . The 0 . 7 FRET state observed for T-boxΔKT in the presence of tRNAGly is consistent with the FRET value for the fully bound state in the WT T-box , indicating that the NCCA/t-box interactions in T-boxΔKT can still be formed . However , tRNAGly bound to T-boxΔKT ( τ0 . 7 = 1 . 6±0 . 3 s ) is at least 15-fold less stable compared to tRNAGly bound to T-box182 ( τ0 . 7 >24 s ) . Furthermore , in contrast to the signal observed for tRNAGly binding to T-box182 , in which fewer than 10% of the FRET traces show transitions back to 0 . 4 FRET , in T-boxΔKT the vast majority of the traces show backward transitions to the 0 . 4 and 0 . 2 FRET states , contributing largely to the instability of the fully bound state . Based on the transition rates between 0 . 2 , 0 . 4 , and 0 . 7 , we estimated the apparent forward ( k2_app ) and reverse ( k-2_app ) transition rates between the partially bound and the fully bound state of the T-boxΔKT to be 1 . 3 ± 0 . 3 s−1 and 1 . 0 ± 0 . 3 s−1 , respectively ( Figure 6E , Figure 6—figure supplement 3C ) . The dramatically reduced k2_app and increased k-2_app in T-boxΔKT leads to a ~ 150 fold change in the equilibrium constant in the second binding step compared to T-box182 , implying that inflexibility of the K-turn region largely inhibits the conformational change in the T-box required to form the NCCA/t-box interaction , and strongly destabilizes the fully bound state . T-box riboswitches represent a unique class of riboswitches as they recognize a macromolecule and require interactions at multiple spatially separated sites on the ligand , unlike other riboswitches that respond to the binding of small ligands . Previous structural studies suggested that tRNA recognition by a T-box riboswitch is a bipartite process ( Grigg and Ke , 2013; Zhang and Ferré-D'Amaré , 2013 ) with Stem I largely responsible for discriminating non-cognate tRNAs while the t-box sequence in the expression platform senses the charged state of the tRNA . We used smFRET to elucidate the binding kinetics of tRNAGly by the glyQS T-box riboswitch . With three FRET pairs between different T-box riboswitch and tRNA ligand constructs , our data collectively reveal a two-step model of uncharged tRNAGly binding to the glyQS T-box riboswitch ( Figure 5 ) . The first binding step involves recognition of the anticodon of the tRNA by the specifier sequence located in Stem I of the T-box riboswitch , leading to a partially bound state . In the second step , the 3’ end of the T-box docks into the NCCA end of the tRNA through interactions with the t-box sequence , which leads to a fully bound state . Without the NCCA interaction , the binding of tRNA is unstable , with a mean lifetime of ~4 s , whereas with interactions both with the anticodon and the NCCA end , the binding of tRNA is very stable , with a mean lifetime >24 s . It should be noted that the latter lifetime is likely to be much longer for two reasons: first , the measurement is limited by fluorophore photobleaching , i . e . the loss of signal is more likely to be due to photobleaching rather than the actual tRNA dissociation; second , the Cy5 label placed at the 5’ end of the tRNA reduces the overall binding affinity by ~4 fold ( Figure 1—figure supplement 2 ) , likely because the fluorophore impairs the NCCA/t-box interactions to some extent . In addition , an intra-T-box FRET pair at the 5’ and 3’ ends of the T-box demonstrates that , while the T-box is largely pre-organized in a folded state before tRNA binding , it still exhibits conformational rearrangement in a tRNA-dependent manner . Specifically , the 3’ half of the T-box ( including Stem III and the anti-terminator ) moves inward relative to the 5’ half ( Stem I ) of the T-box to accommodate the interaction with the NCCA end in the second binding step . While our manuscript was in preparation , Suddala et al . ( Suddala et al . , 2018 ) reported a related single-molecule study on tRNA binding to the glyQS T-box riboswitch and proposed a similar two-step binding model . By using colocalized signals from the bound tRNA and the immobilized T-box , Suddala et al . ( Suddala et al . , 2018 ) uncovered two binding states distinguished by different dissociation rates of the tRNA , aided by using a Stem I-only mutant that cannot interact with the NCCA end of the tRNA . Specifically , in the model of Suddala et al . ( Suddala et al . , 2018 ) binding of the anticodon of the uncharged tRNA results in a relatively unstable state ( with a lifetime of ~4–5 s ) , consistent with the partially bound state in our model , while binding both the anticodon and the NCCA end of the tRNA results in a stable state , consistent with the fully bound state in our model . Although both studies propose consistent kinetic models , in the study by Suddala et al . ( Suddala et al . , 2018 ) , the FRET pair dyes are attached at the variable loop of the tRNA and the 3’ or 5’ end of the glyQS T-box , and do not generate different FRET signals to distinguish the partially bound state from the fully bound state . Therefore neither transitions between the partially bound and the fully bound states , nor the order of events during tRNA binding can be resolved in the study ( Suddala et al . , 2018 ) . In contrast , by employing a FRET pair located at the 5’ end of the tRNA and the 3’ end of the glyQS T-box , we observed directly two FRET states corresponding to the recognition of the anticodon ( 0 . 4 FRET ) and the binding of the NCCA ( 0 . 7 FRET ) , therefore our study allows the discrimination between different states and generates a more complete kinetic framework describing the full trajectory of the tRNA binding process . Our data reveal that anticodon recognition precedes the NCCA end interactions , and that after anticodon recognition the commitment to further establishment of the NCCA/t-box interaction is high . The T-box/tRNAGly complex transits rapidly from the partially bound state to the fully bound state , with a rate constant ( k2 ) on the order of 10 s−1 . Interestingly , our data also reveal that in the fully bound state the NCCA/t-box interaction is not highly-stable or ultra-stable . Brief disruption of the NCCA/t-box interactions can occur , but transition back to the fully bound state is rapid , ~10 fold faster than dissociation of the tRNA from the partially bound state . Therefore tRNA can remain bound during the breaking and reforming of the NCCA/t-box interaction . However , such transient breaking of NCCA/t-box interaction was only observed in ~10% of the total population , suggesting that the reverse transition rate constant ( k-2 ) is overall very small . Two-step binding mechanisms have been observed in a variety of protein or nucleic acid mediated biological process , including T-cell receptor ( TCR ) recognition of the major histocompatibility complex ( MHC ) presenting peptides , where TCRs scan the MHC scaffold first , followed by sensing of specific MHC-presenting peptides ( Wu et al . , 2002 ) ; interaction of the signal recognition particle ( SRP ) receptor with the membrane , where SRP receptors interact with the membrane in a dynamic mode followed by an SRP-induced conformational transition into a stable binding mode ( Hwang Fu et al . , 2017 ) , DNA interrogation by CRISPR Cas9-crRNA , where Cas9-crRNA recognizes the protospacer adjacent motif ( PAM ) on the target DNA followed by sensing of the spacer sequence and triggering R-loop formation ( Sternberg et al . , 2014 ) ; and RNA-induced silencing complexes ( RISCs ) binding to their mRNA targets , where dynamic sampling of the ‘sub-seed’ region occurs before targeting stably across the full seed region ( Chandradoss et al . , 2015; Herzog and Ameres , 2015; Salomon et al . , 2015; Salomon et al . , 2016 ) . In all of these cases , a two-step binding mechanism provides a good balance between sensitivity and specificity . T-box riboswitches fold and function co-transcriptionally in the cell , and hence we propose that a two-step kinetic model is kinetically beneficial during co-transcriptional folding of the T-box and sensing of the tRNA ligand . First , considering that the intracellular concentration of tRNA is on the order of µM in bacteria ( Avcilar-Kucukgoze et al . , 2016; Dong et al . , 1996; Emilsson and Kurland , 1990 ) , the binding of the tRNA ligand to the T-box riboswitch is rate limited by the first step . Pausing at Stem III in the presence of NusG in vitro was estimated to last 30–60 s ( Grundy and Henkin , 2004 ) . Assuming an in vivo concentration of tRNA ~1 μM , the apparent tRNA binding rate in the first step is ~0 . 5 s−1 , which is sufficient for the uncharged tRNA to reach the partially bound state during the pausing . Second , bacterial RNA polymerase ( RNAP ) elongates at a rate of 40–50 nt/s ( Mosteller and Yanofsky , 1970; Vogel and Jensen , 1994 ) , and the length between terminator and anti-terminator sequences is ~40 nts; therefore the decision between termination and anti-termination needs to be made within ~0 . 5–1 s . On the one hand , the relatively slow dissociation ( k-1 ) from the partially bound state can ensure the tRNA ligand stays bound until the completion of the anti-terminator sequence . On the other hand , very rapid transition into the fully bound state ( k2 ) helps secure the interaction between the NCCA end of the tRNA and the t-box sequence and trap the T-box in the anti-terminator conformation before completion of transcription of the terminator sequence , which is immediately downstream . Finally , the fully bound state has a lifetime of at least 24 s , allowing RNAP to elongate more than 1000 nts before the tRNA dissociates . Therefore , the second binding step in the WT T-box with uncharged tRNAGly is close to irreversible in such biological setting . It is worth mentioning that Suddala et al . ( Suddala et al . , 2018 ) revealed that the uncharged T-box/tRNAGly complex exists in two populations: a stable complex , with a tRNA dissociation rate of ~0 . 03 s−1 , and an ultra-stable complex , with a tRNA dissociation rate <2 . 4×10−4 s−1 . Our estimation of the lifetime of the fully bound state was limited by the introduction of Cy5 to the 5’ end of the tRNA and fluorophore photobleaching , and therefore our experiments cannot distinguish between the stable and ultra-stable complexes . Nevertheless , both potential configurations in the fully bound state are stable enough to allow transcription read-through . Based on the known structures of Stem I in complex with tRNAGly ( Grigg and Ke , 2013; Zhang and Ferré-D'Amaré , 2013 ) , it was hypothesized that the intra-T-box conformational change is likely to involve the K-turn region , which sits at the junction between the 5’ and 3’ portions of the T-box . Our observation that the NCCA end of the uncharged tRNA moves towards the anti-terminator stem ( revealed by the T-box182-Cy3 ( 3’ ) and tRNAGly-Cy5 FRET ) , but maintains its relative position to the K-turn region ( revealed by the T-box182-Cy3 ( 5’ ) and tRNAGly-Cy5 FRET ) during the second binding step is consistent with the role of the K-turn region as the hinge of the intra-T-box conformational change . In addition , such conformational changes mediated by the K-turn region are Mg2+ dependent , as suggested by Suddala et al . ( Suddala et al . , 2018 ) . Importantly , our results hint at the critical role of the K-turn region in promoting the fast coordination between anticodon sensing and locking of the NCCA end . Crystal structures of the Stem I/tRNA complex ( Grigg and Ke , 2013; Zhang and Ferré-D'Amaré , 2013 ) show that Stem I flexes around the K-turn region and that this flexing seems to be important to establish the interactions between the anticodon and specifier sequence . Using a mutant where the K-turn region is removed , our data show that in the absence of the K-turn motif both the first and second binding steps are affected , but with a much more dramatic effect on the second step . In this specific K-turn mutant , where the K-turn is replaced by an extended Stem I helix , the conformational change in the T-box that brings closer the 3’ and 5’ portions becomes highly energetically unfavorable , leading to a ~ 10 fold reduction in k2 , and ~15 fold destabilization of the fully bound state . Our kinetic measurement of the K-turn mutant explains the in vivo loss-of-function K-turn mutants ( Winkler et al . , 2001 ) , and emphasizes the importance of the flexing around the K-turn region and the associated conformational changes in the T-box itself in the overall recognition and binding process . Our results with various T-box mutants demonstrate that the T-box structural elements involved in tRNA recognition can drive the two-step binding process in a modular fashion . For example , in the first binding step , interactions with Stem I involve both anticodon recognition and the contacts between the interdigitated T-loops in Stem I and the elbow region ( D- and T-loops ) of the tRNA ( Grigg and Ke , 2013; Zhang and Ferré-D'Amaré , 2013 ) . When introducing a point mutation that impairs the T-loops/elbow region interaction ( Zhang and Ferré-D'Amaré , 2013 ) , we observed a dramatic decrease in the association rate constant and a moderate increase in the dissociation rate constant , leading to an overall ~50 fold reduction on the binding affinity for the first step , however the second step is unaffected . This observation suggests that establishment of the T-loops/elbow interactions is an important part of the Stem I/tRNA recognition process in the first binding step , but does not play any role in NCCA recognition in the second binding step . Similarly , truncation of Stem III has a minor effect on the stability of the fully bound state , but with no influence on the first binding step . Finally , the K-turn region ( discussed above ) , which links the 5’ and 3’ portions of the T-box , plays a key role in coordinating the two binding steps by providing structural flexibility . It is worth mentioning that the impact on the tRNA binding kinetics by mutations in the T-loops , the K-turn , and the Stem III are consistent with the sequence conservation and the in vivo impact on amino acid-mediated transcription read-through using tyrS T-box riboswitches as a model system ( Henkin , 2014; Rollins et al . , 1997; Winkler et al . , 2001 ) . Mutations in the highly conserved T-loop and K-turn region have a more dramatic influence on the tRNA binding kinetics , translating into a larger in vivo impact . While the less conserved Stem III contributes to the stabilization of the anti-terminator conformation in vitro , deletion of it does not appear to significantly impair the tRNA binding process in vitro . Potentially its major function is to create a pause site to coordinate with the co-transcriptional folding of the T-box ( Grundy and Henkin , 2004; Zhang and Landick , 2016 ) . In conclusion , our study provides a comprehensive kinetic framework for describing tRNA recognition by the T-box riboswitch . The two-step binding process is driven by the specific structural elements of the T-box , and is likely to be kinetically beneficial for efficient , co-transcriptional recognition of the cognate tRNA ligand . Specific T-box structural elements drive the two-step binding process in a modular fashion , that is the 5’ and 3’ portions of the T-box are responsible for the first and second binding steps , respectively , with the K-turn region coordinating the two binding steps by allowing structural flexibility , providing a guideline for synthetic biology design of RNA regulatory modules , as well as for the development of new antibiotics using critical T-box structural elements as potential targets . Finally , the glyQS T-box riboswitch represents one of the simplest members of this class of riboswitches . Other T-box riboswitches are larger and have additional structural elements and can even appear in tandem arrangements ( Gutiérrez-Preciado et al . , 2009 ) . While the two-step binding kinetics may be common to all T-box riboswitches , it is likely that the process in other T-box riboswitches shows differences modulated by the additional structural elements . RNA transcription was performed in vitro using His6-tagged T7 RNA polymerase using standard protocols ( Milligan et al . , 1987 ) . Cloning , design of a bicistronic DNA template encoding the B . subtilis glyQS T-box riboswitch and its cognate tRNAGly , and conditions for in vitro transcription were described before ( Chetnani and Mondragón , 2017 ) . For the experiments , all the RNAs from the crude transcription reaction were purified on a 7 . 5% denaturing ( 8 M Urea ) polyacrylamide gel . The RNAs of interest were located on the gel by UV shadowing , the bands were cut out , and the RNAs were eluted into 50 mM sodium acetate ( pH 7 . 0 ) buffer containing 200 mM potassium chloride by overnight rocking at 4°C . The eluted RNAs were precipitated by adding 3 volumes of cold 100% ethanol and stored overnight at −20°C . The precipitated RNAs were pelleted by centrifugation for 15 min at 20 , 000 g . The RNA pellets were washed three times in cold 80% ethanol , dried in a Speedvac and re-suspended in water . The concentration was estimated by its absorbance at 260 nm and was kept frozen at −20°C for long term storage . All mutant constructs were made using a commercial site directed mutagenesis protocol ( Kunkel , 1985 ) ( Quikchange , Stratagene ) and the RNAs were produced and purified by using the same protocol described above . The sequence of all mutant T-box and tRNA constructs were confirmed by sequencing and validated by MFOLD ( Zuker , 2003 ) to ensure that the secondary structural elements were not affected by the mutations . For smFRET experiments , end labelling of RNA molecules was performed by modifying standard labelling protocols ( Rinaldi et al . , 2015 ) . For 3’ end labelling , 50 µg of RNA in 50 µL of reaction volume was incubated with 0 . 1 M Na-periodate in 0 . 1 M Na-acetate buffer at pH 5 . 2 for 90 min in the dark . The reaction was quenched by adding 5 µL of 2 . 5 M KCl and incubating on ice for 10 min . The resultant insoluble KIO4 was removed by centrifugation at 20 , 000 g for 30 min and the supernatant was passed through a P6 column ( Bio-Rad ) to exchange the buffer to 0 . 1 M HEPES ( pH 7 . 0 ) , 40% DMSO . The RNA was incubated with Cy3 hydrazide ( Lumiprobe ) dye for 45 min with a final RNA/dye ratio of ~1:200 . The RNA was then ethanol precipitated as described above . The precipitated RNA was pelleted by centrifugation for 15 min at 20 , 000 g , dried in a Speedvac and re-suspended in water . The final RNA solution was passed through a P6 column to remove any residual free dye . 5’ end labelling of RNA was performed by N- ( 3-Dimethylaminopropyl ) -N′-ethylcarbodiimide hydrochloride ( EDC ) – N-hydroxysuccinimide ( NHS ) coupling through activation of the 5’ monophosphate of the RNA by EDC and imidazole ( Rinaldi et al . , 2015 ) . To improve the overall labelling efficiency of this method , a modified approach was used in which the 5’ triphosphate of 100 µg of RNA was converted to 5’ monophosphate in 100 µL reaction volume by incubating it with 100 units of RNA 5' Pyrophosphohydrolase ( NEB ) at 37°C for 1 hr . The enzyme was removed by phenol chloroform extraction and the supernatant was passed through a P6 column to exchange the buffer to 10 mM HEPES ( pH 7 . 0 ) , 150 mM NaCl , 10 mM EDTA . This was followed by addition of 12 . 5 mg of EDC to the RNA solution along with 50 µL of ethylene diamine and 200 µL of 0 . 1 M imidazole buffer ( pH 6 . 0 ) . The reaction was incubated for 3 hr at 37°C and the RNA was then ethanol precipitated as described above . The resultant RNA pellet was re-suspended in 0 . 1 M sodium carbonate buffer ( pH 8 . 7 ) and residual EDC was removed by passing the solution through a P6 column . The resultant RNA solution in 0 . 1 M sodium carbonate buffer ( pH 8 . 7 ) was incubated with Cy5 NHS ( Lumiprobe ) dye for 45 min with a final RNA/dye ratio of ~1:200 . The RNA was ethanol precipitated as described above and re-suspended in water . The final RNA solution was passed through a P6 column to remove any residual free dye . The overall labelling efficiency of RNA constructs used in this study varied from 75% to 95% . The thermodynamic parameters associated with binding of tRNAGly to glyQS T-box were determined at 25°C by using an ITC-200 Micro-Calorimeter ( MicroCal ) . Prior to the experiment , the two interacting RNAs ( in 50 mM HEPES pH 7 . 0 , 100 mM KCl ) were refolded separately by first heating for 3 min at 90o C followed by incubation on ice for 2 min . At this point , MgCl2 was added to attain a final concentration of either 1 mM or 10 mM . The RNA solution was then heated to 50o C for 10 min and 37o C for 30 min followed by cooling to 25°C . The refolded RNA was then concentrated using a 10 kDa cutoff Amicon filter and washed three times with the final ITC buffer ( 50 mM HEPES pH 7 . 0 , 100 mM KCl and 10 mM MgCl2 or 1 mM MgCl2 ) . For testing the binding of T-box with unlabeled and 5’ Cy5 labelled tRNA in 10 mM MgCl2 buffer , the sample cell was filled with 6 . 6 µM and 11 µM T-box respectively and the corresponding concentration of tRNAGly in the syringe was 111 µM and 152 µM . For the ITC experiment in 1 mM MgCl2 buffer , the cell was filled with T-box at a concentration of 6 . 7 µM and the syringe concentration of unlabeled tRNAGly was 72 µM . For each ITC experiment , the titration was carried out by stepwise ( 2 µL ) injection of tRNAGly from a stirred syringe ( 1000 rev/min ) into the sample cell . Successive injections were spaced by 150 s and values for the change in enthalpy ( ΔHb ) , association constant ( Kb ) , change in enthalpy ( ΔHb ) and stoichiometry ( n ) were determined by nonlinear least-squares fitting of the data using Origin 5 . 0 software ( OriginLab ) . The T-box and tRNA were denatured and refolded using the same protocol as described for the ITC and smFRET experiments . For binding of tRNA , 5 µL of 4 µM of folded T-box and 5 µL of 2 µM folded tRNA samples were mixed together for 30 min in final buffer containing 50 mM HEPES pH 7 . 0 , 100 mM KCl and 15 mM MgCl2 . The folded T-box or T-box +tRNA mixture were loaded on to 6% native polyacrylamide gel containing 15 mM MgCl2 and the gel was run at room temperature for 2–2 . 5 hr . RNAs were stained by SYBRTM green RNA staining dye ( Invitrogen ) . The gel was imaged using ChemiDoc ( Bio-rad ) in SYBR green channel ( for unlabeled T-box constructs ) , Cy5 channel ( for tRNA-Cy5 ) , or Cy3 channel ( for labeled T-box constructs ) . Images were processed and analyzed by ImageJ ( Schneider et al . , 2012 ) . DNA oligos that hybridize to 5’ extension and 3’ extension of the T-box construct were purchased from Integrated DNA Technologies with an amine modification at the 5’ end and 3’ end respectively . 13 . 5 µL of 100 µM DNA oligo was mixed with 1 . 5 µL of 1 M NaHCO3 ( pH 8 . 6 ) . 25 µg of NHS conjugated fluorophore ( Cy3 or Cy5 ) was dissolved with 0 . 5 µL DMSO and mixed with the DNA oligo solution . The mixture was incubated at 37°C overnight . 1 . 67 µL of 3 M NaOAc and 50 µL of pure ethanol was added to the mixture to precipitate the conjugated DNA oligo overnight at −20°C . The precipitated DNA oligo was pelleted by centrifugation for 30 min at 21000 g and re-suspended with 40 µL water . The DNA solution was passed through a P6 column to remove any residue free dye and salt . The overall fluorophore labeling efficiency is ~60% . Slides containing microfluidic channels were prepared as previously described ( Blanchard et al . , 2004 ) . Slides and coverslips were coated with a mixture of poly-ethylene glycol ( PEG , Mw = 500 , 000 ) and PEG-biotin ( Mw = 500 , 000 ) according to previously published protocol ( Blanchard et al . , 2004 ) . The T-box and tRNA were denatured and refolded using the same protocol described for the ITC experiments with the final buffer containing 50 mM HEPES pH 7 . 0 , 100 mM KCl , and 15 mM MgCl2 . T-box RNA was hybridized to the biotinylated DNA oligo at the extension during refolding , and immobilized via biotin-streptavidin interactions to the surface . tRNA was diluted in imaging buffer ( 50 mM HEPES pH 7 . 0 , 100 mM KCl , 15 mM MgCl2 , 5 mM protocatechuic acid ( PCA ) ( Sigma ) , 160 nM protocatechuate-3 , 4-dioxygenase ( PCD ) ( Sigma ) , and 2 mM Trolox ( Sigma ) ) and flowed into the microfluidic channels . smFRET measurements were performed with an objective based total internal reflection fluorescence ( TIRF ) microscope based on a Nikon Ti-E with 100X NA 1 . 49 CFI HP TIRF objective ( Nikon ) . A 561 nm laser ( Coherent Obis at a power density of 4 . 07 × 105 W/cm2 ) was used for the FRET measurement . A 647 nm laser ( Cobolt MLD at a power density of 5 . 88 × 105 W/cm2 ) was used for direct excitation of Cy5 to check the presence of the acceptor . Emissions from both donor and acceptor were passed through an emission splitter ( OptoSplit III , Cairn ) , and collected at different locations on an EMCCD ( iXon Ultra 888 , Andor ) . 1500 frames of time-lapse images were taken with 100 ms exposure time . Each independent measurement in our study should be considered as a technical replicate . The biological samples ( i . e . the in vitro transcribed tRNAs and T-boxes ) were generated once . In each measurement , T-box and tRNA were folded , smFRET images were recorded , and analysis were performed independently . Individual spots were picked from maximum intensity projection of Cy5 emission channel ( Figure 2—figure supplement 1 ) using NIS Elements software , that is only the pixels with the highest intensity value of the same XY coordinates in time-lapse images are displayed in the maximum intensity projection image . Fluorescent intensity trajectories of these spots were generated from Cy3 and Cy5 channels of time-lapse images , and corrected for baseline and bleed-through in MATLAB as previously described ( Fei et al . , 2008 ) . FRET traces were generated by calculating ICy5 / ( ICy5+ICy3 ) at each time point from the intensity trajectories . smFRET traces were idealized by fitting with a Hidden Markov Model using vbFRET ( Bronson et al . , 2009 ) . Dwell time of each FRET state before transition to another FRET state of individual traces was extracted from the idealized traces and the dwell time histograms were fit with exponential decay by Origin 7 . 0 ( OriginLab ) ( Figure 2—figure supplement 2 ) . In most of the cases , dwell time histograms fit well with a single exponential decay ( A . exp ( -t/t0 ) + y0 ) . In a few cases , in which mixed populations with different lifetimes were expected , dwell time histograms were fit with a double exponential decay ( A1 . exp ( -t/t1 ) + A2 . exp ( -t/t2 ) + y0 ) and the population-weighted average lifetime was calculated by ( A1 . t1 +A2 . t2 ) / ( A1 + A2 ) . The latter cases are explicitly mentioned in the text .	Living organisms depend upon a group of chemicals called amino acids to survive . Amino acids are the building blocks of proteins , and proteins have many important roles within and around cells . Bacteria regulate certain genes to ensure they have the right balance of different amino acids to survive . By controlling the availability of certain proteins that help them to make or collect certain amino acids , bacteria can control their overall amino acid balance . Before a protein is made , a molecular machine called RNA polymerase must first copy the information in a gene to make a molecule called a messenger RNA ( mRNA ) . The mRNA is then translated to make the protein from individual amino acids . In this process , each amino acid needs to be first attached to another molecule called a transfer RNA ( tRNA ) . In many bacteria species , the mRNAs involved in making or transporting amino acids contain structures called T-boxes . These structures guide the RNA polymerase to make more of the mRNAs when the levels of the amino acid become too low . A T-box , however , does not sense the level of the amino acid directly . Instead it senses the number of tRNA molecules that do not carry an amino acid . Zhang , Chetnani et al . examined a particular T-box interacting with tRNA using pairs of fluorescent dyes to detect distances between molecules . The T-box first recognizes a part of the tRNA called the anticodon to make sure it binds the correct type of tRNA . It then changes its shape to detect whether the tRNA is attached to an amino acid . This two-step process is driven by multiple structural elements within the T-box , and the flexibility of the T-box plays a critical role . A cell’s survival depends on it keeping amino acid levels under control . Understanding how bacteria do this could lead to new antibiotic drugs that target the T-box to kill cells . This study also provides insights into the workings of mRNA components like T-boxes – a type of riboswitch – which is an unusual means of controlling gene activity .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2018	Specific structural elements of the T-box riboswitch drive the two-step binding of the tRNA ligand
Protein kinases are major drug targets , but the development of highly-selective inhibitors has been challenging due to the similarity of their active sites . The observation of distinct structural states of the fully-conserved Asp-Phe-Gly ( DFG ) loop has put the concept of conformational selection for the DFG-state at the center of kinase drug discovery . Recently , it was shown that Gleevec selectivity for the Tyr-kinase Abl was instead rooted in conformational changes after drug binding . Here , we investigate whether protein dynamics after binding is a more general paradigm for drug selectivity by characterizing the binding of several approved drugs to the Ser/Thr-kinase Aurora A . Using a combination of biophysical techniques , we propose a universal drug-binding mechanism , that rationalizes selectivity , affinity and long on-target residence time for kinase inhibitors . These new concepts , where protein dynamics in the drug-bound state plays the crucial role , can be applied to inhibitor design of targets outside the kinome . Protein kinases have become the number one drug target of the 21th century ( Cohen , 2002; Hopkins and Groom , 2002 ) , due to their central role in cellular processes and involvement in various types of cancer ( Carvajal et al . , 2006; Gautschi et al . , 2008; Katayama and Sen , 2010 ) . Despite their therapeutic significance , the development of specific kinase inhibitors proves to be extremely challenging because they must discriminate between the very similar active sites of a large number of kinases in human cells . One of the biggest success stories is Gleevec: a highly selective drug that specifically targets Abl kinase , providing an efficient treatment of chronic myelogenous leukemia ( CML ) and minimizing side effects ( Iqbal and Iqbal , 2014 ) . Despite being a multi-billion-dollar cancer drug , the mechanism responsible for its impressive selectivity has been elusive until recently . Seminal work by the Kuriyan lab demonstrated that Gleevec can only bind to an inactive DFG ( for Asp-Phe-Gly ) loop conformation in the ‘out-conformation’ due to steric clash of the active , DFG-in conformation ( Nagar et al . , 2002; Schindler et al . , 2000; Seeliger et al . , 2007 ) . Since then it has long been proposed that the conformational state of the fully-conserved DFG loop ( Taylor et al . , 2012 ) dictates the selectivity for Gleevec and other kinase inhibitors ( Lovera et al . , 2012; Nagar et al . , 2002; Schindler et al . , 2000; Treiber and Shah , 2013; Xu et al . , 1997 ) . The orientation of the DFG-motif and its possible steric clashes is indeed important for the ability of a class of inhibitors to bind to the kinase , but proved insufficient to explain drug selectivity and affinity . Earlier elegant work on Src and Abl recognized this and explored other hypotheses ( e . g . , differences in drug-binding pocket , energetic changes remote from the binding site and a conformational-selection mechanism ) to reconcile the differences in Gleevec binding ( Dar et al . , 2008; Levinson et al . , 2006; Seeliger et al . , 2007; 2009 ) , but without conclusive success . Recent quantitative binding kinetics combined with ancestral sequence reconstruction put forward a mechanism where an induced-fit step after drug binding is the key determinant for Gleevec’s selectivity ( Agafonov et al . , 2014; Wilson et al . , 2015 ) , and fully recapitulates the binding affinities . Here we ask the question whether this fundamentally different mechanism is a more general principle for drug efficacy and selectivity not only for Tyr kinases such as Abl , but also for Ser/Thr kinases . To this end , we chose the Ser/Thr kinase Aurora A and investigated the binding kinetics of three distinct kinase drugs: Danusertib , AT9283 , and Gleevec . Aurora A kinase is one of the key regulators of mitotic events , including mitotic entry , centrosome maturation and spindle formation ( Fu et al . , 2007; Lukasiewicz and Lingle , 2009; Marumoto et al . , 2005 ) , as well as assisting in neuronal migration ( Nikonova et al . , 2013 ) . Aurora A has attracted significant attention for the development of targeted agents for cancer because it is overexpressed in a wide range of tumors , including breast , colon , ovary and skin malignancies ( Carvajal et al . , 2006; Gautschi et al . , 2008; Katayama and Sen , 2010; Lok et al . , 2010; Marzo and Naval , 2013 ) . The focus was mainly on ATP-competitive inhibitors , but more recently inhibition by allosteric compounds has also been pursued with the aim of achieving higher selectivity ( Asteriti et al . , 2017; Bayliss et al . , 2017; Burgess et al . , 2016; Janeček et al . , 2016; McIntyre et al . , 2017 ) . So far , only the clinical significance of Aurora A inhibition by ATP-competitive drugs has been established ( Bavetsias and Linardopoulos , 2015; Borisa and Bhatt , 2017 ) , but little is known about their binding mechanisms . Many high-resolution X-ray structures of Aurora A kinase bound to different inhibitors have been solved ( Bavetsias et al . , 2015; Dodson et al . , 2010; Fancelli et al . , 2006; Ferguson et al . , 2017; Heron et al . , 2006; Howard et al . , 2009; Kilchmann et al . , 2016; Martin et al . , 2012; Zhao et al . , 2008 ) , but the selectivity profile of those kinase inhibitors remains very difficult to explain . The drugs used in this study are small , ATP-competitive inhibitors . Danusertib ( PHA739358 ) and AT9283 were developed for Aurora kinases , whereas Gleevec is selective for the Tyr kinase Abl . Danusertib inhibits all members of the Aurora family with low nanomolar IC50 values ( 13 , 79 and 61 nM for Aurora A , B and C , respectively ) ( Carpinelli et al . , 2007; Fraedrich et al . , 2012 ) and was one of the first Aurora kinase inhibitors to enter phase I and II clinical trials ( Kollareddy et al . , 2012; Steeghs et al . , 2009 ) . A crystal structure of Danusertib bound to Aurora A kinase shows an inactive kinase with the DFG-loop in the out conformation ( Fancelli et al . , 2006 ) . AT9283 inhibits both Aurora A and B with an IC50 of 3 nM ( Howard et al . , 2009 ) and has also entered several clinical trials ( Borisa and Bhatt , 2017 ) . Interestingly , the crystal structure of Aurora A with AT9283 shows that this drug binds to the DFG-in , active conformation of the kinase ( Howard et al . , 2009 ) . Both drugs are high-affinity binders that reportedly bind to a discrete kinase conformation and would allow us to probe for a conformational-selection step . Lastly , we selected Gleevec as a drug that is not selective for Aurora A and should , therefore , have a weaker binding affinity . We reasoned that this choice of inhibitors could reveal general mechanisms underlying drug selectivity and affinity . The combination of X-ray crystallography , NMR spectroscopy and comprehensive analysis of drug binding and release kinetics delivered a general mechanistic view . Differential drug affinity is not rooted in the overwhelmingly favored paradigm of the DFG-conformation , but instead in the dynamic personality of the kinase that is manifested in conformational changes after drug binding . Notably , such conformational changes have evolved for its natural substrates , and the drugs take advantage of this built-in protein dynamics . A plethora of X-ray structures and functional assays led to the general notion that dephosphorylated Aurora A and , more universally , Ser/Thr kinases are in an inactive conformation and that phosphorylation or activator binding induces the active structure . A comparison of many X-ray structures of inactive and active forms of Ser/Thr kinases resulted in an elegant proposal of the structural hallmarks for the active state by Taylor and collaborators: the completion of both the regulatory and catalytic spines spanning the N- and C-terminal domains , including the orientation of the DFG-motif ( Kornev and Taylor , 2010; 2015 ) . X-ray structures , however , provide merely static snapshots of possible kinase conformations that do not necessarily reflect the situation in solution . In fact , recent experimental data postulate that phosphorylation of Aurora A does not ‘lock’ the kinase in the active conformation , and that the activation-loop still exhibits conformational dynamics ( Gilburt et al . , 2017; Ruff et al . , 2018 ) . On the other hand , X-ray crystallography provides high-resolution structural data that cannot readily be obtained from FRET or EPR and IR spectroscopy . Two crystals from the same crystallization well capture both the inactive and active conformations of dephosphorylated Aurora A bound with AMPPCP ( Figure 1A , B ) . As anticipated , the first structure ( PDB 4C3R [Zorba et al . , 2014] ) superimposes with the well-known inactive , dephosphorylated Aurora A structure ( PDB 1MUO [Cheetham et al . , 2002] ) and the activation loop is not visible as commonly observed for kinases lacking phosphorylation of the activation loop ( Zorba et al . , 2014 ) . The second structure ( PDB 6CPF; Table 1 ) adopts the same conformation as the previously published phosphorylated , active structure ( PDB 1OL7 [Bayliss et al . , 2003] ) ( Figure 1C ) and the first part of the activation loop could be built , although the B-factors are high . Every hallmark of an active state is seen for this dephosphorylated protein , including the DFG-in conformation that is essential for completing the regulatory spine . In contrast , the DFG-loop is in the out position for the inactive form of Aurora A ( Figure 1D , cyan ) . In the active , non-phosphorylated structure , electron density is seen in the canonical tighter Mg2+-binding site , where the metal ion is coordinated to the α- and β-phosphates of AMPPCP and Asp274 . The presence of the metal is supported by the CheckMyMetal ( Zheng et al . , 2017 ) validation , except that the coordination is incomplete . We surmise that two water molecules , not visible in our data , complete the coordination sphere as is seen in several higher-resolution structures . In the inactive structure , no electron density for Mg2+ can be identified possibly due to the fact that Asp274 is rotated to the DFG-out position and is , therefore , lost as coordination partner . Furthermore , sampling of the active conformation does not depend on AMPPCP binding as dephosphorylated , apo Aurora A also crystallizes in the active form ( PDB 6CPE; Figure 1E , F and Table 1 ) . Our results are consistent with other crystallographic studies on wild-type , dephosphorylated Aurora A in its apo or nucleotide bound state , where the kinase was also found in the active conformation ( Gustafson et al . , 2014; Janeček et al . , 2016; Nowakowski et al . , 2002 ) . We note that in Aurora kinase sequences a tryptophan residue , Trp277 , is immediately following the DFG motif and displays a drastically different orientation whether Aurora A is in an active ( DFG-in ) or inactive ( DFG-out ) conformation ( Figure 1D ) . This Trp moiety is unique for the Aurora kinase family in the Ser/Thr kinome and its position is suggested to be important for tuning the substrate specificity ( Chen et al . , 2014 ) . We used this Trp residue as probe to monitor the DFG flip and drug binding in real time as described below . The fact that the inactive and active states are seen in the crystal implies that both are sampled; however , it does not deliver information about the relative populations or interconversion rates . Therefore , we set out to monitor the conformational exchange of the DFG-in/out flip in solution . Owing to the reported importance of the DFG flip for activity , regulation and drug design , there have been extensive efforts to characterize this conformational equilibrium by computation ( Badrinarayan and Sastry , 2014; Barakat et al . , 2013; Meng et al . , 2015; 2017; Sarvagalla and Coumar , 2015; Shukla et al . , 2014 ) . The general notion of these computational studies is that in the absence of phosphorylation the inactive form of the kinase is most favored , in agreement with experimental evidence . Nevertheless , short-lived excursions to the active state are observed . As an experimental approach , NMR spectroscopy is an obvious choice; however efforts on several Ser/Thr and Tyr kinases led to the general conclusion that the activation loop , including the DFG motif and most of the active-site residues , cannot be detected due to exchange broadening , and at best can only be seen after binding of drugs that stabilize conformations ( Campos-Olivas et al . , 2011; Langer et al . , 2004; Vajpai et al . , 2008; Vogtherr et al . , 2006 ) . [1H-15N]-TROSY-HSQC experiments on uniformly 15N-labeled samples of Aurora A proved to be no exception: many peaks are missing and only three out of four tryptophan side chain indole signals are seen in the 2D spectra of a [15N]-Trp labeled sample ( Figure 2A , B ) . Therefore , we sought a strategy to overcome this general problem of exchange broadening that hampers the detection of the DFG equilibrium . Aurora A was produced containing 5-fluoro-tryptophan residues to allow for one-dimensional 19F spectroscopy to deal with exchange broadening while providing sensitivity close to proton NMR ( Kitevski-LeBlanc and Prosser , 2012 ) . Now , we observe as expected four peaks in our NMR spectra for apo- and AMPPCP-bound wild-type Aurora A ( Figure 2C ) . A deconvolution of the spectrum yields almost identical integral values for all four peaks , whereas the linewidth of one resonance is approximately 5-fold larger ( Figure 2D , purple signal ) . This broad peak is a prime candidate to originate from Trp277 , directly adjacent to the DFG-loop . The W277L mutation confirmed our hypothesis ( Figure 2C ) , and the extensive line broadening of this signal in a one-dimensional spectrum is consistent with its absence in the [1H , 15N]-TROSY-HSQC spectrum . Of note , the W277L mutant is still active , as confirmed by a kinase assay , most likely because this Trp is not conserved in Ser/Thr kinases , where a Leu residue is found at that position for several Ser/Thr family members . Mutating any of the other , more conserved Trp residues resulted in insoluble proteins . The broad line shape for the Trp277 peak hints at severe exchange broadening in the surrounding of the DFG-loop and is consistent with the high B-factors for Trp277 and its neighboring residues observed in all crystal structures described here . Determination of relative populations and rate constants of interconversion is not possible from this data , but this missing piece of information was obtained by stopped-flow kinetics of drug binding . Through groundbreaking experiments on the Tyr kinases Abl and Src , the concept of drug selectivity based on the DFG-loop conformation has received considerable attention in kinase drug discovery ( Lovera et al . , 2012; Treiber and Shah , 2013 ) . A recent report provides kinetic evidence for such conformational selection , but identifies an induced-fit step after drug binding as the overwhelming contribution for Gleevec selectivity towards Abl compared to Src ( Agafonov et al . , 2014 ) . Here , we ask the obvious question if this mechanism of Gleevec binding to Abl might exemplify a more general mechanism for kinase inhibitors . To assess which kinetic steps control drug affinity and selectivity , we first studied the binding kinetics for Gleevec to Aurora A by stopped-flow spectroscopy using intrinsic tryptophan fluorescence under degassing conditions to reduce photobleaching . At 25°C , the binding of Gleevec to Aurora A was too fast to be monitored and , therefore , experiments were performed at 10°C . Binding kinetics of Gleevec to Aurora A exhibited biphasic kinetic traces ( Figure 3A ) . The first , fast phase is characterized by a decrease in the fluorescence intensity ( Figure 3A , B ) , with an observed rate constant , kobs , increasing linearly with Gleevec concentration ( Figure 3C ) . The slope corresponds to the bimolecular rate constant , k2 = 1 . 1 ± 0 . 3 μM−1s−1 , of Gleevec binding to Aurora A and the dissociation of Gleevec is determined from the intercept , k-2 = 31 ± 2 s−1 ( Figure 3C ) . We note that the parameters for the physical binding step are comparable to the ones obtained for Gleevec binding to Abl ( cf . k2 = 1 . 5 ± 0 . 1 μM−1s−1 and k-2 = 25 ± 6 s−1 , measured at 5°C ) ( Agafonov et al . , 2014 ) . The second , slow phase exhibits an increase in fluorescence intensity ( Figure 3A ) , with the observed rate constant decreasing with Gleevec concentration ( Figure 3D ) . The decreasing kobs provides unequivocal evidence of conformational selection , where its rate of interconversion is slower than the rate of ligand dissociation ( k1+k-1≪k-2 ) . The values of k1 and k-1 can be estimated by fitting the data to Equation 1 and are 0 . 014 ± 0 . 001 s−1 and 0 . 011 ± 0 . 002 s−1 , respectively ( Figure 3D ) . These rate constants represent the conformational change from DFG-in to -out and vice versa since Gleevec is a DFG-out selective inhibitor due to steric hindrance ( Nagar et al . , 2002; Schindler et al . , 2000; Seeliger et al . , 2007 ) . In order to more rigorously analyze the data and test the model , all time courses of the fluorescence changes were globally fit using the microscopic rate constants determined above as starting values ( Figure 4 ) to the model in Figure 3G , where also the resulting microscopic rate constants are given . The lack of a conformational transition after drug binding ( i . e . , induced-fit step ) in Aurora A should dramatically decrease drug affinity in comparison to Abl . Indeed , Gleevec binds to Aurora A with a KD of 24 ± 7 µM ( Figure 3F ) compared to the low nM affinity to Abl ( Agafonov et al . , 2014 ) . Two pieces of independent evidence establish that there is indeed no induced-fit step in Gleevec binding to Aurora A: ( i ) the calculated KD from the kinetic scheme is in agreement with the macroscopically measured KD ( cf . Figure 3G and F ) , and ( ii ) the observed koff from the dilution experiment ( Figure 3E ) coincides with the physical dissociation rate ( i . e . , intercept of the binding plot , 31 ± 2 s−1 , in Figure 3C ) . In summary , the lack of an induced-fit step for Gleevec binding to Aurora A is the major reason for Gleevec’s weak binding , and not the DFG-loop conformation or physical drug-binding step , consistent with our earlier results ( Wilson et al . , 2015 ) . Next , we wanted to shed light on why Danusertib , unlike Gleevec , binds very tightly to Aurora A . A high-resolution X-ray structure shows Danusertib bound to Aurora A’s active site with its DFG-loop in the out conformation ( Figure 5A ) ( Fancelli et al . , 2006 ) , and to rationalize Danusertib’s high affinity we measured the kinetics of Danusertib binding to Aurora A directly by stopped-flow experiments at 25°C . An increase in fluorescence intensity was observed at all Danusertib concentrations and showed double-exponential behavior ( Figure 5B ) . The dependence of the two observed rates constants on drug concentration is linear for one of them ( Figure 5C ) and non-linear for the other with an apparent plateau at approximately 16 ± 2 s−1 ( Figure 5D ) . The step with linear inhibitor concentration dependence corresponds to the second-order binding step , whereas a non-linear concentration dependency hints at protein conformational transitions . For a hyperbolic increase of the observed rate with substrate concentrations , one cannot a priori differentiate between a conformational selection and an induced fit mechanism . However , conformational selection happens before drug binding , and the intrinsic slow DFG-in to DFG-out interconversion in Aurora A revealed by Gleevec binding ( Figure 3 ) must , therefore , be unaltered . Since the apparent rate of 16 ± 2 s−1 ( Figure 5D ) is two orders of magnitude faster , it can only reflect an induced-fit step ( i . e . , kobs=k3+k-3 ) . So , what happened to the conformational selection step ? We hypothesize that the lack of this step in our kinetic traces is due to a too small amplitude of this phase , or not observable because of photobleaching having a bigger effect at the longer measurement times . To lessen potential photobleaching , we reduced the enzyme concentration and increased the temperature to 35°C . Indeed , under these conditions , the slow DFG-in to DFG-out kinetics were observed as an increase of fluorescence intensity over time with an observed rate constant of approximately 0 . 1 s−1 ( Figure 5—figure supplement 1A ) . While these experiments clearly establish the three-step binding mechanism , it does not provide accurate rate constants for the conformational selection step and it cannot be observed at 25°C where all the other kinetic experiments are performed . To resolve this issue , we repeated the Aurora A–Gleevec experiment at 25°C ( Figure 5—figure supplement 2A , B ) and obtained reliable rate constants ( k1 = 0 . 09 ± 0 . 01 s−1 and k-1 = 0 . 06 ± 0 . 005 s−1 ) for the conformational selection step in Aurora A , which will be used as ‘knowns’ in what follows . We hypothesize that the conformational selection step reflects the interconversion between inactive/active conformations and is correlated with the DFG-out and -in position ( Figure 1 ) . The following observations support our hypothesis: ( i ) two crystal structures for the apo-protein show Trp277 in very different environments ( Figure 1E ) , ( ii ) Danusertib has been proposed to selectively bind to the DFG-out conformation based on a co-crystal structure ( Figure 5A ) ( Fancelli et al . , 2006 ) , and ( iii ) the same slow step is observed for binding of both Gleevec and Danusertib . Next , the dissociation kinetics for Danusertib was measured by fluorescence and appeared to be extremely slow with an observed slow-off rate of ( 3 . 2 ± 0 . 3 ) × 10−4 s−1 ( Figure 5E ) . Rationalization of complex binding kinetics cannot be done anymore by visual inspection and kinetic intuition , which can , in fact , be misleading . In order to elucidate the correct binding mechanism and obtain accurate kinetic parameters , all kinetic traces were globally fit ( Figure 6 ) to the three-step binding scheme ( Figure 5I ) . Although global fitting of the binding and dissociation kinetics in KinTek Explorer delivered a value for k-2 , evaluation of the kinetic scheme with respect to the time traces exposes that k-2 is not well determined from our experiments . We therefore designed a double-jump experiment to populate the AurAout:D state followed by dissociation to obtain more accurate information on k-2 . Our stopped-flow machine lacks the capability to perform double mixing and , therefore , the double-jump experiment was performed using a Creoptix WAVE instrument . This label-free methodology uses waveguide interferometry to detect refractive index changes due to alteration in surface mass in a vein similar to Surface Plasmon Resonance ( SPR ) . It is an orthogonal technique that sidesteps notable issues associated with fluorescence methods ( e . g . , photobleaching and inner-filter effects ) . In short , after immobilizing Aurora A on a WAVEchip , a high concentration of Danusertib was injected for a short , variable period of time , and dissociation was triggered by flowing buffer through the microfluidics channel to remove the drug . The dissociation kinetics fit to a single exponent with a rate constant , k−2 , of 6 . 8 ± 0 . 4 s−1 ( Figure 5F and Figure 5—figure supplement 1B ) . We want to discuss a few additional kinetic features . First , the observed rate constant measured in the dilution experiment ( Figure 5E , k-3 = ( 3 . 2 ± 0 . 3 ) × 10−4 s−1 ) is slower than k-3 from the global fit ( k-3 = ( 7 . 1 ± 0 . 5 ) × 10−4 s−1 ) , which might seem counterintuitive . The observed rate constant was verified by an additional dilution experiment using Creoptix WAVE ( k-3 = ( 2 . 0 ± 0 . 6 ) × 10−4 s−1 , Figure 5—figure supplement 1C ) . The difference in the observed and microscopic rate constant can , however , be fully reconciled by considering the kinetic partitioning for the proposed scheme , as shown in Figure 6—figure supplement 1 . Second , a powerful and independent validation of the three-step binding mechanism is obtained by comparing the measured overall KD of Danusertib with the calculated macroscopic KD from the microscopic rate constants ( Figure 5G , H , I and Figure 5—figure supplement 1D ) according to Equation 4 , which indeed delivers values that are within experimental error . In addition , our values for k2 , k-3 , and KD are in good agreement with those reported in a recent study using SPR ( Willemsen-Seegers et al . , 2017 ) . Our results illuminate trivial but profound principles of binding affinity and lifetime of drug/target complexes: a conformational selection mechanism always weakens the overall inhibitor affinity , while an induced-fit step tightens the affinity depending on how far-shifted the equilibrium in the enzyme/drug complex is ( Equations 2–4 , Figure 6—figure supplement 2 ) . For DFG-out binders ( e . g . , Danusertib and Gleevec ) , the DFG-in and -out equilibrium weakens the overall affinity 1 . 6-fold; however , the conformational change after drug binding results in a four orders of magnitude tighter binding for Danusertib and is the sole reason for its high affinity to Aurora A compared to Gleevec . The dissociation constants for the bimolecular binding step K2 is very similar for both inhibitors . Finally , the lifetime of Danusertib on the target is very long because of the very slow conformational dynamics within the Aurora A/Danusertib complex ( k-3 = ( 7 . 1 ± 0 . 5 ) × 10−4 s−1 ) . Earlier examples of protein kinases that also show remarkable slow off-rates , presumably caused by conformational changes , include the epidermal growth factor receptors ( Berezov et al . , 2001; Wood et al . , 2004 ) and CDK8 ( Schneider et al . , 2013 ) amongst others ( Willemsen-Seegers et al . , 2017 ) . To the best of our knowledge , we present here for the first time a detailed stopped-flow kinetics analysis for Aurora A that unequivocally shows the slow off-rate is caused by the conformational change within the drug-bound state , and not the dissociation step . We chose AT9283 as a third inhibitor to characterize the binding mechanism because it has been described as a DFG-in binder based on a crystal structure of AT9283 bound to Aurora A ( PDB 2W1G , [Howard et al . , 2009] ) . We , therefore , anticipated that in its binding kinetics one can now detect the DFG-out to DFG-in switch . Rapid kinetic experiments of binding AT9283 to Aurora A at 25°C resulted in biphasic traces and both processes showed an increase in fluorescence over time ( Figure 7A ) . The kobs for the faster phase ( k2 ) was linearly dependent on drug concentration reflecting the binding step ( Figure 7B ) and kobs for the slower phase ( k3 ) has a limiting value of 0 . 8 ± 0 . 2 s−1 and is attributed to an induced-fit step ( Figure 7C ) . For the conformational selection step ( i . e . , DFG-out to DFG-in ) , a decrease in fluorescence is expected because for the reverse flip observed in the Gleevec and Danusertib experiments , a fluorescence increase was seen ( Figure 3A and Figure 5—figure supplement 1A ) . However , we could not find any condition ( e . g . , by varying temperature and ligand concentrations ) where such a phase could be observed . Dissociation is characterized by double-exponential kinetics ( Figure 7D and Figure 7—figure supplement 1A ) . The fast phase ( ~38% of the total amplitude change ) decays with a rate constant of ( 1 . 1 ± 0 . 02 ) × 10−2 s−1 , and the slow phase ( ~62% of the total change in amplitude ) has a rate constant of ( 0 . 1 ± 0 . 01 ) × 10−2 s−1 . To distinguish between the reverse induced-fit step ( k-3 ) and the physical dissociation step ( k-2 ) , a double-jump experiment was performed that unambiguously assigned the faster phase to k-2 ( Figure 7E and Figure 7—figure supplement 1B ) . Our attempts to globally fit all kinetic traces assuming binding to only the DFG-in state and using the rate constants for the DFG-loop flip from the Gleevec experiment failed ( Figure 8—figure supplement 1A ) . An extended model , where AT9283 can bind to both DFGin/out conformations , followed by a common induced-fit step can also not explain the experimental kinetic traces ( Figure 8—figure supplement 1B ) . These failures , together with the lack of a detectable conformational selection step , led to a new model in which both the DFG-in and DFG-out states can bind AT9283 , but only AurAin:AT can undergo an induced-fit step ( Figure 7H ) . All data can be globally fit to this model ( Figure 8 ) and the overall KD calculated from the corresponding microscopic rate constants ( using Equation 5 ) is in good agreement with the experimentally measured KD ( Figure 7F–H ) . Finally , the 10-fold difference between the k-3 from the global fit ( Figure 7H ) and the experimentally observed slow off-rate can be reconciled by kinetic partitioning as shown in Figure 7—figure supplement 1A . In an effort to structurally verify our model we solved a crystal structure of Aurora A with AT9283 bound and indeed observed the DFG-out conformation ( PDB 6CPG , Figure 9B and Table 1 ) , in contrast to the DFG-in conformation as previously reported ( Figure 9A ) ( Howard et al . , 2009 ) . Our structure was obtained by co-crystalizing Aurora A with AT9283 and a monobody that binds to the same site as the natural allosteric activator TPX2 ( Figure 9B ) . Binding of this monobody shifts Aurora A into an inactive conformation , with the DFG-loop in the out conformation . This new structure underscores the plasticity of Aurora A kinase and the ability of AT9283 to bind to a DFG-out state , in addition to the previously reported DFG-in state . Thus , our structural and kinetic data together support that AT9283 can bind to both DFG-in and DFG-out state of Aurora A , and emphasizes the need for caution when interpreting single X-ray structures . We finally compared the binding kinetics of the ATP-competitive inhibitors described above with the natural kinase substrate , ATP ( Figure 10 ) . In order to measure stopped-flow kinetics for ATP binding , FRET was measured by exciting Trp residues in Aurora A and detecting fluorescence transfer to the ATP-analogue mant-ATP ( Lemaire et al . , 2006; Ni et al . , 2000 ) . The binding of mant-ATP to Aurora A showed biphasic kinetic traces ( Figure 10A ) that describe the physical binding step ( i . e . , linear dependence on mant-ATP concentration; Figure 10B ) and the induced-fit step ( Figure 10C ) . The observed rate constant approaches a maximum value defined by the sum of k3+ k−3 ( Figure 10C ) and the intercept can be estimated to be k−3 and is consistent with the value obtained from the koff experiment ( Figure 10D ) . We find that mant-ATP can bind to both the DFG-in and -out conformations , consistent with our nucleotide-bound crystal structures ( Figure 1A–D ) and recent single-molecule fluorescence spectroscopy data that indicates that nucleotide binding does not significantly affect this equilibrium ( Cyphers et al . , 2017 ) . To confirm the model , the kinetic data were globally fit to a two-step binding mechanism ( Figure 10G , H ) . The calculated KD from the corresponding microscopic rate constants ( Figure 10H ) is comparable with the experimental macroscopic KD obtained from a titration experiment ( Figure 10E , F ) . The presence of an induced-fit step for the natural substrate ATP suggests that such conformational change after ligand binding is a built-in property of the enzyme . In other words , inhibitors take advantage of the inherent plasticity of the enzyme that is required for its activity and regulation . The main difference between ATP and inhibitor binding is the rate constant for the reverse induced-fit step ( k−3 ) . In the case of ATP , this rate is much faster and , therefore , does not significantly increase the overall affinity . Faster conformational changes and weaker binding are of course prerequisites for efficient turnover; whereas slow conformational changes , particularly the reverse induced-fit step , are at the heart of action for an efficient drug , because it results in tight binding and a long lifetime on the target . In summary , binding of different ligands to the ATP-binding site , such as nucleotides or ATP-competitive inhibitors , is comprised of the physical binding step followed by an induced-fit step . By definition , it is the nature of the induced-fit step that varies for the different ligands since it happens as a result of ligand binding . Characterizing the detailed kinetic mechanisms of drug binding is not just an academic exercise but delivers fundamental knowledge for developing selective inhibitors with high affinity . An induced-fit step turns out to be key for all tight-binding inhibitors studied . From our results on Aurora A kinase presented here and earlier data on Tyrosine-kinases ( Agafonov et al . , 2014; Wilson et al . , 2015 ) , we propose that this may be a general mechanism for different kinases and multiple inhibitors , thereby providing a platform for future computational and experimental efforts in rational drug design . Albeit , we note that verification of this proposition requires a larger sampling of small molecules and different protein kinases throughout the kinome . The ‘use’ of a highly-skewed equilibrium towards E*:D for a promising drug is logical for the following reasons: ( i ) it increases the affinity for the drug by this coupled equilibrium , ( ii ) it prolongs the residence time of the drug on the target due to the often slow reverse rate , ( iii ) it is specific for each drug as it happens after the drug binding , and ( iv ) it can add selectivity for the targets because it likely involves residues more remote from the active site . An increased drug residence time has significant pharmacological advantages as it can lead to a prolonged biological effect , a decrease of side effects , and a lower risk of metabolic drug modification . Such inhibitors have long been described as slow tight-binding inhibitors ( Copeland , 2016; Copeland et al . , 2006 ) . The concept of the advantageous roles of induced-fit steps is based on simple thermodynamics and protein flexibility , and is , therefore , likely of relevance for drug design to other targets outside of the kinome . Additionally , our data provides unique insight into the extensively discussed DFG flip . Combining X-ray crystallography , NMR spectroscopy and stopped-flow kinetics of drug binding establish the nature of this DFG flip both structurally , thermodynamically and kinetically , and resolves the longstanding question of its role for drug affinity and selectivity . Selective binding of a specific DFG-state by Gleevec has been first proposed as the reason for selectivity towards Abl . This conformational selection principle has ever since been at the center of drug discovery for many kinases , including Aurora A ( Badrinarayan and Sastry , 2014; Liu and Gray , 2006 ) . Based on our results , we argue that conformational selection of the DFG-state by ATP-competitive inhibitors is a mistakenly pursued concept in drug design for the following reasons: ( i ) conformational selection by definition weakens the overall ligand affinity , ( ii ) active site binders are automatically inhibitors , therefore selective binding to a specific DFG-state has no advantage ( Badrinarayan and Sastry , 2014; Liu and Gray , 2006 ) , ( iii ) kinases interconvert between both states . High selectivity gained by DFG-state selective binding could only be achieved in the scenario of a highly skewed population towards the binding-competent state for one kinase relative to all others , which is unfounded . Our results exemplify why rational drug design is so challenging . The characterization of the complete free-energy landscape of drug binding is needed , which will require more sophisticated computational approaches guided by experimental data such as provided in our study . A good illustration of this point are the computational reports that focused on the DFG flip as a key determinant drug selectivity ( Badrinarayan and Sastry , 2014 ) that now have been ruled out by our kinetic measurements . Our data suggest that future design efforts should be focusing on understanding and exploiting induced-fit steps . To this end , the different dynamic personalities of kinases or , more general , drug targets need to be investigated at atomic resolution and used to guide small-molecule design . The findings presented here are encouraging for developing selective inhibitors even for kinases with very similar folds and drug binding pockets since the action does not happen on a single structural element of the protein , but on a complex energy landscape that is unique to each kinase . Dephosphorylated Aurora A proteins were expressed and purified as described before ( Zorba et al . , 2014 ) and analyzed by mass spectrometry to confirm their phosphorylation state . The W227L mutant was generated using the QuickChange Lightning site-directed mutagenesis kit ( Agilent ) . U-[15N] Aurora A was obtained by growing E . coli BL21 ( DE3 ) ( New England Biolabs ) in M9 minimal medium containing 1 g/L 15NH4Cl ( Cambridge Isotope Laboratories , Tewksbury , MA , USA ) and 5 g/L D-glucose as the sole nitrogen and carbon source , respectively . [15N]-Trp labeled wild-type Aurora A was obtained using the standard M9 minimal medium , complemented with all amino acids ( 0 . 5 g/L ) with the exception of tryptophan . One hour prior to induction , 30 mg/L of 15N2-L-Trp ( NLM-800; Cambridge Isotope Laboratories , Tewksbury , MA , USA ) was added to the medium . Similarly , to obtain samples of wild-type and W277L Aurora A containing 5-fluoro-tryptophan , bacterial growth was performed in unlabeled M9 medium containing all amino acids ( 0 . 5 g/L ) except for tryptophan . One hour before protein induction , the medium was supplemented with 30 mg/L of 5-fluoro-DL-tryptophan ( Sigma-Aldrich ) ( Crowley et al . , 2012 ) . NMR samples contained 200–300 µM Aurora A in 50 mM HEPES , pH 7 . 3 , 50 mM NaCl , 20 mM MgCl2 , 5 mM TCEP , 2 M TMAO and 10% ( v/v ) D2O . Inhibiting monobody used for co-crystallization with Aurora A and AT9283 was expressed in E . coli BL21 ( DE3 ) cells harboring the plasmid pHBT containing His6-tagged-Mb . A culture of TB media containing 50 μg/mL kanamycin that was grown overnight at 37°C was added to 1 L of TB media with 50 μg/mL kanamycin to get a starting OD600 of ~0 . 2 . This culture was grown at 37°C until the OD600 reached ~0 . 8 . Protein expression was induced by 0 . 6 mM IPTG at 18°C for 13–15 h and cells were harvested by centrifugation . The cell pellet was resuspended in binding buffer ( 50 mM Tris-HCl , pH 8 . 0 , 300 mM NaCl , 20 mM imidazole , 20 mM MgCl2 , 10% glycerol ) containing 0 . 5 mg/mL lysozyme , 5 μg/mL DNase , and 1x EDTA-free protease inhibitor cocktail . Cells were ruptured by sonication on ice then centrifuged at 18 , 000 rpm at 4°C for 1 h . The supernatant was loaded onto HisTrapTM HP ( GE Healthcare ) after filtration using 0 . 2 μm filtering unit . The pellet was resuspended with GuHCl buffer ( 20 mM Tris-HCl , pH 8 . 0 , 6 M GuHCl ) and allowed to rotate on wheel for 10 min at 4°C and spun down again . The supernatant was passed through 0 . 2 μm filtering unit and loaded onto HisTrap HP column previously loaded with soluble fraction and pre-equilibrated with GuHCl buffer . Refolding monobody on-column was achieved by washing the HisTrap HP column with five column volumes ( CV ) of GuHCl buffer , followed by 5 CV of Triton-X buffer ( binding buffer + 0 . 1% Triton X-100 ) , then 5 CV of β-cyclodextrin buffer ( binding buffer + 5 mM β-cyclodextrin ) , and finally 5 CV of binding buffer . Monobody was eluted with 100% of elution buffer ( binding buffer + 500 mM imidazole ) . The protein was dialyzed overnight in gel-filtration buffer ( 20 mM Tris-HCl , pH 7 . 5 , 200 mM NaCl , 20 mM MgCl2 , 5 mM TCEP , 10% glycerol ) in the presence of TEV protease ( 1:40 TEVP:Mb molar ratio ) . After dialysis , the TEV-cleaved monobody was passed through HisTrap HP column again . The flow-through containing TEV-cleaved monobody was collected and concentrated before loading onto Superdex 200 26/60 gel-filtration column pre-equilibrated with the gel-filtration buffer . The monobody was flash-frozen and stored in −80°C until use . Crystals of dephosphorylated ( deP ) Aurora A122−403 + AMPPCP were obtained by mixing 570 μM ( 18 mg/mL ) deP Aurora A122−403 and 1 mM AMPPCP in a 2:1 ratio with mother liquor ( 0 . 2 M ammonium sulfate , 0 . 2 M Tris-HCl , pH 7 . 50 , 30% ( w/v ) PEG-3350 ) . The crystals were grown at 18°C by vapor diffusion using the hanging-drop method . The protein used for the crystallization was in storage buffer ( 20 mM Tris-HCl , pH 7 . 5 , 200 mM NaCl , 10% ( v/v ) glycerol , 20 mM MgCl2 , 1 mM TCEP ) ; AMPPCP was freshly prepared before use in the same buffer . Crystals were flash-frozen in liquid nitrogen prior to shipping . Crystals of apo , deP Aurora A122−403 were grown at 18°C by vapor diffusion using the sitting-drop method ( 96-well plate ) . A 1:1 ratio of protein to mother liquor was obtained by combining 0 . 5 μL of 300 μM ( 10 mg/mL ) deP Aurora A122−403 in 50 mM HEPES , pH 7 . 3 , 500 mM ammonium acetate , 1 mM MgCl2 , 5 mM TCEP ) with 0 . 5 μL of 0 . 15 M Tris-HCl , pH 7 . 5 , 0 . 15 M ammonium sulfate , 35% ( w/v ) PEG-3350 . Crystals were soaked for 10–20 s in cryo buffer ( 20% ( w/v ) PEG-400 , 20% ethylene glycol , 10% water and 50% mother liquor ) before flash-freezing in liquid nitrogen . The complex between Aurora A122−403 , inhibiting monobody ( Mb ) and AT9283 was crystallized at 18°C by vapor diffusion using the sitting-drop method . In short , a 1:1 ratio of protein mixture to mother liquor was obtained by combining 0 . 5 μL of sample [240 μM deP Aurora A122−403 + 1 . 0 mM AT9283 + 250 μM Mb] with 0 . 5 μL of mother liquor [0 . 1 M Bis-Tris , pH 5 . 5 , 0 . 2 M magnesium chloride , 19% ( w/v ) PEG-3350] . Crystals were soaked for 10–20 s in cryo buffer ( 17 . 5% ( w/v ) PEG-400 , 17 . 5% ethylene glycol , 45% water and 20% mother liquor ) before flash-freezing in liquid nitrogen . Diffraction data were collected at 100 K at the Advanced Light Source ( Lawrence Berkeley National Laboratory ) beamlines ALS 8 . 2 . 1 ( apo-AurA and AurA + Mb + AT9283 ) and 8 . 2 . 2 ( AurA + AMPPCP ) with a collection wavelength of 1 . 00 Å . Data were indexed and integrated using iMOSFLM ( Battye et al . , 2011 ) for apo/AMPPCP-bound Aurora A and Xia2 ( Winter , 2010 ) using XDS ( Kabsch , 2010 ) for the Aurora A/Mb/AT9283 complex , respectively . Data were scaled and merged with AIMLESS ( Evans and Murshudov , 2013 ) , in the case of Aurora A/Mb/AT9283 two data separate data sets were merged . All software was used within the CCP4 software suite ( Winn et al . , 2011 ) . As initial search models 1MQ4 ( Nowakowski et al . , 2002 ) and 3K2M ( Wojcik et al . , 2010 ) were used for Aurora A and monobody , respectively , and molecular replacement was performed using Phaser ( McCoy et al . , 2007 ) . The molecules were placed in the unit cell using the ACHESYM webserver ( Kowiel et al . , 2014 ) . Iterative refinements were carried out with PHENIX ( Adams et al . , 2010 ) , using rosetta . refine ( DiMaio et al . , 2013 ) and phenix . refine ( Afonine et al . , 2012 ) , and manual rebuilding was performed in Coot ( Emsley and Cowtan , 2004; Emsley et al . , 2010 ) . Structure validation was performed using MolProbity ( Chen et al . , 2010 ) and yielded the statistics given below . The Ramachandran statistics for dephosphorylated apo ( AMPPCP-bound ) Aurora A are: favored: 93 . 65 ( 94 . 90 ) % , allowed 5 . 95 ( 4 . 71 ) % , outliers: 0 . 4 ( 0 . 39 ) %; 0 . 48 ( 0 . 0 ) % rotamer outliers and an all-atom clashscore of 4 . 45 ( 2 . 44 ) . For the Aurora A/Mb/AT9283 complex , the Ramachandran statistics are: favored: 92 . 64% , allowed 7 . 06% , outliers: 0 . 3%; 0 . 0% rotamer outliers and an all-atom clashscore of 2 . 81 . We note that the B-factors for the monobodies in the complex of Aurora A/Mb/AT9283 are rather high , indicating significantly flexibility in the parts that are not part of the binding interface with Aurora A . The data collection and refinement statistics are given in Table 1 . Structure factors and refined models have been deposited in the PDB under accession codes: 6CPE ( apo Aurora A ) , 6CPF ( Aurora A + AMPPCP ) and 6CPG ( Aurora A/Mb/AT9283 ) . All figures were generated using Chimera ( Pettersen et al . , 2004 ) . All 19F NMR experiments were performed at 35°C on a Varian Unity Inova 500 MHz spectrometer , equipped with a 1H/19F switchable probe tuned to fluorine ( 90° pulse width of 12 µs ) . All 1D 19F spectra were recorded with a spectral width of ~60 ppm and a maximum evolution time of 0 . 25 s . An interscan delay of 1 . 5 s was used with 5000 scans per transients , giving rise to a total acquisition time of 2 . 5 h per spectrum . To remove background signal from the probe and avoid baseline distortions , data acquisition was started after a ~100 µs delay ( using the ‘delacq’ macro ) and appropriate shifting of the data followed by backward linear prediction was performed . The data were apodized with an exponential filter ( 2 . 5 Hz line broadening ) and zero-filled before Fourier transform . To improve the signal-to-noise ratio several data sets were recorded consecutively and , provided that the sample remained stable , added together after processing ( two for apo Aurora A , four for Aurora A + AMPPCP , and five for W277L + AMPPCP , respectively ) . 19F chemical shifts were referenced externally to trifluoroacetic acid ( TFA ) at −76 . 55 ppm . [1H-15N]-TROSY-HSQC experiments were recorded at 25°C on an Agilent DD2 600 MHz four-channel spectrometer equipped with a triple-resonance cryogenically cooled probe-head . Typically , 115–128 ( 15N ) × 512 ( 1H ) complex points , with maximum evolution times equal to 48 . 5–64 ( 15N ) × 64 ( 1H ) ms . An interscan delay of 1 . 0 s was used along with 32 or 56 scans per transient , giving rise to a net acquisition time 1 . 5–2 . 5 h for each experiment . To improve the signal-to-noise ratio several data sets were recorded consecutively and , provided that the sample remained stable , added together after processing ( typically three data sets per sample ) . All data sets were processed with the NMRPipe/NMRDraw software package ( Delaglio et al . , 1995 ) and 2D spectra were visualized using Sparky ( Goddard , 2008 ) . Deconvolution of the 19F spectra and line shape fitting was performed using the Python package nmrglue ( Helmus and Jaroniec , 2013 ) . In the following equations , K1 , K2 , K3 and K4 are equal to:K1=k−1k1K2=k−2k2=koffkonK3=k−3k3K4=k−4k4 Conformational selection followed by inhibitor binding: ( 2 ) Ein⇌k−1k1Eout+ I ⇌koffkonEout⋅IK1K2KD= ( K1+1 ) ∗K2 Inhibitor binding followed by an induced-fit step: ( 3 ) Eout+ I⇌koffkonEout⋅I⇌k−3k3Eout∗⋅I K2K3KD=K2∗K3 ( K3+1 ) Conformational selection followed by inhibitor binding and an induced-fit step: ( 4 ) Ein⇌k−1k1Eout+I⇌koffkonEout⋅I⇌k−3k3Eout∗⋅IK1K2K3KD= ( K1+1 ) ∗K2∗K3K3+1 Conformational selection mechanism , followed by inhibitor binding to both DFG-in and -out state , but an induced-fit step only occurs in the DFG-in state: ( 5 ) Eout⇌k−1k1Ein+I⇌koffkonEin⋅I⇌k−3k3Ein∗⋅IK4⇃↾Eout⋅IK1K2K3KD= ( K1+1 ) ∗K2∗K3∗K4K1∗K2∗K3+K3∗K4+K4 The uncertainties in the calculated dissociation constant parameter using the equations above are obtained using standard error propagation . FRET using intrinsic tryptophan fluorescence is used to monitor mant-ATP ( obtained from Jena Bioscience ) binding kinetics to Aurora A at 10°C . In the binding experiment or kon , increasing concentration of mant-ATP were quickly mixed to 0 . 5 µM Aurora A ( ratio 1:10 , excitation at 295 nm , emission cut-off at 395 nm ) . In the experiment to measure the release of mant-ATP or koff , 10 µM/10 µM Aurora A/mant-ATP complex was diluted with buffer ( ratio 1:10 ) . A significant decrease in the fluorescence intensity of Aurora A ( excitation at 295 nm , emission cut-off at 395 nm ) can be seen due to the mant-ATP release . Fluorescence titration experiments were measured using FluoroMax-4 spectrofluorometer ( Horiba Scientific ) . Increasing amounts of Aurora A/Danusertib complex ( 4 nM Aurora A and 150 nM Danusertib ) or Aurora A/mant-ATP ( 1 µM Aurora A and 2 mM mant-ATP ) were titrated into an Aurora A solution ( 4 nM and 1 µM Aurora A for experiments with Danusertib and mant-ATP , respectively ) . To measure Danusertib affinity , the excitation wavelength was 295 nm ( 5 nm bandwidth ) and emission spectra were recorded from 310 to 450 nm ( 20 nm bandwidth ) in increments of 2 nm and the temperature was maintained at 25°C . For the mant-ATP experiment , the dissociation constant was measured at 10°C using fluorescence energy transfer from tryptophan residues in Aurora A to mant-ATP by setting the excitation wavelength to 290 nm ( 5 nm bandwidth ) and collecting the emission intensity from 310 to 550 nm ( 5 nm bandwidth ) in increments of 2 nm . A control experiment in the absence of Aurora A was performed using the same experimental settings and used to correct for the mant-ATP interference . In all experiments , a 5 min equilibration time was used after each addition of Aurora A/Danusertib complex or Aurora A/mant-ATP complex . The fluorescence intensity at 368 nm versus Danusertib concentration or the change in fluorescence at 450 nm ( ΔF450 ) versus mant-ATP concentration was fitted to Equation 6 using Levenberg-Marquardt nonlinear fitting algorithm included in KaleidaGraph to obtain the KD . ( 6 ) F=F0+A⋅[I]+[Et]+KD− ( [I]+[Et]+KD ) 2−4⋅[Et]⋅[I]2⋅[Et] F and F0 are the fluorescence and initial fluorescence intensities , respectively . [I] and [Et] are the total concentration of the drug or mant-ATP and the Aurora A , respectively .	Protein kinases are a family of enzymes found in all living organisms . These enzymes help to control many biological processes , including cell division . When particular protein kinases do not work correctly , cells may start to divide uncontrollably , which can lead to cancer . One example is the kinase Aurora A , which is over-active in many common human cancers . As a result , researchers are currently trying to design drugs that reduce the activity of Aurora A in the hope that these could form new anticancer treatments . In general , drugs are designed to be as specific in their action as possible to reduce the risk of harmful side effects to the patient . Designing a drug that affects a single protein kinase , however , is difficult because there are hundreds of different kinases in the body , all with similar structures . Because drugs often work by binding to specific structural features , a drug that targets one protein kinase can often alter the activity of a large number of others too . Gleevec is a successful anti-leukemia drug that specifically works on one target kinase , producing minimal side effects . It was recently discovered that the drug works through a phenomenon called ‘induced fit’ . This means that after the drug binds it causes a change in the enzyme’s overall shape that alters the activity of the enzyme . The shape change is complex , and so even small structural differences can change the effect of a particular drug . Do other drugs that target other protein kinases also produce induced fit effects ? To find out , Pitsawong , Buosi , Otten , Agafonov et al . studied how three anti-cancer drugs interact with Aurora A: two drugs specifically designed to switch off Aurora A , and Gleevec ( which does not target Aurora A ) . The two drugs that specifically target Aurora A were thought to work by targeting one structural feature of the enzyme . However , the biochemical and biophysical experiments performed by Pitsawong et al . revealed that these drugs instead work through an induced fit effect . By contrast , Gleevec did not trigger an induced fit on Aurora A and so bound less tightly to it . In light of these results , Pitsawong et al . suggest that future efforts to design drugs that target protein kinases should focus on exploiting the induced fit process . This will require more research into the structure of particular kinases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics", "cancer", "biology" ]	2018	Dynamics of human protein kinase Aurora A linked to drug selectivity
Formation of a division septum near a randomly chosen pole during sporulation in Bacillus subtilis creates unequal sized daughter cells with dissimilar programs of gene expression . An unanswered question is how polar septation activates a transcription factor ( σF ) selectively in the small cell . We present evidence that the upstream regulator of σF , the phosphatase SpoIIE , is compartmentalized in the small cell by transfer from the polar septum to the adjacent cell pole where SpoIIE is protected from proteolysis and activated . Polar recognition , protection from proteolysis , and stimulation of phosphatase activity are linked to oligomerization of SpoIIE . This mechanism for initiating cell-specific gene expression is independent of additional sporulation proteins; vegetative cells engineered to divide near a pole sequester SpoIIE and activate σF in small cells . Thus , a simple model explains how SpoIIE responds to a stochastically-generated cue to activate σF at the right time and in the right place . How genetically identical daughter cells adopt dissimilar programs of gene expression following cell division is a fundamental problem in developmental biology . A common mechanism for establishing cell-specific gene expression is asymmetric segregation of a cell fate determinant between the daughter cells ( Horvitz and Herskowitz , 1992; Neumüller and Knoblich , 2009 ) . In polarized cells , intrinsic asymmetry can be inherited from generation to generation . For example , the dimorphic bacterium Caulobacter crescentus localizes certain cell fate determinants to the old cell pole , leading to their asymmetric distribution following division ( Iniesta and Shapiro , 2008; Bowman et al . , 2011 ) . However , non-polarized cells such as Bacillus subtilis must generate asymmetry de novo , which is passed on to the daughter cells to differentiate . Bacillus subtilis divides by binary fission to produce identical daughter cells during vegetative growth but switches to asymmetric division when undergoing the developmental process of spore formation ( Piggot and Coote , 1976; Stragier and Losick , 1996 ) . To sporulate , cells place a division septum near a randomly chosen pole of the cell ( Veening et al . , 2008 ) to create two unequally sized daughter cells with dissimilar programs of gene expression . The smaller cell , the forespore , which largely consists of the cell pole , will become the spore , whereas the larger cell , the mother cell , nurtures the developing spore ( Figure 1B ) . An enduring mystery of this developmental system is how stochastically generated asymmetry initiates dissimilar programs of gene expression in the daughter cells resulting from polar division ( Barak and Wilkinson , 2005 ) . 10 . 7554/eLife . 08145 . 003Figure 1 . SpoIIE is compartmentalized in the forespore and activates σF . ( A ) Diagram of the pathway for σF activation . Phorphorylated SpoIIAA ( AA-P ) is dephosphorylated by SpoIIE . Dephosphorylated SpoIIAA ( AA ) then binds to SpoIIAB ( AB ) displacing σF and leading to σF-directed transcription . ( B ) SpoIIE ( green ) , AA , AB and σF are produced in predivisional cells . Prior to completion of asymmetric cell division SpoIIE associates with the polar divisome near one or both cell poles ( the pole at which division initiates is chosen randomly ) . Following completion of cytokinesis , SpoIIE is enriched in the forespore where it dephosphorylates AA-P to activate σF . ( C ) A montage of images taken every 6 min from a single sporulating cell ( strain RL5876 ) producing SpoIIE-YFP . Cells are oriented with the forespore on the right as in the diagram . A movie of this sporulating cell is provided as Video 1 . Scale bar: 0 . 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 00310 . 7554/eLife . 08145 . 004Figure 1—figure supplement 1 . SpoIIE constricts along with FtsZ during asymmetric cell division . ( A ) Montage of images of SpoIIE-YFP and CFP-ZapA in a sporulating cell ( strain RL6042 ) from timelapse microscopy . Images were acquired at 5 min intervals on agarose pads following resuspension to induce sporulation . The lower panel is a kymograph of a slice through the asymmetric FtsZ ring from the images above . The merge shows SpoIIE in green and ZapA in red . An asterisk marks the frame with maximal constriction of the FtsZ ring . A movie of this sporulating cell is provided as Video 2 . ( B ) Montage of images of SpoIIE-YFP , CFP-ZapA , and MalFtm-mNeptune in a sporulating cell ( strain RL6043 ) from timelapse microscopy . Images were acquired at 7 min intervals on agarose pads following resuspension to induce sporulation . The lower panel is a kymograph of a slice through the asymmetric FtsZ ring from the images above . An asterisk marks the frame with maximal constriction of the FtsZ ring . A movie of this sporulating cell is provided as Video 3 . ( C ) Structured illumination microscopy image of a sporulating cell ( strain RL6044 ) expressing SpoIIE-GFP ( green ) and MalFtm-mNeptune ( red ) . Of 16 such cells observed with incomplete polar septa , 81% showed clear constriction of the SpoIIE ring . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 00410 . 7554/eLife . 08145 . 005Video 1 . Movie file of the sporulating cell shown in Figure 1C ( 2fps ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 005 The earliest acting cell-specific regulatory protein in the sporulation program is the transcription factor σF . The σF factor and the proteins that control it – SpoIIAB , SpoIIAA , and SpoIIE – are produced at the onset of sporulation ( Gholamhoseinian and Piggot , 1989 ) , but σF is held inactive until the completion of asymmetric cell division , when it turns on gene expression selectively in the forespore ( Margolis et al . , 1991; Stragier and Losick , 1996 ) ( Figure 1A , B ) . SpoIIAB is an anti-sigma factor that traps σF in an inactive complex ( Min et al . , 1993; Duncan and Losick , 1993 ) . Escape from SpoIIAB is mediated by the anti-anti-sigma factor SpoIIAA ( Diederich et al . , 1994 ) . SpoIIAA is , in turn , activated by SpoIIE , a member of the PP2C family of protein phosphatases ( Bork et al . , 1996; Levdikov et al . , 2011 ) . SpoIIE converts the inactive phosphorylated form of SpoIIAA ( SpoIIAA-P ) to the active dephosphorylated form ( Duncan et al . , 1995 ) ( Figure 1A ) . Dephosphorylation of SpoIIAA-P by SpoIIE is therefore the critical event in activating σF . Understanding how SpoIIE reads out cellular cues to delay dephosphorylation of SpoIIAA-P until after septation and restrict phosphatase activity to the forespore is thus the central challenge in understanding how cell-specific gene transcription is established during sporulation . SpoIIE consists of three domains: a PP2C phosphatase domain at the C-terminus , a ten-pass transmembrane domain at the N-terminus , and a 270-amino acid central domain ( henceforth referred to as the regulatory domain ) that is important for regulating SpoIIE compartmentalization and activity ( Figure 2B; Arigoni et al . , 1999 ) . Prior to asymmetric cell division , SpoIIE localizes to the polar divisome and contributes to its placement ( Arigoni et al . , 1995; Ben-Yehuda and Losick , 2002 ) . After septation is complete , SpoIIE is found principally in the forespore and to a limited extent at a second polar divisome near the distal cell pole ( Figure 1C , Video 1 ) . 10 . 7554/eLife . 08145 . 006Figure 2 . SpoIIE degradation depends on FtsH . ( A ) SpoIIE is compartmentalized to the forespore . A single sporulating cell ( strain RL5874 ) is shown following the completion of asymmetric septation . The membrane stained with FM4-64 is shown in red above SpoIIE-YFP and CFP driven by an in frame fusion to the start of the spoIIE open reading frame . Scale bar: 0 . 5 µm . ( B ) The domain architecture of SpoIIE . The N-terminal cytoplasmic tail ( red ) , followed by 10 transmembrane-spanning segments , the regulatory region ( amino acids 320–589 , gray ) , and the phosphatase domain ( amino acids 590–827 , green ) . ( C ) SpoIIE is degraded during sporulation . Translation was arrested ( by addition of 100 µg/ml chloramphenicol ) in sporulating cells producing SpoIIE-FLAG ( strain RL5877 ) , and samples were withdrawn at the indicated times . SpoIIE was detected by western blot using α-FLAG monoclonal antibody ( left ) . Quantitation of the western ( right ) fit to a single exponential equation . ( D ) The genes for spoIIE and ftsH are near each other in the genome with conserved synteny . The diagram shows genomic organization of diverse endospore forming species . Filled boxes indicate genes with spoIIE in green , ftsH in orange , and other genes in gray . ( E ) SpoIIE degradation requires FtsH , and FtsH mediated degradation requires the N-terminal cytoplasmic tail ( TagSpoIIE ) of SpoIIE . Translation was arrested in vegetatively growing cells producing SpoIIE-FLAG with an IPTG inducible promoter ( strain RL5878 wt , RL5879 ∆ftsH , and RL5880 SpoIIE-∆Tag ( residues 11–37 were deleted ) . Degradation was monitored ( left ) and quantitated ( right ) as in panel C . ( F ) TagSpoIIE is sufficient to target a heterologous protein for FtsH-dependent degradation . Degradation of MalF-TM-FLAG fused at the N-terminus to either TagSpoIIE ( wt RL5888 , ∆ftsH RL5889 ) or the first 10 amino acids of SpoIIE ( wt RL5890 , ∆ftsH RL5891 ) produced during exponential growth was monitored as in panel C . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 00610 . 7554/eLife . 08145 . 007Figure 2—figure supplement 1 . SpoIIE degradation depends on FtsH . ( A ) Translation was arrested ( by addition of 100 µg/ml chloramphenicol ) in sporulating wild-type PY79 cells ( strain RL3 ) , and samples were withdrawn at the indicated times . SpoIIE was detected by western blot using affinity purified rat α-SpoIIE polyclonal antibody . ( B ) Degradation of SpoIIE was monitored in vegetative cells induced to produce untagged SpoIIE ( strain RL6040 , wt ) ( strain RL6041 , ∆ftsH ) as in A . ( C ) Domains of SpoIIE-FLAG were serially deleted or replaced as diagramed at left and expressed in wt or ∆ftsH vegetatively growing cells . Translation was arrested by chloramphenicol addition , and samples were collected at times indicated . SpoIIE was detected by western blotting with α-FLAG monoclonal antibody . ∆Tag removes residues 11–37 ( strain RL5880 , wt ) ( strain RL5884 , ∆ftsH ) , MalF-tm replaces the transmembrane domain ( residues 38–319 ) with the first two transmembrane domains of E . coli MalF ( strain RL5881 , wt ) ( strain RL5885 , ∆ftsH ) , ∆Reg removes the regulatory domain ( residues 320–568 ) ( strain RL5882 , wt ) ( strain RL5886 , ∆ftsH ) , and ∆PP2C removes the phosphatase domain ( residues 590–827 ) ( strain RL5883 , wt ) ( strain RL5887 , ∆ftsH ) . Western blots for strains shown in Figure 2E are duplicated here for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 007 Here we describe three interdependent features of SpoIIE that together explain how SpoIIE links polar septation to the cell-specific activation of σF: ( 1 ) SpoIIE is proteolytically unstable and is degraded dependent on the AAA+ protease FtsH; ( 2 ) SpoIIE is transferred from the polar divisome – the de novo origin of asymmetry – to the proximal cell pole during polar septation; and ( 3 ) SpoIIE forms homooligomeric complexes which promote capture at the pole , protection from proteolysis and activation as a phosphatase , thus linking the cues that direct localization of SpoIIE to its stabilization and activation . Although transcription of spoIIE commences in pre-divisional cells and continues in the mother cell following cytokinesis ( Fujita and Losick , 2003 ) , SpoIIE protein and activity are restricted to the forespore ( Figure 2A ) . This apparent contradiction led us to consider the possibility that spatially restricted proteolysis contributes to compartmentalization of SpoIIE . Selective stabilization of SpoIIE in the forespore coupled to efficient global degradation would enrich SpoIIE in the forespore despite ongoing transcription of spoIIE in predivisional cells and the mother cell . To investigate this hypothesis , we sought to determine if SpoIIE turns over on a timescale commensurate with σF activation and , if so , to identify the responsible protease . To detect SpoIIE degradation during sporulation , we monitored the disappearance of SpoIIE with a C-terminal FLAG tag following inhibition of translation with chloramphenicol ( Figure 2C ) . SpoIIE-FLAG was degraded with a half-life of 7 min , demonstrating that SpoIIE is unstable relative to the approximately 1 hr progression of asymmetric cell division and σF activation ( Figure 1C ) and supporting spatially restricted degradation as a plausible mechanism to compartmentalize SpoIIE . Next , we sought to identify the protease that degrades SpoIIE . We noticed that the gene ( ftsH ) for the transmembrane AAA+ protease FtsH is located near spoIIE in the genome with highly conserved synteny ( Figure 2D ) . Furthermore , FtsH is known to degrade transmembrane protein substrates ( Akiyama , 2009 ) , making it an attractive candidate protease for SpoIIE . FtsH degrades several proteins that block entry into sporulation and prevent the expression of spoIIE , such as the Spo0A inhibitor Spo0E ( Le and Schumann , 2009 ) . We therefore engineered the synthesis of SpoIIE during vegetative growth to bypass the requirement for FtsH in the expression of spoIIE . In exponential phase cells deleted for ftsH , SpoIIE was stable for more than 1 hr after chloramphenicol treatment , whereas in ftsH cells SpoIIE was degraded as rapidly as during sporulation ( t1/2 = 7 . 1 min ) ( Figure 2E ) . ( SpoIIE instability and its dependence on FtsH was also seen with untagged SpoIIE [Figure 2—figure supplement 1A , B] ) . We conclude that SpoIIE is degraded in an FtsH-dependent manner . The simplest explanation for this is that SpoIIE is a direct substrate for the protease . Finally , we attempted to identify the feature or features of SpoIIE that renders it susceptible to proteolysis by FtsH . Truncation of the N-terminal , cytosolic tail of SpoIIE ( removal of residues 11 to 37 ) blocked degradation ( Figure 2E , ∆Tag ) , whereas removal of the regulatory domain or the phosphatase domain or substitution of the transmembrane domain with the first two transmembrane segments of E . coli MalF ( MalF-TM ) did not impede FtsH-dependent degradation ( Figure 2—figure supplement 1C ) . Additionally , the first 37 amino acids of SpoIIE ( TagSpoIIE ) were sufficient to confer FtsH-dependent degradation on a heterologous protein , MalF-TM-FLAG ( Figure 2F ) . Therefore , the N-terminal tail of SpoIIE is a tag that is both necessary and sufficient for FtsH-dependent proteolysis . To test whether SpoIIE degradation is required for compartmentalization of σF activity and SpoIIE , we examined the effect of blocking degradation during sporulation . Here and in the experiments that follow , we removed TagSpoIIE from SpoIIE in cells that were wild type for ftsH to selectively block SpoIIE degradation and circumvent off-target effects from other FtsH substrates had we used an ftsH mutation . Indeed , we observed a dramatic increase in aberrant activation of σF in ∆tag spoIIE cells ( Figure 3A ) . Whereas in wild-type cells σF activity was highly specific for the forespore ( less than 2% non-specific activation ) , ∆tag spoIIE caused non-specific activation of σF in 71% of the cells ( Figure 3B ) . Quantification of σF activity with a lacZ reporter revealed that a strain with ∆tag spoIIE activated σF with a similar time dependence but had 10-fold elevated σF activity ( Figure 3C ) . 10 . 7554/eLife . 08145 . 008Figure 3 . Degradation of SpoIIE is required to compartmentalize SpoIIE and σF activity . ( A ) Images of sporulating cells producing SpoIIE-YFP ( left , strain RL5876 ) or ∆Tag SpoIIE-YFP ( right , strain RL5892 ) . Top images show CFP ( blue ) produced under the control of the σF dependent spoIIQ promoter and FM4-64-stained membrane ( white ) ; bottom images show SpoIIE-YFP . White arrows indicate cells with uncompartmentalized σF activity and SpoIIE in the mother cell . The contrast for images of SpoIIE-YFP has been adjusted to approximately 5X brighter than for ∆Tag-SpoIIE-YFP for display purposes . Scale bar: 1 µm . ( B ) Quantification of the forespore specificity of σF activity from hundreds of cells from images as shown in panel A . ( C ) σF activity was measured during sporulation using a translational fusion of the σF dependent SpoIIQ promoter to LacZ ( wt SpoIIE strain RL5893 , ∆Tag SpoIIE strain RL5894 ) . Time after initiation of sporulation is indicated , and error bars represent the standard deviation from three biological replicates . ( D ) Quantification of the dependence of σF activation on asymmetric cell division driven by ∆Tag-SpoIIE . Hundreds of cells from images as shown in panel A were manually assessed for completion of asymmetric cell division and compartmentalization of σF activity . The percent of cells with each pattern of σF activity is indicated for ∆tag-spoIIE cells ( values for wt spoIIE cells are indicated in parenthesis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 00810 . 7554/eLife . 08145 . 009Figure 3—figure supplement 1 . FtsH is active in the forespore . ( A ) Images of sporulating cells expressing MalF-TM-YFP ( the first two transmembrane segments of E . coli MalF fused to YFP ( strain RL5932 ) or TagSpoIIE-MalF-TM-YFP ( strain RL5933 ) under the control of the σF dependent spoIIQ promoter . Cells were stained with FM4-64 ( middle images ) . Images of PσF-MalF-TM-YFP and PσF-TagSpoIIE-MalF-TM-YFP are shown with identical exposure times and contrast settings; bottom images are adjusted to higher contrast than top images to show limited accumulation of PσF-TagSpoIIE-MalF-TM- YFP in the forespore . ( B ) Sporulating cells expressing MalF-TM-FLAG ( strain RL5934 ) or TagSpoIIE-MalF-TM-FLAG ( strain RL5935 ) under the control of the σF dependent spoIIQ promoter were translationally arrested , and samples were taken at the indicated timepoints . MalF-TM-FLAG was detected by western blotting with α-FLAG monoclonal antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 009 Activation of σF is tightly coupled to the completion of asymmetric cell division , and SpoIIE mutants have been characterized that uncouple cell division and σF activation ( Carniol et al . , 2004; Feucht , et al . , 2002; Hilbert and Piggot , 2003 ) . In contrast to these other cases of σF mis-activation in predivisional cells , stabilization of SpoIIE led to activation of σF primarily in cells that had completed asymmetric cell division ( Figure 3A , D ) . Thus , stabilization of SpoIIE only partially uncouples σF activation from cell division . Consistent with the idea that degradation contributes to compartmentalization of σF activity by helping to restrict SpoIIE to the forespore , we observed a striking correlation between elevated levels of ∆Tag-SpoIIE in the mother cell and mis-activation of σF ( Figure 3A bottom panels ) . Together , these data indicate that SpoIIE degradation is required for compartmentalization both of SpoIIE and σF activity . How is SpoIIE selectively stabilized in the forespore ? We considered two models: ( 1 ) FtsH is not active in the forespore , or ( 2 ) specific features of SpoIIE stabilize it in the forespore . To test the former possibility , we engineered the production of the model FtsH substrate TagSpoIIE-MalF using a forespore specific , σF-dependent promoter . TagSpoIIE-MalF was rapidly degraded in a manner dependent on the TagSpoIIE ( Figure 3—figure supplement 1 ) . Therefore FtsH is active in the forespore , suggesting that SpoIIE is specifically stabilized against FtsH-dependent degradation . To identify features of SpoIIE required for its accumulation in the forespore , we screened for SpoIIE variants defective in compartmentalization . We created amino acid substitutions of the most highly conserved residues in SpoIIE and tested these variants ( and previously described variants ) for function in sporulation ( Figure 4A , Figure 4—source data 1 ) and forespore accumulation ( Figure 4B ) . To monitor accumulation in the forespore of each SpoIIE variant , we compiled average profiles of SpoIIE-YFP along the long axis of hundreds of cells that had undergone polar division . Through this analysis , we identified nine variants of SpoIIE ( for example SpoIIEK356D ) that were absent in the forespore ( Figure 4A , B blue ) and accumulated to reduced levels ( Figure 4C ) . Variants with normal compartmentalization , in contrast , accumulated at approximately wild-type levels ( Figure 4B , C black ) . Supporting the idea that failure to accumulate in the forespore was due to unrestricted , FtsH-dependent degradation , removal of TagSpoIIE restored these SpoIIE mutant proteins to levels several-fold higher than for wild-type SpoIIE and equivalent to ∆Tag SpoIIE ( Figure 4D ) . We conclude that SpoIIE undergoes a transition in the forespore that protects it from FtsH-dependent proteolysis and that this transition is blocked by amino acid substitutions such as K356D . 10 . 7554/eLife . 08145 . 010Figure 4 . SpoIIE stabilization and localization mutants . ( A ) Diagram of SpoIIE mutants with sporulation defects . Variants with localization and σF activation defects are shown above the diagram in blue , and variants with normal localization but defects in σF activation are shown below the diagram in black . ( B ) Localization of SpoIIE mutants . Hundreds of asymmetrically divided cells were aligned at the forespore pole to generate average profiles of SpoIIE-YFP localization for each SpoIIE mutant ( strains RL5895- 5909 ) with a reference plot ( gray ) from wild-type SpoIIE-YFP from σF mutant cells ( strain RL5910 ) . The dashed line represents the approximate position of the asymmetric septum . Images of representative cells with the mislocalized variant SpoIIEK356D ( left , strain RL5895 ) and forespore-localized SpoIIEQ483A ( right , strain RL5904 ) are shown . ( C ) Western blots of protein levels in SpoIIE mutant strains shown in panel B probed for SpoIIE-YFP ( with α-GFP antibody ) , CFP produced under the control of a σF-driven promoter , and σA as a loading control . Levels of each SpoIIE variant were normalized to wt SpoIIE ( strain RL5876 ) . Error bars represent the standard deviation from three biological replicates . All localization mutants ( shown in blue ) are different from the σF mutant control with p values less than 0 . 0025 from a paired t-test . ( D ) Western blots of SpoIIE compartmentalization mutants with TagSpoIIE removed ( strains RL5911-5916 , with strain RL5876 as a reference ) as in panel C . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01010 . 7554/eLife . 08145 . 011Figure 4—source data 1 . Sporulation efficiency of SpoIIE mutants . Cultures of indicated strains with were grown in DSM for 28 hr at 37°C followed by heat killing for 20 min at 85°C . The number of spores was determined by counting viable cells . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01110 . 7554/eLife . 08145 . 012Figure 4—figure supplement 1 . An allele specific suppressor of SpoIIEK356D rescues compartmentalization and stabilization . ( A ) Diagram of SpoIIE with the position of the mutation K356D shown above and the suppressor mutations T353I and V697A shown below . ( B ) SpoIIE variants were detected in extracts of sporulating cells by western blotting using an α-GFP antibody ( wt strain RL5876 , SpoIIEK356D strain RL5895 , SpoIIEK356D , T353I strain RL5936 , SpoIEK356D , V697A strain RL5939 ) . σF activity was measured in these samples using a CFP reporter fused to the σF dependent spoIIQ promoter . ( C ) Localization of SpoIIE mutants . Average profiles of SpoIIE-YFP localization from hundreds of asymmetrically divided cells aligned at the forespore pole are shown for SpoIIEK356D , T353I ( strain RL5943 ) , and for SpoIIEK356D , V697A ( strain RL5944 ) . An image of a representative cell expressing SpoIIEK356D , T353I ( strain RL5943 ) following asymmetric septation is shown above . ( D ) Quantification of the forespore specificity of σF activity based on CFP fused to the σF dependent spoIIQ promoter using strains as in C . ( E ) Multiple turnover measurements of dephosphorylation of SpoIIAA-P . Data were fit to the Michaelis-Menton equation; Vmax = 7 . 0 min–1 for SpoIIE320-827 and 44 . 3 min–1 for SpoIIE320-827 , V697A . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 012 Additionally , we tested whether stabilization of the SpoIIE mutant proteins was sufficient to support σF activation independent of their susceptibility to degradation . We found that even when TagSpoIIE was removed , the mutant proteins failed to activate σF ( Figure 4D middle panel ) . A simple unifying model is that the proposed , K356-dependent conformational rearrangement that protects SpoIIE from proteolysis in the forespore is also required to allow it to activate σF . To investigate the link between stabilization , compartmentalization and activation of SpoIIE , we selected for and isolated several suppressors that restored sporulation to the compartmentalization-defective mutant spoIIEK356D . We chose this mutant because the K356D substitution was located in the regulatory domain of SpoIIE and caused a particularly severe sporulation defect . We isolated intragenic suppressors at two codons in an apparently saturating screen ( see Materials and methods , Figure 4—source data 1 and Figure 4—figure supplement 1 ) . One of them , causing a T353I substitution , was allele-specific ( it suppressed spoIIEK356D but not spoIIES361F or spoIIEV490K , Figure 4—source data 1 ) . The T353I substitution restored SpoIIEK356D to wild-type protein levels ( Figure 4—figure supplement 1B ) , partially restored restriction of SpoIIEK356D to the forespore ( Figure 4—figure supplement 1C ) , and restored compartment-specific σF activation ( Figure 4—figure supplement 1D ) . The coordinated rescue of these phenotypes by a single amino acid substitution supports our model that a common feature of SpoIIE mediates protection from proteolysis , accumulation in the forespore , and activation of σF . The other intragenic suppressors of SpoIIEK356D were substitutions at V697 ( V697A and V697F ) , which is located in the phosphatase domain of SpoIIE . V697A had been independently isolated previously and shown to cause premature activation of σF ( in the absence of the K356D substitution ) ( Hilbert and Piggot , 2003 ) . These suppressors were not allele specific; V697A suppressed all other mutants of SpoIIE , including the compartmentalization defective SpoIIES361F mutant and the compartmentalized SpoIIEQ483A mutant ( Figure 4—source data 1 , [Carniol et al . , 2004] ) . The V697A substitution did not restore compartmentalization or stabilization of SpoIIE . It did restore σF activation but not compartmentalization of σF activity ( Figure 4—source data 1 and Figure 4—figure supplement 1 ) . All together , these results suggest that the V697A substitution locks the phosphatase domain in a high activity state , bypassing the activation defect of SpoIIEK356D . Indeed , biochemical experiments showed that the V697A substitution enhanced the activity of SpoIIE in dephosphorylating SpoIIAA-P ( Figure 4—figure supplement 1E ) . To further investigate the mechanism of SpoIIE compartmentalization , we took advantage of the compartmentalization-defective variant SpoIIEK353D and revisited the localization of stabilized SpoIIE . To isolate events prior to σF activation , and because certain targets of σF ( e . g . spoIIQ ) affect the localization of SpoIIE ( Campo et al . , 2008 ) , we used a mutant lacking σF to analyze the localization of SpoIIE , ∆Tag-SpoIIE , and its K353D mutant derivative ( in contrast to the experiment of Figure 3A in which cells were σF+ ) . Our most striking observation was that ∆Tag-SpoIIE was noticeably enriched at the poles of cells that had not initiated polar division ( Figure 5A ) . Polar enrichment was dependent on stabilization by removal of TagSpoIIE ( Figure 5A gray line ) . Because the forespore is derived from the cell pole , we hypothesized that the pole is a landmark that directs SpoIIE compartmentalization . In support of this idea , polar localization was abolished by the K356D substitution and partially restored by the T353I suppressor ( Figure 5A lower panel ) . Thus , the pole is a cue that directs compartmentalization of SpoIIE , and the same feature ( s ) of SpoIIE that is required for polar recognition is also required for stabilization and σF activation . 10 . 7554/eLife . 08145 . 013Figure 5 . Stabilized SpoIIE localizes to the cell pole . ( A ) Average profiles of SpoIIE-YFP from undivided sporulating cells lacking σF activity are shown ( strains RL5910 , 5917 , 5912 , 5918 ) , with a representative cell with ∆Tag-SpoIIE-YFP ( strain RL5917 ) displayed above . ( B ) Vegetatively growing cells expressing SpoIIE-YFP ( strains RL5919-5922 ) displayed as in panel A; variants as indicated ) were imaged 30 min after induction of expression of the FtsZ polymerization inhibitor MciZ . ( C ) MciZ-expressing cells ( ΔdivIVA [strain RL5923] or otherwise wildtype [strain RL5924] ) were imaged as in panel B and average profiles of ∆Tag-SpoIIE-YFP were generated from 20 randomly selected cell poles . ( D ) DivIVA-FLAG was immunoprecipitated with α-FLAG magnetic beads from extracts of sporulating cells expressing SpoIIE variants as indicated ( strains RL5925 , 5926 ) , and detected by Western blot . The elution ( e ) shown is 100X concentrated relative to the load ( l ) and flowthrough ( ft ) samples . Blots were probed with α-GFP antibody ( top two images; lower image in high contrast* ) and α-DivIVA antisera ( below ) . Because DivIVA oligomerizes , untagged DivIVA is also co-immunoprecipitated . ( E ) SpoIIE preferentially localizes to the divisome rather than the pole . Representative cells expressing SpoIIE-YFP and CFP-ZapA are shown . Left images show exponentially growing divICts cells ( strains RL5927 , 5928 ) , and right images show a sporulating ΔdivIB cell ( strain RL5929 ) . Scale bars indicate 0 . 5 µm in all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01310 . 7554/eLife . 08145 . 014Figure 5—figure supplement 1 . Transcription of spoIIE in the forespore is not required to compartmentalize SpoIIE . ( A ) Representative image of SpoIIE-YFP in a sporulating ∆racA mutant cell ( strain RL6045 ) that failed to capture the chromosome in the forespore . The chromosome was stained with DAPI and the membrane was stained with FM4-64 . ( B ) Representative image of a SpoIIE-YFP in a sporulating cell harboring the spoIIIE36 mutation to block pumping of the chromosome to the forespore and with the spoIIE-YFP gene near the terminus ( strain RL6046 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01410 . 7554/eLife . 08145 . 015Video 2 . Movie file of the sporulating cell shown in Figure 1—figure supplement 1A ( 2fps ) . SpoIIE-YFP is shown in grey , and the divisome marked by CFP-ZapA is shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01510 . 7554/eLife . 08145 . 016Video 3 . Movie file of the sporulating cell shown in Figure 1—figure supplement 1B ( 2fps ) . SpoIIE-YFP is shown in grey , the divisome marked by CFP-ZapA is shown in blue , and the membrane marked by MalFtm-mNeptune is shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 016 We next asked whether SpoIIE has an intrinsic affinity for the cell pole or whether SpoIIE is captured there by features unique to cells undergoing sporulation . To address this question , we engineered the synthesis of SpoIIE-YFP and ∆Tag SpoIIE-YFP in vegetative cells that were blocked in divisome formation through the use of the FtsZ inhibitor MciZ ( Handler et al . , 2008 ) . We observed that ∆Tag SpoIIE-YFP was enriched at the ends of these cells , and this localization recapitulated the features of polar localization seen during sporulation: polar localization was only observed for stabilized SpoIIE , was blocked by K356D substitution , and was restored by the T353I suppressor ( Figure 5B ) . Therefore , a fundamental , constitutive feature of the cell pole mediates SpoIIE polar localization . We next sought to identify the feature of the pole that is responsible for SpoIIE localization . DivIVA recognizes the negative curvature of the cell pole and directs the polar localization of several other proteins during growth and sporulation ( Lenarcic et al . , 2009; Ramamurthi and Losick , 2009 ) . Recently , DivIVA was shown to co-immunoprecipitate with SpoIIE , making it an attractive candidate to anchor SpoIIE to the cell pole ( Eswaramoorthy et al . , 2014 ) . We found that whereas wild-type SpoIIE co-immunoprecipitated with DivIVA from extracts of sporulating cells ( as previously observed ) , SpoIIEK356D did not ( Figure 5D ) . Additionally ∆Tag SpoIIE-YFP polar localization during vegetative growth was abolished by a divIVA deletion ( in the background of a minD deletion to suppress the cell division defect of divIVA deletion ) ( Figure 5C ) . Thus , DivIVA directly or indirectly anchors SpoIIE at the cell pole and can do so independently of sporulation . Based on the result with the K356D mutant , we further propose that this anchoring serves to stabilize , compartmentalize and activate SpoIIE , and that these activities are linked through a common feature of SpoIIE . During sporulation , SpoIIE first accumulates at the polar divisome , constricts along with the septum and then is released into the forespore following the completion of cytokinesis ( as shown by time-lapse and structured illumination microscopy in Figure 1C , Videos 1–3 , Figure 1—figure supplement 1 ) . This suggests that the divisome competes with the pole for SpoIIE binding and that SpoIIE is not free to associate with the pole until the divisome is disassembled . To investigate this model , we monitored SpoIIE localization in the background of a temperature-sensitive allele of divIC that stalls cell division after divisome formation but before cytokinesis ( Levin and Losick , 1994 ) . In this background , SpoIIE-YFP and ∆Tag SpoIIE-YFP localized to the divisome ( as visualized with a ZapA-CFP fusion ) but not to the cell pole ( Figure 5E ) . Similarly , when cytokinesis was blocked during sporulation by a divIB deletion ( Thompson et al . , 2006 ) , ∆Tag SpoIIE-YFP localized to the divisome but not to the cell pole ( Figure 5E ) . Therefore , the divisome sequesters and prevents SpoIIE from associating with the cell pole . We conclude that SpoIIE has affinity for two subcellular sites: the divisome , its dominant binding site , and the pole , where it is captured only after release from the divisome after the completion of cytokinesis . The results discussed above suggest a simple model for how SpoIIE and σF activity are compartmentalized in the forespore . We propose that SpoIIE is sequestered at the asymmetrically positioned divisome and is released and captured at the proximal ( forespore ) pole when cytokinesis is completed . In support of this idea , cells that cannot synthesize additional SpoIIE molecules in the forespore nonetheless robustly compartmentalize SpoIIE ( Figure 5—figure supplement 1 ) . Asymmetric compartmentalization of SpoIIE in the forespore could be achieved by virtue of the close proximity of the divisome and the forespore pole . Weak SpoIIE association with the pole would be compensated for by the small volume of the forespore and reinforced by protection from degradation by FtsH . Finally , any SpoIIE released into the mother cell would be captured at the divisome , preventing capture at the mother cell pole . Thus , a simple model explains how SpoIIE is protected from degradation and compartmentalized in the forespore only after cytokinesis is complete . The heart of this model is that asymmetric positioning of the division septum is all that is necessary for compartment specific stabilization and activation of SpoIIE . To test this prediction , we sought to compartmentalize SpoIIE in cells that had been engineered to undergo polar division independently of sporulation . To do so , we artificially expressed spoIIE and overexpressed ftsAZ in vegetative cells , which was previously demonstrated to reposition the division septum from the mid-cell to near the pole ( Ben-Yehuda and Losick , 2002 ) . To preclude transcription of sporulation-specific genes , we additionally deleted the master regulator for entry into sporulation , spo0A . We then visualized the localization of SpoIIE in these cells . As predicted by our model , cells enriched for SpoIIE were much smaller than average for the entire population ( Figure 6A , B ) . Further , to ask whether this compartmentalization of SpoIIE was sufficient to direct cell-specific activation of σF , we additionally induced synthesis of σF and its regulators SpoIIAA ( the anti-anti σF factor substrate for the SpoIIE phosphatase ) and SpoIIAB ( the anti-σF factor ) in a strain harboring a reporter for σF activity . Remarkably , σF was activated with high selectivity in a subpopulation of the minicells ( Figure 6A , B ) . We conclude that polar division is the only feature of sporulation necessary to restrict SpoIIE protein and activity to the small cell and that this is sufficient to explain compartmentalized activation of σF . 10 . 7554/eLife . 08145 . 017Figure 6 . Repositioning the septum in vegetative cells is sufficient for compartmentalization of SpoIIE . ( A ) In the top image , vegetatively growing cells producing SpoIIE-GFP and overexpressing the ftsZ operon formed minicells enriched for SpoIIE-GFP ( strain RL5930 ) . In the lower image , cells additionally expressed the spoIIA operon and harbored a YFP reporter for σF activity ( strain RL5931 ) . ( B ) Quantification of images as shown in panel A . SpoIIE enriched cells ( representing 56 out of 3168 total cells ) are shown in green in the middle plot and are defined as cells with 2 standard deviations above the mean SpoIIE-GFP intensity . Cells that had active σF were rare in the population ( 12 cells ) and were identified based on detectable levels of YFP fluorescence and their sizes were measured . They are shown in blue at the bottom . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 017 To gain insight into the molecular mechanism of SpoIIE localization and activation , we expressed and purified the C-terminal cytosolic domain of SpoIIE ( residues 320 to 827 , SpoIIE320-827 ) for biochemical characterization . A striking feature of SpoIIE320-827 was that it multimerizes , and analytical ultracentrifugation revealed these multimers to be hexamers and higher order assemblies of hexamers ( Figure 7A ) . 10 . 7554/eLife . 08145 . 018Figure 7 . Multimerization is required for compartmentalization of SpoIIE and σF activation . ( A ) Purified soluble SpoIIE320-827 was analyzed by sedimentation velocity analytical ultracentrifugation detected by absorbance at 280nm and fitted to c ( s ) using Sedfit . Peaks corresponding to the predicted sedimentation coefficient for monomeric SpoIIE , hexamers , and multimers of hexamers were observed . ( B ) Diagram of the mutations and truncations in SpoIIE analyzed in panel C . ( C ) Multimerization of SpoIIE variants was analyzed by gel filtration with a 24 ml Superose 6 column . 1 ml fractions from 7-–18 ml of the run were collected , run on SDS-PAGE gels , and stained with SYPRO Ruby . ( D ) Model for handoff of SpoIIE from the divisome to the adjacent cell pole . SpoIIE ( green ) initially accumulates at the divisome and constricts along with FtsZ during cytokinesis . Prior to the completion of cell division , SpoIIE is degraded by FtsH through its TagSpoIIE ( red ) . ( We cannot distinguish whether association with the divisome protects SpoIIE from proteolysis or if SpoIIE turns over while associated with the divisome . ) Upon the completion of cytokinesis , SpoIIE transfers to the adjacent cell pole where multimerization protects it from proteolysis ( as depicted by the light orange Tags ) and leads to phosphatase activation . We propose that close proximity favors transfer to the immediately adjacent pole and that concentration of SpoIIE in the forespore , which is almost entirely derived from the pole , promotes multimerization . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 01810 . 7554/eLife . 08145 . 019Figure 7—figure supplement 1 . SpoIIE variants defective for sporulation multimerize . Multimerization of SpoIIE variants was analyzed by gel filtration with a 24 ml Superose 6 column . 1 ml fractions from 7–18 ml of the run were collected , run on SDS-PAGE gels , and stained with SYPRO Ruby . SpoIIE variants that fail to accumulate in the forespore are marked in blue . Gels of wt SpoIIE and SpoIIEK356D are reproduced from Figure 7C for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 08145 . 019 To identify determinants of multimerization and its role for SpoIIE function , we made serial N-terminal truncations of SpoIIE starting at residue 320 and determined the oligomeric state of each by gel filtration ( Figure 7B , C ) . We found that the 13-amino acid interval from residues 345 to 358 contained a feature required for multimerization; although SpoIIE truncated to residue 345 ( SpoIIE345-827 ) multimerized , all truncations extending to 358 ( SpoIIE358-827 ) and beyond did not ( Figure 7C ) . This region encompasses K356 , raising the possibility that K356D might block stabilization and activation of SpoIIE at the pole by blocking multimerization , and that T35I might suppress the phenotypes of K356D by restoring multimermization . Indeed , gel filtration experiments confirmed that SpoIIE320-827 , K356D did not multimerize and that T353I partially restored multimerization ( Figure 7C ) . Of all the SpoIIE variants analyzed , the K356D substitution uniquely blocked multimerization , highlighting a key role for the N-terminal region of the regulatory domain in mediating multimerization ( Figure 7—figure supplement 1 ) . In sum , these biochemical experiments in conjunction with the in vivo experiments described above lead us to propose that multimerization of SpoIIE is the critical transition that stabilizes SpoIIE , enabling it to recognize the cell pole and leading to its activation as a phosphatase . First , we found that multimerization is required for stabilization of SpoIIE and compartmentalization of SpoIIE to the forespore ( Figure 4B , C—figure supplement 1 ) . Second , removal of TagSpoIIE uncoupled multimerization from degradation , and revealed an additional link between SpoIIE activation and multimerization ( Figure 4D ) . Finally , we found that recognition of the cell pole by SpoIIE also depended on multimerization , even when SpoIIE was synthesized in vegetative cells uncoupled from degradation and σF activation ( Figure 5B ) . A hallmark of sporulation is a process of asymmetric division that creates a septum near one randomly selected pole of the cell . Polar placement of the septum directs the protein phosphatase SpoIIE to activate σF in the resulting forespore . How SpoIIE activates σF at the right time and in the right place has been one of the enduring mysteries of this developmental system . As the end of the cell used for asymmetric division is chosen without regard to whether it is the old or new pole ( Veening et al . , 2008 ) , the cues that SpoIIE interprets to achieve cell-specific activation of σF must arise de novo , that is , from the position of the septum rather than from preexisting asymmetry . Indeed , our results show that no feature of the sporulation process other than polar placement of the septum is necessary for compartmentalizing SpoIIE and for cell-specific activation of σF ( Figure 6 ) . In addition , as transcription of spoIIE commences prior to asymmetric division ( Fujita and Losick , 2003 ) ( Figure 2A ) , SpoIIE must also respond to temporal cues to ensure it is not active prior to the completion of cytokinesis . Here we have provided evidence for a model in which SpoIIE leverages the asymmetric position of the septum to selectively associate with the adjacent cell pole of the forespore where it is stabilized and activated ( Figure 7D ) . Three key features of our model are as follows: Together these three features provide a simple mechanism for how cues derived from asymmetric cell division restrict SpoIIE to the forespore and couple σF activation to the completion of cytokinesis . At the same time our model raises several unanswered questions important both for understanding sporulation and diverse related biological systems . How does SpoIIE localize to the cell pole and the divisome ? Localization to the divisome depends on FtsZ and FtsA , the earliest assembling proteins to define the divisome ( Levin et al . , 1997 ) . But whether SpoIIE interacts with these proteins directly , what features of SpoIIE mediate divisome association , and how SpoIIE influences FtsZ polymerization and divisome maturation are unknown . Answering these questions will help us to understand how SpoIIE is transferred from the divisome to the cell pole as well as how SpoIIE influences the position of the division septum . Similarly , localization to the pole depends on DivIVA , which directly senses the shape of the pole and acts as an organizing center for other pole-associated proteins ( Lenarcic et al . , 2009; Ramamurthi and Losick , 2009 ) . But it is not known whether this interaction is direct or depends on an accessory protein . How does oligomerization of SpoIIE promote σF activation ? Genetic and biochemical evidence are consistent with a model in which stabilization , compartmentalization , and activation of SpoIIE are linked by oligomerization of SpoIIE molecules and that this oligomerization takes place in the forespore after asymmetric division . We cannot exclude the possibility , however , that oligomerization commences earlier in sporulation and that some other unrecognized feature of SpoIIE is additionally required for its transition to a stable and active state in the forespore . Structural information about the organization of SpoIIE oligomers and an in vivo assay for oligomerization may help distinguish between these possibilities and yield new insights into how it contributes to compartment specific σF activation . How is activation of σF coordinated with the completion of asymmetric cell division ? Our model proposes two mechanisms to prevent predivisional activation of σF: First , the features of the forespore ( small size , high concentration of cell pole , and proximity to the divisiome ) that promote SpoIIE stabilization and σF activation are all emergent properties that depend on completion of cell division . Second , competition between the divisome and cell pole for binding to SpoIIE prevents premature accumulation and activation of σF . Although there has been uncertainty about when SpoIIE is released from the divisome and when asymmetry in SpoIIE compartmentalization is established ( Eswaramoorthy et al . , 2014; Lucet et al . , 2000; Wu et al . , 1998 ) , our time-lapse imaging and structured illumination microscopy indicate that SpoIIE constricts along with the FtsZ ring during cytokinesis ( Figure 1 , Figure 1—figure supplement 1 ) . This is consistent with our model that SpoIIE remains sequestered at the divisome until cytokinesis is completed . In the future it will be important to determine just how association with the divisome prevents SpoIIE from oligomerizing and activating σF . How is SpoIIE protected from degradation in the forespore ? We have shown that the N-terminal tail of SpoIIE is necessary and sufficient for FtsH-dependent degradation ( Figure 2E , F ) and that stabilization in the forespore is mediated by features of SpoIIE that are required for interaction with the cell pole ( Figure 3 , 4 ) . Additionally , we have presented evidence that multimerization of SpoIIE is required for both stabilization and interaction with the pole . One possibility is that multimerization shields the TagSpoIIE from FtsH as depicted in Figure 7D . Alternatively , as FtsH has been shown to have weak unfoldase activity ( Herman et al . , 2003 ) , multimerization might render SpoIIE resistant to FtsH unfolding and hence proteolysis . Finally , although we favor the view that SpoIIE is directly recognized by FtsH , it is conceivable that it requires an adaptor as is the case for some substrates of AAA+ proteases ( Gottesman , 2003 ) . If so , SpoIIE could be protected from degradation by negative regulation of the adaptor . How is the phosphatase activity of SpoIIE regulated ? Our genetic analysis provides clues for how activation occurs . We found that the V697A substitution locks SpoIIE in a high activity state in vitro , and restores σF activity in mutants defective for compartmentalization and σF activation . V697 is in an active site proximal loop ( Levdikov et al . , 2011 ) ; in many PP2C phosphatases this loop coordinates a third manganese ion that is critical for activity ( Su et al . , 2011 ) . However , SpoIIE lacks the aspartate that coordinates this manganese , which could indicate that V697A locks the phosphatase in a conformation that compensates for the missing manganese ion . Additionally , we found that the stimulation of phosphatase activity ( and binding to the pole ) is genetically linked to multimerization: oligomerization , and σF activation were blocked by the substitution K356D and restored by T353I . We therefore speculate that multimerization induces a conformational change that organizes the catalytic center , compensating for the missing manganese and activating the phosphatase . A test of our hypothesis for a multimerization-dependent conformational change in the active site will require reconstituting multimerization-dependent activation of SpoIIE in vitro . Other PP2C phosphatases , such as the tumor suppressor protein PHLPP ( Gao et al . , 2005 ) , similarly lack the aspartate to coordinate a third magnesium ion , suggesting that our speculation , if correct , could represent a more general regulatory mechanism for PP2C phosphatases . In summary , we propose that the asymmetrically positioned division machinery – the de novo-generated source of asymmetry – positions SpoIIE to be captured at the adjacent cell pole , triggering σF-directed gene expression in the forespore . Capture at the pole , proteolytic stabilization and stimulation of the phosphatase all depend on oligomerization of SpoIIE . Thus , three interlinked regulatory events are sufficient to explain how SpoIIE exploits a stochastically generated spatial cue to the cell-specific activation of a transcription factor . B . subtilis strains were constructed in PY79 using standard molecular genetic techniques ( Harwood and Cutting , 1990 ) . Full details of strain genotypes , and construction are provided in Supplementary file 1 . For IPTG dependent expression ( Plac ) , the hyperspank promoter was used ( from pDR111a , gift of David Rudner ) , and for σF dependent expression ( PσF ) , the spoIIQ promoter was used . Constructs were made by Gibson Assembly ( New England Biolabs , Ipswitch , MA ) , and point mutations were introduced using QuikChange mutagenesis ( Agilent Technologies ) . Suppressors of the spoIIE-K356D mutation were isolated by growing 100 ml cultures of strain RL5895 in DSM sporulation medium at 37°ºC for 28 hr . 11 ml of cells were heat killed at 80°C and used to re-inoculate a new 100 ml DSM culture . Heat killed cells from the second round culture were plated on DSM agar plates . Genomic DNA was prepared from the strain and retransformed to strain RL5875 , lacking spoIIE , to confirm linkage to spoIIE , and the spoIIE locus was sequenced . Finally , suppressor mutations were then reconstructed by quickchange mutagenesis . Mutations T353I , V697A , V697F , and pseudorevertants K356T and K356Y were each isolated from multiple independent cultures , suggesting that the screen was near saturation . Protein degradation rates were measured by shutting off translation by addition of chloramphenicol ( 100 µg/ml ) to cultures . Samples were removed at indicated timepoints and immediately put on ice . Cells were lysed by mechanical disruption in a FastPrep ( MP-BIO , Santa Ana , CA ) . Western blots were conducted by standard procedures and imaged on a BioRad ChemiDoc imager using chemiluminescence . Antibodies used were polyclonal anti-GFP ( Rudner and Losick , 2002 ) , polyclonal anti-σA ( Fujita , 2000 ) , polyclonal anti-DivIVA ( Eswaramoorthy et al . , 2014 ) , and monoclonal anti-FLAG M2 ( Sigma Aldrich , St . Louis , MO ) . Standards were used to determine linearity in each experiment . Immunoprecipitation of DivIVA was performed as published ( Eswaramoorthy et al . , 2014 ) . All micrographs were acquired on an Olympus BX61 upright fluorescence microscope with a 100X objective , with the exception of timelapse images taken on a Nikon ti inverted microscope . Cells were immobilized on 2 . 5% agarose pads made with sporulation resuspension medium . To quantatitatively analyze micrographs , cells were segmented from phase images using either MicrobeTracker ( Sliusarenko et al . , 2011 ) or SupperSeggerOpti ( Kuwada et al . , 2015 ) , and analyzed with custom MatLab scripts ( scripts for quantitative image analysis are included as a Source code 1 ) . Cell specificity of σF activity was determined by analyzing the distribution of PσF-CFP along the long axis of cells . SpoIIE localization profiles were calculated for each cell as the normalized ratio of SpoIIE-YFP to FM4-64 along the long axis of the cell . Because SpoIIE is a transmembrane protein , FM4-64 accounts for differences in membrane area along the cell axis . For sporulating cells that had undergone division , division septa were detected using FM4-64 and cells were oriented based on the position of the polar septum . Vegetative cells ( strains RL5930 and RL5931 ) were induced to produce minicells by dilution from a log phase overnight culture to OD 0 . 05 and addition of 1mM IPTG , and 0 . 25% xylose as appropriate . Cells were imaged after 4 hr of growth at 37°C . Segmentation was performed based on phase images subtracted for FM4-64 to identify division septa in chained cells . Cells enriched for SpoIIE-GFP were defined as cells with average SpoIIE-GFP intensity greater than two standard deviations above the mean . Structured illumination microscopy was performed on a Zeiss Elyra microscope in the Harvard Center for Biological Imaging . SpoIIE was expressed as an N-terminal Sumo-6His fusion in BL21 ( DE3 ) cells following overnight induction with 0 . 5 mM IPTG at 14°C . Cells were lysed in 50 mM Tris pH 8 . 5 , 200 mM NaCl , 1mM beta-mercaptoethanol and purified using HisTrap columns ( GE Healthcare , Pittsburg , PA ) eluting with a gradient of imidazole . The Sumo-6His tag was removed by cleavage with Ulp1 ( Sumo Protease ) followed by Ni-NTA subtraction . For velocity analytical ultracentrifugation , SpoIIE was dialyzed to 20 mM Tris pH8 . 5 , 100 mM NaCl , 2 mM DTT overnight and data was collected at 280 nM spinning at 20 , 000 RPM at The Biophysical Instrumentation Facility ( NSF-0070319 ) at MIT . Data were fit to a continuous model using SedFit ( Schuck , 2000 ) . Gel filtration was conducted on a 24 ml Superose 6 column ( GE Healthcare , Pittsburg , PA ) , loading 100 µl of 1 µM SpoIIE . Phosphatase assays of soluble fragments of SpoIIE lacking the transmembrane domain ( SpoIIE320-827 and SpoIIE320-827 , V697A ) were performed using 32P phosphorylated SpoIIAA ( phosphorylated by purified SpoIIAB ) as a substrate . Multiple turnover reactions were performed with 0 . 05 µM SpoIIE and varying concentrations of SpoIIAA-P as indicated . Dephosphorylation of SpoIIAA was detected by TLC chromatography on PEI-Cellulose plates developed in 1 M LiCl , 0 . 8 M Acetic acid .	An important question in biology is how genetically identical cells activate different sets of genes . This is particularly perplexing for cells that rely on random events to specify the genes they switch on . Normally , cells of a bacterium called Bacillus subtilis divide symmetrically to produce two identical cells that express identical sets of genes . However , B . subtilis cells can also undergo a developmental program to form a spore to help it survive periods of extreme conditions . To do this , first a B . subtilis cell divides asymmetrically by placing the site of division close to a randomly selected end of the cell . This creates a smaller cell that becomes the spore and a larger cell that nurtures the developing spore . Each cell must turn on different genes to play its role in spore development , but how asymmetry in the position of cell division leads to these differences in gene expression has been a longstanding mystery . Bradshaw and Losick studied a regulatory protein called SpoIIE , which is responsible for switching on genes in the small cell . SpoIIE is made before cells divide asymmetrically , but only accumulates in the small cell . The experiments revealed that an enzyme broke down the SpoIIE protein if it wasn’t in the small cell . This prevented SpoIIE from incorrectly switching on genes before division was completed or in the large cell . Protection of SpoIIE from being broken down in the small cells was then shown to be linked to the placement of cell division; SpoIIE first accumulates at the asymmetrically positioned cell division machinery and then is transferred to a secondary binding site at the nearby end of the cell . Capture of SpoIIE at the end of the cell was coupled to its stabilization as SpoIIE molecules interacted with one another to form large complexes . Together these findings provide a simple mechanism to link the asymmetric position of cell division to differences in gene expression . Future studies will focus on understanding how SpoIIE is captured at the end of the cell and how this prevents SpoIIE from being degraded .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2015	Asymmetric division triggers cell-specific gene expression through coupled capture and stabilization of a phosphatase
Elucidating signaling pathways that regulate cellular metabolism is essential for a better understanding of normal development and tumorigenesis . Recent studies have shown that mitochondrial pyruvate carrier 1 ( MPC1 ) , a crucial player in pyruvate metabolism , is downregulated in colon adenocarcinomas . Utilizing zebrafish to examine the genetic relationship between MPC1 and Adenomatous polyposis coli ( APC ) , a key tumor suppressor in colorectal cancer , we found that apc controls the levels of mpc1 and that knock down of mpc1 recapitulates phenotypes of impaired apc function including failed intestinal differentiation . Exogenous human MPC1 RNA rescued failed intestinal differentiation in zebrafish models of apc deficiency . Our data demonstrate a novel role for apc in pyruvate metabolism and that pyruvate metabolism dictates intestinal cell fate and differentiation decisions downstream of apc . Mutations in the adenomatous polyposis coli ( APC ) gene are responsible for Familial Adenomatous Polyposis ( FAP ) , a genetic predisposition to colorectal cancer , and are also found in the majority of sporadic colonic tumors ( Fearnhead et al . , 2001 ) . Critical roles for APC in colon carcinogenesis are attributed to its ability to negatively regulate the proliferative consequences of Wnt signaling through degradation of β-catenin , and maintain normal intestinal differentiation by controlling the biosynthesis of retinoic acid ( RA ) ( Jette et al . , 2004; Nadauld et al . , 2006a , 2004 , 2005; Rai et al . , 2010; Schneikert et al . , 2007; Shelton et al . , 2006 ) . Although tremendous progress has been made in understanding the role of APC , its full battery of functions continue to expand . Altered energy metabolism is an emerging hallmark in cancer ( Hanahan and Weinberg , 2011 ) . The observation that cancer cells produce energy to support cell growth and proliferation differently than normal cells is known as the Warburg effect , and refers to neoplastic cells favoring aerobic glycolysis , even in the presence of ample oxygen ( Vander Heiden et al . , 2009; Warburg , 1956 ) . One of the major molecular mechanisms contributing to Warburg effect is mitochondrial dysfunction through impaired pyruvate metabolism ( Diaz-Ruiz et al . , 2011 ) . Pyruvate lies at the junction of glycolysis and the tricarboxylic acid ( TCA ) cycle . Contingent on the metabolic needs of the cell , pyruvate can be transported into the mitochondria , and through the action mainly of pyruvate dehydrogenase ( PDH ) , it can be used to drive ATP production and generate building blocks for macromolecule biosynthesis through oxidative phosphorylation . Alternatively , pyruvate can be converted to lactate via lactate dehydrogenase ( LDH ) and exported out of the cell . Aberrations in genes involved in pyruvate metabolism and transport have been reported in human diseases , particularly in cancer ( Gray et al . , 2014 ) . For example , monocarboxylate transporter 4 ( MCT4 ) and LDHA are overexpressed in cancer ( Kim et al . , 2013; Rong et al . , 2013 ) . An isoform of pyruvate kinase 2 ( PKM2 ) is preferentially expressed in numerous cancer types including pancreatic , colon and lung , and has been shown to promote aerobic glycolysis in HeLa cells by functioning as a transcriptional coactivator for HIF-1 ( Cerwenka et al . , 1999; Christofk et al . , 2008; Luo and Semenza , 2011; Schneider et al . , 2002; Yeh et al . , 2008 ) . Restoration of the pyruvate dehydrogenase complex activity through inhibition of pyruvate dehydrogenase kinase 1 ( PDK1 ) in head and neck squamous cell carcinoma cell lines led to reduced HIF-1a expression and tumor growth ( McFate et al . , 2008 ) . The recently identified mitochondrial pyruvate carrier subunit MPC1 is part of the MPC complex that is responsible for the uptake of pyruvate into the inner mitochondrial matrix ( Bricker et al . , 2012; Herzig et al . , 2012 ) . Recent work has revealed that MPC1 is downregulated in various human cancers and that this correlates with poor survival ( Schell et al . , 2014 ) . Consistent with a causative role in tumorigenesis , re-expression of MPC1 repressed the Warburg effect in colon cancer cell lines ( Schell et al . , 2014 ) . It is not clear how MPC1 is regulated or how its activities relate to the known genetic events that contribute to colon cancer development . Given the potential role for MPC1 in colorectal cancer and the importance of APC mutation , we investigated the mechanistic relationship between the mutational status of apc and mpc1 . Herein , we report that apc regulates pyruvate metabolism by controlling the levels of mpc1 via RA . Further , mpc1 is required and sufficient for initiating normal intestinal differentiation downstream of apc . Our findings strongly suggest that changes in metabolic profile can drive cell fate and differentiation decisions . To investigate the relationship between apc and mpc , we utilized the apcmcr zebrafish , which is homozygous for a truncating mutation in the Mutation Cluster Region ( MCR ) of apc and similar to what is found in human colon tumors ( Hurlstone et al . , 2003; Miyoshi et al . , 1992a , 1992b ) . In parallel , we also knocked down the expression of apc in wild type ( WT ) embryos using antisense morpholino ( apc mo ) ( Figure 1—figure supplement 1 ) . Evaluating gene expression of mpc1 and mpc2 by qRT-PCR , we found that both genes were significantly downregulated in apcmcr and apc mo embryos compared to WT/het siblings and control mo , respectively ( Figure 1A , B ) . This was confirmed by whole mount in situ hybridization for mpc1 and mpc2 ( Figure 1C , D ) . Additional in situ analyses for mpc1 and mpc2 in WT embryos revealed staining in the head , eyes , vasculature and somites at 24–48 hr post-fertilization ( hpf ) ( Figure 1E ) . At later time points , expression in the pectoral fin buds , liver and gut emerged ( Figure 1E ) . Cross sections of 72 hpf WT embryos previously probed with mpc1 and mpc2 confirmed gut expression for both genes ( black arrows ) ( Figure 1F ) . 10 . 7554/eLife . 22706 . 003Figure 1 . mpc1 and mpc2 are downregulated in apcmcr and apc morphant embryos . ( A , B ) Quantitative RT-PCR analysis of mpc1 and mpc2 gene expression in apcmcr ( A ) and apc mo ( B ) embryos . Values represent mean ± SD . Graph shown above is representative of at least three independent experiments . Statistical significance was analyzed using unpaired t-test . ( C , D ) Whole mount in situ hybridization for mpc1 and mpc2 in 72 hpf apcmcr ( C ) and apc mo ( D ) embryos . ( E ) Whole mount in situ hybridization for mpc1 and mpc2 in wild type ( WT ) embryos . head ( h ) , eyes ( e ) , somite ( som ) , vasculature ( vas ) , gut ( g ) , liver ( l ) . ( F ) Cross sections from 96 hpf WT embryos probed with either mpc1 or mpc2 confirmed gut-specific expression of both genes . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 003 10 . 7554/eLife . 22706 . 004Figure 1—source data 1 . Fold change calculations for Figure 1A , BDOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 004 10 . 7554/eLife . 22706 . 005Figure 1—figure supplement 1 . PCR analysis confirming apc knockdown . cDNA from 48 hpf embryos injected with either control ( cont mo ) or apc morpholino ( apc mo ) was used to amplify a 264 bp band corresponding to apc WT . A negative control with no reverse transcriptase ( - ) was also included . Amplification of 18s served as control for input cDNA ( 67 bp ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 005 Previous studies have established phenotypes associated with impaired apc function in the developing zebrafish including malformation of the gut , eyes , pancreas and jaw , arrested fin buds and failed heart looping ( Nadauld et al . , 2004; Hurlstone et al . , 2003; Nadauld et al . , 2006b ) . To determine whether loss of mpc1 would recapitulate morphological defects related to apc deficiency , we knocked down the expression of mpc1 in WT embryos with a splice-blocking morpholino which we confirmed by PCR ( Figure 2—figure supplement 1A ) . Microinjection of 0 . 75 mM mpc1 morpholino into WT embryos at the one- to two-cell stage resulted in about 87% of injected embryos appearing morphant ( n = 228 ) ( Figure 2—figure supplement 1B ) . Consistent with downregulation of mpc1 in apcmcr , mpc1 morphants ( mpc1 mo ) exhibited a range of phenotypes consisting of smaller head and eyes , enlarged hindbrain vesicle ( black arrows ) , pericardial edema , body curvature , and loss of pectoral fins ( blue arrows ) ( Figure 2A ) . 10 . 7554/eLife . 22706 . 006Figure 2 . Knock down of mpc1 expression phenocopies loss of apc . ( A ) Gross phenotype associated with mpc1 knock down . ( B ) Whole mount in situ hybridization analysis for organ-specific markers in mpc1 mo . Alcian blue staining revealed improper cartilage development ( * ) and confirmed loss of pectoral fins in mpc1 mo . pectoral fin bud ( pfb ) , heart ( h ) , gut ( g ) , pancreas ( p ) . ( C ) Cross section of the eye and gut from control or mpc1 mo . Prior to sectioning , embryos were previously stained with eye and gut-specific markers , irbp ( red arrow ) and fabp2 ( black arrow ) , respectively . ( D ) Co-injection with human MPC1 RNA led to increased percentage of mpc1 mo with normal pectoral fins as determined by in situ staining for idi1 . Statistical significance was analyzed using Fisher’s exact test . See also Figure 2—figure supplements 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 00610 . 7554/eLife . 22706 . 007Figure 2—figure supplement 1 . mpc1 morphant phenotype is rescued by human MPC1 RNA . ( A ) PCR analysis amplifying mpc1 WT ( 322 bp ) or morphant band ( 189 bp ) . Un-injected embryos ( UI ) served as a second control group . ( B ) Observed penetrance for mpc1 morpholino . ( C ) Observed phenotype for MPC1 mRNA overexpression in WT embryos . ( D ) Rescue of pectoral fins ( pfb ) and underdeveloped midbrain ( mb ) in 72 hpf mpc1 mo co-injected with human MPC1 RNA ( mpc1 mo + MPC1 RNA ) was determined by in situ hybridization for id1 and otx2 , respectively . ( E ) Quantification of embryos with normal midbrain and eye development as measured by otx2 staining . Statistical significance was analyzed using Fisher’s exact test . ( F ) PCR analysis confirming presence of human MPC1 transcript ( 300 bp ) in 48 hpf embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 00710 . 7554/eLife . 22706 . 008Figure 2—figure supplement 2 . mpc2 morphants phenocopy loss of mpc1 . ( A ) Knockdown of mpc2 expression in WT embryos ( mpc2 mo ) resulted in phenotypes previously described for mpc1 morphants including enlarged hindbrain ( black arrows ) , body deformities and absence of pectoral fins ( blue arrows ) . ( B ) PCR analysis using primers specific for amplifying mpc2 WT ( 137 bp ) or morphant band ( 1675 bp ) confirmed knockdown of mpc2 expression in 48 hpf embryos . ( C ) Observed penetrance for mpc2 morpholino . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 008 In situ hybridization with gata6 revealed that the primordial gut formed ( 93% , n = 55 ) in mpc1 mo but developed abnormally , as shown by reduced staining for fabp2 ( 100% , n = 36 ) , which marks the differentiated gut ( Figure 2B ) . Histological analyses on mpc1 mo gut confirmed these findings , there were fewer cells comprising the gut tube and they appeared cuboidal and non-polarized ( Figure 2C ) . Additionally , intestinal folds were visibly lacking in the mpc1 mo gut ( black arrow , Figure 2C ) . In contrast , cross-section of the gut from control mo showed polarized columnar intestinal cells , with the nuclei lined up clearly against the basal membrane ( Figure 2C ) . Since APC has also been reported to play a crucial role in congenital hypertrophy of retinal pigment epithelium ( CHRPE ) in humans and normal ocular development in the zebrafish embryo , we examined the eyes of mpc1 mo and found that irbp , a marker for photoreceptor and retinal pigmented epithelial cells , was severely reduced in mpc1 mo ( 95% , n = 40 ) ( Figure 2B ) ( Nadauld et al . , 2006b; Chapman et al . , 1989 ) . Cross-section of mpc1 mo eye revealed small lens and disorganized cell layers ( Figure 2C ) . The retinal cells appeared to be undifferentiated as supported by the loss of irbp expression ( red arrow , Figure 2C ) . Further phenotypic analyses of mpc1 mo by in situ hybridization using tissue-specific markers exposed diminished maturation for brain ( 95% , n = 22 ) and fin buds ( 100% , n = 28 ) as indicated by ascl1a and id1 expression , respectively ( Figure 2B ) . Also , mpc1 mo hearts failed to loop as determined by myl7 staining ( 100% , n = 39 ) ( Figure 2B ) . As with the gut , terminal differentiation of the pancreas in mpc1 mo was severely reduced as assessed by trypsin expression , a marker for exocrine pancreas ( 100% , n = 28 ) ( Figure 2B ) . Insulin , denoting the endocrine pancreas , remained normal ( 100% , n = 25 ) ( Figure 2B ) . Cartilage staining with alcian blue confirmed the absence of pectoral fins and revealed improper jaw formation in mpc1 mo ( 100% , n = 152 ) ( Figure 2B ) . We verified that the morphological defects we observed in mpc1 mo were specifically due to knock down of mpc1 by co-injecting with 0 . 5 ng of full length human MPC1 mRNA and analyzing the embryos by in situ hybridization for id1 and otx2 , a marker for both the midbrain and eyes . Overexpression of MPC1 mRNA alone resulted mostly in normal-appearing embryos , a small percentage exhibited cyclopia , severe body curvature and truncated tail ( 29% , n = 187 ) ( Figure 2—figure supplement 1C ) . However , in co-injected embryos ( mpc1 mo + MPC1 RNA ) , we found that MPC1 mRNA restored fin development as indicated by id1 staining ( 50% , n = 49 ) ( Figure 2D , Figure 2—figure supplement 1D ) . We obtained similar results with otx2 , MPC1 mRNA was able to rescue normal midbrain and eye development in mpc1 mo ( Figure 2—figure supplement 1D and E ) . The presence of MPC1 transcript was confirmed by PCR ( Figure 2—figure supplement 1F ) . mpc1 and mpc2 form a heterodimer complex that is responsible for transporting pyruvate from the inner mitochondrial space into the inner mitochondrial matrix ( Bricker et al . , 2012; Herzig et al . , 2012 ) . To examine whether loss of mpc2 would result in similar phenotypes as mpc1 mo , we knocked down its expression in WT embryos using antisense morpholino ( Figure 2—figure supplement 2A and B ) . We found that mpc2 mo exhibited similar developmental defects as mpc1 mo , such as smaller head and eyes , enlarged hindbrain vesicle ( black arrows ) , body curvature , and absence of pectoral fins ( blue arrows ) ( Figure 2—figure supplement 2B ) . In contrast to mpc1 mo , only a third of mpc2 mo-injected embryos appeared morphant , the majority of which exhibited a mild phenotype ( Figure 2—figure supplement 2C , Figure 2—figure supplement 1A , data not shown ) . MPC2 expression is inconsistently altered in cancer and variably correlated with survival ( Schell et al . , 2014 ) . We therefore focused further studies on mpc1 . In light of our data relating reduced mpc1 levels to failed intestinal differentiation , we sought to determine if re-expression of MPC1 would rescue intestinal defects in apc-deficient zebrafish embryos . We injected 0 . 75 mM apc mo with or without 0 . 1 ng MPC1 mRNA into WT embryos and evaluated intestinal differentiation by in situ hybridization for fabp2 . Compared to apc mo ( 6% , n = 265 ) , there was a significant increase in embryos with differentiated gut in the morpholino plus mRNA group ( 29% , n = 312 ) ( Figure 3A , C ) . Control embryos all displayed normal fabp2 expression ( data not shown ) . 10 . 7554/eLife . 22706 . 009Figure 3 . MPC1 rescues gut phenotype of apc mo and apcmcr . ( A , B ) In situ hybridization for fabp2 , in 72 hpf WT embryos injected with cont mo , apc mo or both apc mo and human MPC1 RNA ( apc mo + MPC1 RNA ) ( A ) . In situ hybridization for fabp2 in 72 hpf apc WT , apcmcr or apcmcr injected with human MPC1 mRNA ( apcmcr + MPC1 RNA ) ( B ) . ( C , D ) Quantification of injected embryos with differentiated gut as determined by fabp2 staining . Statistical significance was analyzed using Fisher’s exact test . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 009 We confirmed this finding by injecting 0 . 5 ng MPC1 mRNA into apcmcr and observed similar results , 40% of injected embryos re-expressed fabp2 ( n = 52 ) ( Figure 3B , D ) . However , only 3% of un-injected apcmcr showed fabp2 staining ( n = 29 ) , while WT/het siblings were all positive ( n = 80 , data not shown ) . MPC1 was also able to rescue cardiac defects in apcmcr , as we saw improved blood circulation in injected mutants as well ( data not shown ) . These results suggest that re-introduction of mpc1 can drive intestinal differentiation . Because of the integral role of MPC1 in pyruvate metabolism , we next investigated if mpc1 mo harbor metabolic defects as a consequence of diminished mpc1 function . We assessed mitochondrial respiration by measuring oxidative consumption rates ( OCR ) in 72 hpf embryos and there was a significant reduction in OCR in mpc1 mo compared to control ( Figure 4A ) . We also looked at triglyceride ( TG ) levels as an indicator of disturbance in normal energy utilization and observed a similar trend ( Figure 4B ) . Moreover , there was an extensive dysregulation of pyruvate metabolism upon loss of mpc1 , as we discovered a profound upregulation of pyruvate metabolic genes in mpc1 mo , suggesting a compensatory mechanism to account for reduced pyruvate transport across the inner mitochondrial membrane ( Figure 4—figure supplement 1A and B ) . Knock down of mpc2 did not affect mpc1 transcript level ( Figure 4—figure supplement 1C ) . 10 . 7554/eLife . 22706 . 010Figure 4 . Knock down of mpc1 or apc leads to altered mitochondrial respiration and pyruvate metabolism . ( A ) Mitochondrial respiration was evaluated by measuring oxygen consumption rates ( OCR ) in 72 hpf embryos . ( B ) Triglyceride ( TG ) levels were determined in lysates prepared from 72 hpf embryos using a colorimetric assay . ( C , D ) Quantitative RT-PCR analysis of enzymes involved in pyruvate metabolism in apc mo and ( C ) apcmcr ( D ) embryos . pyruvate dehydrogenase alpha 1a ( pdha1a ) ; pyruvate dehydrogenase kinase , isozyme 1 ( pdk1 ) ; pyruvate kinase , muscle , a ( pkma ) ; citrate synthase ( cs ) . ( E ) Lactate levels in apc wild type ( WT ) , un-injected apcmcr ( UI ) or apcmcr embryos injected with human MPC1 mRNA ( MPC1 RNA ) . For figures A–E , values represent mean ± SD . Graph shown above is representative of at least three independent experiments . Statistical significance was analyzed using unpaired t-test . ( F , G , H ) Gross phenotype ( F ) , alcian blue staining ( G ) and in situ hybridization for fabp2 ( H ) in pdha1a , pcxb , and pcxb + pdha1a mo . pcxb ( pyruvate carboxylase b ) . See also Figure 4—figure supplements 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 010 10 . 7554/eLife . 22706 . 011Figure 4—source data 1 . Mean and standard deviation values for Figure 4A , B , E; fold change calculations for Figure 4C , D . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 011 10 . 7554/eLife . 22706 . 012Figure 4—figure supplement 1 . Knockdown of mpc1 leads to dysregulated pyruvate metabolism . ( A ) A simplified scheme showing key enzymes ( in blue ) involved in pyruvate transport and metabolism ( adapted from Schell and Rutter [2013] ) . ( B ) Quantitative RT-PCR analysis of pyruvate metabolism enzymes in mpc1 mo . ( C ) Quantitative RT-PCR analysis of mpc1 in mpc2 mo . For B–C , values represent mean ± SD . Graph shown above is representative of at least three independent experiments . Statistical significance was analyzed using unpaired t-test . pyruvate dehydrogenase alpha 1a ( pdha1a ) ; pyruvate carboxylase b ( pcxb ) ; lactate dehydrogenase A4 ( ldha ) ; pyruvate dehydrogenase kinase , isozyme 1 ( pdk1 ) ; citrate synthase ( cs ) ; solute carrier family 16 , member 1 ( slc16a1 ) ; solute carrier family 16 , member 3 ( slc16a3 ) ; pyruvate kinase , muscle , a ( pkma ) ; mitochondrial pyruvate carrier 2 ( mpc2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 012 10 . 7554/eLife . 22706 . 013Figure 4—figure supplement 1—source data 1 . Fold change calculations for Figure 4—figure supplement 1B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 013 10 . 7554/eLife . 22706 . 014Figure 4—figure supplement 2 . PCR analysis confirming knockdown of pdha1a , pcxb . ( A , B ) cDNA from 48 hpf embryos injected with cont mo , pcxb mo or pdha1a mo was used to amplify a band corresponding to pcxb WT ( 355 bp ) ( A ) or pdha1a WT ( 254 bp ) ( B ) . A negative control with no reverse transcriptase ( - ) was also included . Amplification of 18s served as control for input cDNA ( 67 bp ) . ( C ) Microinjection of WT embryos with pdha1a morpholino ( pdha1a mo ) resulted in a higher penetrance compared to pcxb morpholino ( pcxb mo ) . Co-injection of both morpholinos gave the highest percentage of embryos with phenotype ( pcxb mo + pdha1a mo ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 014 Earlier studies have reported that MPC1 is downregulated in colon cancer and its expression positively correlates with APC ( Schell et al . , 2014 ) . Together with our previous data showing that mpc1 is downregulated in apc-deficient zebrafish and that mpc1 mo exhibit impaired oxidative respiration , we hypothesized that apc regulates mpc1 and therefore , pyruvate metabolism overall . To test this , we evaluated mitochondrial respiration and TG levels in apcmcr and apc mo and compared to appropriate controls , we observed significant defects in mitochondrial function upon loss of apc ( Figure 4A , B ) . Several enzymes in the pyruvate pathway were also differentially regulated in apcmcr or apc mo ( Figure 4C , D ) . To further validate that impaired mitochondrial function in apcmcr is facilitated by reduced mpc1 expression , we injected MPC1 mRNA into apc mutant embryos and looked at lactate levels as indicator of improved mitochondrial function . Compared to un-injected apcmcr , mutant embryos overexpressing MPC1 showed a significant reduction in lactate ( Figure 4E ) . apc WT/het sibs ( WT ) represent basal lactate levels in normal embryos ( Figure 4E ) . Pyruvate , after passing through the inner mitochondrial membrane , is converted to oxaloacetate and acetyl-CoA by pyruvate carboxylase ( PC ) and pyruvate dehydrogenase ( PDH ) , respectively . To ascertain if the knock down of enzymes downstream of mpc1 would result in a phenotype similar to mpc1 mo , we targeted pcxb and pdha1a , separately and in tandem , with antisense morpholinos ( Figure 4—figure supplement 2A and B ) . Loss of either metabolic gene or both , resulted in morphant embryos that lacked pectoral fins ( blue arrows ) , jaw ( red arrows ) and appeared identical to mpc1 mo ( Figures 4F–H , 2A ) . Interestingly , knock down of pcxb resulted in only 22% of embryos with phenotype ( n = 27 ) while targeting pdha1a gave a higher percentage of morphant embryos ( 63% , n = 19 ) ( Figure 4—figure supplement 2C ) . A synergistic effect was observed when both enzymes were diminished ( 83% , n = 18 ) ( Figure 4—figure supplement 2C ) . fabp2 in situ staining ( black arrows ) also revealed intestinal developmental defects in the pdha1a mo , pcxb mo and pdha1a + pcxb morphant embryos ( Figure 4H ) . Knock down of pcxb expression , however , not only resulted in low penetrance but mild phenotype as well ( Figure 4F–H ) . This could be due to the activation of an alternative pathway where oxaloacetate can be derived from glutamine instead of pyruvate ( DeBerardinis et al . , 2007 ) . To further elucidate how apc is controlling mpc1 , we initially looked at Wnt signaling as one of the major roles of APC is to regulate degradation of β-catenin ( Fearnhead et al . , 2001 ) . Perturbation of the Wnt pathway by treatment of apc mo with 10 uM NS-398 , a COX-2-specific inhibitor that has been shown to impair β-catenin activity in an apc-deficient background , did not affect mpc1 or mpc2 levels ( Figure 5A ) ( Eisinger et al . , 2007 ) . We did see a dramatic reduction in expression of known β-catenin target gene , mmp9 , implying that apc regulation of mpc1 is independent of Wnt ( Figure 5A ) . 10 . 7554/eLife . 22706 . 015Figure 5 . RA deficiency results in dysregulated pyruvate metabolism that is independent of Wnt pathway . ( A ) apc mo treated either with DMSO control or 10 uM NS-398 were analyzed by qRT-PCR to determine expression levels of mpc1 , mpc2 and mmp9 . ( B , C ) WT embryos treated with either DMSO control or 5 uM DEAB were analyzed by qRT-PCR to determine expression levels of enzymes involved in pyruvate metabolism . ( D ) Lactate levels in 72 hpf WT embryos treated with either DMSO or 5 uM DEAB . For figures A–D , values represent mean ± SD . Graph shown above is representative of 3 independent experiments . Statistical significance was analyzed using unpaired t-test . matrix metallopeptidase 9 ( mmp9 ) ; solute carrier family 16 , member 1 ( slc16a1 ) . . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 015 10 . 7554/eLife . 22706 . 016Figure 5—source data 1 . Fold change calculations for Figure 5A , B , C; mean and standard deviation values for Figure 5D . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 016 In addition to regulating Wnt signaling , apc controls the program of intestinal differentiation through regulation of RA levels , as we have previously shown ( Jette et al . , 2004; Nadauld et al . , 2006a , 2004 , 2005; Rai et al . , 2010; Shelton et al . , 2006 ) . To interrogate the involvement of RA in pyruvate metabolism downstream of apc , we treated WT embryos with DMSO or 5 uM DEAB , a known inhibitor of RALDH which catalyzes the second step in the conversion of Vitamin A to the active metabolite , RA ( Marill et al . , 2003; Russo et al . , 1988 ) . By qRT-PCR , we found that mpc1 and mpc2 transcript levels went down significantly with inhibition of RA ( Figure 5B ) . Expanding our analyses to include other components of pyruvate metabolism , there were five other genes that were either up- or downregulated upon DEAB treatment , supporting the notion that the pyruvate metabolic program is altered at various points when RA levels are perturbed ( Figure 5C ) . The DEAB-treated embryos also had low lactate levels , further establishing that RA inhibition results in mitochondrial dysfunction ( Figure 5D ) . To further understand the implication of our findings in zebrafish , we employed publicly available curated databases to investigate mutations and gene expression alterations of pyruvate metabolism genes in human cancers . Using samples deposited at The Cancer Genome Atlas ( TCGA ) , we selected for colon adenocarcinomas with mutations upstream of codon 1600 of APC , a region encompassing the MCR , and resulting in a truncated protein similar to those found in a majority of patients with FAP ( Fearnhead et al . , 2001 ) . Consistent with our previous data , we found that mpc1 expression is significantly downregulated in colon adenocarcinomas with APC deletions ( n = 91 ) compared to normal colon ( n = 19 ) ( Figure 6A ) . We also looked at other genes involved in pyruvate transport and metabolism and interestingly , MPC1 , MPC2 , PDHA1 and PC showed consistent downregulation in a specific subset of colon adenocarcinomas known as colon mucinous adenocarcinomas ( AC ) ( n = 22 ) ( Figure 6—figure supplement 1A ) . In addition to altered gene expression , we also found colorectal cancer samples in COSMIC that had mutations both in APC and pyruvate metabolism enzymes . There were five samples with somatic APC deletions that had mutations in multiple pyruvate metabolism enzymes as well , most of which are predicted to be probably ( ** ) or possibly ( * ) damaging , further supporting a genetic link between APC mutation and dysregulation of pyruvate metabolism ( Figure 6—figure supplement 1B ) . 10 . 7554/eLife . 22706 . 017Figure 6 . In silico analyses of APC and pyruvate metabolism gene alterations in cancer . ( A ) MPC1 expression levels in TCGA normal colon and colon adenocarcinoma samples harboring truncating mutations in APC upstream of codon 1600 . Statistical significance was analyzed using Mann Whitney test . ( B ) ONCOMINE database was analyzed for gene expression alterations in pyruvate metabolism genes in all cancer types . A control group composed of Uniprot random genes was used for comparison . Graph shows percentage of up- and downregulated genes with respect to the total unique analyses for each gene tested for all cancer groups . Statistical significance was analyzed using Mann Whitney test . ( C ) cBioportal analysis to estimate Kaplan-Meier overall survival of TCGA colorectal adenocarcinoma and kidney chromophobe patients with or without alterations in pyruvate metabolism ( PM ) genes . For B–C , pyruvate metabolism genes used in meta-analyses: MPC1 , MPC2 , CS , PDK1 , PDHA1 , PC , PKLR , LDHA , SLC16A1 , GYS1 . See also Figure 6—figure supplement 1 , Supplementary files 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 01710 . 7554/eLife . 22706 . 018Figure 6—figure supplement 1 . Pyruvate metabolism genes are mutated in human colon carcinomas . ( A ) Gene expression levels of pyruvate metabolism genes in TCGA normal colon and colon mucinous adenocarcinoma ( MA ) . Statistical significance was analyzed using Mann Whitney test . ( B ) Colon carcinomas harboring APC mutations were identified from the COSMIC database . Of the 35 samples that were found , five had multiple mutations in pyruvate metabolism genes that are predicted to be at least possibly damaging based on Polyphen2 in silico protein analysis . ( * ) – deletion; Polyphen2 prediction: ( ) – probably damaging , ( * ) – possibly damaging , ( italics ) – benign . DOI: http://dx . doi . org/10 . 7554/eLife . 22706 . 018 Several studies have shown that individual genes in the pyruvate metabolism pathway are altered in various cancer types ( Gray et al . , 2014; Schell et al . , 2014 ) . Using Oncomine , we extended these studies by treating the genes involved in pyruvate transport and metabolism as a group ( n = 10 ) . We discovered that this pathway is significantly dysregulated in cancer compared to a group of randomly-generated Uniprot genes ( n = 55 ) ( Figure 6B , Supplementary file 1 ) . We then looked at overall survival for patients , with or without mutations and/or gene expression changes in the pyruvate metabolism gene set using TCGA samples in cBioportal . Out of 21 cancer types that we analyzed , only colorectal adenocarcinoma and kidney chromophobe carcinoma showed a significant difference in overall survival between the two groups ( Figure 6C , Supplementary file 2 ) . Recent identification of MPC1 and MPC2 , genes responsible for pyruvate uptake into the mitochondrial matrix , has added a new complexity to targeting pyruvate metabolism in human disorders , including cancer ( Gray et al . , 2014; Bricker et al . , 2012; Herzig et al . , 2012 ) . How dysregulation of metabolism relates to the accumulation of genetic hits that cause tumor suppression has largely been unstudied . Here , we demonstrate a direct relationship between loss of a key tumor suppressor gene , APC , and dysregulation of MPC1 . Our finding that mpc1 expression is downregulated in embryos harboring a genetic mutation ( apcmcr ) or knocked down expression ( apc mo ) of apc is reflected in human tumors as well , where we not only discovered a significant downregulation of MPC1 expression in colon adenocarcinomas with APC deletions , but also samples possessing mutations in both APC and several pyruvate metabolism enzymes ( Figure 1 , Figure 6 , Figure 6—figure supplement 1 ) . Our meta-analyses of human cancer data sets also revealed extensive dysregulation of pyruvate metabolism , at multiple points in the pathway , and this incidence of altered gene expression of pyruvate metabolism genes in cancer is significantly more prevalent compared to a group of randomly selected genes ( Figure 6 , Figure 6—figure supplement 1 , Supplementary file 1 ) . It is remarkable how MPC1 and MPC2 , PDHA1 and PC—genes that are involved in the transport and conversion of pyruvate in the inner mitochondrial matrix , respectively—are all significantly downregulated in a subset of colon adenocarcinomas categorized as colon mucinous adenocarcinomas , as loss of these genes essentially shuts down oxidative phosphorylation in the mitochondria ( Figure 6—figure supplement 1 ) . Interestingly , additional in silico analyses suggest a potential use of pyruvate metabolism genes as prognostic markers for colorectal adenocarcinoma and kidney chromophobe carcinoma , as we found a strong correlation between aberrations in pyruvate metabolism genes with poor overall survival in these cancer types ( Figure 6 ) . It is interesting to note that there are differences in the dysregulation of pyruvate metabolism genes in apcmcr/apc mo and mpc1 mo ( Figure 4 , Figure 4—figure supplement 1 ) . APC is a multifunctional protein that has critical roles in various cellular processes ( Fodde , 2003 ) . In addition , regulation of other metabolic genes by APC could occur in parallel with regulation of mpc1 . mpc1 knock down alone , therefore , would not alter the expression of these genes in the same way as knock down of apc . Confirming a functional epistatic relationship between apc and mpc1 , knock down of mpc1 in WT embryos resulted in phenotypes that have been previously reported for impaired apc function ( Figure 2 ) ( Nadauld et al . , 2004; Hurlstone et al . , 2003 ) . APC has been shown to positively regulate glycogen synthase kinase-3 ( GSK-3 ) activity , an enzyme that inhibits glycogen synthase ( GS ) which is involved in converting glucose into glycogen for storage ( Bouskila et al . , 2010; Valvezan et al . , 2012 ) . Taken together , these findings suggest a major role for APC in controlling cellular bioenergetics and homeostasis , as it can affect glycogen synthesis and oxidative phosphorylation through GSK-3 and MPC1 , respectively . The exact role of metabolism in cancer as a driver versus passenger process has been unclear . Indeed , roles for metabolism in directing cell fate and differentiation decisions are only now being considered ( Schell et al . , 2014; Bracha et al . , 2010; Sperber et al . , 2015; Yanes et al . , 2010; Zhou et al . , 2012 ) . The rescue of intestinal differentiation defects in embryos with impaired apc function by exogenous MPC1 mRNA establishes a clear role for metabolic programming as a switch that can control cell fate decisions ( Figure 3 ) . Mechanistic insights into regulatory pathways that control metabolism and how perturbations in cellular bioenergetics effect cell differentiation and proliferation can lead to a better understanding of normal development and tumorigenesis . In this respect , the role of retinoic acid in promoting cell fate and differentiation remains undefined . Our studies suggest that RA may control a program of metabolism that is permissive for intestinal differentiation . The actions of RA are complex , and it is likely that the effects of RA on metabolism are indirect . Consistent with this , treatment of either WT or apcmcr embryos with RA did not result in an immediate induction of mpc1 ( data not shown ) . To conclude , we present a novel role for apc in controlling a metabolic program driving intestinal differentiation through regulation of mpc1 . Our data strongly support the notion that metabolic changes are a major part of the decision process in determining cell fate and provide a better understanding of how cancer genetics is linked with biochemical metabolic pathways . Wild-type ( WT ) TU ( RRID:ZIRC_ZL57 ) and apcWT/mcr ( RRID:ZFIN_ZDB-ALT-050914-2 ) Danio rerio ( zebrafish ) were maintained as previously described ( Westerfield , 1995 ) . Fertilized embryos were collected following natural spawnings in 1 × E3 medium ( 286 mg/L NaCl , 13 mg/L KCl , 48 mg/L CaCl2·2H2O , 40 mg/L MgSO4 , 0 . 01% methylene blue ) and allowed to develop at 28 . 5°C . Morpholino oligonucleotides were obtained from Gene Tools LLC ( Philomath , OR ) and solubilized to 1 mM or 3 mM stock solutions in 1x Danieau buffer . For microinjections , 1 nl of morpholino was injected into WT embryos at the 1- to 2-cell stages ( Draper et al . , 2001 ) . Knock down of gene expression was assessed by PCR . Primers were designed according to guidelines recommended by Gene Tools ( www . gene-tools . com ) to amplify WT and splice-blocked morphant bands . z18s was amplified as a control gene . For RNA rescue experiments , full length human RNA transcripts were transcribed from linearized DNA using mMESSAGE mMACHINE transcription kit ( Ambion - Waltham , MA ) . For microinjections , 1–2 nl of RNA was injected into embryos at the 1- to 2-cell stages . Overexpression of mRNA transcript was assessed by PCR . Statistical analyses were performed using Fisher’s exact test ( GraphPad Prism v 6 . 04 , RRID:SCR_002798 ) . A complete list of morpholinos , PCR primers and working concentrations used are provided in Supplementary files 3–4 . Wild type embryos were given DEAB ( VWR International - Radnor , PA ) at 5 µM . apc morphants were treated with 10 µM NS398 ( Cayman Chemical - Ann Arbour , MI ) . Embryos were harvested at 72 hpf in RNAlater ( Ambion ) for RNA/cDNA prep . In situ hybridizations were performed as previously described using digoxigenin-labeled riboprobes for ascl1a ( achaete-scute family bHLH transcription factor 1a ) , fabp2 ( fatty acid binding protein 2 , intestinal ) , gata6 ( GATA binding protein 6 ) , id1 ( inhibitor of DNA binding 1 ) , insulin , irbp ( interphotoreceptor retinoid-binding protein ) , mpc1 ( mitochondrial pyruvate carrier 1 ) , mpc2 ( mitochondrial pyruvate carrier 2 ) , myl7 ( myosin , light chain 7 , regulatory ) , otx2 ( orthodenticle homeobox 2 ) and trypsin ( Thisse and Thisse , 2008 ) . Embryos were cleared in 2:1 benzyl benzoate/benzyl alcohol solution and documented using an Olympus SZX12/DP71 imaging system ( Olympus Corporation - Japan ) . RNA Reference Sequences deposited in ZFIN ( zfin . org , RRID:SCR_002560 ) were used in designing the riboprobes . RNA from zebrafish embryo lysates was isolated using the RNeasy kit ( Qiagen - Germany ) . cDNA was synthesized from 1 µg of total RNA using iScript ( Bio-Rad - Hercules , CA ) . Intron-spanning primers , when possible , were designed using the Universal ProbeLibrary Assay Design Center ( Roche Applied Science ) . A complete list of primer sets is provided in Supplementary file 5 . PCR master mix was prepared with the FastStart Essential DNA Probe Master kit and Universal ProbeLibrary probes according to the manufacturer’s protocols ( Roche Applied Science - Germany ) . PCR was performed in triplicate using the LightCycler 96 System ( Roche Applied Science ) with 45 cycles of amplification and annealing temperature of 60°C . Fold change in gene expression was measured by normalizing against 18S rRNA and comparing test group with control . Cartilage of 96 hpf embryos was stained with alcian blue as previously described ( Neuhauss et al . , 1996 ) . Briefly , embryos were fixed in 4% sucrose-buffered paraformaldehyde , bleached with 30% hydrogen peroxide for 2 hr and stained with alcian blue overnight . The embryos were then cleared in acidic ethanol for 4 hr , dehydrated stepwise in ethanol and stored either in glycerol or 2:1 benzyl benzoate/benzyl alcohol solution . Stained embryos were examined using an Olympus SZX12/DP71 imaging system ( Olympus Corporation ) . Metabolic respiration in 72 hpf embryos , expressed as oxygen consumption rate ( OCR ) , was measured using XF24 Extracellular Flux Analyzer ( Seahorse Bioscience - North Billerica , MA ) as previously described ( Stackley et al . , 2011 ) . As a minor modification , mixing step was omitted during measurement cycle . Statistical analyses were performed using unpaired t-test ( GraphPad Prism v 7 . 02 , RRID:SCR_002798 ) . Embryos were harvested at 72 hpf and homogenized in 0 . 05% PBST +1X protease inhibitor . TG levels were determined using the Infinity Triglycerides Liquid Stable Reagent ( Thermo Scientific - Waltham , MA ) by measuring absorbance at 540 nm . Total protein concentration was determined using the DC Protein Assay ( Bio-Rad ) to normalize TG levels . Statistical analyses were performed using unpaired t-test ( GraphPad Prism v 7 . 02 , RRID:SCR_002798 ) . Lactate levels were measured in 72 hpf embryos using the EnzyChrom L-Lactate Assay kit ( BioAssay Systems - Hayward , CA ) as previously described ( Bestman et al . , 2015 ) . Groups of 25–50 embryos were used in the assay . Statistical analyses were performed using unpaired t-test ( GraphPad Prism v 7 . 02 , RRID:SCR_002798 ) . Embryos were fixed in 10% neutral buffered formalin , dehydrated in 70% ethanol and embedded in paraffin . Five-micron sections were cut using a Shandon Finesse E Microtome ( Thermo Scientific ) and stained with hematoxylin and eosin ( H and E ) . Sections were analyzed using a Nikon Eclipse 80i/DS-Fi1 imaging system ( Nikon Instruments Inc - Japan ) . Publicly available curated databases and analysis software were utilized to examine mutations and gene expression alterations in APC and pyruvate metabolism enzymes ( MPC1 , MPC2 , CS , PDK1 , PDHA1 , PC , PKLR , LDHA , SLC16A1 , GYS1 ) . COSMIC ( http://cancer . sanger . ac . uk/cosmic , RRID:SCR_002260 ) was mined for mutations that are found in human cancers ( Forbes et al . , 2015 ) . Polyphen2 ( http://genetics . bwh . harvard . edu/pph2/index . shtml , RRID:SCR_008584 ) was used to predict functional and structural consequences of amino acid substitutions in proteins mentioned above ( Adzhubei et al . , 2010 ) . For MPC1 expression analysis , Oncomine ( www . oncomine . org , RRID:SCR_010949 ) was utilized to determine the colon adenocarcinoma subset ( n = 101 ) in the TCGA ( http://cancergenome . nih . gov , RRID:SCR_003193 ) colorectal carcinoma sample set ( n = 237 ) , which was further selected , using COSMIC , for truncating mutations in APC upstream of codon 1600 , encompassing the MCR region ( n = 91 ) . MPC1 expression level in these samples was compared with normal colon ( n = 19 ) . Colon mucinous adenocarcinomas ( n = 22 ) within the same TCGA colorectal carcinoma sample set were analyzed for gene expression of pyruvate metabolism genes , with normal colon as control . Statistical analysis was performed using Mann Whitney test ( GraphPad Prism v 6 . 04 , RRID:SCR_002798 ) . Oncomine also allowed for gene expression analysis of individual pyruvate metabolism genes ( n = 10 ) and randomly selected Uniprot genes ( n = 55 ) ( http://www . uniprot . org/uniprot/ ? query=reviewed:yes+AND+organism:9606&random=yes , RRID:SCR_002380 ) in multiple cancer types employing these thresholds: p value = 0 . 001; fold-change = 1 . 5; gene rank = top 10%; data type = all . A plot was generated to show percentage of datasets meeting set thresholds with respect to total unique analyses for each gene tested in both groups . Statistical analysis was performed using unpaired t-test ( GraphPad Prism v 6 . 04 , RRID:SCR_002798 ) . Overall survival in 21 different TCGA cancer types , segregated by presence of mutations in pyruvate metabolism genes , was analyzed with cBioportal ( http://www . cbioportal . org/index . do , RRID:SCR_014555 ) ( Cerami et al . , 2012 ) . To verify the specificity of our pyruvate metabolism gene set as predictor of overall survival , four groups of ten random genes from Uniprot were utilized as a negative control gene set . Unpaired t-test was used to compare two unmatched , independent groups . Fisher’s exact test was used to determine if outcome is related to a categorical condition by more than chance . Mann-Whitney test was used to compare distribution of two unmatched groups . For fold change data , statistical significance was determined from t-test analyses of relative gene expression . For sample size calculations , the minimum number of samples per group ( 95% power ) was determined by assuming the probability of the defect in the control group is 5% or lower and 80% in the experimental group .	Colon cancer remains an important problem in healthcare . Cancer researchers are looking for new ways to detect the disease earlier and treat it more effectively . This is challenging because many of the genetic and molecular causes of colon cancer are still poorly understood . Mutations in the gene that encodes a protein called APC are one of the major causes of the disease . The APC protein normally keeps cells from growing and dividing too fast or in an uncontrolled way and is hence referred to as a tumor suppressor . For example , APC induces stem cells in the intestine to develop into specialized cells that keep the gut working normally . Mutations in tumor suppressor genes are common in many cancers . Other research has shown that cancer cells must reprogram their own metabolism – in other words , all the chemical processes that keep the cell alive – to meet the demands of proliferating rapidly . In particular , recent studies reveal that colon cancer cells produce less of a protein called mpc1 , which is involved in metabolism . These discoveries raised the following questions: does APC have an additional role in maintaining normal metabolism in cells by controlling how much mpc1 is produced ? Do mutations in the gene for APC lead to colon cancer because they alter the cell’s metabolism ? Sandoval et al . have now discovered a connection between APC and changes in cancer cells that help them to adapt to a new metabolic program . Experiments with zebrafish – a model animal that is now commonly used in the field of cancer biology – showed that APC acts via mpc1 to regulate how the cell uses energy . This regulation goes awry in colon cells that have abnormal APC activity; however , restoring the cell’s metabolism back to normal was enough to induce cells in the intestine to develop properly . Together , these findings suggest that restoring the normal balance of energy production in colon cancer cells may be an effective way to make the cells behave normally . This hypothesis remains to be tested and , if confirmed , further studies will be needed to determine whether it will lead to new treatments for colon cancer in humans .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cancer", "biology" ]	2017	A metabolic switch controls intestinal differentiation downstream of Adenomatous polyposis coli (APC)
Defecation allows the body to eliminate waste , an essential step in food processing for animal survival . In contrast to the extensive studies of feeding , its obligate counterpart , defecation , has received much less attention until recently . In this study , we report our characterizations of the defecation behavior of Drosophila larvae and its neural basis . Drosophila larvae display defecation cycles of stereotypic frequency , involving sequential contraction of hindgut and anal sphincter . The defecation behavior requires two groups of motor neurons that innervate hindgut and anal sphincter , respectively , and can excite gut muscles directly . These two groups of motor neurons fire sequentially with the same periodicity as the defecation behavior , as revealed by in vivo Ca2+ imaging . Moreover , we identified a single mechanosensitive sensory neuron that innervates the anal slit and senses the opening of the intestine terminus . This anus sensory neuron relies on the TRP channel NOMPC but not on INACTIVE , NANCHUNG , or PIEZO for mechanotransduction . Defecation is important for food processing that provides nourishment to the animal . It eliminates waste ( feces ) from the digestive tract via the anus ( Thomas , 1990; Heaton et al . , 1992; Lembo and Camilleri , 2003 ) , an unglamorous but essential body function . Compared to the extensively studied feeding behavior , defecation has received relatively little attention . Malfunction of defecation can lead to constipation and other diseases ( Lembo and Camilleri , 2003 ) , and abnormal development of neural circuits governing defecation may underlie birth defects such as Hirschsprung's disease due to elimination of intestinal ganglion cells required for bowel peristalsis ( Romeo et al . , 1994; Passarge , 2002 ) , one of the major birth defects of the digestive system afflicting one in 4000 of the population . Drosophila larvae provide a useful model system for the studies of feeding behavior and nutrition intake ( Ikeya et al . , 2002; Rulifson et al . , 2002; Hwangbo et al . , 2004; Bader et al . , 2007 ) . With an array of feeding assays and powerful genetic tools , these animals have yielded valuable information regarding the basis of the feeding behavior ( Shen , 2012; Zhao and Campos , 2012; Bhatt and Neckameyer , 2013 ) . However , modulation of defecation behaviors has received much less attention until recently ( Edgecomb et al . , 1994; Cognigni et al . , 2011 ) . Harnessing the experimental resources of this model system for the study of gut movements and the underlying neural basis should also help us understand the mechanisms of the defecation behavior . In the larval intestines , peristaltic movements of the digestive tract push food from the anterior towards the posterior end . The rate of flow depends on various signals from gut cells and associated neurons ( Benoit and Tracy , 2008; Schoofs et al . , 2009 ) . In Caenorhabditis elegans two groups of excitatory GABAergic motor neurons have been identified with partially redundant functions in activating enteric muscle cells ( EMCs ) ( McIntire et al . , 1993 ) . Little is known about the motor control of gut movements in Drosophila larvae or any involvement of sensory neurons for defecation . Mechanosensation is essential for many activities of Drosophila . Studies in adult flies have demonstrated that internal sensory neurons are important in regulating behaviors such as feeding , defecation , and egg laying ( Yang et al . , 2009 ) . Whereas recent studies have identified mechanosensitive channels in specific sensory neurons in the larval body wall for harsh or gentle touch ( Kim et al . , 2012; Yan et al . , 2013 ) , whether and how a larva senses stretches of its internal organs is unknown nor have the neurons and channels mediating such mechanosensation been identified . In this study we establish Drosophila larvae as a model system to study defecation behavior by performing studies of larvae 96 hr after egg laying ( AEL ) . First , we show that Drosophila larvae exhibit rhythmic cycles of sequential contractions of the hindgut and the anal sphincter to expel feces . Second , we identify the motor neurons that innervate the hindgut and anal sphincter and show that these two groups of neurons fire sequentially with the same periodicity as the defecation cycle . Unexpectedly , we found that a single sensory neuron innervates the anal slit to sense its opening . Finally , we show that the TRP channel NOMPC but not other known mechanosensitive channels in Drosophila is required for the mechanosensation of this anus sensory neuron . The Drosophila larval hindgut is the last part of the intestine , posterior to the Malpighian tubule , on the dorsal side under larval cuticle . At the posterior end of the hindgut is anal sphincter , which has a layer of thick sphincter muscles and a much narrower canal ( Figure 1A , Figure1—figure supplement 1 ) ( Murakami and Shiotsuki , 2001 ) . Because the Drosophila larval body wall is transparent , contractions of the hindgut and anal sphincter can be monitored in vivo . Fluorescent markers , expressed with a hindgut-specific byn-Gal4 ( Johansen et al . , 2003 ) , allowed visualization of contractions of the larval hindgut and anus sphincter in whole-mount of living larvae ( Figure 1B ) . The defecation behavior consists of sequential contractions of the posterior hindgut and anal sphincter in a very stereotypical manner ( Figure 1B , D ) , leading to opening of the anal slit to expel feces out of the lumen . This defecation process is repeated every 38 s at 25°C ( Figure 1D ) . To demonstrate those gut movements triggered defecation , we fed the larvae with yeast laced with blue food dye and video taped their defecation cycle . As shown in Figure 1C and Video 1 , each sequential contraction of hindgut and anus sphincter triggered a defecation cycle to expel feces out of the body . 10 . 7554/eLife . 03293 . 003Figure 1 . The periodical defecation process of the Drosophila larvae and the innervation of hindgut and anal sphincter by motor neurons . ( A and B ) Sequential contractions of the posterior hindgut and anal sphincter . ( A ) Schematic representation of the posterior hindgut and anal sphincter . The posterior hindgut: dashed line; anal sphincter: solid line . ( B ) Posterior hindgut and anal sphincter contract sequentially in the defecation process ( visualized with byn-Gal4 > GFP ) . From left to right: quiescent state , contraction of the posterior hindgut , contraction of the anal sphincter , and back to resting state ( scale bar: 100 µm ) . ( C ) Posterior hindgut and anal sphincter contract sequentially to expel the feces out . From left to right: quiescent state , feces pushed to anus via hindgut contraction , contraction of anal sphincter , and end of defecation cycle ( scale bar: 100 µm ) . ( D ) Time course of defecation cycle measured by GFP fluorescence intensity as in ( B ) . A region of interest ( ROI ) was drawn on the posterior hindgut; the fluorescence intensity in ROI increased due to the tissue compression . ( E ) Cell bodies of PDF neurons ( green , PDF-GFP ) and HGN1 neurons ( red , HGN1 > UAS-tdTomato ) in the terminal segment of the ventral nerve cord ( VNC ) ( scale bar: 20 µm ) . ( F–H ) Innervations in the gut of neurons labeled with PDF ( green , PDF-GFP ) and HGN1 ( red , HGN1 > UAS-tdTomato ) . PDF ( F ) and HGN1 ( G ) axons ( scale bar: 100 µm ) . ( I and J ) Buttons of PDF ( I ) and HGN1 ( J ) neurons labeled by Brp-GFP ( green dots ) along axons ( red ) ( scale bar: 10 µm ) . ( K and L ) Anti-vGlut staining of PDF and HGN1 axons on hindgut . ( K ) Anti-vGlut staining on hindgut ( red ) in PDF-GFP larvae ( green ) . Blue arrow heads indicate axon terminals ( scale bar: 50 µm ) . ( L ) Anti-vGlut staining on anal sphincter ( green ) in HGN1 > tdTomato larvae ( red ) . Blue arrow heads indicate axon terminals ( scale bar: 50 µm ) . ( M ) Schematic representation of the PDF and HGN1 neurons and their innervations on gut in a whole animal lateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 00310 . 7554/eLife . 03293 . 004Figure 1—figure supplement 1 . The muscle structures of hindgut and anal sphincter . Hindgut is with a thin layer of muscles and a wide lumen , while the anal sphincter muscles are thicker and form a much narrower canal ( scale bar: 100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 00410 . 7554/eLife . 03293 . 005Figure 1—figure supplement 2 . Glutamatergic innervations of motor neurons on the hindgut . ( From left to right ) Motor neuron axons labelled with vGLUT-Gal4 driven tdTomato; PDF neuron specific GFP; merge of two channels ( scale bar: 50 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 00510 . 7554/eLife . 03293 . 006Video 1 . Defecation behavior visualized with dyed feces . A larva was placed lateral side up on a slide . The animal was fed with dyed food and its feces can be seen when moving along the intestine . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 006 To investigate the neural basis for the gut movements , we searched for the neuronal innervation of the hindgut and anal sphincter muscles . Since the axons that innervate the hindgut are from the most posterior pair of the axon bundles in the ventral nerve cord ( VNC ) , the cell bodies of the neurons that innervate the hindgut and anal sphincter are most likely in the terminal segments of the VNC . We identified two groups of neurons , labeled by PDF-Gal4 and HGN1-Gal4 ( Nassel et al . , 1993; Edgecomb et al . , 1994; Renn et al . , 1999; Landgraf et al . , 2003; Cognigni et al . , 2011 ) , which innervate the posterior hindgut and anal sphincter , respectively ( Figure 1E–G ) . These neurons have their cell bodies in the terminal segments of VNC ( Figure 1E ) and send their axons along the midline of the ventral body wall to the posterior end of the larva , where they enter the hindgut . Within the hindgut , the HGN1 axons extend posteriorly to the anus sphincter surface to form dense arborizations over the muscles , while the PDF axons arborize over the posterior two-third of the hindgut with refined branches ( Figure 1F–H ) . The PDF and HGN1 neurons are glutamatergic , as they could be labeled with antibody staining against Drosophila vGlut ( Daniels et al . , 2008 ) ( Figure 1K , L ) . The axonal branches of PDF neurons on the hindgut can also be labeled with vGlut-Gal4 ( Mahr and Aberle , 2006 ) ( Figure 1—figure supplement 2 ) , indicating that they are likely glutamatergic motor neurons . The labeled neurons in both cases have their axon terminals in close proximity of the gut muscles and form abundant bouton structures ( Figure 1I , J ) . These results suggest that PDF and HGN1 neurons , which are likely motor neurons , might play a role in regulating hindgut contractions ( Figure 1M ) . In order to explore the functional connection between HGN1 neurons and anal sphincter muscles , we expressed Channlerhodopsin-2 ( ChR2 ) , a light activated cation channel , in the HGN1 neurons and recorded the excitatory junction potentials ( EJPs ) in the gut muscles before and after activating ChR2 by light . The gut muscles received tonic excitatory inputs ( Figure 2A ) . Due to the fillet recording methods we used to gain access of anus sphincter muscles , this firing pattern might differ from those in intact animals . Illumination of the larval VNC with blue light caused a dramatic increase of EJPs in the anus sphincter muscles in the larva with ChR2 expression in the HGN1 neurons but not in the control animals ( Figure 2B , C ) , providing evidence for HGN1 innervation of sphincter muscles . The PDF neurons have been previously shown to promote visceral muscle contractions ( Talsma et al . , 2012 ) . Activation of PDF neurons expressing ChR2 with blue light also triggered a dramatic increase of EJPs in the anus sphincter muscles ( Figure 2B , C ) , indicating that PDF neurons also play a role in regulating anal sphincter contractions , although PDF neurons do not directly form synapse with these muscles . This light-induced activation was absent in UAS or Gal4 control larvae and dependent on retinal , which is the chromosphere for ChR2 channels ( Figure 2B , C ) . 10 . 7554/eLife . 03293 . 007Figure 2 . Excitatory output from VNC neurons to gut muscles . ( A ) Spontaneous EJP activity of the anus sphincter muscles . Top panel: tonic input to the muscles; middle panel: burst inputs recorded in the muscles; lower panel: zoom-in of the spikes in a single burst . ( B ) Light induced activation in the gut muscles . Top panel: light-triggered EJPs increase in the anus sphincter muscles of HGN1 > ChR2 larva . Grey bars indicate the blue light . Lower panel: light stimulation on UAS-ChR2 larva as control . ( C ) Paired plot of the EJP frequency in dark and light condition ( n = 15 , *p < 0 . 001 , paired t test ) . ( D ) Light-triggered gut contractions in larvae carrying ChR2 in their PDF or HGN1 neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 007 To test whether direct activation of PDF or HGN1 neurons could trigger gut muscle contraction , we first expressed ChR2 in the PDF neurons and use a Minos insertion line to label the hindgut . Because the blue light used to excite GFP in the hindgut can also activate ChR2 , we could activate the neurons expressing ChR2 and monitor the gut movements at the same time . Both the hindgut and anal sphincter contracted strongly ( Figure 2D ) upon stimulation . In contrast , with ChR2 expression in the HGN1 neurons , the anal sphincter but not hindgut contracted upon blue light stimulation of HGN1 neurons ( Figure 2D and Video 2 ) . The contractions were absent in larvae with only UAS-ChR2 or larvae fed with regular food without retinal . 10 . 7554/eLife . 03293 . 008Video 2 . Light-induced defecation via activation of ChR2 in HGN1 neuron . A larva was placed ventral side up on a slide . The defecation could be observed shortly after blue light illumination ( as visualized by the auto-florescence of the internal organs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 008 To test whether PDF and HGN1 neurons are important for the normal defecation behaviors , we expressed Kir2 . 1 in these neurons to inhibit their activities . Silencing PDF neurons caused the interval of the defecation to increase from 38 s to 94 s ( Figure 3A ) . Silencing the HGN1 neurons did not significantly alter the interval of anus sphincter opening ( Figure 3A ) . Conceivably the peristaltic hindgut movements driven by PDF neuronal activity could have generated sufficient pressure to force open the anus sphincter . Indeed , silencing both the PDF neurons and HGN1 neurons caused the larva to display barely any hindgut movement over 5 min thus rendering it difficult to estimate the defecation interval , in contrast to the nearly eight cycles of contraction over 5 min—corresponding to a defecation interval of 38 s—in control animals ( Figure 3B ) . These results suggest that the PDF neurons and HGN1 neurons are required for the normal defecation behavior . 10 . 7554/eLife . 03293 . 009Figure 3 . PDF and HGN1 motor neurons are essential for the defecation behavior . ( A ) Silencing PDF but not HGN1 neurons increased the defecation intervals ( p < 0 . 001 , one-way ANOVA ) . ( B ) Silencing both PDF and HGN1 neurons reduced the defecation frequency dramatically ( p < 0 . 001 , paired t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 009 To test whether the periodic defecation cycle of Drosophila larvae is associated with rhythmic firing of the PDF and HGN1 neurons innervating the hindgut and anal sphincter , we performed in vivo whole-mount Ca2+ imaging by monitoring the neuronal activity with a genetically encoded Ca2+ indicator GCaMP5 ( Tian et al . , 2009 ) driven by neuronal-specific Gal4s . Indeed , the PDF neurons displayed a stereotypical periodic firing pattern as revealed by Ca2+ elevation in both soma and dendrite area ( Figure 4A , C ) . The average interval between peak Ca2+ signals is 38 s , which is highly consistent with the temporal pattern of contraction of the hindgut ( Figure 1D and Figure 4M ) . We also monitored the Ca2+ signals of the HGN1 neurons and found that the HGN1 neurons also exhibited oscillation of Ca2+ levels with a periodicity of 38 s ( Figure 4B , D , M , Video 3 ) . Furthermore , Ca2+ signals of HGN1 neurons spread from dendrites to soma with a fixed latency ( Video 3 ) , indicating that the HGN1 neurons may receive excitatory inputs with a periodicity of 38 s . 10 . 7554/eLife . 03293 . 010Figure 4 . Sequential firing of the PDF and HGN1 neurons . ( A ) Spontaneous Ca2+ oscillation imaged by GCaMP5 in PDF neurons in vivo . From left to right , Ca2+ signal of PDF neurons in quiescent state; peak Ca2+ signal of PDF neurons; Ca2+ intensity increase ( scale bar: 20 µm ) . ( C ) Ca2+ oscillates over time . Solid line: GCaMP5; dashed line: RFP , the two black arrow heads indicate the time points of left and middle panel in A . ( B ) Spontaneous Ca2+ oscillation imaged by GCaMP5 in HGN1 neurons in vivo . From left to right , Ca2+ signal of HGN1 neurons in quiescent state; peak Ca2+ signal of HGN1 neurons; Ca2+ intensity increase ( scale bar: 20 µm ) . ( D ) Ca2+ oscillates over time . Solid line: GCaMP5; dashed line: RFP , the two black arrow heads indicate the time points of left and middle panel in ( B ) . ( E ) PDF ( green ) and HGN1 ( red ) neurons at the lateral view of the VNC . d: dorsal; v: ventral . ( F ) Neuropil of PDF and HGN1 neurons overlap ( scale bar: 20 µm ) . ( G ) Grasp signal that resembles the location and shape of neuropil co-localization ( scale bar: 20 µm ) . ( H ) Spontaneous Ca2+ oscillation imaged by GCaMP5 in PDF ( purple ) and HGN1 ( green ) neurons simultaneously in vivo . Dashed line: RFP recorded at the same time . Arrows indicate the time points of the images as shown in ( I–L ) . ( I–L ) , representative images from H . Color bar shows the range ( 1–256 ) ( scale bar: 20 µm ) . ( M ) Form left to right; intervals of defecation and Ca2+ oscillation of PDF and HGN1 neurons ( error bar: S . E . M . , n = 8 , 7 , and 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01010 . 7554/eLife . 03293 . 011Figure 4—figure supplement 1 . HGN1 neurons' rhythmic activity in dissected VNC . ( A ) Ca2+ oscillation of HGN1 neurons in a dissected VNC . ( B ) Interval between Ca2+ peaks of HGN1 neurons in dissected VNC was decreased . ( C ) HGN1 neurons received periodical excitatory inputs ( indicated by grey bars ) . ( D ) HGN1 neurons exhibited burst firing . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01110 . 7554/eLife . 03293 . 012Figure 4—figure supplement 2 . Single component of split-GFP did not show fluorescence . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01210 . 7554/eLife . 03293 . 013Figure 4—figure supplement 3 . PDF neurons have axonal projections inside the VNC . ( A ) PDF neurons' dendrites labelled with Denmark . ( B ) PDF neurons' axons visualized with sytGFP . ( C ) Merged image showing the intense co-localization between PDF axons and the HGN1 neuron dendritic area . Arrowheads highlighting the PDF axons out of the VNC ( scale bar: 20 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01310 . 7554/eLife . 03293 . 014Video 3 . Rhythmic activity of HGN1 motor neurons . A larva was placed ventral side up . GCaMP5 was driven by HGN1-Gal4 to monitor the neuronal activity . The video is at 6× speed . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 014 To further investigate the properties of HGN1 neurons , we performed Ca2+ imaging and whole-cell patch clamp recording of these neurons in dissected VNC . HGN1 neurons exhibited periodic Ca2+ activities though with less regularity in this isolated preparation ( Figure 4—figure supplement 1A , B ) . They also displayed clusters of EPSCs ( Figure 4—figure supplement 1C ) , indicating that they received excitatory input from upstream neurons . Whole-cell patch recording of HGN1 neurons revealed that they fired bursts of action potentials ( Figure 4—figure supplement 1D ) , similar to what was seen in other Drosophila larval motor neurons ( Cattaert and Birman , 2001; Fox et al . , 2006; Imlach et al . , 2012 ) . The arborizations of PDF neurons and HGN1 neurons overlap extensively along the midline of the posterior VNC ( Figure 4E , F and Video 4 ) , raising the question whether they interact with each other . Green fluorescent protein reconstitution across synaptic partners ( GRASP ) has been developed as a technique to indicate synaptic connection of two neurons , each expressing one component of the split GFP ( Feinberg et al . , 2008; Gordon and Scott , 2009; Gong et al . , 2010; Han et al . , 2012 ) . We expressed the two GFP components separately in PDF neurons and HGN1 neurons by using two different binary expression systems , PDF-LexA and HGN1-Gal4 . An intense GFP signal was observed in the area where the processes of these two groups of neurons overlap ( Figure 4G ) , while neither PDF neuron nor HGN1 neuron expressing one part of the split GFP of GRASP generated any fluorescent signals by itself ( Figure 4—figure supplement 2 ) , suggesting that a functional connection might exist between PDF and HGN1 neurons . By labeling the PDF neurons simultaneously with the dendritic RFP marker DenMark and the axonal GFP marker sytGFP , we found that the PDF neurons send their axons to the area where their processes overlap with HGN1 neuron dendrites ( Figure 4—figure supplement 3 ) . 10 . 7554/eLife . 03293 . 015Video 4 . 3-D reconstructions of PDF and HGN1 neurons in the VNC . PDF neurons are labelled with PDF-GFP . HGN1 neurons are labelled with HGN1-Gal4 driven tdTomato . The Z-stack images were taken with 2-µm optical slice and projected along the Y-axis to get the stereo visualization of the structures . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 015 To test whether the sequential contractions of the hindgut and the anal sphincter are associated with sequential firings of the PDF neurons and HGN1 neurons , we employed two Gal4 drivers to express GCaMP5 in both PDF neurons and HGN1 neurons at the same time . The cell bodies of PDF neurons are near the ventral surface of the VNC , while the cell bodies of HGN1 neurons are more dorsal and posterior ( Figure 4E ) , making it possible to distinguish the two groups of cell bodies when monitored laterally ( Figure 4E ) . By monitoring the Ca2+ signals in PDF neurons and HGN1 neurons simultaneously , we found that the Ca2+ level began to rise in PDF neuron cell bodies and then spread to the area occupied by arborizations of both groups of neurons , followed by Ca2+ elevation in the cell bodies of HGN1 neurons with a very short delay , indicating that these two groups of neurons have coordinated firing patterns ( Figure 4H–L and Video 5 ) . 10 . 7554/eLife . 03293 . 016Video 5 . Dual imaging of PDF and HGN1 neurons . The larva was placed lateral side up to visualize both groups of neurons . The neurons expressed GCaMP5 under the drive of PDF and HGN1 Gal4s . Left side: ventral; right side: dorsal . A Ca2+ signal propagates from PDF neurons to HGN1 neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 016 Next , we asked whether there are sensory neurons for sensing movements of the gut or anus . The dendritic arborization ( da ) neurons are primary sensory neurons , which cover the entire body wall of a Drosophila larva . They are important for sensing chemical , thermal , light , and mechanical stimulations ( Tracey et al . , 2003; Xiang et al . , 2010; Kim et al . , 2012; Yan et al . , 2013 ) . In the vicinity of the anal slit , we found a specialized PPK-Gal4-labeled neuron . The cell body of this neuron resides on the anterior side of the anal slit , and its dendritic arbors surround the entire anal slit ( Figure 5A ) . The majority of its dendrites forms a thin layer of arbors and covers the body wall around the anal slit ( Figure 5A ) . There are also some arborizations extending along the anus sphincter ( Figure 5A ) . The axon of this sensory neuron joins the nerve bundle that includes the axons of other da neurons and projects to the terminal segment of the VNC . 10 . 7554/eLife . 03293 . 017Figure 5 . Sensory feedback from anus sensory neuron to the VNC motor neurons . ( A ) Anus sensory neuron's ( ASN ) location and morphology . Left panel: filled arrowhead indicates the location of the ASN; open arrowheads indicate other PPK-Gal4 neurons . Upper right panel: arrowhead: cell body; lower right panel: lateral view of the dendritic extension along the anus sphincter . Dashed line indicating the border of the intestine ( scale bar: 50 µm ) . a: anterior; p: posterior . ( B ) Images and schematic drawing of the anus opening , showing the stretching of the ASN dendrites . ( C ) Ca2+ elevates in ASN when anus opens . Left panel: resting state when the anus is closed; right panel: anus is open and the dendrites are stretched . ( Color range: 1–256 . Scale bar: 50 µm ) . ( D ) Plot of the Ca2+ intensity of the cell body region . Black trace: GCaMP5 signal; grey line: tomato signal as control in the same neuron . ( E ) Peak-to-peak interval of ASN Ca2+ activity ( error bar: S . E . M . , n = 7 ) . ( F ) Dual labeling of PPK and PDF neurons . Red: PPK neuron axon terminals ( PPK–GFP ) ; green: PDF neuron cell bodies and dendrites ( PDF-Gal4 > td-Tomato ) ( scale bar: 20 µm ) . ( G ) GRASP signals between ANS/PDF neurons ( scale bar: 20 µm ) . ( H ) Increase of EJPs in the anus sphincter muscles when the ASN carrying ChR2 is activated by blue light ( indicated by grey bars ) . ( I ) Paired plots of the EJPs frequency in dark and light condition ( n = 14 , p < 0 . 001 , paired t test ) . ( J ) Direct ASN stimulation triggered PDF/HGN1 neurons' activity ( arrow head ) ; grey bars indicate spontaneous oscillations of PDF/HGN1 neurons . ( K ) Laser ablation of ASN eliminated the single neuron ( ASN ) while other PPK neurons remained intact . Larval genotype: HGN1-Gal4; UAS-GCaMP5/PPK-tdTomato . ( L ) ASN ablation increased HGN1 neurons' activity interval . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01710 . 7554/eLife . 03293 . 018Figure 5—figure supplement 1 . Single component of split-GFP did not show fluorescence . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 01810 . 7554/eLife . 03293 . 019Figure 5—figure supplement 2 . Direct ASN stimulation triggered asymmetric activation of PPK neurons' axon terminals . ( A ) GCaMP intensity of ASN axon terminals before and after ASN ( labelled with PPK > GCaMP ) stimulation with a probe . Dashed lines outline the ROI areas . ( B ) GCaMP intensity of ROIs in ( A ) over time showing asymmetric Ca2+ elevation of ASN axon terminals . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 019 This single anus sensory neuron ( ASN ) has its cell body stochastically located on one side of the midline but its dendrites are highly symmetric . When the anal slit opens , the dendrites are dramatically stretched ( Figure 5B ) . To investigate whether this stretch could activate the ASN , we carried out in vivo imaging of the ASN Ca2+ level with GCaMP5 . We found that opening of the anal slit is accompanied with Ca2+ elevation in both dendrites and soma of this neuron ( Figure 5C , D and Video 6 ) . The periodicity of this Ca2+ response is similar to that of the movements of the anus and the oscillations of PDF neurons and HGN1 neurons in the VNC ( Figure 5E ) . This result indicates that the ASN is mechanosensitive and participates in sensing the radial stretch caused by opening the anus . 10 . 7554/eLife . 03293 . 020Video 6 . ASN's response to anus opening in a wild-type larva . The ASN was labelled with PPK-Gal4 driven GCaMP5 . A dramatic increase of Ca2+ over dendrites and soma was observed when the anus opened . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 020 We next tested whether the activation of the ASN could affect the firing patterns of motor neurons in the VNC . This sensory neuron projects to the terminal segments of the VNC and its axon terminals overlap with the area occupied by the dendrites of the HGN1 neurons and PDF neurons ( Figure 5F ) . To investigate whether there might be direct synaptic contact between these neurons , we carried out GRASP analysis by expressing components of the split GFP in PDF neurons and the ASN . We found that there is very strong GRASP signal at the site where ASN axons and PDF dendrites overlap ( Figure 5G ) , while neither PDF neurons nor ASN expressing one part of the split GFP of GRASP generated any fluorescent signals ( Figure 5—figure supplement 1 ) . Interestingly , the GRASP signal is asymmetric , in concordance with the localization of the ASN cell body to one side of the midline leading to a more intense axonal projection to the ipsilateral VNC , which is also evident with asymmetric Ca2+ elevation of the ASN axon terminals in the VNC ( Video 7 ) . 10 . 7554/eLife . 03293 . 021Video 7 . Asymmetric activity of ASN axon projection in the VNC . The sensory neuron axons were labelled with PPK-Gal4 driven GCaMP5 . Note the last segment exhibited intense activity increase and the signal on the right side was stronger that the left side , resulted from the asymmetric projection of the ASN . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 021 To further determine whether there is feedback from the ASN to activate the motor neurons in the VNC , we employed PPK-Gal4 to drive expression of ChR2 in the ASN so as to activate this neuron by blue light and monitored the EJPs of the gut muscles innervated by these motor neurons . Activation of ASN via blue light illumination induced large increases of EJPs in the majority of the anus sphincter muscles ( Figure 5H , I ) . To confirm that ASN rather than other PPK-Gal4-labelled neurons provides the feedback , we imaged PDF neurons and HGN1 neurons with GCaMP5 while inserting a tapered glass probe and advancing it to split open the anus sphincter , in a manipulation that mimicked the anus opening during defecation . We found that both PDF and HGN1 neurons responded to this local stimulation ( Figure 5J ) . Stimulation of the anus with this glass probe induced an asymmetric Ca2+ increase in the VNC that is consistent with the asymmetric projection of ASN axons ( Figure 5—figure supplement 2 ) . To study the functional importance of the ASN feedback to motor neurons , we used 2-photon laser to ablate the cell body of ASN at 48 hr after egg laying ( AEL ) . The ASN was completely abolished 48 hr after laser ablation , while the other PPK neurons remained intact ( Figure 5K ) . We then imaged the cell bodies of HGN1 neurons in the VNC and found abnormality of their rhythmic firing pattern , which displayed a much longer interval compared to control animals ( Figure 5L ) . These results suggest that the central neurons in the VNC receive excitatory feedback from ASN to increase motor neuron firing . Together with the GRASP analysis revealing the physical proximity of the processes of ASN and PDF neurons , our results indicate that the sensory neuron in the anus might sense the stretch when the anal slit is open and respond by provide feedback modulation of the PDF neurons and HGN1 neurons in the central nervous system . To search for the putative mechanotransduction channel in the ASN , we examined the defecation behavior of several mechanotransduction channel mutants . We found that the defecation rhythm remains normal in the iav , nan , and piezo mutants . However , the defecation interval in the nompC mutants was significantly increased ( Figure 6A ) . NOMPC is a TRPN channel essential for adult hearing and larval gentle touch ( Gong et al . , 2004; Yan et al . , 2013 ) . We found that NOMPC is highly expressed all over the dendrites and soma of the ASN , as revealed by staining with antibody against the NOMPC protein ( Figure 6—figure supplement 1 ) , raising the possibility that it might play a role in ASN sensing of radial stretch . Indeed , we found that the Ca2+ response is absent in the ASN of nompC null mutant larvae ( Figure 6B , C , Video 8 ) . This defect could be rescued by expressing wild-type NOMPC in the ASN with PPK-Gal4 ( Figure 6B , C , Video 9 ) . Our study thus identified a new role of the NOMPC channel , namely for sensing redial stretch of the intestinal terminus . NANCHUNG ( NAN ) and INACTIVE ( IAV ) , two other TRP channels that often work in concert with NOMPC in other sensory neurons , are thought to form heterodimers and function in the mechanotransduction in Drosophila ( Gong et al . , 2004 ) . We also tested the role of IAV in the ASN's stretch sensing . We found the ASN of iav1 mutant larvae exhibited Ca2+ response comparable to that in the wild-type larvae ( Figure 6C and Video 10 ) , suggesting that IAV is not required for the mechanotransduction of ASN . 10 . 7554/eLife . 03293 . 022Figure 6 . NOMPC is required for the mechanotransduction of ASN . ( A ) nompC but not iav , nan , and piezo mutant affects the defecation cycle of the larvae ( error bar: S . E . M . , n = 6 , 7 , 7 , and 6 . p < 0 . 001 , one-way ANOVA ) . ( B ) Ca2+ activity in the ASN responding to anus opening in nompC null mutant and NOMPC rescue larvae . Arrowheads indicate the anus opening for the rescue larva . ( C ) Group data of the Ca2+ response of ASN ( error bar: S . E . M . , n = 7 , 10 , 9 , and 9 . *p < 0 . 001 , one-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 02210 . 7554/eLife . 03293 . 023Figure 6—figure supplement 1 . NOMPC expression over the ASN revealed by NOMPC staining . The ASN was labelled by PPK-Gal4 driven GFP ( scale bar: 10 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 02310 . 7554/eLife . 03293 . 024Video 8 . nompC mutant eliminated ASN's response to anus opening . The ASN in the nompC mutant stayed quiescent even when it received a similar intensity of stretch . The genotype of the larva is: nompC1/nompC3; NOMPC-Gal4/UAS-GCaMP5 , UAS-mCherry . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 02410 . 7554/eLife . 03293 . 025Video 9 . NOMPC channel rescued the nompC mutant phenotype in ASN . A dramatic Ca2+ response could be detected when the wild-type NOMPC was expressed in the ASN with a nompC null background . The genotype of the larva is: nompC1/nompC3; NOMPC-Gal4 , UAS-NOMPC/UAS-GCaMP5 , UAS-mCherry . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 02510 . 7554/eLife . 03293 . 026Video 10 . IAV is not required for the ASN's stretch sensing . The ASN of iav1 mutant larva exhibited similar response pattern to anus opening compared with wild-type larvae . The genotype of the larva is: iav1/y; NOMPC-Gal4/UAS-GCaMP5 , UAS-mCherry . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 026 This study establishes the Drosophila larva as a model system for studying the defecation behavior . We found that Drosophila larvae exhibit periodic defecation cycles , involving sequential contractions of the hindgut and the anal sphincter . We also found two groups of neurons , which innervate the hindgut and anal sphincter respectively , and can excite the hindgut and anal sphincter muscle in a sequential manner . In addition , we found a single sensory neuron that could sense the opening of the anal slit and send feedback to the motor neurons ( Figure 7 ) . Studies of C . elegans as a model system have investigated the defecation circuit ( Thomas , 1990; Avery and Thomas , 1997; Branicky and Hekimi , 2006; Kwan et al . , 2008 ) . Studies of the adult fly have identified neurons regulating defecation behaviors subject to dietary and reproductive modulation ( Cognigni et al . , 2011 ) . In this study of the defecation behavior in Drosophila larvae , we have identified not only the motor neurons innervating gut muscles but also a sensory neuron strategically located to sense radial stretch during defecation and provide feedback to the central nervous system . 10 . 7554/eLife . 03293 . 027Figure 7 . Graphic summary of the motor neurons and a mechanosensitive sensory neuron in the defecation circuitry . DOI: http://dx . doi . org/10 . 7554/eLife . 03293 . 027 Previous studies of the defecation behaviors of the adult fly ( Cognigni et al . , 2011 ) have revealed that its defecation rate is regulated by both the internal state and environment , rather than showing a robust rhythm . However , at the larval stage , the motor neurons and gut muscles as well as the sensory neuron responding to anus movement , all show very robust rhythmic activities . Given that feeding and defecation are dominant behaviors for third-instar larvae , conceivably robust rhythmic feeding and defecation behaviors may facilitate their nutrition intake and waste expulsion . In contrast , adult flies will likely encounter more complex environments and may need to conduct their defection behaviors in a more controllable manner . Mechanosensation serves a number of important physiological functions in Drosophila larvae . The radial stretch sensation is a special type of mechanosensation essential for the function of many organs with luminal structures such as the digestive system and the blood vessels . However , how the organs sense radial stretch remains unclear . We have identified a sensory neuron that can sense radial stretch with its highly specialized morphology in Drosophila larvae . In addition , we found that the TRP channel NOMPC but not other TRP channels tested , such as IAV that is often associated with NOMPC function , is required for normal ASN mechanotransduction . Interestingly , the ASN could be labeled by both class III da neuronal marker and class IV da neuronal maker , raising the question whether it might have the dual functions to sense different stimuli . The ASN may provide a neuronal model to study the distinct and cooperative roles of different channels in a single neuron in the sensing of different intensity of stimulation . The two motor neurons and the sensory neuron ASN provide an entry point to elucidate defecation circuitry . The two motor neurons appear to be functionally connected , possibly involving synaptic connections between them , although we cannot exclude the possibility of multiple neurons being engaged in their functional connections . It remains to be determined as to how they are entrained with this rhythmic firing pattern , and whether it involves a central pattern generator upstream of PDF neurons . Interestingly , PDF is a peptide that has important roles in multiple neuropeptide signaling pathways in the fruit fly ( Renn et al . , 1999; Kim et al . , 2013 ) ; it would be interesting to test whether this neuropeptide also plays a role in the regulation of defecation behaviors by PDF neurons in the VNC . It is also of interest to explore possible contributions of indirect effects of PDF over muscle contraction , such as an influence of tracheal branching in the hindgut that may affect muscle contractions ( Linneweber et al . , 2014 ) . Recently , a study has suggested a novel role of HGN1 neurons in regulating the long-term food intake behaviors of adult flies ( Olds and Xu , 2014 ) . In our study we found that HGN1 neurons control the rhythmic pattern of larval defecation . These two studies suggest that Drosophila HGN1 neurons at different developmental stages might have multiple functions in regulating feeding and defecation behaviors . Though separated in evolution millions years ago , the structures of Drosophila gut and human gut share striking similarity . There are circular and longitudinal muscles lining the gut ending with the anal sphincter that controls defecation ( Netter , 1997; Murakami and Shiotsuki , 2001 ) . It remains an open question as to the extent of similarity of the mechanisms that control the gut movements . Diseases such as Hirschsprung's disease and anorectal malformation with failure to pass meconium ( Loening-Baucke and Kimura , 1999 ) are caused by developmental abnormality related to the gut and its innervation . Several genes and specific regions on the chromosomes have been shown or suggested to be associated with Hirschsprung's disease . Mutations in two human genes could lead to the absence of certain nerve cells in the colon ( Puri and Shinkai , 2004 ) . With the powerful genetic tools , further study of the Drosophila larval gut rhythmicity and its neural modulation will help us identify evolutionarily conserved features as well as strategies that may have been adopted by different organisms for their fitness . All the larvae were raised in the normal fly medium ( for the light activation assay , 100 µM all-trans retinal was added to the food ) . Flies are kept in 12 hr/12 hr dark/light circle at 25°C . PDF-Gal4 , HGN1-Gal4 , and UAS-ChR2 are from Bloomington stock center . GRASP was done using lines: PDF-loxA > loxAop-mCherry and HGN1-Gal4 > UAS-GFP or PPK-Gal4 > UAS-GFP . w[1118]; Gr28b[MB03888] is a Minos insertion which is from stock center ( #24190 ) , UAS-GCaMP5 fly line is from Loren L Looger lab in Janelia Farm . piezoKO is from A Patapoutian lab in Scripps . GRASP between PDF neurons and HGN1 neurons: w; PDF-Lexa/UAS-CD4-sp1-10; HGN1-Gal4/LexAOp-CD4-sp11 . GRASP between PPK neurons and HGN1 neurons: w; +/UAS-CD4-sp1-10; HGN1-Gal4/PPK-tdTomato-sp11 . UAS-CD4-sp1-10; LexAOp-CD4-sp11 are from K Scott lab ( UC Berkeley ) . For the whole-mounting imaging , a freely moving larva was picked up and rinsed with distilled water . Then the larva was transferred into 4% PFA overnight at 4°C . The larva was put between cover glass and images were taken by Zeiss confocal microscopy . In some cases , the whole VNC or different part of the gut was dissected out and mounted on a cover glass in PBS for imaging . For immunostaining of Drosophila larvae , third instar larvae were dissected in PBS . The whole hindgut and anus were isolated from their bodies . The tissues were then fixed in 4% PFA solution for 20 min at room temperature and treated with the primary antibody ( NOMPC antibody from J Howard [Yale] , vGlut antibody from G Davis [UCSF] ) overnight at 4°C and secondary antibody for 2 hr at room temperature . Images were acquired with Leica SP5 confocal microscope . Free moving third instar larva was pinned onto Sylgard coated chamber dorsal side up and filleted along the dorsal body wall . The larva was dissected in a saline containing: ( in mM ) : 103 NaCl , 3 KCl , 5 TES , 10 trehalose , 10 glucose , 7 sucrose , 26 NaHCO3 , 1 NaH2PO4 , and 4 MgCl2 , adjusted to pH 7 . 25 and 310 mOsm . 2 mM Ca2+ was added to the saline fresh before use . Fat bodies were gently removed from the gut surface . Additional pins were used to immobilize the gut . The preparation was visualized by Zeiss axioscope microscopy with 40× water lens . Sharp electrode with resistance around 80 MΩ was filled with 3 M KCl . The electrode tip was approached to the gut surface under the control of the MP-285 manipulator ( Sutter , USA ) . The signal was acquired by the Axon 200B amplifier and filter at 2 kHz . The electrode was moved forward until the voltage suddenly dropped to around −40 mV . Blue light was generated by mercury lamp with multiple filters . For ChR2 activation , a GFP filter was used to give out blue light with wavelength around 488 nm . The light application was controlled by a shutter equipped on the microscope . For pulse light activation , 2 s light pulse was repeated for three times . Recording data were analyzed with Clampfit and Matlab . The frequency before and during light application were calculated and compared as the index of activation . The recordings were performed following the protocol described by Hu et al . ( 2010 ) with slight modifications . Briefly , the entire VNC of a third instar larva was dissected , and the peri-neural sheath was gently removed in recording saline containing 103 mM NaCl , 3 mM KCl , 5 mM TES , 10 mM trehalose , 10 mM glucose , 7 mM sucrose , 26 mM NaHCO3 , 1 mM NaH2PO4 , 1 . 5 mM CaCl2 , and 4 mM MgCl2 ( adjusted to 280 mOsm , pH 7 . 3 ) . The dissected VNC were transferred to a glass-bottom recording chamber containing recording saline and immobilized with a platinum frame . The HGN1 neurons were identified by their GFP signals under a 40× water objective . Current-clamp and voltage-clamp recordings were performed using patch-clamp electrodes ( 9–10 MΩ ) filled with internal solution ( 140 mM potassium D-gluconate , 10 mM HEPES , 4 mM MgATP , 0 . 5 mM Na3GTP , 1 mM EGTA , adjusted to 265 mOsm , pH 7 . 3 ) . Cells were used for recording if the Rm value was greater than 500 MΩ and the membrane potential value was lower than −50 mV . A small constant hyperpolarizing current was injected during recording , immediately after break-in , to bring the membrane potential of neurons to approximately −60 mV . Cells were held at −60 mV in voltage-clamp mode for EPSC recordings . Signals were acquired with an Axon-700B multiclamp amplifier and were digitized at 10 kHz and filtered at 2 kHz using a 1322A D-A converter . Data were analyzed using Clampex 9 . 0 software ( Molecular Devices ) . A larva was mounted between two cover glasses with ventral side up . Its tail end was exposed for access of probe stimulation . A glass pipette was pulled and polished to form a taper-shaped probe with a diameter around 20 µm . The probe was spread on with grease to reduce friction . A piezo-controller was used to control the movement of the probe with fixed angle and increment . For imaging of ASN axons and PDF/HGN1 cell bodies , the probe was advanced 10 µm and the stimulation lasted for 1 s . ASN ablation was carried out as previously described ( Song et al . , 2012 ) . Briefly , a single second instar larvae 48 hr after egg laying ( AEL ) was mounted dorsal side up , and the cell body of the ASN was targeted using a focused 930-nm two-photon laser ( ∼350–700 mW ) mounted on a custom-built Zeiss fluorescence microscope . Following lesion , animals were recovered on grape juice agar plates and imaged live at the appropriate stages . A third instar larva was gently picked up from the food surface and rinsed with distilled water several times . The larva was transferred into a drop of PBS on the slide . A cover glass was put on the larva and pressed slightly to reduce larval movement . The preparation was then put under the Zeiss Pascal 510 confocal microscopy equipped with a 20× air objective . Time series images were acquired and used for analysis . For HGN1/PDF dual imaging , the larva was mounted with the lateral side up , so that both groups of neurons could be visualized simultaneously . For PPK neurons imaging , larvae were transferred to a piece of filter paper saturated with 100 mM sucrose for 4–6 hr to remove the food debris in the gut which could have potentially covered the neuron's images during experiments . An automatic alignment was made to most image series since the larvae tend to move slightly during image acquisition . An ImageJ plugin ‘registration ROI’ was utilized to correct the movements of the images during recording .	In animals , the final stage in the digestion of food is the removal of waste from the body . Until recently , however , defecation has received less attention than other aspects of digestion such as feeding behavior and nutrition . Fruit flies , also known as Drosophila , are commonly used in research as a model of animal biology . Food is moved through the digestive tract of fruit fly larvae when the muscles that circle the wall of the intestine contract . This process continues until the waste reaches the anus and is expelled from the body . Now Zhang et al . have found that when fruit fly larvae defecate , the muscles at the end of the intestine contract just before the muscles in the anus contract . The nervous system controls these muscles via sequential firing of two sets of nerve cells that connect to the intestine and anus muscles , respectively . Zhang et al . also identified a nerve cell that can sense when the anus is opened and relay this information back to the nervous system . The nerve cell is activated when stretched by the opening of the anus in a process that requires a protein called NOMPC . Problems with defecation can lead to constipation and other diseases . For example , Hirschsprung's disease—a birth defect that affects one in 4000—is caused by abnormal development of the nerve cells that control muscles in the gut . Experiments on fruit flies could help us to understand how defecation works in humans and to develop new treatments for disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Identification of motor neurons and a mechanosensitive sensory neuron in the defecation circuitry of Drosophila larvae
The nucleus of the solitary tract ( NTS ) is a key gateway for meal-related signals entering the brain from the periphery . However , the chemical mediators crucial to this process have not been fully elucidated . We reveal that a subset of NTS neurons containing cholecystokinin ( CCKNTS ) is responsive to nutritional state and that their activation reduces appetite and body weight in mice . Cell-specific anterograde tracing revealed that CCKNTS neurons provide a distinctive innervation of the paraventricular nucleus of the hypothalamus ( PVH ) , with fibers and varicosities in close apposition to a subset of melanocortin-4 receptor ( MC4RPVH ) cells , which are also responsive to CCK . Optogenetic activation of CCKNTS axon terminals within the PVH reveal the satiating function of CCKNTS neurons to be mediated by a CCKNTS→PVH pathway that also encodes positive valence . These data identify the functional significance of CCKNTS neurons and reveal a sufficient and discrete NTS to hypothalamus circuit controlling appetite . Obesity has emerged as one of the global healthcare challenges of the 21st century . Common obesity is primarily a consequence of food intake beyond the body’s energetic requirements , with the excess energy consequently stored as fat . Mechanistically , the brain integrates and responds to multiple homeostatic hormones , neurotransmitters , nutrients and peripherally generated neural signals to maintain energetic balance ( Dietrich and Horvath , 2013; Morton et al . , 2014 ) . One of the primary integration nodes within the brain for meal-related and metabolic signals from the periphery is the nucleus of the solitary tract ( NTS ) ( Grill and Hayes , 2012; Rinaman , 2010; Schwartz , 2010; Wu et al . , 2012 ) . The NTS hosts a variety of factors to control homeostatic functions , including a small subset of neurons that express cholecystokinin ( CCKNTS ) ( Garfield et al . , 2012; Herbert and Saper , 1990; Vitale et al . , 1991 ) . CCK consists of a family of peptides , the best characterized of which is a 33 amino acid peptide secreted from endocrine cells in the jejunum in response to nutrients in the intestinal lumen . This gut-derived peptide has a number of gastrointestinal ( GI ) functions including the promotion of satiation/satiety ( Gibbs et al . , 1973; Saito et al . , 1981; Smith and Gibbs , 1994 ) . Peripherally derived CCK does not readily penetrate the brain ( Fan et al . , 1997; Passaro et al . , 1982 ) . However , its short-term effect on appetite has been in part attributed to stimulation of vagal sensory neurons influencing the brainstem ( Fan et al . , 1997; Gibbs et al . , 1973; Saito et al . , 1981; Smith and Gibbs , 1994 ) . CCK is also synthesized within the brain , where it is post-translationally processed into an 8 amino acid peptide ( CCK-8 ) that reduces food intake when centrally infused ( Blevins et al . , 2000; Hirosue et al . , 1993 ) . However , the source of brain CCK that controls appetite has not been established . Recently developed chemogenetic and optogenetic approaches now provide a means to decipher the real-time contribution of discrete neuronal populations and networks to behavior ( Sternson and Roth , 2014; Tye and Deisseroth , 2012 ) , although , no direct functional assessment of CCKNTS neurons has yet been undertaken . Although CCK is produced in the NTS ( Garfield et al . , 2012; Herbert and Saper , 1990 ) , little is known about the function of this discrete source of brain CCK . To facilitate visualization and characterization of CCKNTS , we employed a knock-in mouse line expressing Cre recombinase at the Cck locus ( Cck-iCre ) crossed with a Cre-dependent enhanced yellow fluorescent protein reporter ( Ai3 ) line ( CckYFP; Figure 1A ) . CCK-eYFP cells were most abundantly expressed within the caudal aspect of the NTS ( Figure 1B , C ) , a brain region innervated by gastrointestinal vagal afferents and involved in nutrient sensing ( Appleyard et al . , 2005; Blouet and Schwartz , 2012; Ritter , 2011; Schwarz et al . , 2010 ) . 10 . 7554/eLife . 12225 . 003Figure 1 . CCKNTSneurons are activated by feeding . ( A ) Schematic of mouse crossing to generate a CCK-iCre::R26-loxSTOPlox-eYFP mouse line ( CckYFP ) . ( B ) Representative expression of CckYPF in a NTS coronal section . ( C ) Quantification of CckYPF-expressing cells across the rostral-to-caudal extent of the NTS . ( D ) Representative c-Fos-IR in CckYPF NTS cells in ad libitum fed , fasted or fasted then re-fed mice ( white arrows denote colocalized neurons ) and ( E ) quantification of c-FOS-positive CckYPF NTS cells ( n = 3–4; one-way ANOVA F ( 2 , 7 ) = 39 . 82 , p = 0 . 0001; Sidak’s post hoc comparison ) . ( F ) Quantification of c-FOS-IR across the rostral-to-caudal extent of the NTS in CckYPF cells by bregma level following intragastric delivery of water , amino acids ( aa ) or sucrose ( n = 3–5 per group; one-way ANOVA , F ( 2 , 21 ) = 7 . 280 , p = 0 . 0040; Tukey’s post hoc comparison ) . p<0 . 05 , p< 0 . 01 , p< 0 . 001 . Scale bar in B and D represents 200 μm . AP , area postrema; DMX , dorsal motor nucleus; NTSco , nucleus of the solitary tract , commissural part; NTSm , nucleus of the solitary tract , medial part; NTSl , nucleus of the solitary tract , lateral part . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 003 To determine whether CCKNTS cells are responsive to food intake , CckYFP mice were exposed to either ad libitum fed , dark cycle fasted or dark cycle fasted followed by 2 hr re-feeding conditions , and a surrogate marker of neuronal activation , c-Fos immunoreactivity ( IR ) , was assessed . In contrast to light cycle ad libitum fed and the dark cycle fasted conditions , ingestion of food on an empty stomach significantly increased c-Fos-IR within CCKNTS cells , indicating responsiveness to food consumption ( Figure 1D , E ) . To clarify whether this response is related to the nutrients , as opposed to stomach stretching or orosensorial aspects of feeding , we next investigated whether CCKNTS cells are responsive to nutrients if directly delivered to the stomach . Dark cycle fasted CckYFP mice were intragrastrically delivered isovolumetric ( 0 . 5 ml ) non-nutritive water or isocaloric ( 1 kcal ) sucrose or amino acids . As observed with chow intake , gavage of sucrose or amino acids significantly increased c-Fos-IR within CCKNTS cells compared with water ( Figure 1F ) . These results suggest that CCKNTS cells are activated by nutrient intake . We next considered whether activation of CCKNTS neurons could promote satiety by communicating a nutrient consumption signal . Cck-iCre mice were bilaterally injected into the NTS with AAVs that mediate the Cre-dependent expression of designer receptors exclusively activated by designer drugs ( DREADDs; expressed as DREADD-mCherry fusion proteins , hM3Dq; Cck-iCre-hM3Dq-mCherryNTS ) ( Figure 2A ) . DREADDs are designer muscarinic receptor variants that can only be activated by an otherwise biologically inert designer drug , clozapine-N-oxide ( CNO ) ( Alexander et al . , 2009; Krashes et al . , 2011 ) . Using this approach , we achieved expression of the DREADD-fused mCherry reporter protein in the caudal aspect of the NTS ( Figure 2B ) . CNO activated Cck-iCre-hM3Dq-mCherryNTS cells in vivo as shown by increased c-Fos-IR ( Figure 2C ) and ex vivo using electrophysiology in NTS slices ( Figure 2D ) . Activation of CCKNTS neurons in ad libitum fed Cck-iCre-hM3Dq-mCherryNTS mice injected with CNO ( 0 . 3 mg kg-1 , IP ) suppressed food intake for 24 hr ( Figure 2E ) . Likewise , CCKNTS neuron activation suppressed re-feeding in food deprived mice ( Figure 2F ) . 10 . 7554/eLife . 12225 . 004Figure 2 . Activation of CCKNTS neurons reduces feeding and body weight . ( A–D ) Bilateral stereotaxic injection of Cre-dependent excitatory hM3Dq-mCherry virus into the NTS of male Cck-iCre mice facilitated activation of CCKNTS neurons . ( A ) Schematic and Cre-mediated recombination of hM3Dq-mCherry allele . ( B ) Representative image of Cre-dependent expression of hM3Dq-mCherry within the caudal aspect of the NTS of a Cck-iCre mouse ( coronal sections; numbers indicate bregma levels , scale bar represents 200 μm ) . ( C ) c-Fos-IR in the NTS and co-expression ( green ) in hM3Dq-mCherry-transduced CCKNTS neurons ( red ) ( scale bar represents 200 μm ) . ( D ) Membrane potential and firing rate of Cck-iCre-hM3Dq-mCherryNTS neurons increased upon 5 μM CNO application . ( E ) Cck-iCre-hM3Dq-mCherryNTS mice exhibited a significant reduction of spontaneous feeding following CNO , compared to saline , administration ( n = 6; RM ANOVA , main effect of treatment [F ( 1 , 5 ) = 22 . 41 , p = 0 . 0052] , main effect of time [F ( 47 , 235 ) = 101 . 6 , p<0 . 0001] , and interaction [F ( 47 , 235 ) = 1 . 807 , p = 0 . 0023] ) ; tick marks on x-axis represent 3 hr , measurements collected with 30-min intervals ) and ( F ) a reduction of post-fast re-feeding following CNO compared to saline administration ( n = 6; paired t test , t ( 5 ) = 4 . 769 , p = 0 . 005 ) . ( G ) CNO-induced reduction in spontaneous feeding was attenuated by pre-treatment with the CCK-A-receptor antagonist , devazepide ( DEV; 1 mg/kg ) , ( n = 6; ANOVA , F ( 3 , 20 ) = 16 . 81 , p<0 . 0001; Sidak’s post hoc comparison p<0 . 001 , p< 0 . 05 ) . ( H ) CNO did not change fasting glucose level or ( I ) glucose disposal rate following a systemic glucose load ( 1 g/kg , IP ) . ( J ) Repeated CNO administrations over 4 days reduced body weight ( n = 6; RM ANOVA , main effect of treatment [F ( 1 , 100 ) = 60 . 78 , p<0 . 0001] , main effect of time [F ( 9 , 100 ) = 8 . 877 , p<0 . 0001] , and interaction [F ( 9 , 100 ) = 7 . 483 , p<0 . 0001] ) and decreased food intake ( main effect of treatment [F ( 1 , 100 ) = 16 . 13 , p = 0 . 0001] , main effect of time [F ( 9 , 100 ) = 4 . 106 , p = 0 . 0002] , and interaction [F ( 9 , 100 ) = 3 . 307 , p = 0 . 0014]; Sidak’s post hoc comparisons , p<0 . 05 , p<0 . 01 , *p<0 . 001 ) in Cck-iCre-hM3Dq-mCherryNTS as compared to Cck-iCre-mCherryNTS mice . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 004 We next tested whether CCK itself mediates these food supressing effects of CCKNTS neuron activation . Cck-iCre-hM3Dq-mCherryNTS mice were pre-treated with the CCK-A receptor antagonist devazepide which blunted CNO-induced appetite suppression both 4 hr and 24 hr after treatment ( Figure 2G ) , with no effects in control mice . These results demonstrate CCK-A receptors as downstream effectors of CCKNTS neuron-mediated reduction in feeding . Given our recent findings that parabrachial CCKergic ( CCKPBN ) transmission regulates hepatic glucose production ( Flak et al . , 2014; Garfield et al . , 2014 ) , we next investigated whether CCKNTS neurons impact glucose homeostasis . Unlike CCKPBN neurons , chemogenetic activation of CCKNTS neurons in Cck-iCre-hM3Dq-mCherryNTS mice altered neither circulating blood glucose levels ( Figure 2H ) nor glucose disposal following a systemic glucose load ( Figure 2I ) . These results reveal a neuroanatomical functional divergence in the energy-intake- versus energy-metabolic features of CCKergic neurotransmission within the brain , with CCKNTS neurons modulating food intake , but not glucose homeostasis . We next tested whether long-term stimulation of CCKNTS neurons promotes a sustained reduction in food intake and body weight loss . We treated Cck-iCre-hM3Dq-mCherryNTS and control Cck-iCre-mCherryNTS mice with CNO for 4 days ( two injections per day ) . Repeated activation of CCKNTS neurons resulted in a pronounced reduction in food intake with a concomitant reduction in body weight ( Figure 2J ) . Following withdrawal of CNO treatment , mice returned to the original body weight within approximately 48 hr and to baseline food intake after a transient phase of rebound feeding ( Figure 2J ) . These data demonstrate that sustained activation of CCKNTS neurons is sufficient to promote pronounced and reversible anorexia and body weight loss in mice . Together , these data reveal that activation of CCKNTS neurons promotes anorexia and weight loss , attenuating the homeostatic drive to eat even in face of marked negative energy balance ( as occurs in food deprivation ) . As a first step to delineate the circuits through which CCKNTS neurons modulate food intake , we surveyed the neuronal activation-like profile of Cck-iCre-hM3Dq-mCherryNTS mice treated with CNO or saline . Chemogenetic activation of CCKNTS neurons elicited striking c-Fos-IR in the paraventricular nucleus of the hypothalamus ( PVH ) in food-deprived mice compared with saline-treated mice ( Figure 3—figure supplement 1 ) . The PVH represents a crucial hub for the brain’s regulation of energy balance and integrates a diverse range of nutritionally related hormonal and synaptic inputs , including inputs from the NTS ( Grill and Hayes , 2012; Riche et al . , 1990 ) . To evaluate whether CCKNTS neurons directly innervate the PVH , a Cre-inducible AAV vector expressing Channelrhodopsin-2-eYFP ( ChR2-eYFP ) was stereotaxically injected into the NTS of Cck-iCre mice ( Figure 3A , B ) . Cre-mediated recombination of the vector led to expression of eYFP in CCKNTS cells and their projection fields ( due to axonal localization of the ChR2-eYFP fusion protein ) , revealing robust CCKNTS → PVH innervation ( Figure 3C ) , with no substantial projections at other anterior hypothalamic regions ( Figure 3—figure supplement 2A–C ) . CCKNTS efferent projections were also observed to innervate the ventral and more caudal aspect of the dorsomedial nucleus of the hypothalamus and , to a lesser extent , the arcuate nucleus ( Figure 3—figure supplement 2D–F ) . 10 . 7554/eLife . 12225 . 005Figure 3 . Activation of CCKNTS neurons efferent to the PVH suppresses appetite . ( A ) Schematic of CCKNTS→ PVH targeting strategy using bilateral NTS delivery of Cre-dependent ChR2-eYFP expressing vector in Cck-iCre mice . ( B ) Selective eYFP expression following Cre-mediated recombination in the caudal aspect of the NTS ( scale bar represents 200 μm ) . ( C ) CCKNTS efferents ( green ) innervate the PVH ( scale bar represents 400 μm ) . ( D ) CCKNTS axon targeting for photostimulation , positioning of the optic fiber and photostimulation parameters ( scale bar represents 400 μm ) . ( E ) Current clamp recording of a CCKNTS neuron expressing ChR2 ( scale bar 20 mV/100 ms ) . ( F ) Bilaterally transduced CCKNTS axons in the PVH and c-Fos-IR following PVH photostimulation ( scale bar represents 200 μm ) . ( G ) In vivo optogenetic photostimulation of NTSCCK→PVH terminals in Cck-iCre-ChR2-eYFPNTS significantly reduced nocturnal feeding ( n = 6 , RM ANOVA: main effect of treatment ( F ( 1 , 10 ) = 8 . 663 , p = 0 . 0147 ) , main effect of time ( F ( 18 , 180 ) = 97 . 25 , p<0 . 0001 ) and interaction ( F ( 18 , 180 ) = 2 . 788 , p = 0 . 0003 ) Sidak’s post hoc comparisons , at least p<0 . 05 ) ( H ) without altering locomotor activity ( RM ANOVA: main effect of treatment ( F ( 1 , 10 ) = 1 . 510 , p = 0 . 0 . 2472 ) , main effect of time ( F ( 18 , 180 ) = 1 . 797 , p = 0 . 0285 ) and interaction ( F ( 18 , 180 ) = 1 . 198 , p = 0 . 2671 ) or ( I ) water consumption ( RM ANOVA: main effect of treatment ( F ( 1 , 10 ) = 0 . 0924 , p = 0 . 7673 ) , main effect of time ( F ( 18 , 180 ) = 50 . 68 , p<0 . 0001 ) and interaction ( F ( 18 , 180 ) = 0 . 2666 , p = 0 . 9989 ) as compared to control Cck-iCre-eYFPNTS . Tick marks on x-axis represent 10-min intervals . ( J ) Real time food intake reduction following optogenetic activation of NTSCCK→PVH terminals in fasted Cck-iCre-ChR2-eYFPNTS and reversion following injection of CCK-A receptor antagonist ( devazipide; DEV ) ( n = 8; RM one-way ANOVA , treatment F ( 1 . 809 , 12 . 66 ) = 16 . 15 , p = 0 . 0004; individual F ( 7 , 14 ) = 0 . 4241 , p = 0 . 8714 , Sidak’s post hoc comparison *p<0 . 005 , p<0 . 05 ) . ( K ) Neither photostimulation nor DEV treatment alter food intake in fasted Cck-iCre-eYFPNTS control mice ( n = 6; RM one-way ANOVA , treatment F ( 1 . 294 , 6 . 469 ) = 1 . 486 , p = 0 . 2780; individual F ( 5 , 10 ) = 5 . 089 , p = 0 . 014 ) . ( L ) Photostimulation of NTSCCK→PVH terminals reduces total intake of chocolate pellets over 30 min following 18–20 hr of food deprivation ( n = 5 , paired two-tailed t-test , t ( 4 ) = 6 . 949 , p = 0 . 0023 ) , as compared to no photostimulation . ( M ) Representative confocal image of NTSCCK→PVH mCherry fibers and varicosities in close apposition to putative PVH MC4R-GFP neurons ( scale bar represents 20 μm; 10 μm stack , maximum intensity projection ) . ( N ) Top , IV relationship of a PVH MC4R-GFP neuron produced by the superimposition of membrane potential deflection in response to a series of current injections of constant increment ( scale bar 20 mV/200 ms ) ; bottom , current clamp recording of the above neuron following bath application of CCK-8 ( 500 nM; scale bar 20 mV/30 s ) . ( O ) Percentage of PVH MC4R-neurons expressing CCK-A receptor mRNA as assessed by single-cell qPCR . NTS , nucleus of the solitary tract; PVH , paraventricular nucleus of the hypothalamus; AP , area postrema; cc , central canal; 3v , third ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 00510 . 7554/eLife . 12225 . 006Figure 3—figure supplement 1 . Chemogenetic activation of CCKNTS neurons elicits c-Fos in the paraventricular nucleus of the hypothalamus ( PVH ) . ( A ) Schematic and Cre-mediated recombination of hM3Dq-mCherry vector . ( B ) Cck-iCre-hM3Dq-mCherryNTS mice were food deprived overnight and then treated with a single systemic injection of CNO ( 0 . 3 mg/kg; IP ) or vehicle . 2 hr following CNO injection , mice were transcardially perfused with saline then fixative , brains extracted and processed for c-Fos immunohistochemistry ( c-Fos-IR ) . Compared to saline , CNO administration increased c-Fos-IR in the ( C ) NTS and ( D ) PVH of Cck-iCre::hM3Dq-mCherryNTS mice . AP , area postrema; NTS , nucleus of the solitary tract; cc , central canal; DMX , dorsal motor nucleus of the vagus; PVH , paraventricular nucleus of the hypothalamus . Scale bar represents 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 00610 . 7554/eLife . 12225 . 007Figure 3—figure supplement 2 . Hypothalamic projections of CCKNTS neurons . Representative hypothalamic coronal sections from a Cck-iCre-ChR2-eYFPNTS mouse . Scale bar represents 200 μm . 3v , third ventricle; AHA , anterior hypothalamic ( hyp ) area , anterior part; AHP , anterior hyp area , posterior part; Arc , arcuate hyp nucleus; DM , dorsomedial hyp nucleus; DMD , dorsomedial hyp nucleus dorsal; DMV , dorsomedial hyp nucleus ventral; f , fornix; LA , lateroanterior hyp nucleus; LH , lateral hyp area; ME , median eminence; mt , mamillothalamic tract; opt , optic tract; Pe , periventricular hyp; PVH , paraventricular nucleus of the hypothalamus; SCh , suprachiasmatic nucleus; VMH , ventromedial hyp nucleus; ZI , zona incerta . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 00710 . 7554/eLife . 12225 . 008Figure 3—figure supplement 3 . Subset of PVH MC4R neurons express CCKA receptor and are responsive to CCK . ( A ) Schematic of mouse crossing to generate a Mc4r-t2a-Cre::tdTomato mouse line . ( B ) Schematic of PVH location in a coronal hypothalamic section . ( C ) CT values for individual Mc4r-t2a-Cre::tdTomato PVH neurons used for single-cell qPCR of isolated cells . Cckar mRNA was detected in 4 out of 15 analyzed cells . Cell No . 9 was considered negative . ( D ) Representative traces from current clamp recording of a Mc4r-t2a-Cre::tdTomato PVH neuron activated following bath application of CCK-8 ( 100 nM; 6/20 cells ) and blockade in presence of the CCK-A receptor antagonist SR 27 , 897 ( 250 nM; 0/11 cells ) . Scale bar 20 mV/1 min . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 008 To probe for the physiological significance of this CCKNTS → PVH circuit , Cck-iCre mice injected into the NTS with AAVs encoding for ChR2-eYFP ( a blue-light-sensitive cation channel ) were implanted with an optic fiber above the PVH to stimulate these CCKNTS efferents ( Figure 3A–D ) . Ex vivo application of blue light ( 473 nm ) to CCKNTS neurons expressing ChR2 using slice electrophysiology resulted in faithful action potential discharge with stimulation/discharge fidelity preserved at both 10 Hz and 30 Hz ( but not 50 Hz , data not shown ) in Cck-iCre-ChR2-eYFPNTS cells ( Figure 3E ) . In vivo photostimulation of CCKNTS axon terminals in the PVH elicited c-Fos-IR in PVH neurons in close proximity of transduced CCKNTS axons ( Figure 3F ) . We investigated whether specifically activating CCKNTS axon terminals within the PVH ( 10 ms , 30 Hz , 1 s on , every 4 s ) prior the initiation of mouse nocturnal feeding could communicate a post-prandial-like signal , and thereby reduce subsequent spontaneous feeding . Monitoring home-cage food intake revealed that optogenetic activation of the CCKNTS→PVH circuit produced an almost complete , and reversible , suppression of spontaneous feeding ( Figure 3G ) . Importantly , optogenetic activation of the CCKNTS→PVH pathway neither suppresses locomotor activity ( Figure 3H ) nor water consumption ( Figure 3I ) . Furthermore , real-time optogenetic stimulation of this circuit ( 10 ms , 30 Hz , 1 s on , every 4 s ) significantly suppressed re-feeding in food deprived Cck-iCre-ChR2-eYFPNTS mice ( Figure 3J ) , but not control Cck-iCre-eYFPNTS mice ( Figure 3K ) . The reduction in food intake in Cck-iCre-ChR2-eYFPNTS stimulated mice was blunted by pre-treatment with the CCK-A receptor antagonist devazepide ( Figure 3J ) , which did not affect food intake in control mice ( Figure 3K ) . Next , we tested the role of the CCKNTS→PVH circuitry to supress feeding in response to palatable food ( chocolate pellets ) . Food deprived Cck-iCre-ChR2-eYFPNTS mice were offered chocolate pellets via a computer-controlled delivery system using a within-subjects design ( optogenetic activation or no activation ) . Optogenetic activation of CCKNTS→PVH efferent reduced the total number of chocolate pellets consumed ( Figure 3L ) , indicating that the appetite-suppressing properties of the CCKNTS→PVH circuit maintains salience even when the homeostatic drive to eat following food deprivation is boosted with a hedonic component . Converging pharmacological and genetic data has established the importance of melanocortin-4 receptor ( MC4R ) -expressing neurons in the regulation of energy balance ( Fan et al . , 1997; Huszar et al . , 1997; Krashes et al . , 2016 ) , with the PVH having been identified as a principal site of their satiety-promoting action ( Balthasar et al . , 2005; Garfield et al . , 2015; Shah et al . , 2014 ) . To determine whether MC4R-expressing neurons are positioned to be a functional exponent of CCKNTS→PVH efferents , we first employed a modified bacterial artificial chromosome ( BAC ) transgenic MC4R-GFP reporter mouse line ( Ghamari-Langroudi et al . , 2015; Liu et al . , 2003 ) and crossed it with the Cck-iCre line . CCKNTS neurons were transduced with a mCherry-conjugated ChR2 AAV vector in the Cck-iCre::MC4R-GFP double transgenic line . We observed CCKNTS→PVH mCherry-positive fibers and varicosities in close apposition to MC4R-expressing neurons ( Figure 3M and Figure 3—figure supplement 3 ) . Moreover , using patch clamp electrophysiology , we observed that approximately 23% ( 5/23 cells ) of PVH MC4R-expressing neurons were excited by CCK-8 application ( Figure 3N ) . To increase the stringency of genetic identification of MC4R-expressing neurons , we also used a second and recently described knock-in mouse line ( Mc4r-t2a-Cre ) that express Cre recombinase from the endogenous and transcriptionally active Mc4r locus ( Garfield et al . , 2015 ) and in which MC4R-expressing neurons are fluorescently labeled by means of Cre-enabled tdTomato expression ( Mc4r-t2a-Cre::tdTomato; Figure 3—figure supplement 3A ) . Single-cell qPCR of manually sorted PVH Mc4r-t2a-Cre::tdTomato-expressing neurons revealed Cckar mRNA to be expressed in 27% ( 4/15 ) of MC4R PVH cells analyzed ( Figure 3O and Figure 3—figure supplement 3C ) . Likewise , electrophysiological recordings revealed that approximately 30% ( 6/20 cells ) of PVH Mc4r-t2a-Cre::tdTomato neurons were excited by CCK-8 application , in a CCK-A receptor-dependent manner ( 0/11 cells; Figure 3—figure supplement 3D ) . Thereby , these data identify CCK as a novel peptide neurotransmitter activating the appetite-controlling PVH MC4R neurons . In addition to the homeostatic regulation of energy balance , the NTS is also associated with negative valance and anorexia related to nausea ( Lachey et al . , 2005; Rinaman , 2004; Swank and Bernstein , 1994 ) . We next considered whether the CCKNTS→PVH circuit is associated with the induction of an aversive physiological state . The motivational valence of this circuit was assessed using a real-time place preference test ( Stamatakis and Stuber , 2012 ) . We observed no place preference when Cck-iCre-ChR2-eYFPNTS or Cck-iCre-eYFPNTS controls were tested under normal energy balance conditions , an important indication that activation NTSCCK→PVH efferents is not aversive ( Figure 4A–C ) . However , while Cck-iCre-eYFPNTS controls maintained no chamber preference when food deprived , calorie depleted Cck-iCre-ChR2-eYFPNTS mice exhibited a significant preference for the photostimulation-paired chamber ( Figure 4B , C ) . This behavioral response is reminiscent of that observed following direct optogenetic activation of PVH MC4R-expressing neurons ( Garfield et al . , 2015 ) . We further examined this observation using a T-shaped maze . In this test , exploration of one of the three maze’s arms was paired with the photostimulation of the NTSCCK→PVH efferent ( Figure 4D ) . Over the three testing trials Cck-iCre-ChR2-eYFPNTS mice , but not Cck-iCre-eYFPNTS controls , developed a clear preference for the photostimulation-paired arm ( Figure 4E , F ) . Thus , when hungry and in the absence of food , mice sought out activation of the NTSCCK efferent to PVH , revealing the CCKNTS→PVH circuit encodes positive valence . These data suggest that the activation of this circuit provides relief from the unpleasantness of energy deficit by mimicking a post-prandial phase . 10 . 7554/eLife . 12225 . 009Figure 4 . CCKNTS neurons signal positive valence via the PVH . ( A and D ) Experimental strategies for the interrogation of the valence encoded by NTSCCK→PVH terminals . ( B ) Food-deprived Cck-iCre-ChR2-eYFPNTS mice exhibited a significant place preference for the photostimulation-paired chamber during a real-time place preference assay , as compared to control Cck-iCre-eYFPNTS mice ( n = 5–6; Two-way ANOVA – fed: no effect of photostimulation ( F ( 1 , 18 ) = 0 . 3244 , p = 0 . 5760 ) , no effect of ChR2 ( F ( 1 , 18 ) = 0 . 001145 , p = 0 . 9734 ) or interaction ( F ( 1 , 18 ) = 0 . 5007 , p = 0 . 4883 ) ; food deprived: main effect of photostimulation ( F ( 1 , 18 ) = 5 . 289 , p = 0 . 0336 ) , no effect of ChR2 ( F ( 1 , 18 ) = 0 . 08811 , p = 0 . 7700 ) and interaction ( F ( 1 , 18 ) = 12 . 63 , p = 0 . 0023 ) ; Sidak’s post hoc comparisons , p = 0 . 05 ) . ( C ) Representative real-time place preference location plots one representative mouse per condition . ( E ) Cck-iCre-ChR2-eYFPNTS developed preference for the photostimulation-paired arm in a three-trial T-maze test ( n = 5–6 , main effect of treatment ( F ( 1 , 27 ) = 15 . 93 , p = 0 . 0005 ) main effect of trials ( F ( 2 , 27 ) = 3 . 36 , p = 0 . 0498 ) and interaction ( F ( 2 , 27 ) = 4 . 36 , p = 0 . 0228 ) ; Sidak’s post hoc comparisons , **p<0 . 001 ) , as compared to Cck-iCre::eYFPNTS . ( F ) Representative T-maze location plots from one representative mouse per condition . NTS , nucleus of the solitary tract; PVH , paraventricular nucleus of the hypothalamus . DOI: http://dx . doi . org/10 . 7554/eLife . 12225 . 009 While further studies attempting site-specific Cck loss of function are needed to fully clarify the physiological necessity of NTS CCKergic transmission in eating behavior and body weight regulation , here we reveal that activation of CCKNTS produces a prolonged effect on appetite and a rapid reduction in body weight . We did not find evidence that this reduction in energy intake is due to an aversive state , but rather , reveal that the CCKNTS→PVH circuit transmits positive valence in an energy-state-dependent manner; a particularly attractive prospect considering that in patient populations the negative valence of energy deficit is a major factor contributing to low compliance on weight-loss diets . CCK-ires-Cre ( Cck-iCre; Ccktm1 . 1 ( cre ) Zjh/J; Stock No . 012706 ) , R26-loxSTOPlox-eYFP ( Ai3; B6 . Cg-Gt ( ROSA ) 26Sortm3 ( CAG-EYFP ) Hze/J; Stock No . 007903 ) , MC4R-GFP ( B6 . Cg-Tg ( Mc4r-MAPT/Sapphire ) 21Rck/J; Stock No . 008323 ) or R26-loxSTOPlox-tdTomato ( Ai9; B6 . Cg-Gt ( ROSA ) 26Sortm9 ( CAG-tdTomato ) Hze/J; Stock No . 007909 ) were obtained from Jackson Laboratories ( Bar Harbor , ME ) . Mc4r-t2a-Cre mice were previously described ( Garfield et al . , 2015 ) . Mice were provided with standard mouse chow and water ad libitum , unless otherwise noted and housed at 22–24°C with a 12-hr light/12-hr dark cycle . All experimental procedures were performed in accordance with the UK Animals ( Scientific Procedures ) Act 1986 or the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee . Drugs for in vivo use were prepared in sterile saline and administered intraperitoneally ( IP ) . Clozapine-N-oxide ( CNO; Tocris Bioscience , Bristol , UK; Cat . No . 4936 ) was administered at 0 . 3 mg/kg . Devazepide ( Tocris Bioscience , Cat . No . 2304 ) was dissolved in DMSO , further diluted with sterile saline and administered at 1 mg/kg , IP , 40 min before CNO injection or optogenetic photostimulation . CCK-8 ( Tocris , Cat . No . 1166 ) and SR 27 , 897 ( CCK-A antagonist; Tocris Bioscience , Cat . No . 27897 ) were dissolved in artificial cerebrospinal fluid ( aCSF ) . DREADD and Optogenetics viruses employed were obtained from University North Carolina Vector Core ( Chapel Hill , NC ) . DREADD constructs were packaged in AAV serotype-8 and injected at a titer of 1 . 3 × 1012 vg/ml . AAV-EF1a-DIO-EYFP , AAV/EF1a-DIO-ChR2 ( E123T/T159C ) -eYFP , AAV/EF1a-DIO-ChR2 ( E123T/T159C ) -mCherry were packaged in AAV serotype-2 and injected at a titer of about 1 × 1012 vg/ml . Nucleus-specific delivery of AAVs was achieved through stereotaxic injections . NTS delivery of AAVs was achieved through a modified stereotaxic procedure . Briefly , 5–8 week-old male mice were anesthetized with a mixture of ketamine and xylazine dissolved in saline ( 100 and 10 mg/kg , respectively; 10 ml/kg ) . Mice were placed in a stereotaxic frame ( Kopf Instruments ) using ear bar with head angled at about 45° . Under magnification , an incision was made at the level of the cisterna magna and neck muscles carefully retracted . A 33G needle was used for dura incision . The obex served as reference point for injection . Injections were performed using a glass micropipette ( diameter 20–40 μm ) . NTS coordinates were A/P , -0 . 2; M/L , ±0 . 2; D/V , -0 . 2 from the obex . Virus was delivered under air pressure using a PLI-100A Pico-Injector ( Harvard Apparatus , Cambridge , UK ) . About 150 nl of virus/side were delivered with multiple microinjections over 4–5 min . The pipette remained in place for a minimum of 3 min after injection . Animals were administered an analgesic ( 5 mg/kg Carprofen , subcoutaneously ) for 2 days post-operatively and given a minimum of 14 days recovery before being used in any DREADD experiments . During this time , mice were acclimated to handling and IP injections . For optogenetic experiments , optic fibers were placed above the PVH at least 4 weeks after the NTS viral delivery to ensure labelling of distal projection sites . Based on the Franklin and Paxinos Mouse Brain Atlas ( Franklin and Paxinos , 2008 ) , the following coordinates were used for targeting optic fibers ( 200 μm diameter core; CFMLC22U-20; Thorlabs , Newton , NJ ) at the PVH ( mm from Bregma ) : A/P , −0 . 65; M/L , ± 0; D/V , −4 . 20 mm . After placement of the optic fiber , mice were allowed additional 14 days post-surgical recovery , as described above . Mice were excluded from behavioral analysis if post hoc histological validation showed either no viral transduction or misplaced optical fibers . Dark-cycle food intake . Mice were injected with saline or CNO ( 0 . 3 mg/kg; IP Tocris Bioscience ) 30 min prior to the onset of the dark . At onset of dark , food was returned and intake recorded automatically via the TSE Phenomaster system ( TSE , Bad Homburg , Germany ) or manually ( Figure 2g ) . Post-fast re-feed . Mice were fasted overnight . The following morning , mice were treated IP with saline or CNO and food returned 30 min later and intake recorded manually . Daily treatment . Mice were treated IP twice daily for 4 days with 0 . 3 mg/kg of CNO or saline near the onset of the dark and light cycles . Food intake and body weight were recorded daily prior to the dark cycle . Nutrient gavage . Mice were fasted overnight and the following morning , received intragastric 0 . 5 ml volume gavage of water , amino acids ( Peptone enzymatic digest from casein , Sigma-Aldrich , Dorset , UK; Cat . No . 70172 ) or sucrose ( Sigma-Aldrich , Cat . No . 84100 ) . Amino acids and sucrose were diluted in drinking water and delivered in an amount of 1kcal . Blood glucose studies . Mice were food deprived overnight and blood glucose concentration was determined from tail bleeds using OneTouch Ultra glucometer and test strips ( LifeScan , Johnson and Johnson Medical Limited , Livingstone , UK ) . Basal blood glucose concentration was determined prior to injection of any substances . Fiber optic cables ( 1 . 5 m long , 200 μm diameter , 0 . 22 NA; Doric Lenses , Franquet , Quebec , Canada ) were firmly attached to the implanted fiber optic cannulae with zirconia sleeves ( Doric Lenses ) . Photostimulation was programmed using a pulse generator software ( Prizmatix , Givat-Shmuel , Israel ) that controlled a blue light laser ( 473 nm; Laserglow , Toronto , Canada ) via a USB-TTL interface ( Prizmatix ) . Photostimulation for feeding experiments: light pulse trains 10-ms pulse width , 30 Hz , 1 s on , 4 s off . Photostimulation for place preference and T-maze tests: light pulse trains 10-ms , 30 Hz , 1 s on , 0 . 5 s off . Light power was adjusted such that the light power exiting the fiber optic cable was at least 10 mW using a digital optical power meter ( PM100D , Thorlabs ) and an online light transmission calculator for brain tissue ( http://web . stanford . edu/group/dlab/cgi-bin/graph/chart . php ) . After the completion of photostimulation experiments , mice were perfused and the approximate locations of fiber tips were identified according to the atlas coordinates ( Franklin and Paxinos , 2008 ) . After removal , optic fiber were connected to optic fiber cable and tested for light transmission . Mice were tested for real-time place preference in a model in which one chamber was paired with 30-Hz photostimulation and the other , identical chamber resulted in no photostimulation . Total test duration was 20 min . A similar protocol was used for the T-maze experiments . For the T-maze test mice were tested in three 10-min trials with an inter-trail interval of 15–20 min , during which mice were returned to the home cage . Time spent in the photostimulation versus non-photostimulation zones was recorded via a CCD camera interfaced with Any-maze software ( Stoelting , Wood Dale , IL ) . Following deep terminal anesthesia with pentobarbital , mice were transcardially perfused with phosphate-buffered saline ( PBS ) followed by 10% neutral buffered formalin ( Fisher Scientific , Loughborough , UK ) . Brains were extracted , cryoprotected in 30% sucrose in PBS , sectioned coronally on a freezing sliding microtome ( Bright solid state freezer series 8000 , Bright Instruments , Luton , UK ) at 30 μm and collected in four equal series . IHC was performed using standard methods and as previously described ( Garfield et al . , 2012 ) . Briefly , sections were washed in PBS before blocking in 0 . 5% BSA/0 . 25% Triton X-100 in PBS for 1 hr at room temperature . Tissue was incubated overnight at room temperature in blocking buffer containing the primary antibodies: rabbit anti-c-FOS ( EMD Millipore , Livingston , UK; Cat . No . PC05; diluted 1/5000 ) , or chicken anti-GFP ( Abcam , Cambridge , UK; Cat . No . ab13970; diluted 1/1000 ) . The next day , sections were washed in PBS then incubated in blocking buffer containing appropriate secondary antibody ( Alexa Fluor , Life Technologies; diluted 1/500 ) for 1 hr . 3 , 3′-Diaminobenzidine ( DAB ) staining was used for used for detection of c-Fos-IR using: a rabbit anti-c-FOS ( EMD Millipore Cat . No . PC05; diluted 1/5000 ) , secondary biotin-SP donkey anti-rabbit ( Jackson Immunoresearch , West Grove , PA; Cat . No . 711-065-152; diluted 1/250 ) , an avidin/biotin-based peroxidase system ( Vectastain Elite ABC Kit , Vector Laboratories , Burlingame , CA; Cat . No . PK-6100 ) and a developing kit ( DAB Peroxidase Substrate Kit , Vector Laboratories , Cat . No . SK-4100 ) . c-Fos-IR was pseudocolored using Photoshop software to appear colored in images . Sections were mounted onto microscope slides and coverslipped in an aqueous mounting medium ( Vectashield Antifade Mounting Medium , Vector Laboratories , Cat . No . H-1000 ) . Slides were imaged on a VS120 slide scanner ( Olympus , Essex , UK ) or AXIOSKOP2 ( Zeiss , Oberkochen Germany ) . For counting of c-FOS-IR nuclei , the boundaries of the nucleus were defined using neuroanatomical landmarks and the Franklin and Paxinos Mouse Brain Atlas ( Franklin and Paxinos , 2008 ) . For electrophysiological validation mice were anesthetized with sodium pentobarbital ( Euthatal ) and decapitated . The brain was rapidly removed and placed in cold ( 0–4°C ) , oxygenated ( 95%O2/5%CO2 ) ‘slicing’ solution containing ( in mM ) sucrose ( 214 ) , KCl ( 2 . 5 ) , NaH2PO4 ( 1 . 2 ) , NaHCO3 ( 26 ) , MgSO4 ( 4 ) , CaCl2 ( 0 . 1 ) , D-glucose ( 10 ) . The brain was glued to a vibrating microtome ( Campden Instruments , Loughborough , UK ) and 200-μm thick coronal sections of the brainstem containing the NTS were prepared . Slices were immediately transferred to a’ recording’ solution containing ( in mM ) NaCl ( 127 ) , KCl ( 2 . 5 ) , NaH2PO4 ( 1 . 2 ) , NaHCO3 ( 26 ) , MgCl2 ( 1 . 3 ) , CaCl2 ( 2 . 4 ) , D-glucose ( 10 ) , in a continuously oxygenated holding chamber at 35°C for a period of 25 min . Subsequently , slices were allowed to recover in ‘recording’ solution at room temperature for a minimum of 1 hr before recording . For whole-cell recordings , slices were transferred to a submerged chamber and a Slicescope upright microscope ( Scientifica , Uckfield , UK ) was used for infrared - differential interference contrast and fluorescence visualization of cells . During recording slices were continuously perfused at a rate of ca . 2 ml/min with oxygenated ‘recording’ solution ( as above ) at room temperature . All pharmacological compounds were bath applied . Whole cell current-clamp recordings were performed with pipettes ( 3–7 MΩ when filled with intracellular solution ) made from borosilicate glass capillaries ( World Precision Instruments , Aston , UK ) pulled on a Zeitz DMZ micropipette puller ( Zeitz Instruments GmBH , Martinsried , Germany ) . The intracellular recording solution contained ( in mM ) K-gluconate ( 140 ) , KCl ( 10 ) , HEPES ( 10 ) , EGTA ( 1 ) , Na2ATP ( 2 ) , pH 7 . 3 ( with KOH ) . Recordings were performed using a Multiclamp 700B amplifier and pClamp10 software ( Molecular Devices , Sunnyvale , CA ) . Liquid junction potential was 16 . 4mV and not compensated . The recorded current was sampled at 10 kHz and filtered at 2 kHz unless otherwise stated . Photostimulation of channelrhodopsin2 was achieved by 470 nm blue light delivered via the microscope objective . Light was generated by a pE-4000 LED illumination system ( CoolLED , Andover , UK ) driven via clampex 10 . 4 software ( Molecular Devices , Sunnyvale , CA ) . For CCK-8 application on MC4R PVH neurons , brain coronal slices from 6–8 week-old Mc4r-GFP or Mc4r-t2a-Cre::tdTomato mice were prepared . Extracted brains were immediately submerged in ice-cold , carbogen-saturated ( 95% O2 , 5% CO2 ) high-sucrose solution ( 238 mM sucrose , 26 mM NaHCO3 , 2 . 5 mM KCl , 1 . 0 mM NaH2PO4 , 5 . 0 mM MgCl2 , 10 . 0 mM CaCl2 , 11 mM glucose ) . Then , 300-µm thick coronal sections were cut with a Leica VT1000S vibratome and incubated in oxygenated aCSF ( 126 mM NaCl , 21 . 4 mM NaHCo3 , 2 . 5 mM KCl , 1 . 2 mM NaH2PO4 , 1 . 2 mM MgCl2 , 10 mM glucose ) at 34˚C for 30 min . Then , slices were maintained and recorded at room temperature ( 20–24˚C ) . The intracellular solution for current clamp recordings contained the following ( in mM ) : 128 K gluconate , 10 KCl , 10 HEPES , 1 EGTA , 1 MgCl2 , 0 . 3 CaCl2 , 5 Na2ATP , 0 . 3 NaGTP , adjusted to pH 7 . 3 with KOH . CCK-8 ( 100 nM ) and SR 27 , 897 ( CCK-A antagonist; 250 nM ) were applied to the bath through perfusion . Synaptic blockers ( 1 mM kynuerenate and 100 µM picrotoxin ) were added to the aCSF to synaptically isolate MC4R PVH neurons . Adult ( 4–5 week old ) Mc4r-t2a-Cre::tdTomato male mice ( n=2 ) were anesthetized with isoflurane . Brains were extracted and immediately chilled in ice-cold , carbogen-saturated ( 95% O2 , 5% CO2 ) high-sucrose solution ( 238 mM sucrose , 26 mM NaHCO3 , 2 . 5 mM KCl , 1 . 0 mM NaH2PO4 , 5 . 0 mM MgCl2 , 10 . 0 mM CaCl2 , 11 mM glucose ) . Next , 300-µm thick coronal sections were cut with a Leica VT1000S Vibratome and incubated in oxygenated aCSF ( 126 mM NaCl , 21 . 4 mM NaHCO3 , 2 . 5 mM KCl , 1 . 2 mM NaH2PO4 , 1 . 2 mM MgCl2 , 2 . 4 mM CaCl2 , 10 mM glucose ) at 34°C for 30 min . The PVH was visualized by fluorescence stereoscope then micro-dissected and enzymatically dissociated according to a published protocol ( Saxena et al . , 2012 ) , except that trituration was performed with fire-polished Paster pipettes . From the resulting cell suspension , tdTomato+ cells were individually washed and collected ( Hempel et al . , 2007 ) , frozen at -80°C and then processed for cDNA synthesis and amplification ( Picelli et al . , 2014 ) . To control for mRNA contamination during cell collection , an equivalent volume of cell-picking buffer was sampled and processed along with each batch of cell samples . After 20 cycles of amplification by polymerase chain reaction ( PCR ) , cDNA was purified ( Picelli et al . , 2014 ) and eluted in 30 μl PCR-grade water , and then analyzed by quantitative PCR ( qPCR ) for Actb , tdTomato , and Cckar in duplicate reactions . Actb and Cckar qPCR assays were obtained from Integrated DNA Technologies ( IDT , Coralville , IA; Cat . No . Mm . PT . 58 . 28904620 . g and Mm . PT . 58 . 12665706IDT ) . The tdTomato assay was custom synthesized by IDT from the following sequences: left , ACCCAGACCGCCAAGCTGAA; right primer , AGTTCATCACGCGCTCCCACT; internal probe , GCCCCCTGCCCTTCGCCTGG . Statistical analyses were performed using Prism 6 ( Graphpad Software , La Jolla , CA ) . Data were analyzed using t-test , one-way ANOVA , two-way or repeated measures ( RM ) ANOVA with post hoc comparisons , where appropriate . N represents independent biological replicates . No statistical methods were used to predetermine sample sizes . Sample size was computed based on pilot data and published literature . Data are presented as mean ± SEM and statistical significance was set at p<0 . 05 .	Obesity primarily results from eating more food than the body requires , the energy from which is then stored as fat . In recent years obesity has become increasingly common , with the resulting health problems presenting one of the major healthcare challenges of the twenty-first century . New ways to tackle the obesity epidemic are therefore required to improve human health on a global scale . To regulate how much food is eaten , the gut sends chemical messengers to the brain about how much food has been consumed . These messengers activate particular cells in the brain that signal to other brain regions to trigger a decision about whether we’ve had enough food to eat . This raises a question: if we can artificially activate these cells , can we ‘trick’ the brain into thinking that food has been consumed ? A brain region called the nucleus of the solitary tract ( NTS ) is known to play a key role in receiving signals from the gut about meals . By studying mice , D’Agostino et al . found that cells in the NTS that make a brain hormone called cholecystokinin ( CCK ) are particularly activated by food . Further experiments then used a technique called optogenetics to activate these cells in mice that had free access to different types of food . This activation significantly reduced how hungry the mice were , causing them to eat less food and lose weight . D’Agostino et al . also showed that CCK cells relay the signal about food intake to a brain region called the hypothalamus . Overall , D’Agostino et al . have found a way to trick the brain into thinking that food has been eaten when it actually hasn’t , and for this reason mice eat less without feeling hungry and lose weight . The next step is to try and find a way to activate the CCK cells in obese humans who have health complications associated with excess body weight .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2016	Appetite controlled by a cholecystokinin nucleus of the solitary tract to hypothalamus neurocircuit
The computational principles by which the brain creates a painful experience from nociception are still unknown . Classic theories suggest that cortical regions either reflect stimulus intensity or additive effects of intensity and expectations , respectively . By contrast , predictive coding theories provide a unified framework explaining how perception is shaped by the integration of beliefs about the world with mismatches resulting from the comparison of these beliefs against sensory input . Using functional magnetic resonance imaging during a probabilistic heat pain paradigm , we investigated which computations underlie pain perception . Skin conductance , pupil dilation , and anterior insula responses to cued pain stimuli strictly followed the response patterns hypothesized by the predictive coding model , whereas posterior insula encoded stimulus intensity . This novel functional dissociation of pain processing within the insula together with previously observed alterations in chronic pain offer a novel interpretation of aberrant pain processing as disturbed weighting of predictions and prediction errors . Classic bottom-up views construe perception as a feedforward stream of sensory information that is passed along the neural hierarchy from receptors to high-level brain regions ( Hubel and Wiesel , 1959 ) . Accordingly , neurons are thought of as feature detectors and cortical responses to sensory stimuli are expected to scale with the presence of stimulus features , that is , activity in pain processing brain regions should reflect the activation level of nociceptors . This basic account has been extended by a wealth of findings demonstrating that top-down expectations play an important role in modulating both the pain experience and the activity in pain processing brain regions ( Sawamoto et al . , 2000; Koyama et al . , 2005; Lorenz et al . , 2005; Brown et al . , 2008; Atlas et al . , 2010; Bingel et al . , 2011; Wiech et al . , 2014b ) . Other neuroimaging studies have shown that stimulus-response functions differ between brain regions ( Davis et al . , 1998; Coghill et al . , 1999; Apkarian et al . , 2001; Bornhövd et al . , 2002; Davis et al . , 2002; Porro et al . , 2004 ) and that brain activation is modulated by concurrent task demands ( Bantick et al . , 2002; Valet et al . , 2004; Wiech et al . , 2005; Seminowicz and Davis , 2007; Villemure and Bushnell , 2009 ) . However , these theories cannot explain the reduction in sensory cortical activity for expected compared to unexpected stimuli ( Alink et al . , 2010; Egner et al . , 2010; Todorovic et al . , 2011; Kok et al . , 2012 ) . In contrast , theories of Bayesian perceptual decision making , as formalized in predictive coding models ( Knill and Pouget , 2004; Friston , 2005; Summerfield and de Lange , 2014 ) , can explain such expectation suppression effects . Their proposal is that perception arises from the integration of sensory input with predictions about upcoming stimuli continuously generated by an internal model . More formally , the percept is determined by the posterior probability as computed by Bayes’ theorem from the predictions ( prior ) and the sensory input ( likelihood of a given stimulus ) . Within this framework , measurements of brain activity are composed of the activity of two distinct neuronal populations – one population encoding the expected stimulus based on an internal model of the world ( prediction ) and one population encoding the mismatch between sensory input and the prediction ( prediction error; PE ) ( Rao and Ballard , 1999; Friston , 2005 ) . A direct hypothesis derived from this framework is that sensory brain responses should be reduced when the brain’s prediction was accurate . In this situation , the resulting PE is small and regional brain activation is lower for accurate than for inaccurate predictions . This has been observed for primary visual cortex ( Alink et al . , 2010; Kok et al . , 2012 ) , early auditory electrophysiological responses ( Todorovic et al . , 2011 ) , and the fusiform face area ( Summerfield et al . , 2008; den Ouden et al . , 2010; Egner et al . , 2010 ) . The organization of cortical pain processing differs from other sensory modalities in that many cortical pain processing areas receive direct thalamic input and thus avoid a clear hierarchical organization ( Craig , 2002; Dum et al . , 2009 ) . It is therefore unclear whether the same computational principles apply to pain as well . If pain processing is also based on predictive coding principles , this framework would offer an elegant and general computational mechanisms of perception across modalities ( Wiech , 2016 ) and could help explain several expectation-related effects , including placebo effects ( Petrovic et al . , 2010; Büchel et al . , 2014; Tabor et al . , 2017 ) . In order to arbitrate between possible mechanisms underlying pain perception , we used a probabilistic heat pain task to formally compare a predictive coding model against a stimulus intensity model and a stimulus plus expectation model , respectively ( Figure 1A–C ) . Three different visual cues manipulated expectations about an upcoming cutaneous heat stimulus ( Figure 1D ) . Each cue was associated with a different probability of receiving painful or non-painful heat on the forearm ( 25 , 50 , or 75% chance of receiving pain and referred to as low , medium , and high cue , respectively; Figure 1E ) . Using functional magnetic resonance imaging ( fMRI ) in combination with model-based analyses in this task , we quantified evidence for all models in skin conductance responses ( SCR ) , pupil diameter , and across the brain . 10 . 7554/eLife . 24770 . 003Figure 1 . Hypotheses and design . ( A ) The stimulus intensity coding model is insensitive to predictive cues and postulates only a main effect of temperature . ( B ) Expectation may have an additive effect on brain responses in that a higher expectation of receiving pain results increased pain and increased physiological responses . ( C ) The predictive coding model has two components; prediction and prediction error ( PE ) . Pain processing regions increase activity with increasing predictions of pain ( from low to high pain probability ) . If the stimulus is painful , a PE signaling the difference between sensory input and the prediction occurs . In accordance with previous studies , we modeled the error for warm stimuli as zero ( see Materials and methods , Results ) . The hypothesized predictive coding response is a weighted sum of the two components . The model has two free weight parameters; both are required to be positive . Solid lines represent equal weighting , while dashed lines represent a higher weighting for the PE . ( D ) Subjects saw a central fixation dot during a 12 s inter-trial-interval ( ITI ) . A cue indicating the probability of a painful stimulus in the current trial appeared 300 ms before the heat stimulus started . Duration of heat stimulation was 1 . 5 s during which the cue was still visible . After a variable delay of 3–5 s , a rating screen appeared for 2 s and subjects reported whether the last stimulus had been painful or not . The fixation dot changed its color in 12 . 5% of the trials and participants responded to this change with a button press . ( E ) Cues predicted pain with 25 , 50 , or 75% probability and were counterbalanced across subjects . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 003 Before comparing the different models against each other , we verified that the two stimuli were clearly distinguishable . Pain ratings obtained after each run showed that the 28 participants distinguished between the two stimulus intensities ( t ( 27 ) = 20 . 9; p<0 . 001 ) , that intensity ratings were close to the calibrated intensities of 30 and 75 , respectively ( mean warm: 29 . 0 ± 9 . 1 std . ; mean pain: 75 . 0 ± 10 . 3 std . ) , and that warm stimuli were not perceived as painful ( Figure 2A ) . Trial-by-trial ratings classifying stimuli as either painful or non-painful matched the stimulus intensity with 94 . 3% accuracy , further supporting the qualitative difference between the two stimuli . Target reaction times to color changes of the fixation dot did also not differ between two stimulus intensities ( warm 520 . 3 ± 94 ms; pain 516 . 1 ± 104 . 8 ms; t ( 27 ) =0 . 51; p=0 . 61; Figure 2B ) , suggesting a similar attention allocation for both stimulation intensities . 10 . 7554/eLife . 24770 . 004Figure 2 . Behavioral and physiological results . ( A ) Intensity ratings reported at the end of each block for warm and painful stimuli , respectively . Intensity ratings were significantly higher for pain stimuli ( t ( 27 ) = 20 . 9; p<0 . 001 ) and correspond well to the stimulation levels chosen during calibration ( 30 and 75 ) . ‘Pain threshold’ was marked at the center ( 50 ) of the visual analogue scale ( VAS ) used for these ratings . Error bars in all plots show the standard error of the mean . ( B ) Target reaction time did not differ between stimulation intensities ( t ( 27 ) =0 . 51; p=0 . 61 ) . ( C ) Skin conductance responses ( SCR ) for pain ( red ) and warm ( blue ) stimuli . SCR responses reflect the pattern hypothesized by the predictive coding model . ( D ) Pupil dilation amplitudes shows the same response pattern as SCR , also supporting the predictive coding model . ( E ) Evoked skin conductance responses ( SCR ) for warm ( blue ) and painful ( red ) stimuli are plotted for each condition and followed the rank order hypothesized by the predictive coding model . ( F ) Pupil diameter responses plotted using the same groupings as in ( D ) . SCR and pupil traces are aligned to cue onset at 0 s , stimulus onset is at 300 ms ( unlabeled tick mark ) , and shaded areas indicate standard errors . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 004 From an ANOVA perspective , the stimulus intensity model predicts a main effect of stimulus , whereas the stimulus plus expectation model predicts an additional main effect of cue ( Figure 1A , B ) . By contrast , the summation of predictions and PE in the predictive coding model should result in a cue × stimulus interaction ( Figure 1C ) . We thus computed ANOVA’s for SCR , pupil dilation and brain data before conducting formal model comparisons . Peak amplitudes of both SCR and pupil dilation showed the expected interaction effect ( SCR: F ( 2 , 40 ) =27 . 7; p<0 . 001; pupil: F ( 2 , 38 ) = 9 . 5; p<0 . 001 ) . Responses of both measures increased with higher probability for pain when the stimulus was non-painful , but responses were lower expected pain compared to unexpected pain ( Figure 2C , D ) . This response profile mirrors the profile hypothesized by the predictive coding model ( Figure 1C ) . Plotting the grand means of the evoked SCR and pupil responses confirmed the rank-order of conditions observed in the peak amplitude analyses ( Figure 2E , F ) . In addition to the interaction effects , the main effect of stimulus was also significant for SCR ( F ( 1 , 20 ) =7 . 5; p=0 . 012 ) and pupil dilation ( F ( 1 , 19 ) =32 . 5; p<0 . 001 ) . In both cases the overall response was stronger for the painful than for the non-painful stimuli ( Figure 2C , D ) . The main effect of cue was also significant for the SCR ( F ( 2 , 40 ) =4 . 6; p=0 . 015 ) , but was not significant for the pupil dilation ( F ( 2 , 38 ) =2 . 7; p=0 . 078 ) . Hence , the ANOVA results are compatible with both the predictive coding and stimulus intensity model , while the SCR cue effect is also predicted by the stimulus plus expectation model . However , a post-hoc t-test comparing SCRs to painful and warm stimuli following a high cue did not reveal the difference proposed by the stimulus plus expectation model ( t ( 20 ) =1 . 54; p=0 . 14; Figures 1B and 2C ) . We next computed ANOVA’s on brain activity extracted from anatomically defined a priori regions of interest ( ROI ) . Among those ROIs , bilateral anterior insula ( left: F ( 2 , 58 ) =5 . 5; p=0 . 007; right: F ( 2 , 58 ) = 7 . 5; p=0 . 001 ) and right amygdala ( F ( 2 , 58 ) =5 . 4; p=0 . 007 ) showed the expected cue × stimulus interaction ( Figure 3A , Table 1 ) . Importantly , the response pattern matched the pattern expected by the predictive coding model , that is , responses in anterior insula and amygdala increased with pain expectation for warm stimuli and decreased for pain stimuli . Furthermore , all regions except for the postcentral gyrus , amygdala and PAG showed a significant main effect of stimulus ( Table 1 ) , but no ROI showed a significant main effect of cue . 10 . 7554/eLife . 24770 . 005Figure 3 . Parameter estimates for regions of interest . ( A ) Mean parameter estimates ( ± standard error ) are plotted for left ( L ) and right ( R ) hemispheres in each panel , except for the midline structure PAG . Blue indicates warm stimuli , red indicates painful stimuli . Cues are on the x-axis , with ‘l’ designating low , ‘m’ designating medium , and ‘h’ designating high probability of pain . PAG = periaqueductal gray . ( B ) Pattern expression for the neurological pain signature ( NPS; Wager et al . , 2013 ) . interaction effect significant FDR corrected q < 0 . 05 . #interaction: p<0 . 05 , uncorrected . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 00510 . 7554/eLife . 24770 . 006Table 1 . ANOVA results for brain ROI and NPS . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 006StimulusCueCue X stimulusRegionF ( 1 , 27 ) PF ( 2 , 54 ) PF ( 2 , 54 ) PACCL13 . 930 . 00091 . 210 . 30531 . 230 . 3017R15 . 990 . 00040 . 950 . 39231 . 110 . 3372anterior insulaL8 . 30 . 00770 . 410 . 66515 . 460 . 0069R9 . 690 . 00431 . 580 . 21557 . 480 . 0014posterior insulaL15 . 730 . 00050 . 280 . 75381 . 580 . 2145R12 . 120 . 00170 . 150 . 86370 . 090 . 9111parietal operculumL18 . 30 . 00021 . 20 . 30890 . 020 . 9779R23 . 35<0 . 00010 . 170 . 84080 . 720 . 4918post central gyrusL6 . 140 . 01981 . 10 . 34090 . 180 . 839R2 . 570 . 12060 . 180 . 83870 . 410 . 6675amygdalaL0 . 10 . 75060 . 10 . 90464 . 940 . 0107R0 . 830 . 3690 . 510 . 60335 . 390 . 0074thalamusL80 . 00870 . 390 . 67611 . 40 . 2545R7 . 60 . 01041 . 320 . 2752 . 620 . 0823PAG0 . 020 . 90271 . 020 . 36754 . 340 . 0178NPS47 . 73<0 . 00010 . 140 . 87082 . 180 . 1228ACC: anterior cingulate cortex , PAG: periaqueductal gray , NPS: neurological pain signature . FDR q<0 . 05 . Although the above ROIs are associated with pain processing , a recently developed multivariate pattern , termed neurological pain signature ( NPS; Wager et al . , 2013 ) , provides a more specific and sensitive estimate of heat pain intensity ( Wager et al . , 2013; Krishnan et al . , 2016 ) . We thus computed an ANOVA on the pattern expression values as indicators of overall pain intensity for the NPS ( Figure 3B ) . Stimulus intensity had an effect on NPS expression ( F ( 1 , 27 ) =47 . 7; p<0 . 001 ) , but neither cue nor the interaction were significant ( both p>0 . 12; Table 1 ) . Since NPS responses are strongly correlated with experimental heat pain reports ( Wager et al . , 2013; Krishnan et al . , 2016 ) , they can potentially serve as an indicator of trial-by-trial pain reports in this context to test for effects of correct predictions on pain reports . A post-hoc t-test revealed that unexpected pain tended to elicit stronger responses than expected pain ( t ( 27 ) =2 . 2; p=0 . 036 ) . In order to test for the proposed effects in brain regions outside of the a priori defined ROIs , we computed a whole brain analysis for the effects of stimulus ( pain > warm ) , effects of cue ( cue high > cue low ) , and for the interaction contrast ( ( cue high , warm ) > ( cue low , warm ) ) > ( ( cue high , pain ) > ( cue low , pain ) ) . The stimulus intensity contrast revealed activations in classical pain processing areas , including posterior and mid-insula , secondary somatosensory cortex , parietal operculum , and midcingulate cortex ( Figure 4A ) . A cue × stimulus interaction was observed again in left ( peak MNI coordinates: x=−30 , y = 24 , z=-4 ) and right ( x = 46 , y = 20 , z=−8 ) anterior insula ( Figure 4B ) . No other brain region showed the interaction effect at a family wise error rate of p<0 . 05 . Testing for the main effect of cue did not reveal any significant voxels . 10 . 7554/eLife . 24770 . 007Figure 4 . Whole brain results . ( A ) A main effect of stimulus was observed in pain processing regions including insula , parietal operculum , and midcingulate cortex . ( B ) Anterior insula showed a significant interaction between cue and stimulus . Maps are displayed at p<0 . 05 , whole brain FWE corrected using nonparametric permutation testing resulting in pseudo-t maps . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 007 After observing that the response profiles of SCR , pupil , bilateral anterior insula , and right amygdala were as expected by a concurrent representation of predictions and PE , we conducted formal model comparisons using Bayes factors ( BF ) ( Kass and Raftery , 1995; Rouder and Morey , 2012 ) to identify the best explanatory model . Bayes factors are computed as the ratio of marginal likelihoods of the data under each of two models and thus quantify the evidence for one model over the other given the data . This metric thus allows the identification of the best model while implicitly controlling for the number of free parameters . Following Kass and Raftery ( 1995 ) , we consider log-BF >3 as strong evidence for the predictive coding model and values of log-BF < −3 as strong evidence for the alternative model . Comparing the predictive coding model against the stimulus intensity model revealed strong evidence in favor of the predictive coding model for both SCR and pupil responses ( log-BFSCR = 6 . 65; log-BFpupil = 6 . 75; Figure 5A ) . Comparing it against the stimulus plus expectation model revealed similarly decisive evidence in favor of the predictive coding model ( log-BFSCR = 8 . 05; log-BFpupil = 6 . 09; Figure 5B ) . These log-BF values indicate that the predictive coding model was at least 400 times more likely than each of the two alternatives . 10 . 7554/eLife . 24770 . 008Figure 5 . Formal model comparison . ( A ) log-BF comparing the predictive coding model against the stimulus intensity model for SCR , pupil , NPS , and ROIs . SCR , pupil and right anterior insula show strong evidence for predictive coding ( log-BF >3 ) , while NPS and posterior insula favor the stimulus intensity model ( log-BF < −3 ) . ( B ) log-BF comparing the predictive coding model against the stimulus plus expectation model . Results are similar to ( A ) , but evidence for the stimulus plus expectation model is weaker . ( C ) Voxel-wise log-BF comparing the predictive coding model against the stimulus intensity model and in ( D ) against the stimulus plus expectation model . Maps are thresholded at \|log-BF\|>3 . Warm colors indicate support for the predictive coding model , cold colors indicate support for the alternative model . Surface projections of unthresholded log-BF insula maps reveal an anterior-posterior gradient . AI , anterior insula; PI , posterior insula; PO , parietal operculum; PCG , post-central gyrus; ACC , anterior cingulate cortex; amy , amygdala; thal , thalamus; PAG , periaqueductal gray . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 008 In contrast , NPS expression was better explained by the stimulus intensity model compared to the predictive coding model ( log-BF = −3 . 98 ) , mirroring the previously observed main effect of stimulus ( Figure 5A ) . Computing log-BF’s for the individual ROIs confirmed the results of the ANOVA interaction tests in that the anterior insula showed strong evidence for the predictive coding model compared to the two alternative models ( Figure 5 ) . Generally , the right hemisphere yielded a clearer picture in terms of model evidence , potentially because of stronger signal in the hemisphere contralateral to the heat stimulation on the left arm . For example , evidence for the predictive coding model against the stimulus intensity model in the left anterior insula ROI was below threshold , while the evidence was above threshold for the right anterior insula ( Figure 5A , B ) . Interestingly , the right parietal operculum ROI showed strong evidence for the stimulus intensity model ( log-BF = −3 . 02; Figure 5A ) . Comparing the stimulus intensity against the stimulus plus expectation model did not reveal decisive evidence for one over the other model on the ROI level . Although no comparison reached the threshold of \|log-BF\|>3 , all ROIs and physiological measures weakly supported the simpler , stimulus intensity model ( log-BF range: 0 . 12–2 . 16 ) . In order to obtain a spatially more detailed picture of the computational processes of pain processing across the brain , we computed voxel-wise log-BF’s comparing the predictive coding model against the stimulus intensity model and the stimulus plus expectation model , respectively . Again , responses in bilateral anterior insula strongly supported the predictive coding model ( Figure 5C , D ) . Within the posterior insula and parietal operculum , this more fine-grained analyses revealed bilateral evidence for the simpler , stimulus-intensity model , which was less evident on the ROI level . Similar results were obtained when comparing the predictive coding model against the stimulus plus expectation model ( Figure 5D ) , but evidence for the stimulus plus expectation model was weaker . Directly comparing the stimulus intensity model against the stimulus plus expectation model revealed modest support for the stimulus intensity model ( log-BF >2 ) in midcingulate cortex , posterior insula , and parietal operculum . A surface projection of the non-thresholded , voxel-wise log-BF maps comparing the predictive coding model against the two alternative models within the insula , demonstrated a gradual change in evidence from anterior to posterior insula ( Figure 5C , D ) . This gradient is also evident when the average log-BF from the insula is plotted over the anterior-posterior dimension ( Figure 6A ) . Importantly , and in line with anatomical considerations , the gradient is also steeper in the right hemisphere ( contralateral to stimulation ) based on the more decisive evidence in both anterior and posterior insula . 10 . 7554/eLife . 24770 . 009Figure 6 . Insula results . ( A ) Plotting the average log-BF for left and right insula against y-coordinates shows the anterior-posterior gradient from predictive coding to stimulus intensity coding . ( B ) Weight parameters are positive for prediction and PE terms , as postulated . The PE contributes approximately two times as much to the anterior insula signal as the prediction does . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 009 Since we expect both predictions and PEs to contribute positively to the measured brain signal , we extracted the weight parameters for predictions and PEs from the left and right anterior insula regions in which log-BF >3 . Firstly , the weights for both predictions and PEs were positive as postulated by the model . Interestingly , the weight for the PE was approximately two-times as strong as the prediction ( left anterior insula: 1:2 . 1; right: 1:2 . 2 ) , which is very similar to the ratio of 1:2 reported in a previous study in the fusiform face area ( Egner et al . , 2010 ) . A stronger weighting of the PE results in reduced responses for expected compared to unexpected painful stimuli as illustrated in Figure 1C and observed here in SCR ( Figure 2C ) , pupil dilation ( Figure 2D ) , and anterior insula and amygdala activation ( Figure 3A ) as well as in other studies ( Alink et al . , 2010; Todorovic et al . , 2011 ) . While neuronal coding of reward prediction errors is well understood ( Schultz et al . , 2015 ) , the specifics of aversive PE coding are currently debated ( Belova et al . , 2007; Seymour et al . , 2007; Boll et al . , 2013; Fiorillo , 2013; Klavir et al . , 2013; Roy et al . , 2014; Matsumoto et al . , 2016 ) . We therefore compared three versions of the predictive coding model that differed in their PE specification . The original model , presented above , builds on a pain PE in which a warm stimulus does not elicit a PE within pain regions ( Figures 1A and 7A ) . Alternative models incorporated an absolute PE ( i . e . , the absolute difference between stimulus and prediction ) and a signed PE ( i . e . , the difference between stimulus intensity and prediction ) , respectively ( Figure 7A ) . All models share the same prediction term . 10 . 7554/eLife . 24770 . 010Figure 7 . Comparing different PE types . ( A ) Different variants of the predictive coding model . All variants share the same prediction term ( as in Figure 1A ) , but differ in the computation of the PE . The original model used here , specifies a pain PE , which equals zero for warm stimuli ( second panel , solid lines ) . An alternative model specifies an absolute PE ( third panel , dotted lines ) . The third alternative model uses a signed PE ( fourth panel , dash-dot lines ) . Please note that all three alternatives result in the same PE for painful stimuli . They only differ in the PE for warm stimuli . The right-most panel shows the expected response profile for each of the three PE definitions when prediction and PE are equally weighted , i . e . simple sum of both terms . Please note that the signed PE ( dash-dot line ) model does not capture any factorial interactions between cues and stimuli . ( B ) log-BF comparing the absolute PE model against the original , pain PE model for ROIs and autonomic measures . No evidence stronger than log-BF < −3 is available for the absolute PE model . ( C ) log-BF comparing the signed PE model against the pain PE model shows no decisive evidence for the signed PE model . DOI: http://dx . doi . org/10 . 7554/eLife . 24770 . 010 The original , pain PE model provided a better fit of the pupil , NPS , and ROI responses when compared against the absolute PE model ( Figure 7B ) . Only the PAG and bilateral amygdala tended to favor the absolute PE model , but the model evidence did not pass the threshold of log-BF < −3 . The pain PE model also provided a better fit than the signed PE model for SCR and pupil diameter ( Figure 7C ) . For the ROIs , the pain PE also provided better fits , while the signed PE tended to fit the NPS response slightly better . But again , none of the ROI comparisons revealed above threshold evidence . In summary , neither the absolute nor the signed PE model provided compelling evidence for a better fit than the pain PE model . In fact , the pain model explained the responses better than the alternatives in most of the ROIs . Combining a probabilistic heat pain paradigm and Bayesian model comparison , we identified a functional dissociation between anterior and posterior insula suggesting that these regions implement different computations supporting pain perception . While posterior insula and parietal operculum employ stimulus intensity coding , activity in the anterior insula reflects the summation of pain expectation and prediction errors resulting from unexpected pain , thus conforming to a predictive coding account ( Egner et al . , 2010; Büchel et al . , 2014 ) . This functional dissociation was evident at a ROI level as well as on a voxel-wise analysis . The combination of expectation and prediction error was also by far the best model explaining SCR and pupil diameter responses to painful and non-painful heat stimuli . By contrast , the response profile of a multivariate brain pattern predictive of pain ratings ( NPS; Wager et al . , 2013 ) reflected only stimulus intensity . The insula as a whole is involved in a multitude of different processes – at least 20 according to one review ( Nieuwenhuys , 2012 ) – from somatosensory to emotional and conflict processing ( Kurth et al . , 2010; Chang et al . , 2013; Wiech et al . , 2014a ) . Interestingly , the cytoarchitectonic organization of the insula and its functional connectivity with other brain regions vary smoothly along an anterior-posterior gradient ( Cerliani et al . , 2012; Nieuwenhuys , 2012 ) . A similar functional gradient was evident in the present study , thereby linking the underlying cytoarchitectonic and connectivity gradients to distinct pain processing modes . The posterior insula is strongly involved in somatosensory and pain perception ( Mazzola et al . , 2012; Nieuwenhuys , 2012 ) , it receives direct spinothalamic input ( Craig , 2002; Dum et al . , 2009 ) , and it is functionally and structurally connected to somatosensory cortices ( Wiech et al . , 2014a ) . This connectivity pattern is well suited to support a stimulus intensity coding role of the posterior insula as observed here using phasic pain stimuli and as also indicated by a more tonic model of pain ( Segerdahl et al . , 2015 ) . It should be noted that the sensory role of the posterior insula includes non-painful stimuli , too ( Davis et al . , 1998 ) . By contrast anterior insula activity represents a combination of pain prediction and PE . Its predictive function fits nicely with previous reports that pre-stimulus activity in the anterior insula modulates the perception of subsequent stimuli ( Ploner et al . , 2010; Wiech et al . , 2010 ) . Furthermore , the anterior insula flexibly connects to emotional and attentional brain regions ( Taylor et al . , 2009; Ploner et al . , 2010 ) and integrates information from a diverse set of prefrontal and limbic brain regions ( Critchley , 2005; Seminowicz and Davis , 2007; Kurth et al . , 2010; Cerliani et al . , 2012; Wiech et al . , 2014a ) . In addition to its predictive function , the anterior insula also encodes mismatches between predictions and aversive outcomes during reinforcement learning ( Seymour et al . , 2004; Pessiglione et al . , 2006; Boll et al . , 2013 ) and other concurrent task demands ( Seminowicz and Davis , 2007 ) . The strong connections between anterior insula and prefrontal regions involved in contextual processing as well as its ‘hub-like’ , evaluative function ( Baliki et al . , 2009; Chang et al . , 2013; Uddin et al . , 2014 ) render the anterior insula particularly suitable for the evaluation of predictions against sensory input . The anterior insula could thus represent a mediator between somatosensory signals in posterior insula and contextual representations in prefrontral cortex , integrating those representations for perceptual decisions and behavioral responses ( Kong et al . , 2006; Seminowicz and Davis , 2007; Baliki et al . , 2009 ) . Integration of multiple information streams in this brain region could thus be crucial for the construction of pain experiences that are shaped by learning and external feedback ( Wiech , 2016; Geuter et al . , 2017 ) . In addition to the anterior insula , both SCR and pupil diameter in this study – and reaction times in a similar paradigm ( Wiech et al . , 2014b ) – showed a pattern predicted by our model . Expectations of certain visual stimuli can also sharpen their cortical representation ( Kok et al . , 2012 ) , but it is unknown how this would translate to pain reports — whether predicted pain would be more or less intense . A recent study did observe no difference in pain ratings between correctly and incorrectly cued pain stimuli ( Zeidan et al . , 2015 ) . Here , we opted for a time-efficient design and did not collect trial-by-trial pain ratings to address this question . However , investigating potential indicators for perceived pain , responses of the NPS ( Wager et al . , 2013; Krishnan et al . , 2016 ) and autonomic measures ( Geuter et al . , 2014 ) , can offer insights . All three measures – NPS , SCR , and pupil – showed stronger responses to unexpected compared to expected pain , hinting at a potential enhancement of unexpected pain stimuli that needs to be investigated more thoroughly in futures studies . Taken together , the results show that principles of predictive coding are relevant for behavioral responses in the context of pain . Other stimulus attributes than painfulness , e . g . general aversiveness , salience , or motivational demands , co-vary with stimulus intensity . Previous studies have correlated brain activity with pain reports and stimulus intensity in order to dissociate the two ( Coghill et al . , 1999; Büchel et al . , 2002; Davis et al . , 2002 , 2004; Porro et al . , 2004; Baliki et al . , 2009 ) . Interestingly , anterior insula activity correlated with perceived heat even in the absence of heat stimulation ( Davis et al . , 2004 ) . Studies by Downar and colleagues ( Downar et al . , 2000 , 2003 ) also found that anterior insula and anterior cingulate cortex responded to unexpected or novel stimuli . Within a predictive coding framework , the overall response is decomposed into two distinct functional components – a prediction term and PE term – that are key components of learning theory . Interestingly , the anterior insula also showed a prediction error response in the present study in line with previous work ( Downar et al . , 2000 , 2002; Seymour et al . , 2004; Boll et al . , 2013 ) . From a psychological perspective , the decomposition is important because directing attention towards expected aversive events ( high probability of pain , prediction term ) and towards unexpected events ( PE ) is adaptive . The prediction error is assumed to drive learning ( Rescorla and Wagner , 1972 ) and is thus critical for adaption to the environment . By contrast , salience – understood as the difference to preceding sensory events ( Mouraux et al . , 2011 ) – emerges after stimuli have been processed and the elicited surprise or PE has been computed . Salience can thus be considered a secondary stimulus property resulting from a high PE that in turn can modulate subsequent updating of beliefs . This process is formalized in the Pearce-Hall model of reinforcement learning in which a surprising , salient outcome , affects the learning rate in the next trial ( Pearce and Hall , 1980; Boll et al . , 2013; Atlas et al . , 2016 ) . Predictive coding theories offer a parsimonious computational implementation of cross-modal , Bayesian perceptual decision making ( Knill and Pouget , 2004; Friston , 2005; Summerfield and de Lange , 2014 ) . These accounts can explain several effects within a single framework including extra-classical receptive field effects in visual cortex ( Rao and Ballard , 1999 ) , repetition suppression ( Summerfield et al . , 2008; Todorovic et al . , 2011 ) , and have been suggested as a framework to understand placebo effects ( Büchel et al . , 2014 ) . In support of a domain general integration process of expectations and PEs , the ratio of the contributions of both processes to physiological signals observed here , mirrored the ratio previously reported for the fusiform face area in visual perception ( Egner et al . , 2010 ) . Interestingly , in both studies , the PE was weighted stronger than the prediction , which suggests that learning and updating of the internal model are crucial for perception . In addition , the observation that the anterior insula also processes PEs in other modalities ( Downar et al . , 2000; Iglesias et al . , 2013 ) hints towards a cross-modal role of the anterior insula within a predictive coding framework . Another feature of predictive coding models is their hierarchical organization: At each level of the neural hierarchy , predictions and PEs will be computed for the specific features encoded in this region ( Rao and Ballard , 1999; Friston , 2005 ) . For example , early visual and auditory areas process multimodal stimuli under the assumption of independent physical sources and only higher areas form joint representations using Bayesian inference ( Rohe and Noppeney , 2015 ) . Although anterior insula activity matched the response pattern proposed by the predictive coding model , other brain regions and the NPS followed a stimulus intensity model . This discrepancy could be either due to inherently distinct computations implemented in those regions or due to the fact that the nature of predictions and PEs changes across regions ( Iglesias et al . , 2013 ) . Because pain is an inherently multi-faceted experience that includes sensory-discriminative , emotional , and motivational components , the present predictive coding model could capture certain aspects of this multi-faceted experience better than others . The computational difference observed between anterior and posterior insula could , at least in part , reflect such functional differences . In fact , studies investigating visual processing within a predictive coding framework also observed regionally restricted effects based on the manipulated stimulus features . For example , activity of the fusiform face area and parahippocampal place area is well described by a predictive coding model , but each region responds selectively to their respective preferred stimuli , i . e . faces and houses ( den Ouden et al . , 2010; Egner et al . , 2010; Jiang et al . , 2013 ) . Similarly , expectations of certain low-level visual features such as grating orientation , selectively attenuate primary visual cortex activity , but not activity in higher visual areas ( Alink et al . , 2010; Kok et al . , 2012 ) . Based on these results , the observed computational differentiation between anterior and posterior insula indicates that both regions process distinct features of painful stimuli and these could be related to different psychological and behavioral outcomes in healthy and patient populations . The representation of aversive prediction errors in the brain is still not fully understood . Important open questions include whether aversive PE are represented on a continuous dimension along with reward prediction errors and whether particular brain regions represent absolute , signed , or pain PE . Activity reflecting absolute aversive PE has been observed in the amygdala ( Boll et al . , 2013; McHugh et al . , 2014 ) , while signed aversive PE have been observed in the striatum and PAG ( Seymour et al . , 2005 , 2007; Roy et al . , 2014; Zhang et al . , 2016 ) . Within sensory cortices , unexpected omissions of visual or auditory stimuli lead to enhanced activity in auditory or visual areas , reminiscent of absolute prediction errors ( den Ouden et al . , 2009 , 2010; Todorovic et al . , 2011; Todorovic and de Lange , 2012 ) . Comparing models incorporating different PE specifications , showed that the model based on an asymmetric , pain PE explained brain responses in the present study best . Our results thus suggest that PE encoding in the anterior insula differs between situations when the outcome is more pain than expected compared to unexpected pain omissions . The differentiation of PEs observed here is similar to a distinction observed in visual processing: in two studies , activity in the fusiform face area , a face-selective brain region , reflected prediction errors for face stimuli , but not for house stimuli ( den Ouden et al . , 2010; Egner et al . , 2010 ) . In summary , the observed responses in SCR , pupil dilation , and anterior insula activation demonstrate that at least part of the pain experience can be explained by a domain general predictive coding framework . The parallels observed between pain and visual processing ( Egner et al . , 2010 ) hint towards a general processing principle based on internal predictions and PE . An interesting question for future research is how the contributions of predictions and PE shift in states of altered or chronic pain conditions that are also related to altered learning processes ( Vlaeyen , 2015 ) . As anterior insula structure and function changes profoundly in chronic pain conditions ( Bushnell et al . , 2013; Ceko et al . , 2013; Hong et al . , 2014; Flodin et al . , 2015 ) , it is possible that the precision or influence of the prediction is strongly enhanced in chronic pain conditions or that PEs are incorrectly computed ( Edwards et al . , 2012 ) . If the underlying computations are domain general , this would also explain the hyper-sensitivity observed in certain chronic pain populations to non-painful tactile and visual stimuli ( López-Solà et al . , 2014 ) . This framework could hence open up new ways to investigate pain processing in clinical populations . Twenty-eight healthy subjects ( 17 female ) with an average age of 25 . 9 years ( range: 21–33 years ) participated in this study . No subject reported any psychiatric , neurological , dermatological , or pain conditions . Due to equipment malfunction , skin conductance data from seven subjects could not be analyzed ( resulting in a sample size of N = 21 for SCR analyses ) and technical issues prohibited pupil data collection for eight subjects ( leaving N = 20 for pupil analyses ) ; only one participant had neither SCR nor pupil data . Other behavioral and fMRI data analyses are based on the full sample of 28 participants . The sample size was determined as 1 . 5 times the sample of a seminal fMRI study on pain expectations that tested 19 subjects ( Atlas et al . , 2010 ) . The Ethics committee of the Medical Chamber Hamburg approved the study . After arrival at the laboratory , subjects were informed about the procedures of the experiment and provided written informed consent . The experiment was divided into three parts – a temperature calibration phase , a behavioral training session , and the functional magnetic resonance imaging ( fMRI ) experiment . First , we calibrated the temperatures to be used in the experiment individually for each subject ( outside of the MR-scanner ) . For calibration , subjects rated 36 cutaneous heat stimuli ( total duration: 1 . 5 s , ramp-up: 70°C/s , ramp-down: 40°C/s ) with temperatures ranging from 42°C to 49 . 5°C ( in steps of 0 . 5°C ) in a pseudo-randomized order using a computerized visual analogue scale ( VAS ) . Sixteen different temperatures between 42°C and 49 . 5°C in steps of 0 . 5°C were presented two times each during calibration ( except for 44 , 45 , 46 , and 47°C , which were repeated three times each ) , resulting in a total of 36 stimuli . The stimulus interval was 13–17 s plus the time participants needed for their VAS rating ( mean: 5 . 04 s , standard deviation: 1 . 01 s ) . Heat stimuli were applied to the left volar forearm and different skin sites were used for calibration , behavioral training and fMRI scanning . The extremes of the VAS were labeled ‘no sensation at all’ and ‘unbearable pain’ . The center of the VAS was labeled ‘pain threshold’ . This VAS partition was necessary because we needed to determine one painful and one non-painful , but clearly noticeable level of stimulation for the main experiment . Subjects were instructed to only rate stimuli as above the pain threshold if the stimulus induced any painful sensation . For stimuli that were perceived as different from baseline but not painful , subjects rated the intensity of the warmth on the lower half of the VAS . ‘Unbearable pain’ was explained to the subjects as the intensity at which they would have to lift the thermode from the arm . VAS ratings were converted to numerical values ranging from 0 to 100 . Intensity ratings did not differ between men and women ( t ( 26 ) = 1 . 32; p=0 . 2 ) . The average correlation across subjects between temperature and rating was high: r¯ = 0 . 78 ( standard deviation: 0 . 13 ) . We used linear regression to determine one temperature that was clearly noticed by the subject but not painful ( VAS 30 ) and a second temperature that was perceived as painful but tolerable ( VAS 75 ) . We next applied the selected temperatures to the subjects’ forearm to ensure that the warm stimulus was not painful , but clearly distinguishable from baseline and that the painful stimulus was bearable – this was the case for every subject . The average temperature for the warm stimulus was 45 . 0°C ( standard deviation: 1 . 2°C ) and the average temperature for the painful stimulus was 49 . 4°C ( standard deviation: 1 . 3°C ) with a maximum temperature of 49 . 5°C . Following calibration , subjects were informed about the cues and the contingencies between cues and heat stimuli ( Figure 1E ) . The explicit information and the training block ensured that subjects knew the contingencies . The training also minimized learning taking place during the fMRI session . Cue-intensity contingencies were counterbalanced across subjects and subjects were shown their respective pairings on a computer screen . The behavioral training session consisted of one block of 48 trials ( see Task , below ) . After the training block , subjects were presented with each of the cues separately on the screen and reported which cue was associated with high , medium , and low probability of pain , respectively . All subjects associated each cue with its correct probability of receiving pain . After training , subjects were positioned in the MRI scanner and completed 4 blocks of the experiment for a total of 192 trials . The design was identical to the training session , except that each block had a different , pseudo-randomized trial order . The order of blocks was randomized across subjects . The thermode was moved to a different position after each block to prevent sensitization of the skin . During each block , we measured BOLD responses , skin conductance , and pupil diameter . After the end of the fMRI experiment we acquired a high-resolution anatomical image of each subject’s head . The whole experiment lasted about 2 h per subject . During each trial , a fixation dot was presented centrally on the screen . One of three cues then appeared 300 ms before the heat stimulus started . Heat stimulus duration was 1500 ms ( including ~200 ms ramp up and down , respectively ) . The cue was visible during heat stimulation and remained on display until the end of the heat stimulation . After a variable interval of 3–5 s , a rating screen appeared asking subjects whether the last stimulus had been painful . Subjects answered ‘yes’ or ‘no’ by pressing either the left or right arrow key of a button-box ( Figure 1D ) . A fixation dot was presented again during the inter-trial interval ( ITI ) of 12 s duration . At the end of the training block and after each fMRI block , subjects rated the perceived intensity of the warm and the painful stimuli ( on the same VAS as used during calibration ) . Ratings were in good agreement with the calibrated target ratings of VAS 30 and VAS 75 , respectively ( Figure 2A ) . Each cue was presented 16 times in each experimental block . The high pain probability cue was followed by the painful stimulus in 75% of the 16 trials and by the warm stimulus in 25% of the trials . Probabilities for the medium cue were 50% for each stimulus . For the low pain probability cue , the chance for a painful stimulus was 25% and 75% for a warm stimulus ( Figure 1E ) . Gray-scale versions of abstract symbols ( kindly provided by Dr . Philippe Tobler [Tobler et al . , 2006] ) served as cues ( Figure 1E ) . We included a basic target detection task to ensure that subjects paid attention to the task ( Egner et al . , 2010 ) . In 12 . 5% of the trials , the fixation dot changed its color to red at the beginning of the somatosensory stimulation . Subjects were asked to respond to the color change by pressing a third key . They were informed that the color change was completely unrelated to the main experimental task . Cues were not related to the color change , as target trials were evenly distributed across cues . During the fMRI experiment , subjects were rewarded with 50 cents for each correct target hit . Detection performance was at ceiling with a minimum of 23 out of 24 correct detections ( mean: 23 . 8 ) . Importantly , the main effect of stimulus on target reaction time was non-significant ( F ( 1 , 27 ) = 0 . 295; p=0 . 591 ) , indicating that subjects were similarly attentive during pain and warm trials . Stimulus presentation , response logging and thermode triggering were carried out using the Psychophysics Toolbox 3 ( http://www . psychtoolbox . org ) . Thermal stimulation was delivered via a MRI compatible 3 cm diameter Peltier thermode ( CHEPS Pathway , Medoc , Israel ) . Skin conductance was recorded using a Biopac EDA100C MRI system ( Biopac Systems , Inc . , Goleta , CA , USA ) and a CED1401 A/D converter ( Cambridge Electronic Design , Cambridge , UK ) at a sampling rate of 100 Hz . Electrodes were attached to the thenar and hypothenar eminences of the left hand . Pupil diameter was recorded from the right eye using an MR-compatible EyeLink 1000 system ( SR Research , Ottawa , ON , Canada ) at a sampling rate of 1000 Hz . The lights in the MRI room were dimmed and luminance was kept constant across subjects . This setup provided a balance between eye-tracking quality and participant comfort . Functional magnetic resonance imaging ( fMRI ) data were acquired on a Siemens Trio 3 Tesla system equipped with a 32-channel head coil ( Siemens , Erlangen , Germany ) . Thirty-eight transversal slices ( voxel size 2 × 2 × 2 mm , 1 mm inter-slice gap ) were acquired within each volume using a T2* sensitive echo planar imaging ( EPI ) sequence ( TR = 2 . 34 s , TE = 26 ms , flip angle: 80° , field of view: 220 × 220 mm , parallel acceleration factor = 2 ) . Slices were tilted about 30° relative to the AC–PC line to improve coverage in the brainstem . Additionally , T1 weighted structural images ( 1 × 1 × 1 mm resolution ) were obtained using a MPRAGE sequence ( TR = 2300 ms , TE = 9 ms , flip-angle = 9° ) .	All over the human body , there are receptors that help to alert the brain to potential harm . For example , intense heat on the skin elicits a signal that travels to the brain and activates many parts of the brain . Some of the same brain regions that are switched on by signals of potential bodily harm also help the brain to form expectations about events . A person’s expectations may have a strong influence on how they experience pain . For example , if a person expects that taking a pill will reduce their pain , they may feel less pain even if the pill is a fake . Exactly how the brain processes pain signals and expectations remains unclear . Does the brain activity simply reflect how intense the heat is ? Some scientists think there may be two separate processes going on: one that predicts what will happen and another that calculates the difference between the prediction and what the receptors actually detect . This difference is called a prediction error . If every unpredicted sensory signal elicits a calculation of the prediction error that would help improve the brain’s future predictions . Now , Geuter et al . show that the predictions are a key part of how the brain perceives pain induced by heat . In the experiments , 28 people had heat applied to skin on their forearm at temperatures that were either noticeable but not painful or painful . Their brain activity was recorded using functional magnetic resonance imaging ( fMRI ) , and measurements were taken of the pupils in their eyes and their skin’s response to heat . The fMRI scans showed that activity in the back part of a brain region called the insular cortex reflects the intensity of the heat that is applied to the person’s arm , while the front part of the same region signals pain predictions and the prediction error . This suggests that scientists are correct that pain predictions and prediction error calculations are an integral part of the pain response . More studies are needed to determine if these brain processes might contribute to chronic pain and whether a similar process occurs in response to other types of unpleasant experiences .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2017	Functional dissociation of stimulus intensity encoding and predictive coding of pain in the insula
Many cells contain non-centrosomal arrays of microtubules ( MTs ) , but the assembly , organisation and function of these arrays are poorly understood . We present the first theoretical model for the non-centrosomal MT cytoskeleton in Drosophila oocytes , in which bicoid and oskar mRNAs become localised to establish the anterior-posterior body axis . Constrained by experimental measurements , the model shows that a simple gradient of cortical MT nucleation is sufficient to reproduce the observed MT distribution , cytoplasmic flow patterns and localisation of oskar and naive bicoid mRNAs . Our simulations exclude a major role for cytoplasmic flows in localisation and reveal an organisation of the MT cytoskeleton that is more ordered than previously thought . Furthermore , modulating cortical MT nucleation induces a bifurcation in cytoskeletal organisation that accounts for the phenotypes of polarity mutants . Thus , our three-dimensional model explains many features of the MT network and highlights the importance of differential cortical MT nucleation for axis formation . Microtubules ( MTs ) are polar cytoskeletal filaments that can adopt different global network organisations to fulfil different functions . The vast majority of studies have focussed on understanding the architecture and function of MT arrays organised by centrosomes , such as radial arrays or the mitotic spindle . In contrast , much less is known about the organisation , assembly and function of non-centrosomal MT arrays despite their ubiquity in differentiated cell types , such as neurons , epithelia , fission yeast and plants ( Bartolini and Gundersen , 2006; Carazo-Salas and Nurse , 2006 ) . In a variety of cell-types , non-centrosomal MT arrays play an essential role in directing the subcellular localisation of mRNAs to spatio-temporally control gene expression . In neuronal dendrites , for example , MTs form bidirectional arrays with MT plus-ends both pointing away and towards the cell body ( Conde and Caceres , 2009; Kapitein and Hoogenraad , 2011 ) . MTs and associated motor proteins were implicated in the transport of mRNAs along dendrites ( Bramham and Wells , 2007 ) , where their activity-dependent translation contributes to long term changes in synaptic function , neuronal circuitry and memory ( Miller et al . , 2002; Sutton and Schuman , 2006; Holt and Schuman , 2013 ) . Similar overlapping , bidirectional MT arrays in the Xenopus oocyte have been proposed to mediate the transport of Vg1 mRNA to the vegetal cortex , where it orchestrates germ layer patterning ( Messitt et al . , 2008; Gagnon et al . , 2013 ) . The Drosophila oocyte is probably the best-studied example of mRNA transport along non-centrosomal MTs . In this system , a diffuse gradient of MTs of mixed polarity is required for the localisation of bicoid and oskar mRNAs to opposite ends of the cell ( Becalska and Gavis , 2009 ) . Despite the large amount of work , however , the organisation of the non-centrosomal MT cytoskeleton underlying this mRNA localisation is controversial ( MacDougall et al . , 2003; Januschke et al . , 2006; Zimyanin et al . , 2008 ) , and its assembly and function are not understood . In stage 9 oocytes , MTs grow from most parts of the cell cortex into the volume ( Theurkauf et al . , 1992; Parton et al . , 2011 ) thereby giving rise to a complex , three-dimensional MT network without pronounced polarity along the anterior-posterior ( AP ) axis ( Theurkauf et al . , 1992; Cha et al . , 2001 ) . In contrast to this apparent disordered organisation , bicoid and oskar mRNAs become reliably localised by Dynein and Kinesin to the anterior corners and to the posterior pole of the oocyte , respectively , thereby defining the AP axis ( Brendza et al . , 2000; Duncan and Warrior , 2002; Januschke et al . , 2002; Weil et al . , 2006; Zimyanin et al . , 2008 ) . The most pronounced feature of the MT cytoskeleton at stage 9 is a gradient of cortical MTs from the anterior to the posterior pole , where MT nucleation is suppressed by the polarity protein PAR-1 ( Doerflinger et al . , 2006; Roth and Lynch , 2009 ) . Live imaging of oskar mRNAs and direct measurements of growing MTs showed that the MT network is mostly disordered with only a weak statistical bias of about 8% more plus ends pointing towards the posterior pole ( Zimyanin et al . , 2008 ) . This bias vanishes in absence of PAR-1 when MTs nucleate from all over the cortex ( Parton et al . , 2011 ) . While these findings established a directionality of the MT meshwork for the first time , several questions remain unanswered . For example , mRNA transport and localisation is highly reproducible , raising doubts about whether the underlying cytoskeletal organisation can be mostly disordered . Moreover , Dynein-dependent transport to MT minus ends localises injected , so-called naive bicoid mRNA to the cortex closest to the injection site ( Cha et al . , 2001 ) , which is difficult to reconcile with a cytoskeleton that is simply biased towards the posterior everywhere . Similarly , the different behaviour of so-called conditioned bicoid mRNA , which localises specifically to the anterior cortex irrespective of the injection site ( Cha et al . , 2001 ) , has remained unexplained . Finally , motor-driven cytoskeletal transport is not the only transport mechanism in oocytes . Kinesin moves only 13% of oskar mRNA at any given time , and the remaining 87% is subject to diffusion and slow cytoplasmic flows that are driven indirectly by Kinesin activity on the MT network ( Zimyanin et al . , 2008 ) . This raises the question if cytoskeletal transport alone is sufficient to account for mRNA localisation ( Glotzer et al . , 1997; Forrest and Gavis , 2003 ) and highlights the importance of distinguishing its contribution to localisation from the contribution of flows and diffusion ( Serbus et al . , 2005 ) . Here , we present the first theoretical model for stage 9 Drosophila oocytes . Based on the distribution of MT nucleation sites around the cortex , this model accurately reproduces the observed distribution of MTs and cytoplasmic flows in the oocyte . It reveals that the MT cytoskeleton is compartmentalised and more ordered than previously thought . By modelling the movement of mRNAs on this network , we also show that this MT organisation is sufficient to explain the localisation of oskar and naive bicoid mRNA . Finally , we show that modulation of MT nucleation gradients causes a bifurcation in cytoskeletal organisation that explains mutant phenotypes . Thus , our results explain many features of the assembly , organisation and function of the non-centrosomal MT array in Drosophila oocytes and highlight the key role of differential MT nucleation or anchoring at the cortex ( Lüders and Stearns , 2007 ) . MTs in the oocyte are nucleated or anchored at the cortex ( Theurkauf et al . , 1992; Parton et al . , 2011 ) and grow from the membrane into the volume . The cortical MT density follows a steep gradient along the posterior-lateral cortex from high density at the anterior corners to low densities at the posterior pole ( Figure 1D , ‘Materials and methods’ MT nucleation probability ) . In our model , we emulated these features by selecting seeding points for MTs at random positions along the oocyte cortex , comprised of two parabolic , rotationally symmetric caps that capture the typical shape of a stage 9 Drosophila oocyte ( Figure 1A , ‘Materials and methods’ Coordinates and oocyte geometries ) . The density of seeding points decreases steeply along the posterior-lateral cortex to zero at the posterior pole , and decreases weakly along the anterior cortex towards the anterior centre ( Figure 1A , B , ‘Materials and methods’ MT nucleation probability ) . Each seeding point nucleates a MT polymer that grows until it either hits a boundary or reaches a target length imposed by the aging of MTs before catastrophe ( Gardner et al . , 2011 ) ( ‘Materials and methods’ MT growth ) . In total , we computed more than 55 , 000 MTs for each realisation of a 3D wild-type cytoskeleton . 10 . 7554/eLife . 06088 . 003Figure 1 . Models for the microtubule ( MT ) meshwork show local order in the cytoskeleton . ( A ) 3D geometry of a stage 9 Drosophila oocyte ( grey , anterior to the left , posterior to the right ) containing more than 55 , 000 MT seeding points . From the corners to the centre , MT seeding density decreases weakly along the anterior ( kA = 1000 μm , h0A=0 . 8 ) and strongly along the posterior-lateral cortex ( kP = 150 μm , h0P=0 , see ‘Materials and methods’ MT nucleation probability ) . Nucleated MT polymers are stiff random walks , initially pointing in a random direction . Only MT segments in a cross section are shown ( green ) to emulate confocal images . MT target lengths are chosen from a probability distribution that accounts for the MT aging process . The mean target length is set to a fraction ϵ of the anterior-posterior ( AP ) axis length , here ϵ = 0 . 5 ( ‘Material and methods’ MT growth ) . The inset shows the 3D angular distribution of 0 . 5% of all MT segments with 3D statistical bias . ( B ) Cross section through the MT cytoskeleton shown in A with 2D directional bias ( top right ) . The inset shows 2D posterior bias ( in percent ) as function of depth ( bottom right ) . ( C ) Local vector sum of MT segments from the cross section in panel B on a coarse-grained grid shown as streamlines that visualize local directionality . ( D ) Staining of α-tubulin ( green ) shows MT density distribution in a fixed stage 9 oocyte . Nuclei in blue ( DAPI ) , scale bar is 30 μm . ( E ) Schematic detailing the work flow in the model and comparisons to experiments . ( F ) Local directionality of MT cross section as in panel C for an average over 50 independent realizations of the cytoskeleton . ( G ) Local directionality computed from the rod model with the same parameters as in panels A , B but shortened MT lengths ( ‘Material and methods’ Motor velocity field ) . Orange arrows show the separatrices between subcompartments . ( H ) MT density distribution computed from 50 realizations of the polymer model . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 00310 . 7554/eLife . 06088 . 004Figure 1—figure supplement 1 . Compartmentalization of the MT cytoskeleton is robust to changes in oocyte geometry . ( A ) Alternative geometry for a stage 9 oocyte comprised of a posterior parabolic cap and an anterior disc . MT segments intersecting a cross section in one realization of the polymer model with nondimensional seeding density parameters kP = 150 μm , h0P=0 , kA=1000 μm , h0A=0 . 8 and MT length parameter ϵ = 0 . 5 from more than 55 , 000 seeding points are shown in green . Inset shows directionality of 0 . 5% of all MT segments with posterior bias . ( B ) Cross section through 3D MT cytoskeleton shown in panel A with 2D directional bias ( top right ) . Inset shows 2D posterior bias ( in percent ) as function of depth ( bottom right ) . ( C ) Local vectorial sum of MT segments visualized as a streamplot . One individual realization of the MT cytoskeleton shows poor spatial order . ( D ) α-tubulin staining of an early stage 9 oocyte . Scale bar is 25 μm . ( E ) Schematic of the work flow in our model . ( F ) Streamplot of the local vectorial sum of 50 realizations of the polymer MT cytoskeleton with parameters as in A , B . ( G ) Local net orientation of straight rod MTs in the rod model with parameters as in A , B but reduced mean MT target length ϵ = 0 . 28 , showing the same compartmentalization of the oocyte as the average in the polymer model . ( H ) MT density distribution in the ensemble of 50 realization of the polymer MT cytoskeleton . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 004 A cross section through the computed MT meshwork ( Figure 1A , B ) bears a striking resemblance to confocal images of Tau- ( Micklem et al . , 1997; Ganguly et al . , 2012 ) and EB-1- ( Parton et al . , 2011 ) tagged MTs , correctly showing the pronounced AP gradient of MT density . Taking into account the orientations of all MT segments in the 3D volume , it also reproduces the experimentally measured directional bias ( Parton et al . , 2011 ) with 8 . 5% more MT segments pointing posteriorly than anteriorly ( Figure 1A , inset ) . The directional bias in a 2D slice varies depending on the depth of the slice in the oocyte , ranging from 50 . 7% of posteriorly oriented MT segments at the cortex to 60 . 6% in the mid plane ( Figure 1B , inset ) . This demonstrates that measurements in 2D confocal slices poorly characterise the fully extended 3D system . Central to understanding mRNA localisation is the question of MT orientations . Calculating the local vectorial sum of MT segments on a coarse-grained grid gives the local MT orientations in the computed cytoskeleton . Local orientations of an individual realisation of the cytoskeleton ( Figure 1A , B ) show poor global network order ( Figure 1C ) . However , cargo localisation in the oocyte does not involve only a single cytoskeletal realisation . MTs disappear after 5–10 min in the presence of Colchicine ( Theurkauf et al . , 1992; Zhao et al . , 2012 ) ( V Trovisco , personal communication ) , a drug that blocks MT growth and destabilizes dynamic MTs , indicating that the whole network turns over within minutes . Thus , the oocyte samples many tens to a hundred of independent MT organisations over the 6–9 hr of stage 9 . Summation of local orientations over an ensemble of 50 independent realisations of the computed MT network reveals a striking spatial partitioning of the cytoskeleton into several subcompartments ( Figure 1F ) . At the anterior , the mean orientation points posteriorly , while MTs at the lateral sides on average point inwards toward the AP-axis . Counting the number of MT segments that contribute to each grid box in the ensemble also shows the distribution of MT density ( Figure 1H ) . In agreement with experiments ( Figure 1D ) , the MT density exhibits a pronounced AP gradient with highest values in the anterior corners , even when MT seeding is uniform on the anterior surface . In the limit of a large ensemble , the MT polymers sample all possible initial directions and possible lengths at every point along the cortex . In this limit , we can test the predicted compartmentalised MT organisation by constructing a second model in which MTs are represented as straight rods . The net orientation of the cytoskeleton at a given point inside the oocyte is then computed as weighted sum of MT contributions from each point along the boundary ( ‘Materials and methods’ Model setup ) . With a shorter MT target length to compensate for the effectively longer length of straight rods compared to curved polymers , computation of net MT orientations in the rod model shows a topology ( Figure 1G ) that confirms the mean topology in the polymer model ( Figure 1F ) . The three compartments of a posterior-pointing anterior section and two inwards-pointing lateral sections are effectively bounded by separatrices ( Figure 1G , orange arrows ) . Cargo molecules that are transported on MTs to their plus ends move along the arrows and converge at the separatrices , eventually leading to the posterior pole as the sole point of attraction in the entire oocyte volume ( the attractor ) . By contrast , cargo that is transported to MT minus ends moves opposite to the arrows and diverges away from the separatrices . This mean topology of the MT cytoskeleton is insensitive to the exact choice of the MT nucleation probability and to the choice of the MT length distribution ( ‘Materials and methods’ MT nucleation probability , MT growth ) . It also remains unchanged in a differently shaped oocyte geometry in both the polymer model and the rod model ( Figure 1—figure supplement 1 ) . In summary , both models show that even disordered non-centrosomal MT arrays can feature well-defined mean organisations , and that the MT cytoskeleton in oocytes is organised in a compartmental fashion . The suitably scaled local vectorial sum of MT segments ( Figure 1C ) for an individual realisation of the polymer model ( Figure 1A , B ) is a vector field vm which represents active Kinesin-driven transport on the cytoskeleton ( ‘Materials and methods’ Motor velocity field ) . In vivo during stage 9 , the oocyte cytoplasm undergoes slow cytoplasmic flows that are abolished in kinesin heavy chain mutants . This indicates that flows are driven by kinesin-dependent transport of an unknown cargo through the viscous cytoplasm ( Palacios and St Johnston , 2002; Serbus et al . , 2005 ) , thus making cytoplasmic flows a secondary read-out of cytoskeletal organisation ( Khuc Trong et al . , 2012 ) . Therefore , we next tested if our computed polymer MT cytoskeleton can produce flows that are consistent with observed cytoplasmic streaming . Measurements of speeds of autofluorescent yolk granules in live stage 9 oocytes showed that mean flow velocities are slow ( Figure 2F ) . The physics of slow incompressible fluid flows u driven by forces f is described by the Stokes equations ( 1 ) 0=−∇p+μ∇2u+f , ∇⋅u=0 . 10 . 7554/eLife . 06088 . 005Figure 2 . Computed cytoplasmic flow fields capture key elements of in vivo flows . ( A ) Streamlines ( light blue lines ) visualize the 3D cytoplasmic flow field computed from the realization of the cytoskeleton shown in Figure 1A . The horizontal plane shows a 2D cross-section through the 3D field . Anterior to the left , posterior to the right . ( B ) Cross-section through the 3D field shown in panel A with arrows indicating flow directions and colouring indicating flow speeds . ( C ) Confocal image of a live stage 9 oocyte . Arrows show the flow field computed from particle image velocimetry ( PIV ) of streaming yolk granules and averaged over ≈5 min . Scale bar is 25 μm . ( D ) Same as A , but showing the mean flow organization for an average of 100 individual 3D flow fields , analogous to the mean organization of the MT cytoskeleton in Figure 1F . ( E ) Same as B for the average in panel D . ( F ) Mean fluid flow speeds were obtained by PIV ( red , 13 . 7 ± 0 . 8 nm/s , mean ± sem ) and automatic particle tracking ( orange , 15 . 3 ± 0 . 7 nm/s , mean ± sem ) from 48 oocytes . Experimentally measured flow speeds were used to calibrate the forces f in the Stokes Equation 2 such that the computed mean speeds in 3D ( blue , mean: 14 . 5 nm/s ) or in 2D cross sections ( green , mean: 14 . 8 nm/s ) match the measured values . The larger spread in experimental flow speeds may reflect greater variability of motor activity , cytoplasmic composition , geometry or age in vivo . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 00510 . 7554/eLife . 06088 . 006Figure 2—figure supplement 1 . The oocyte nucleus covers a negligible fraction of the oocyte volume and disturbs the flow field only locally . ( A ) Streamlines visualizing the flow field computed in the 3D oocyte geometry with additional no-slip boundary condition on a sphere representing the nucleus . Same forces as in Figure 2A , B , main text . The nucleus occupies 2 . 2% of the oocyte volume . Horizontal plane shows a cut through the flow field . ( B ) Young stage 9 oocyte stained with DAPI ( blue ) , phalloidin ( red ) and showing oskar MS2 GFP ( green ) in the process of reaching the posterior pole . Scale bar is 25 μm . ( C ) Extracted and rotated shape of the oocyte in panel B ( red ) , and symmetrized geometry resulting from averaging the shape above and below the AP-axis ( blue ) . Black circle indicates the nucleus . ( D ) Ratio of nucleus volume to oocyte volume computed for Ne = 9 early ( red ) and Nm = 7 mid stage 9 oocytes ( blue ) . Arrowhead marks the oocyte shown in panel B . ( E , F ) Cross sections through the 3D flow fields for the same forces as in panel A with ( F ) and without ( E ) nucleus . ( G , H ) Same as E and F , but vertical cross section that does not contain the nucleus . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 006 We make the simplest possible assumption that the forces f are proportional to the motor-velocity field , and use experimentally measured flow speeds to calibrate the scalar factor of proportionality . By solving the Stokes equations , we then computed the full 3D fluid flow field ( Figure 2A ) corresponding to an individual realisation of the MT cytoskeleton ( Figure 1A ) , and 2D cross sections through the 3D field ( Figure 2B ) were compared to in vivo flow patterns visualised by particle image velocimetry ( PIV , Figure 2C ) . Despite large variability , both computed and in vivo flows show very similar patterns , generally being strongest in the anterior half of the oocyte , and weaker in the posterior half ( Figure 2B , C ) . Computed flows occasionally reach further into the posterior half than typically seen in PIV flow fields , thereby slightly overestimating the range of flows . However , except in rare cases , computed flows do not reach the posterior pole . This result is largely independent of the presence of the oocyte nucleus which occupies less than 2 . 5% of the oocyte volume , even for small oocytes at the beginning of stage 9 ( Figure 2—figure supplement 1 , ‘Materials and methods’ Nucleus to oocyte volume ratio , Impact on flow field ) . Thus , computed flows appear consistent with observed slow cytoplasmic streaming . We tested next if our computed cytoskeleton in combination with the derived cytoplasmic flows and diffusion can account for dynamic mRNA transport and localisation . We described the mRNA distributions as continuous concentration fields . oskar mRNA is assumed to reside in either one of two states: the Kinesin-bound state with concentration cb , in which cargo is transported actively on the cytoskeleton-derived motor-velocity field vm ( Figure 1C ) , or the unbound state with concentration cu , in which cargo is transported by the cytoplasmic flows u ( Figure 2A ) and diffuses with diffusion constant D . Cargo can exchange between both states by chemical reactions , thereby resulting in the reaction-advection-diffusion equations: ( 2 ) ∂t cb+∇⋅ ( vm cb ) =kb cu−ku cb , ∂t cu+∇⋅ ( u cu ) =−kb cu+ku cb+D∇2cu . Parameter values were constrained with experimentally measured values ( ‘Materials and methods’ Parameter values ) . Furthermore , throughout the simulation we cycled through the pairs of fluid flow u and motor-velocity fields vm ( Figure 3 ) to account for the dynamic nature of the MT network ( Parton et al . , 2011 ) and flow patterns , which change over time scales of minutes compared to the 6–9 hr during which oskar mRNA is localised ( ‘Materials and methods’ Autocorrelation function ) . 10 . 7554/eLife . 06088 . 007Figure 3 . The model recapitulates oskar and bicoid mRNA transport , implying dominance of cytoskeletal transport . ( A–C ) Shown are fixed oocytes with oskar MS2 GFP ( green ) and stained with DAPI ( blue ) and Phalloidin ( red ) . oskar mRNA forms a central cloud at late stage 8 ( A ) , a collimated channel while moving to the posterior ( B ) , and a posterior crescent at stage 9 ( C ) . Scale bars are 25 μm . ( D ) 3D oocyte showing the initial oskar mRNA cargo distribution . ( E , F ) Cross sections through a simulation of oskar mRNA transport with diffusion , motor-transport and flows showing the distribution of total cargo cb + cu at the indicated time points . No posterior anchor is present . Simulations of three 6 hr cycles through 50 ( vm , u ) -pairs in random order once ( twice , see Figure 4 ) . For simulations of 1 . 5 hr , 25 ( vm , u ) -pairs were chosen at random . Compare to experimental observations in panels B and C . ( G ) Same simulation as in D–F , but without cytoplasmic flows , showing largely identical localisation as in F . ( H ) Simulation of bicoid mRNA transport shows the mRNA quickly accumulating at the nearest cortex ( arrowheads ) when injected in the posterior ( inset ) , corresponding to the behavior of naive bicoid . ( I ) Same as in H , corresponding to naive bicoid mRNA injection at the anterior-dorsal region ( inset ) . ( J ) Same as in H for injection at the anterior middle ( inset ) , showing that over long times simulated bicoid mRNA localises to the anterior corners ( arrowheads ) . Localization to the anterior depends on sufficient proximity of the injection site as observed for injections of naive bicoid mRNA . ( K ) Same simulation for oskar mRNA as in D–F , but with a posterior anchor ( arrowhead ) and without active motor-driven transport . Compare to panels F and G . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 007 Starting from an initial diffuse cloud of oskar mRNA in the centre of the oocyte ( Figure 3A , D , E ) , simulations show that the mRNA quickly concentrates in the centre before forming a channel from the centre of the oocyte towards the posterior pole ( Figure 3E ) . Formation of this channel reflects the locally inwards orientation of MTs in the posterior half of the oocyte ( Figure 1E , F ) . This transient state closely resembles transient patterns of fluorescently-tagged oskar mRNA during the transition between stages 8 and 9 ( Figure 3B ) , thereby supporting the notion that our computed MT cytoskeleton correctly captures key aspects of the in vivo MT network . Upon reaching the posterior cortex , oskar mRNA is translated to produce Long Osk ( Vanzo and Ephrussi , 2002 ) , which is involved in anchoring oskar mRNA . However , oskar mRNA localises normally at stage 9 when anchoring is disrupted ( Micklem et al . , 2000; Vanzo and Ephrussi , 2002 ) . Anchoring is therefore not necessary at stage 9 , and we did not include it in our model at this point . Despite the lack of any anchoring , oskar mRNA forms a posterior crescent after 1 . 5 hr and becomes highly concentrated by the end of the simulation ( Figure 3F ) , correctly capturing observations of in vivo localisation ( Figure 3C ) . A 4 μm thick slice at the posterior pole contains 12 . 5% of total cargo in the oocyte ( Figure 3F ) , showing that continual transport combined with slow diffusion is sufficient to reach and maintain high concentrations of mRNA . In our description of transport ( Equation 2 ) , cargo acts as an unspecific passive tracer . Passive tracers merely follow the transport fields vm and u , and these two fields must contain all information about the localisation sites . We therefore asked which field contains most information about localisation . We first tested localisation in the absence of cytoplasmic flows by setting the fluid flow field identical to zero u ≡ 0 . This situation is similar to but more severe than in slow Kinesin mutants ( Serbus et al . , 2005 ) . Starting from the central oskar mRNA cloud ( Figure 3D ) , we found that oskar mRNA still localises to and forms a concentrated crescent at the posterior pole ( Figure 3G ) . A 4 μm thick slice off the posterior pole contains 11 . 3% of total available oskar mRNA cargo , only marginally less than localisation with cytoplasmic flows present . This suggests that the cytoskeletal transport alone localises the majority of mRNA . Cargo localisation in the absence of cytoskeletal transport can be tested by either setting the motor-velocity field identical to zero vm ≡ 0 , or by setting the binding constant kb to zero . In either case , cargo disperses throughout the oocyte without forming a central channel and completely fails to produce any posterior crescent . We further asked whether the flow could localise cargo to wild-type levels if an anchor captured cargo at the posterior . To this end , we replaced the bound state in the model by an anchored state to which cargo can bind only at the very posterior boundary . Binding to the anchor occurs with a reaction rate constant that is 10 times larger than the reaction rate constant kb used for oskar mRNA binding to the MT cytoskeleton ( ‘Materials and methods’ Parameter values ) . Moreover , anchored cargo cannot be released , thus making this anchor a perfect sink with stronger trapping properties than any realistic anchor . Under these idealised conditions , oskar mRNA accumulates slightly at the anchor ( Figure 3K ) . However , a 4 μm thick posterior slice contains only 2 . 7% of total cargo in the oocyte , about 78% less cargo than in the wild-type simulations , and only 25% more cargo compared to the purely diffusive case , which localises 2 . 2% of cargo . Thus , even under the most favourable conditions , the cytoplasmic flows cannot localise mRNA to wild-type levels , arguing against a mixing and entrapment mechanism for oskar mRNA localisation at stage 9 . Despite the small fraction of actively transported cargo ( 13% ) , motor-driven mRNA transport is both necessary and sufficient to account for oskar localisation . In contrast to oskar mRNA , bicoid mRNA is believed to be transported by Dynein ( Duncan and Warrior , 2002; Januschke et al . , 2002; Weil et al . , 2006 ) , but other parameters such as the fraction of bound cargo and the active transport speeds are similar to oskar mRNA ( Belaya , 2008 ) ( V Trovisco , personal communication ) . To test whether the proposed cytoskeletal organisation also captures transport and localisation of injected bicoid mRNA , we inverted the directions of the motor-velocity fields ( Figure 1C ) to account for the minus end directed transport by Dynein instead of the plus end directed transport by Kinesin . Other parameters including cytoplasmic flow fields remained unchanged . Simulations show that bicoid mRNA accumulates at both posterior-lateral sides ( Figure 3H ) when initially placed in the posterior half of the oocyte ( Figure 3H , inset ) . This is in very good agreement with experimental observations when naive bicoid mRNA is injected into this posterior region ( Cha et al . , 2001 ) . When bicoid mRNA is placed initially in the anterior-ventral region ( Figure 3I , inset ) , it accumulates at the anterior and lateral cortex ( Figure 3I ) , again in concordance with the experimental observations ( Cha et al . , 2001 ) . This simulated localisation of bicoid mRNA remains virtually identical in the absence of cytoplasmic streaming . Thus , splitting of a bulk amount of injected bicoid mRNA occurs when the RNA is placed on the border ( separatrices ) between two diverging subcompartments of the MT cytoskeleton , each one transporting part of the cloud towards the adjacent cortex . The agreement between simulations and experiments therefore further supports an on average compartmentalised MT cytoskeleton , and naive bicoid mRNA , like oskar mRNA , unspecifically traces out the MT cytoskeleton . In contrast to naive bicoid mRNA , endogenous bicoid mRNA is transported into the oocyte after being transcribed in neighbouring nurse cells where it also becomes modified in presence of the protein Exuperantia . Endogenous bicoid mRNA can be approximated by so-called conditioned bicoid mRNA that is first injected into the nurse cells , then sucked out and injected into the oocyte . Conditioned bicoid mRNA localises specifically to the anterior surface within 30 min , even if this surface is not closest to the injection site ( Cha et al . , 2001 ) . In the model , starting from an initial distribution at the anterior surface where endogenous bicoid mRNA enters the oocyte , cargo quickly moves to the anterior cortex before concentrating in the anterior corners after several hours of simulated time ( Figure 3J ) . This resembles the anterior ring-like localisation of bicoid mRNA in wild-type stage 9 oocytes ( St Johnston et al . , 1989; Weil et al . , 2006 ) , and occurs independently of cytoplasmic flows . However , the model does not reproduce transport specifically to the anterior surface when injection is further away from the anterior . This suggests that Exuperantia activity somehow either enables bicoid mRNA to move along an unidentified population of MTs ( Cha et al . , 2001; MacDougall et al . , 2003 ) that are not included in the computed cytoskeleton , or that it allows localisation of the mRNA by an unknown MT-independent mechanism . We have shown so far that cortical gradients of MT nucleation are sufficient to assemble a functional , compartmentalised MT cytoskeleton that successfully localises oskar mRNA and naive bicoid mRNA . Next , we asked if cortical MT nucleation can also produce cytoskeletal organisations that explain mutants with partial or complete polarity defects . Mutations interfering with the posterior follicle cells that surround the oocyte , for example , can disrupt the proper positioning of mRNAs ( Gonzalez-Reyes et al . , 1995; Roth et al . , 1995 ) . Specifically , follicle cell clones ( rasΔC40b ) that are adjacent to only one side of the oocyte posterior repel cargo from that side of the cortex ( termed clone adjacent mislocalisation [Poulton and Deng , 2006] ) , likely due to an altered MT organisation . To mimic this phenotype , we used an ensemble of wild-type cytoskeletons and artificially added MT nucleation sites to a patch on one side of the posterior pole . Simulations of cargo transport with diffusion , motor transport and flows show that oskar mRNA is repelled from this patch and localises to the adjacent posterior boundary , in agreement with experiments ( Figure 4—figure supplement 2 ) . Mislocalisation of mRNAs also occurs in mutants of the polarity protein PAR-1 , which acts to suppress MT nucleation at the posterior pole of the oocyte ( Shulman et al . , 2000; Doerflinger et al . , 2010 ) . In strong par-1 hypomorphs with strongly reduced PAR-1 expression , MT minus ends occupy the whole cortex including the posterior pole ( Parton et al . , 2011 ) , thereby causing oskar mRNA to mislocalise to a dot in the centre of almost 90% of mutant oocytes ( Doerflinger et al . , 2006 ) . To test this phenotype in the model we distributed MT seeding points uniformly on the posterior surface ( Figure 4G , inset ) while keeping the seeding density on the anterior surface constant . The AP gradient of MT density then vanishes and the directional bias of MT segments evens out ( Figure 4G , ensemble-averaged 3D-bias: 50 . 1%:49 . 9% ) . The ensemble-averaged local orientations of MTs show that MTs point towards a single focal point in the centre of the oocyte ( Figure 4G ) that acts as a stable fixed point of the system . Computation of 3D flow fields for each realisation in the new ensemble ( 3D mean: 13 . 5 nm/s; 2D mean: 14 . 5 nm/s , N = 50 ) and simulations of transport with and without cytoplasmic flows show oskar mRNA concentrating in a cloud in the centre of the oocyte ( Figure 4C ) . oskar mRNA accumulation to a central dot also remains unchanged if the posterior MT seeding density is decreased slightly ( Figure 4H , inset ) as might be expected for hypomorphs that do not abolish PAR-1 expression completely . 10 . 7554/eLife . 06088 . 008Figure 4 . The cortical MT seeding density determines sites of mRNA localisations and gives rise to a bifurcation in the cytoskeleton . ( A–E ) Simulations of oskar ( A–C ) and bicoid ( D ) mRNA transport with diffusion , cytoskeletal transport and flow for the cytoskeletal architectures shown in panels E–H , and their corresponding flow fields . For oskar mRNA , simulations reproduce wild-type localisation ( A , same as Figure 3F ) , and partial ( B ) or complete mislocalisation ( C ) . Initial condition as in Figure 3D . Simulation times were occasionally increased to 6 hr ( B ) to rule out transient concentration patterns . For bicoid mRNA , simulations capture mislocalisation to both anterior and posterior as in gurken/torpedo/cornichon mutants and in strong par-1 hypomorphs ( D , arrowheads ) . Time points as indicated . The inset in D shows initial condition . ( E–H ) Average local MT orientations in ensembles of 50 realizations of the polymer model for varying MT seeding densities along arclength s ( see panel E ) of the posterior-lateral cortex ( insets ) . A MT seeding density that increases from a wild-type gradient ( E , h0P=0 , kP=150 μm ) laterally towards the posterior ( F , h0P=0 , kP=17 . 5 μm ) to a near-uniform distribution ( G , h0P=0 , kP=1 μm ) shows a saddle-node bifurcation by creating a pair of stable ( green ) and unstable ( red ) fixed points . ( H ) A MT seeding density that is slightly lower at the posterior pole ( h0P=0 . 7 , kP=40 μm ) produces mean MT orientations virtually indistinguishable from uniform seeding density ( compare to G ) . ( I–K ) Vector field of the mathematical normal form of the 2D saddle-node bifurcation ( ‘Materials and methods’ Bifurcation normal form ) . Values of the bifurcation parameter λ as indicated . Fixed points are located at positions x=±λ . ( L ) Fluorescence in-situ hybridization to bicoid mRNA in a strong par-1 hypomorph . The mRNA ( red ) localises around the cortex , with most accumulation at the anterior corners and at the posterior pole ( compare to panel D , arrowheads ) . Scale bar is 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 00810 . 7554/eLife . 06088 . 009Figure 4—figure supplement 1 . Either the MT seeding density or the MT length can act as bifurcation parameter . ( A–J ) Each panel shows the mean local MT orientations for an ensemble of 50 realizations of the polymer MT cytoskeleton under variations of the seeding density parameter kP and MT length parameter ϵ as indicated . Approximate location of the stable fixed point is shown in green , with region of attraction in blue , and domain of attraction of the posterior pole in red . The point on the AP-axis between red and blue arrow region marks the unstable fixed point . Percentages indicate the ensemble-averaged 3D directional bias . Experimentally , the directional bias was measured as 57 . 97%:42 . 03% ( Parton et al . , 2011 ) . Cytoskeleton in panel I was used as wild-type cytoskeleton ( Figure 1F , main text ) . Anterior seeding density is unchanged in all panels kA = 1000 μm , h0A=0 . 8 . ( K–O ) Same as panels A–J but for ensembles with 100 realizations of the cytoskeleton to ensure reliable visualization of vector field even when fewer MTs reaching the oocyte center . ( P–T ) Vector fields computed from a saddle-node bifurcation normal form . The critical bifurcation point is λ = 0 , and for λ < 0 , fixed points are located at x±=±−λ . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 00910 . 7554/eLife . 06088 . 010Figure 4—figure supplement 2 . Lateral MT growth produces the clone-adjacent-mislocalisation phenotype ( see main text ) . ( A–D ) Addition of lateral MTs reproduces the clone-adjacent-mislocalisation phenotype . ( A ) Cross section through one realization of the wild-type MT cytoskeleton in the polymer model ( h0A=0 . 8 , kA=1000 μm , h0P=0 , kP=150 μm ) . Additional MTs have been added dorsal to the posterior pole ( black arrow ) to mimick posterior follicle cell clones ( RASΔC40b MARCM ) that overexpress dystroglycan ( Poulton and Deng , 2006 ) . ( B ) Density of MTs for 50 realizations of the cytoskeleton shows enrichment on one side of the posterior pole ( white arrow ) . ( C ) Streamplot showing the local vectorial orientation averaged over 50 realizations of the cytoskeleton . Note the upwards tilt away from the central posterior pole . ( D ) Simulations of oskar mRNA transport by diffusion , cytoskeletal transport and corresponding cytoplasmic flows show that cargo is repelled from the site of additional MT nucleation on one side of the posterior pole , thereby capturing the CAM phenotype ( Poulton and Deng , 2006 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 010 Interestingly , bicoid mRNA simulated either with slightly decreased ( Figure 4H ) or with uniform ( Figure 4G ) posterior MT nucleation not only localises to the anterior corners as in wild-type but also enriches at the posterior pole ( Figure 4D , arrowheads ) , despite injection close to the anterior ( Figure 4D , inset ) . This matches experiments showing bicoid mRNA mislocalisation to both anterior and posterior in gurken , torpedo and cornichon mutants ( EGFR mutants ) in which posterior follicle cells fail to signal a MT reorganisation in the oocyte ( Gonzalez-Reyes et al . , 1995; Roth et al . , 1995 ) . Failure of the external signal to the oocyte is thought to have the same effect as par-1 hypomorphs , but previous experiments with bicoid mRNA in par-1 hypomorphs resulted in ambiguous localisations ( Shulman et al . , 2000; Benton et al . , 2002 ) . Here , using confocal fluorescence in situ hybridisation , our experiments show that bicoid mRNA localises primarily to the anterior corners and to the posterior pole in strong par-1 hypomorphs ( Figure 4L ) , thereby mirroring EGFR mutants and agreeing with the model predictions . Thus , ( near- ) uniform MT nucleation along the posterior and lateral cortex is sufficient to explain the polarity phenotypes observed for oskar and bicoid mRNAs in gurken , torpedo and cornichon mutants and in strong par-1 hypomorphs . In 10% of strong par-1 hypomorphs and 30% of weak par-1 hypomorphs with weakly reduced PAR-1 expression , oskar mRNA is only partially mislocalised , combining a wild-type-like posterior crescent with a centrally mislocalised dot ( Doerflinger et al . , 2006 ) . We therefore modulated the MT nucleation along the posterior-lateral cortex to test if this is sufficient to produce intermediate cytoskeletal organisations between wild-type ( Figure 4F ) and strong par-1 hypomorphs ( Figure 4H ) . Increasing MT nucleation laterally towards the posterior pole ( Figure 4G , inset ) again creates a stable focus in the centre of the oocyte . However , its domain of attraction does not cover the entire oocyte . Instead , a small basin of attraction towards the posterior pole persists ( Figure 4G , red ) , separated from the basin of attraction of the stable focus by an unstable saddle-node point . Further increasing the seeding density towards the posterior pole shows that the pair of stable and unstable fixed points move further apart ( Figure 4—figure supplement 1 ) until the unstable point no longer resides inside the oocyte geometry , thereby giving rise to the strong par-1 hypomorph topology ( Figure 4H , I , J ) . Cargo simulated on cytoskeletons with stable and unstable fixed points and their corresponding flow fields can split into two separate accumulations , first at the posterior pole and second in a dot along the AP-axis ( Figure 4B ) . This agreement with experimental results suggests that intermediate levels of PAR-1 at the posterior lead to a cytoskeleton with a potential barrier between the oocyte centre and its posterior pole . It also shows that modulation of MT nucleation along the posterior-lateral cortex alone is sufficient to capture this . Because the potential barrier only affects the bound state , diffusion and flows can aid posterior localisation in this context by pushing unbound cargo across the unstable point and increasing the amount of oskar mRNA that reaches the posterior pole . In summary , variations of MT nucleation along the cortex not only explain mRNA localisations in wild-type but can also account for the mislocalisations of bicoid and oskar mRNAs in polarity mutants . The creation of stable and unstable fixed points in the transition from wild-type to strong par-1 hypomorph topology constitutes a classical saddle-node bifurcation ( Figure 4K–M , ‘Materials and methods’ Bifurcation normal form ) with the cortical MT seeding density as the bifurcation parameter . Interestingly , in both the polymer model and the rod model , the mean length of MTs acts as a second bifurcation parameter . For example , for sufficiently short MT filaments ( Figure 4K ) , few MTs from the anterior can reach and contribute to orientations in the posterior half of the oocyte . In the posterior half , MTs from the lateral side either point towards the anterior , thereby combining with anterior MTs to create a stable node , or towards the posterior , to form a domain of attraction at the posterior pole . Therefore , for sufficiently weak contributions from anterior MTs , MTs from the lateral sides create the unstable tipping-point . To understand more generally how the MT lengths and the posterior-lateral distribution of seeding points together influence the three different cytoskeletal topologies ( Figure 4E–G ) we computed the parameter space of the straight rod model ( Figure 5 ) . If MTs are absent at the posterior pole , the wild-type topology ( Figure 5Di , vi ) covers a large fraction of parameter space ( Figure 5A , red ) for sufficiently long MTs ( large ϵ ) and cortical MT nucleation gradient ( large kP ) . The creation of two fixed points splits the oocyte into different domains of attraction ( Figure 5A , blue , Figure 5Di , ii ) for either shorter MTs ( Figure 5A , vertical dashed arrow; Figure 4—figure supplement 1I , N ) or for increasingly uniform MT nucleation ( Figure 5A , horizontal dashed arrow; Figure 4—figure supplement 1I , H ) . For almost uniform nucleation along the entire posterior-lateral cortex , the unstable fixed point exits the geometry , leaving behind only a single stable focus ( Figure 5A , green , arrowhead , Figure 5Diii , iv ) . If some MT nucleation is allowed at the posterior pole this region of strong par-1 hypomorph topology expands substantially ( Figure 5B , C , green ) . Interestingly , if MTs are sufficiently long , MTs from the lateral and anterior cortex can reach and overpower those from the posterior pole , thus restoring wild-type-like cytoskeletal topology that allows oskar mRNA localisation . Therefore , a low level of posterior nucleation of MTs does not necessarily lead to mislocalisation of oskar mRNA . Instead , and in addition to the distribution of cortical MT nucleation , the length of MT filaments emerges as another key regulator for polarisation and function of the non-centrosomal network . 10 . 7554/eLife . 06088 . 011Figure 5 . The parameter space of the rod model shows the relation and interconversion between all three distinct cytoskeletal architectures . ( A ) The regions corresponding to wild-type ( red ) , weak par-1 hypomorph ( blue ) and strong par-1 hypomorph topologies ( green , arrowhead at far left ) are shown as a function of the mean MT length ϵ and extent of the posterior seeding density kP . The bifurcation line between wild-type and weak par-1 hypomorph topologies can be crossed by either changing the seeding density laterally ( horizontal dashed arrow ) , or by shortening the MTs ( vertical dashed arrow ) . ( B , C ) Parameter spaces as in A for increasing MT nucleation at the posterior pole ( B: h0P=0 . 03 , C:h0P=0 . 1 ) . Inset in panel C shows a magnification of the triple point at which small changes in parameters can convert each cytoskeletal architecture into any other . ( D ) Local net orientations of MT rods for parameter values indicated in the inset of panel C ( yellow circles ) . In all panels , parameter values for anterior MT nucleation were kept constant ( h0A=0 . 8 , kA=1000 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 011 Non-centrosomal MT networks represent a large , yet poorly understood class of MT arrangements that often fulfil specialised functions ( Keating and Borisy , 1999 ) . Non-centrosomal MTs are frequently aligned in parallel , thereby forming linear arrays as in epithelia or neurons ( Bartolini and Gundersen , 2006; Conde and Caceres , 2009; Kapitein and Hoogenraad , 2011 ) . The non-centrosomal MT cytoskeleton in Drosophila oocytes was also first hypothesised to form a highly polarised linear array with MTs growing from the anterior surface towards the posterior pole ( Clark et al . , 1994; Brendza et al . , 2000 ) . Instead , the MT cytoskeleton is a complex network without a visibly discernible polarity along the AP axis ( Theurkauf et al . , 1992; Cha et al . , 2001 , 2002 ) , and its organisation remained ambiguous ( MacDougall et al . , 2003; Januschke et al . , 2006; Zimyanin et al . , 2008; Parton et al . , 2011 ) . We showed that in two different theoretical models , in which MTs grow from the oocyte cortex into the volume , the cytoskeleton features an on average compartmental organisation and is therefore more ordered than previously thought . In combination with cytoplasmic flows and diffusion , this cytoskeletal organisation successfully reproduces localisations of oskar and naive bicoid mRNAs in wild-type oocytes . Such a spatially varying , compartmental organisation suggests that a single statistical measure of polarity , derived by averaging data from the entire oocyte ( Zimyanin et al . , 2008; Parton et al . , 2011 ) , may not be a sufficient metric to characterise mRNA localisation . Cytoplasmic flows may generally contribute to mRNA transport . Flows at later stages of oogenesis are fast and well-ordered , and occur concurrently with a late-phase enhancement of oskar mRNA accumulation ( Sinsimer et al . , 2011 ) . This is consistent with a contribution of flows to mRNA localisation ( Glotzer et al . , 1997; Forrest and Gavis , 2003 ) . At stage 9 of oocyte development , which is the focus of the work presented here , flows are slow and chaotic . At this stage , mutants with reduced flows showed oskar mRNA localising normally ( Serbus et al . , 2005 ) , thus suggesting that slow and chaotic flows may not play the same role as in late oogenesis . However , flows in these mutants were merely reduced rather than abolished completely , and the measurements underestimated flow speeds ( Serbus et al . , 2005 ) , hence leaving the interpretation of flows unclear . We here find that the effects of slow cytoplasmic flows on mRNA transport are negligible , and that cytoskeletal transport alone is sufficient for localisations of oskar and naive bicoid mRNAs . In this view , slow cytoplasmic flows arise primarily as inevitable physical byproduct of active motor-driven transport on the cytoskeleton rather than as an evolutionarily selected trait . This appears to mirror findings in the Caenorhabditis elegans zygote in which P-granules segregate by dissolution and condensation rather than via transport by cytoplasmic flows ( Brangwynne et al . , 2009 ) . Interestingly , neither MT-based transport alone nor combined with cytoplasmic flows and diffusion is sufficient to reproduce the anterior localisation of nurse cell conditioned bicoid mRNA irrespective of the position of injection into the oocyte ( Cha et al . , 2001 ) . This Exuperantia-dependent mechanism may rely either on an unobserved population of MTs in the oocyte that can be specifically recognised by conditioned bicoid mRNA ( Cha et al . , 2001; MacDougall et al . , 2003 ) , or on an unknown MT-independent mechanism . The central finding of our work is that gradients of cortical MT nucleation are sufficient for the assembly of a functional compartmentalised MT cytoskeleton in wild-type oocytes . While many non-centrosomal MT arrays are linear and emphasise questions about establishment and maintenance of parallel filament orientations ( Ehrhardt , 2008; Lindeboom et al . , 2013 ) , our result stresses the need to understand gradients in MT nucleation as an alternative strategy for the assembly of functional non-centrosomal arrays . Whether MTs in Drosophila oocytes are differentially nucleated along the cortex itself or created elsewhere and then differentially anchored at the cortex remains an interesting open question , but both scenarios are compatible with our model . Understanding how this gradient is established will therefore depend on discovering how PAR-1 regulates MT interactions with the cortex . Cortical MT gradients cannot only account for the wild-type cytoskeletal configuration but also for the phenotypes observed in a hierarchy of par-1 hypomorphs . Modulation of MT gradients along the posterior-lateral cortex alone are sufficient to explain the splitting of oskar mRNA between the centre and the posterior pole of the oocyte ( Doerflinger et al . , 2006 ) via a saddle-node bifurcation , suggesting that the anterior and posterior-lateral surfaces of the oocyte are functionally decoupled . Generally , bifurcations in temporal behaviour govern important qualitative transitions in many biological systems , such as the lactose network in Escherichia coli ( Ozbudak et al . , 2004 ) , cell cycle in yeast ( Charvin et al . , 2010 ) , and collapses of bacterial populations ( Dai et al . , 2012 ) . In Drosophila , one important qualitative change is the temporal transition between stages 7/8 and 9 . During this transition the MT cytoskeleton reorganises from uniform nucleation around the cortex and oskar mRNA in the centre to the AP MT nucleation gradient with oskar mRNA at the posterior . It is therefore tempting to speculate that the sequence of PAR-1 mutants and the underlying bifurcation represent static snapshots of this dynamic developmental transition in wild-type . In conclusion , the present work provides a model that describes the assembly , organisation and function of the non-centrosomal MT array in Drosophila oocytes and directs future attention to the molecular mechanisms that enable differential MT nucleation or anchoring at the cortex . For calculation of the MT cytoskeleton , the AP axis is aligned with the z-axis of a cartesian coordinate system , and results are later shifted and rotated to align with the x-axis for visualization , for subsequent computations of cytoplasmic flows and simulations of cargo transport . Dimensional coordinates cover the ranges x ∈ [−L , L] , y ∈ [−L , L] and z ∈ [0 , z0 L] with length scale L = 50 μm . After nondimensionalization with scale L , coordinates are ranged as x′ ∈ [−1 , 1] , y′ ∈ [−1 , 1] and z′ ∈ [0 , z0] . Similarly , nondimensionalization of the shape parameter kP in the MT seeding density leads to kP′=kP/L , and all primes will be dropped subsequently . Methods figures are shown in nondimensional spatial coordinates , and reported parameter values are nondimensional unless noted otherwise . The 3D geometry of a typical stage 9 Drosophila oocyte is defined as two parabolic , rotationally-symmetric caps ( Figure 1 ) . Anterior ( i = A ) and posterior ( i = P ) parabolic caps are parameterized in cylindrical coordinates ρ=x2+y2∈[0 , 1] and ϕ ∈ [0 , 2π ) as ( 3 ) σi ( ρ , ϕ ) =ρ e^ρ+z0i ( 1−ρ2 ) e^z , where e^ρ= ( cos ( ϕ ) , sin ( ϕ ) , 0 ) and e^z= ( 0 , 0 , 1 ) are the unit vectors in radial and z-direction , respectively . The line element along the parabola is calculated asdsi=‖∂σi∂ρ‖dρ=1+ ( 2z0i ρ ) 2dρ , with arclength ( 4 ) si ( ρ ) =ρ21+ ( 2z0i ρ ) 2+14z0isinh−1 ( 2z0i ρ ) , defined such that si ( ρ = 0 ) = 0 denotes the tip of the parabolic cap while si ( ρ=1 ) =s0i denotes the distance from the tip to the anterior corners ( Figure 6A , G ) , and hence 0≤si ( ρ ) ≤s0i . The surface element for parabolic caps ( in units of L ) is given by ( 5 ) dΣi=‖∂σi∂ρ×∂σi∂ϕ‖=ρ 1+ ( 2z0iρ ) 2dρ dϕ , with total surface area ( 6 ) Σi=π6 ( z0i ) 2 ( [1+ ( 2z0i ) 2]3/2−1 ) . 10 . 7554/eLife . 06088 . 012Figure 6 . Seeding point density in the polymer model . Shown are N randomly drawn seeding points on the posterior cap of the standard oocyte geometry , according to Equation 15 distributed either ( near- ) uniformly on the cap ( A , G ) or in a parabolic gradient from the posterior pole to the anterior corners ( F , L ) . Between uniform and parabolic distribution , the seeding density can vary either by laterally reducing the density ( B , C–E , parameter kP ) , or by reducing the density at the posterior pole ( H , I–K , parameter h0P ) . Total number of points is calculated with a fixed number of anterior points NA = 4000 ( Equation 20 ) . For actual computations of wild-type cytoskeletons ( kP=3 , h0P=0 , kA=20 , h0A=0 . 8 , see main text for dimensional parameters ) , more than 55 , 000 seeding points are used . DOI: http://dx . doi . org/10 . 7554/eLife . 06088 . 012 The inward pointing normals for posterior and anterior caps are given by ( 7 ) n^P ( ρ , ϕ ) =−2z0P ρ e^ρ+e^z1+ ( 2z0Pρ ) 2 , n^A ( ρ , ϕ ) =2z0A ρ e^ρ+e^z1+ ( 2z0Aρ ) 2 . A special case arises for z0A=0 when the anterior parabolic cap becomes a flat disc . In this case , the arclength ( Equation 4 ) reduces to sA ( ρ ) = ρ , and the surface area ( Equation 6 ) simplifies to ΣA = π . The inwards pointing normals for this case are given by n^P ( Equation 7 ) for the posterior cap and n^A=e^z for the anterior disc , respectively . To investigate the robustness with respect to changes in oocyte shape , we test our model for the cytoskeleton , cytoplasmic flows and mRNA transport in two different geometries for a stage 9 Drosophila oocyte . Geometry-1 is used as standard geometry , and is comprised of two parabolic caps ( Equation 3 ) with z0P=1 . 48 for the posterior cap and z0A=0 . 2 for the anterior cap ( see Figure 1 ) . This results in a length of the AP axis of z0P−z0A=1 . 28 with an aspect ratio of 1 . 56 that qualitatively resembles a typical stage 9 oocyte . Geometry-2 is tested as alternative geometry , and consists of a posterior parabolic cap with z0P=1 , and an anterior flat disc with z0A=0 ( Figure 1—figure supplement 1 ) , giving an AP-axis length of 1 and aspect ratio of 2 . We find that results are robust with respect to this change in geometry .	Cells contain a network of filaments known as microtubules that serve as tracks along which proteins and other materials can be moved from one location to another . For example , molecules called messenger ribonucleic acids ( or mRNAs for short ) are made in the nucleus and are then moved to various locations around the cell . Each mRNA molecule encodes the instructions needed to make a particular protein and the network of microtubules allows these molecules to be directed to wherever these proteins are needed . In female fruit flies , an mRNA called bicoid is moved to one end ( called the anterior end ) of a developing egg cell , while another mRNA called oskar is moved to the opposite ( posterior ) end . These mRNAs determine which ends of the cell will give rise to the head and the abdomen if the egg is fertilized . The microtubules start to form at sites near the inner face of the membrane that surrounds the cell , known as the cortex . From there , the microtubules grow towards the interior of the egg cell . However , it is not clear how this allows bicoid , oskar and other mRNAs to be moved to the correct locations . Khuc Trong et al . used a combination of computational and experimental techniques to develop a model of how microtubules form in the egg cells of fruit flies . The model produces a very similar arrangement of microtubules as observed in living cells and can reproduce the patterns of bicoid and oskar RNA movements . This study suggests that microtubules are more highly organised than previously thought . Furthermore , Khuc Trong et al . 's findings indicate that the anchoring of microtubules in the cortex is sufficient to direct bicoid and oskar RNAs to the opposite ends of the cell . The next challenge will be to find out how the microtubules are linked to the cortex and how this is regulated .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "physics", "of", "living", "systems" ]	2015	Cortical microtubule nucleation can organise the cytoskeleton of Drosophila oocytes to define the anteroposterior axis
Despite continual renewal and damages , a multicellular organism is able to maintain its complex morphology . How is this stability compatible with the complexity and diversity of living forms ? Looking for answers at protein level may be limiting as diverging protein sequences can result in similar morphologies . Inspired by the progressive role of apical-basal and planar cell polarity in development , we propose that stability , complexity , and diversity are emergent properties in populations of proliferating polarized cells . We support our hypothesis by a theoretical approach , developed to effectively capture both types of polar cell adhesions . When applied to specific cases of development – gastrulation and the origins of folds and tubes – our theoretical tool suggests experimentally testable predictions pointing to the strength of polar adhesion , restricted directions of cell polarities , and the rate of cell proliferation to be major determinants of morphological diversity and stability . There are three key elements that allow us to bridge the scale from cellular level to macroscopic stable morphologies . One of the implications of the coupling between position and polarity is that in a sheet of cells , turning AB polarity in one cell will cause a force on its neighbors . In case of two cells ( Figure 2—figure supplement 3 and Figure 2—video 1 ) , the pair relaxes the imposed stress by rotating both the polarity and their positions . In biological terms , turning AB polarity in one cell ( e . g . by apical constriction , illustrated in Figure 2—figure supplement 2 ) of an epithelial sheet will induce bending of the sheet as is the case with bottle cells in invagination . The present formulation of PCP has several implications . First , we restrict the effects of PCP to directed ( anisotropic ) cell–cell adhesion and do not consider its other possible roles , in for example asymmetric cell differentiation , thus primarily focusing on its role on CE . Second , in our current formulation , AB polarity and PCP influence each other’s orientation on equal footing ( Equation 7 ) . PCP , however , is typically constrained to the apical plane and thus is expected not to influence the orientation of AB polarity . Disabling PCP’s effect on AB polarity ( see Materials and methods ) does not influence our main results on tube formation and gastrulation . However , the symmetry in polarities is appealing for its simplicity and is indirectly supported by the following experimental observations: First , cells can acquire PCP without AB polarity present ( Baer et al . , 2009; Zorn and Wells , 2009 ) . Second , proteins required for AB polarity can be planar-polarized ( Warrington et al . , 2013; Aigouy and Le Bivic , 2016; Beati et al . , 2018; Choi and Sokol , 2009; Dollar et al . , 2005; Kaplan and Tolwinski , 2010 ) . Third , changes in cell shapes during invagination ( e . g . sliding of adherens junctions and formation of bottle cells ) are regulated by PCP in neural tube closure ( Ossipova et al . , 2014; Nishimura et al . , 2012; Kinoshita et al . , 2008 ) , gastrulation in C . elegans ( Lee et al . , 2006 ) , sea urchin ( Croce et al . , 2006 ) , and Xenopus ( Choi and Sokol , 2009 ) . These changes in cell shape effectively reorient AB polarity ( Figure 2—figure supplement 2 ) . Adult organismal shapes are stable over long time , maintaining sizes and relative positions of lumens and folds , despite continual local damages and cell renewal . To test if cellular polarity could enable such stability in time and to random local perturbations , we first performed a series of tests with AB polarized cells ( Figure 3 and Figure 3—video 1 ) . When starting a bulk of cells with AB polarities pointing randomly , an initial rapid expansion ( Figure 3A–C ) stabilizes into a complex morphology of interconnected channels ( Figure 3C–E ) . The shape remains unchanged for at least 10 times longer than the initial expansion ( Figure 3C–E ) . The stability of the shape is illustrated by the time evolution of the average energy per cell ( Figure 3F ) that after an initial fast drop converges to a constant value . As expected , this value is higher than the energy of a hollow sphere ( yellow dot in Figure 3F ) – a configuration obtained if we start with radially , instead of random , polarized cells and preserve radial polarization at all times . The observed behavior is not sensitive to the shape of the potential ( Figure 2—figure supplement 1A ) but is sensitive to how the neighborhood is defined ( Figure 2—figure supplement 1B–D ) . Rerunning the simulation in Figure 3 with different initial conditions results in a different stable shape ( Figure 2—figure supplement 1E and Figure 3—figure supplement 1 ) . The macroscopic features of the shapes are robust to noise ( Figure 2—figure supplement 1E–F and Figure 3—figure supplement 1 ) . While the shapes emerging under high and low noise are not identical , the relative position and sizes of the majority of channels and lumens are preserved . The changes caused by noise stem from perturbations during initial expansion stage . If the same level of noise is applied after the system reached stable state , after time t = 10000 , noise does not cause any major macroscopic changes ( Figure 2—figure supplement 1G ) . The obtained shapes have self-sealing features , as an initial cut and unwrapping of a section of a surface refolds and seals back into the original morphology ( Figure 3—figure supplement 2A–C ) . Furthermore , the shapes ( Figure 3D–E ) are also robust to overall growth ( Figure 3—figure supplement 2D–F ) retaining the same macroscopic features , just scaled to a larger size . Robustness to noise and cell proliferation further support the link between polarity and stability of morphologies , for example organ shapes , as they expand from infant to adult . The orientation of polarities in a subpopulation of cells may be set by the environment that the cells are embedded in , for example signaling molecules deposited into extracellular matrix can influence orientation of the AB polarity ( Overeem et al . , 2015 ) or signals from neighboring cells of a different type can orient PCP ( Chu and Sokol , 2016 ) . We will refer to these constraints as boundary conditions . To investigate sensitivity to boundary conditions , we consider three cases where polarities are fixed at all times and point either radially out from the center of mass ( Figure 4A ) , radially out from a central axis ( Figure 4B ) , or pointing away from a central plane ( Figure 4C ) . As anticipated , the difference in symmetries of boundary conditions results in a sphere , a cylinder , or two parallel planes ( Figure 4—video 1 ) . At the same time , in these symmetric cases , the differences in initial conditions but without imposed boundary conditions are not sufficient to generate different structures; they all converge to the nested ‘Russian doll’-like hollow spheres ( Figure 4D ) . In development , this highlights the importance of the neighboring tissues for defining boundary conditions . Our results thus support the idea that polar adhesion enables stable and robust macroscopic shapes . The closest biological parallels would be the complex luminal morphologies emerging in reaggregation experiments on for example Hydra ( Seybold et al . , 2016 ) or in vitro culture of purjunkie brain cells ( Muguruma et al . , 2015 ) . Together with our simulations , these experiments highlight how stable and complex morphologies can develop in non-proliferating populations from cell rearrangements alone . Transitions from spheres to folded shapes are ubiquitous in development . Folds are an important part of in vivo organ development , and the composition of cell types in the folded organoids is closer to that in real organs ( Greggio et al . , 2013 ) . To date , it is unclear what drives the transition from spheres to folded lumens . One possibility is that it is driven by the mechanical properties of the matrigel that effectively may place the growing organoid under pressure . Alternatively , data from 3D brain organoids suggests that the rapid cell proliferation leads to the emergence of surface folding ( Li et al . , 2017 ) . The simplicity of our tool allows to explore both of these scenarios . To model dividing cells , we pick a random cell from the entire population and introduce a new daughter cell with inherited polarity direction placed in a random location a half cell radius away from the mother cell . This event introduces dynamic perturbation by locally increasing cell density and requires some time to relax back to equilibrium . If proliferation is slow , and the time between two cell divisions anywhere in the system is longer than relaxation time of the whole system ( the time it takes to reach equilibrium ) , the system approaches global equilibrium and will expand as a sphere . However , if proliferation is increased , the system will be pushed out of equilibrium and folds will emerge ( Figure 5A , Materials and methods ) . In more quantitative terms , our forces are such that a single cell can move up to 0 . 2 cell diameter per time unit . For cells that divide every 1000 time units , the transition to non-equilibrium buckling happens when the system has grown to about 5000 cells ( Figure 5—video 1 ) . As cells divide faster , our simulations predict a transition from a smooth spherical shell to an increasingly folded structure with multiple pronounced folds , in line with the observation of brain organoids proliferating at different rates ( Li et al . , 2017 ) . In comparison with the model for cortical convolutions by Tallinen et al . , 2016 in which folding is a result of expanding cortical sheet adhered to the non-expanding white matter core , our mechanism does not require a bulk core . Instead folds emerge in a fast-expanding sheet when the growth is faster than the global relaxation to dynamical equilibrium . While we find that the external pressure is not necessary for folding , pressure alone can also drive folding ( Figure 5B , Materials and methods ) . However , this scenario contradicts the observation that pancreatic organoids can grow as spheres or folded morphologies in gels with the same stiffness but different media composition ( Greggio et al . , 2013 ) . In principle , both scenarios may contribute to folding , but visually the fold morphologies are different . To differentiate between the two , we have quantified the final folded structures in terms of their local minima ( Figure 5 , Materials and methods , see also Figure 5—video 1 ) . Our simulations predict that in the pressure-driven case , the number of local minima will reach an upper limit as organoids increase in size ( Figure 5B ) . In the case of out-of-equilibrium proliferation , new folds can continue forming as organoids grow ( Figure 5A ) . Increased proliferation causes more and shallower folds . These folds are different than obtained with pressure which causes fewer but deeper minima . Quantitatively , both the depth and the horizontal extension of the folds are about double as large with pressure than with growth-induced folding ( Figure 5—figure supplement 1 ) . Despite the numerous evidences supporting the role of PCP in tubulogenesis , it remains unresolved whether oriented cell division or the extent of CE controls tube length and width ( Karner et al . , 2009; Carroll and Yu , 2012 ) . It is also debated if it is important for the tubes to maintain regular shape , or if it is only important for tube initiation and growth ( Kunimoto et al . , 2017 ) . The simplicity of our approach allows us to address these questions by introducing cell–cell interactions through PCP . This term favors front-rear cell alignment in the interaction potential with only two additional parameters: the strength of the orientational constraint of AB polarity with respect to PCP , λ2 , and the strength of PCP , λ3 ( see Equations 4–8 ) . For simplicity , we focus on the stability ( ability to maintain regular diameter over time ) and tube morphogenesis in systems without cell division . Inducing PCP in a spherical lumen leads to two significant events . First , independent of initial orientation , after some transient time PCPs becomes globally ordered and point in direction parallel to an emerging equator that self-organizes around the sphere ( inset in Figure 6 ) . This arrangement has the lowest energy . Second , cells start intercalating along the axis perpendicular to PCP orientation , gradually elongating the lumen ( Figure 6—video 1 ) . During intercalations , cells exchange their neighbors through T1-like transitions as reported experimentally , Figure 6—figure supplement 3 ( Nishimura et al . , 2012; Sanchez-Corrales et al . , 2018 ) . The intercalations along the axis continue until the force balance between AB polarity and PCP is restored at a new equilibrium . Thus , our model predicts that the strength of PCP ( λ3 ) relative to AB polarity ( λ1 ) determines the width and the length of the tube ( Figure 6 ) . We obtain similar results if we constrain PCP to always remain in the apical plane and thus does not allow PCP to reorient AB polarity ( Figure 6—figure supplement 1 , Materials and methods ) . Note , that this result is very different in nature from the tube presented in Figure 4B as both AB polarity and PCP can now reorient in each cell at any time . These results support the observations that stable tubes can emerge without cell proliferation . In addition , when first the tube is formed , loss of PCP does not lead to cyst formation as recently shown by Kunimoto et al . ( 2017 ) . However , localized cysts could result if the lumen is initialized with varying strength of PCP along the axis ( Figure 6—figure supplement 2 ) . Currently invagination in neurulation and gastrulation is understood and quantitatively modeled as a process driven by changes in cell shapes or the mechanical properties of cells with AB polarity ( Rauzi et al . , 2013; Tamulonis et al . , 2011; Misra et al . , 2016; Hočevar Brezavšček et al . , 2012 ) . This process is often assumed to be driven by apical constriction and decoupled from the eventual tube formation and elongation . However , emerging data suggests that PCP drives both invagination and tube elongation ( Nishimura et al . , 2012; Croce et al . , 2006; Long et al . , 2015 ) likely because apical constriction is controlled by PCP ( Ossipova et al . , 2014; Nishimura et al . , 2012; Croce et al . , 2006 ) and cell intercalations , similar to those in CE , contribute to invagination ( Nishimura et al . , 2012; Rembold et al . , 2006; Sanchez-Corrales et al . , 2018 ) . To probe the limits of our approach , we investigated if AB polarity , PCP , and boundary conditions reminiscent of posterior organizer ( Loh et al . , 2016 ) are sufficient to recapitulate the main stages of sea urchin gastrulation: invagination , tube formation , elongation ( by CE ) , and finally merging of the gastrula tube with the pole opposite to invagination site . The current understanding is that sea urchin gastrulation consists of primary invagination , driven by swelling of the inner layer of extracellular matrix beneath invaginating cells ( Lane et al . , 1993 ) , and formation of a ring of bottle cells due to apical constriction ( Kimberly and Hardin , 1998 ) , and secondary invagination where tube elongates due to CE ( Lyons et al . , 2012 ) . PCP is necessary for both invagination , possibly through its effect on apical constriction in bottle cells ( Nishimura et al . , 2012 ) , and tube extension ( Croce et al . , 2006 ) . Motivated by these observations , we set boundary condition such that PCP of the invaginating cells are oriented around the anterior-posterior ( top-bottom ) axis , and are always in the apical plane . This constrain on PCP orientation allows for CE . While this particular configuration is not documented , it is consistent with observed effects of WNT orienting PCP within the apical plane ( Humphries and Mlodzik , 2018 ) . Second , we simulate the combined effect of bottle cells ( Figure 2—figure supplement 2 ) and bending by swelling of the extracellular matrix by applying an external force , F , on AB polarity ( see Materials and methods ) . This force gradually reorients AB polarity away from the anterior-posterior axis , thus leading to bending of the epithelial sheet . The effect is maximal for cells closest to the anterior-posterior axis . This external force is a phenomenological description aiming at capturing the observed effects of how change in AB polarity results in tissue bending and does not aim at capturing mechanisms driving the reorientation of AB polarity . As a result , cells start to rearrange , the bottom flattens ( Figure 7B ) and bends inward ( Figure 7C ) . Subsequently , the CE-driven by PCP causes the invaginated cells to rearrange , tube elongates , and merges with the top of the sphere ( Figure 7F and Figure 7—video 1 ) . In line with experimental observations , the tube elongates due to cells moving in the tube ( Martik and McClay , 2017 ) . The contribution of extracellular matrix swelling is specific to sea urchin invagination . As bottle cells alone are sufficient to drive invagination in other systems , we have tested a cell-intrinsic scenario where AB polarities of bottle cell neighbors prefer to be tilted towards each other ( by e . g . modifying the potential in Equation 6 to capture AB polarities as in Figure 2—figure supplement 2 , unpublished results ) . This also resulted in successful invagination . Several observations suggest that radial intercalation movements towards the center of invagination may drive tissue bending ( Sanchez-Corrales et al . , 2018; Panousopoulou and Green , 2016; Rembold et al . , 2006; Chung et al . , 2017 ) . When tested in our model , PCP alone ( omitting external or cell-intrinsic reorientation of AB polarity ) results in exvagination and tube elongating outside of the sphere ( this is more energetically favorable than extending inwards ) . A similar , exogastrulating , phenotype was observed in a PCP mutant ( Long et al . , 2015 ) . While the similarity may be accidental , it is possible that this mutation abrogates PCP-driven apical constriction ( Ossipova et al . , 2014; Nishimura et al . , 2012 ) and since cellular apices face outside , abrogating apical constriction eliminates the bias in direction of tube formation . We have also tested the consequences of apical constriction ( reorienting AB polarity ) in the model without PCP and while this was sufficient for invagination , the cavity remained spherically symmetric and failed to form a tube . To address the origins of morphological diversity and stability across species and organs , we focused on a phenomenological description of polarized cell–cell interactions . This allowed us to bridge local single-cell symmetry breaking events to global changes in morphologies spanning tens of thousands of interacting cells . With this tool at hand , we find that with only a few parameters , we can recapitulate the two global symmetry breaking events: formation of epithelial sheets and folds by cells with AB polarity , and emergence of global axial symmetry ( tubes ) among cells with PCP . Remarkably , our results show that interactions among AB polarized cells lead to stable morphologies , that after initial relaxation remain indefinitely in their final configuration . The morphologies are robust to noise , growth , and local damage ( Figure 3—figure supplement 2 ) . These results may explain how organs and embryos preserve their architecture while growing . Polar cell–cell interactions not only provide clue to the morphological stability , but also point to a simple explanation to the origin of the diversity . We find that the exact morphological details are defined by initial conditions , for example initial positions and orientation of polarities , and boundary conditions , for example polarities restricted to certain direction for a fraction of the cells . It is thus tempting to speculate that diverse shapes do not require multiple interacting morphogen gradients , but can be a result of differences in initial and/or boundary conditions: as for example presence of yolk cells at start and boundary constraints by vitelline membrane ( Schierenberg and Junkersdorf , 1992; Wu et al . , 2010 ) . The diversity of shapes and forms is further enriched by a second symmetry breaking event , PCP , oriented perpendicular to AB polarity . Within our phenomenological framework , addition of PCP component is simple , and requires only two additional parameters: one favoring perpendicular orientation of AB polarity and PCP within a cell , and another , favoring parallel PCP alignments between neighbor cells . These constraints are the coarse-grained representation of the well-established experimental and computational results on intracellular symmetry breaking events and global ordering of planar polarities mediated by cell–cell coupling ( Le Garrec et al . , 2006; Amonlirdviman et al . , 2005; Wang et al . , 2006 ) . The first constraint allowed formation of axial symmetry and in combination with AB polarity , stable tubes , with length and diameter remaining constant with time . The second constraint resulted in cell rearrangements and intercalations consistent with the cell-autonomous CE typically associated with PCP . The patterns of neighbor exchanges during cell-rearrangements are in line with the ubiquitous T1 exchanges through formation and resolution of four cell vertex , Figure 6—figure supplement 3 ( Nishimura et al . , 2012; Sanchez-Corrales et al . , 2018 ) . The mechanism of the CE in our model is in line with the results by ‘filopodia tension model’ where elongated structures of many cells emerge from local cell–cell interactions in a direction defined by PCP ( Belmonte et al . , 2016 ) . The presented formulations of our model captures only some of the known events contributing to CE and does not include PCP-driven changes in cell shape mediated by for example apical constriction ( Nishimura et al . , 2012 ) or contribution of external forces ( Lye and Sanson , 2011 ) . Combining AB polarization and a local induction of PCP in a subpopulation of cells was sufficient to obtain main stages of sea urchin gastrulation: invagination , tube formation , and elongation through CE as well as merging of the tube with the animal pole at the top of the blastula . It is important to notice that the model for gastrulation uses an external force ( Equation 25 ) acting on AB polarity , and the model for neurulation uses an imposed external constraint on PCP . This is not fully satisfying , and suggests an extension of the model to capture tissue bending induced by changes in cell shapes at the edge of the region that invaginates . The existing in silico models treat invagination and CE-driven tube elongation as independent processes ( Figure 1—figure supplement 1 ) . Recent data , however , suggests that multiple mechanisms ( intracellular apical constriction , intercellular directed cell division and cell intercalations , and supracellular actomyosin cables ) act simultaneously and contribute to both invagination and tubulogenesis ( Chung et al . , 2017; Nishimura et al . , 2012; Ossipova et al . , 2014 ) . Within our approach apical constriction ( modeled as reorientation of AB polarity ) and CE can act in parallel ( Figure 7 ) . It will be interesting to parallel recent experimental work ( Chung et al . , 2017 ) and computationally investigate how a combination of intra- , inter- and supracellular mechanisms contribute to the robustness of tubulogenesis , and to what extent the model can capture the range of observed phenotypes . It has been proposed that the above mechanisms may all be coordinated by PCP ( Nishimura et al . , 2012 ) . Besides the reported molecular links , a simple logic suggests that these mechanisms cannot be isotropic as in this case the initial bending will result in spherical structures . Thus , apical constriction , cell intercalations , and actomyosin cables have to be anisotropic ( planar polarized ) in directions consistent with the eventual tube orientation . This anisotropy is reported for both ‘wrapping’ tubes forming parallel to the epithelial plane , for example neurulation ( Nishimura et al . , 2012 ) , and ‘budding’ tubes forming orthogonally to the epithelial plane , for example salivary glands ( Chung et al . , 2017; Sanchez-Corrales et al . , 2018 ) . As organizing signals such as WNT can induce and orient PCP ( Chu and Sokol , 2016 ) within the apical plane , we asked if it is in principle possible to design PCP constraints ( not limited to the apical plane ) that would result in ‘wrapping’ and ‘budding’ . First , pointing PCP out of the plane was sufficient for a sheet of cells to bend ( Figure 7—figure supplement 1 ) . This is because in our formulation , PCP can drive reorientation of AB polarity and that in its turn is able to bend the sheet ( Figure 7 ) . Both ‘budding’ and ‘wrapping’ were qualitatively captured by the model when the axial and radial anisotropy were set by constraining orientation of PCP for cells within a circle ( ‘budding’ in sea urchin ) and two stripes of cells ( mimicking hinge points in neurulation ‘wrapping’ ) . While CE was needed for proper tube forming in sea urchin example , the tube formed without CE in neurulation ( Figure 7—figure supplement 1 ) . Thus , the simulations suggest that ‘wrapping’ in neurulation and gastrulation in Drosophila ( Figure 7—figure supplement 1 ) vs . ‘budding’ in sea urchin and organogenesis ( Andrew and Ewald , 2010; Zegers , 2014 ) may be outcomes of different constraints imposed on PCP . While it is intriguing to speculate that PCP may be oriented out of epithelial plane directly by organizing signals , this may also be an indirect effect of a sequence of intermediate steps . As the organizing signals not only induce and orient PCP but also drive apical constriction ( effectively reorienting AB polarity ) the PCP may be gradually oriented out of the original plane of epithelium by the following sequence of events: PCP → apical constriction → tilt in AB polarity → tilt in apical plane → PCP out of original epithelial plane . This less precise but simpler interpretation highlights the fact that PCP may drive many of the alternative mechanisms of tubulogenesis and shifts the focus from the differences in mechanisms driving tubulogenesis to the differences in boundary conditions – a set of constraints imposed on cell polarities by neighboring tissue ( e . g . notochord in neurulation and organizer in gastrulation ) . In addition to our conceptual findings , we propose three testable predictions . First , we predict that two potential mechanisms behind the emergence of folds in pancreatic organoids – matrigel resistance and rapid , out-of-equilibrium , cell proliferation – will result in distinct morphologies . Our results suggest that in case of rapid proliferation , the growing structure will develop many shallow folds close to the surface which later tend to deepen . In contrast , external pressure causes fewer but deeper and longer folds ( Figure 5—figure supplement 1 ) . And further , as organoids grow in size , the number of folds will reach an upper limit when under pressure , however , in case of rapid proliferation , the number of folds will keep growing ( Figure 5 ) . Visual inspection of published morphologies seems to support the out-of-equilibrium growth ( Greggio et al . , 2013; Li et al . , 2017 ) . To assess if the growth is out-of-equilibrium in 3D organoids , one can quantify the distributions of cell shapes ( Cerruti et al . , 2013 ) . Our model thus predicts that quantitative counting of folds and measurements of the fold depth and length relative to the size of the growing structure may discriminate between the alternative hypothesis . Quantification of the folds can be done in in vitro organoids by either phase or confocal fluorescence microscopy of whole-mount immunostained samples ( Greggio et al . , 2013; Li et al . , 2017 ) . The fold depth and length can be quantified with the same approach as applied to simulated shapes ( Materials and methods ) in binarized images of the 3D organoid surfaces . As oganoids cannot be cultured without gel supporting 3D growth , it will be necessary to vary both gel stiffness and generation time to uncouple their respective contribution to the folding . The work by Little , 2017 shows that in brain organoids generation time can be both slowed down and speeded up by either genetic manipulation or by adding small molecule inhibitors of pathways regulating cell proliferation . Unfortunately , changing stiffness of matrigel also changes its biochemical composition and may affect cell proliferation and differentiation . One will have to turn to synthetic hydrogels , where it is now possible to uncouple mechanical and biochemical clues ( Gjorevski et al . , 2016 ) . To illustrate possible applications of our approach , we have only focused on two out of several possible mechanisms that may contribute to folding . The other likely alternative is that folding may result from differences in biomechanical interactions or generation times characteristic to the different cell types . These alternative scenarios are straightforward to consider in our model and will be an exciting venue to explore when more quantitative data on differences in organoid morphologies is available . Our second prediction is that in case of tubes formed by non-proliferating cells , the length and width of the tubes are controlled by the relative strength of AB polarity and PCP . This result calls for quantification of adhesion proteins along the AB polarity and PCP axes . In PCP mutants with shorter and wider tubes , one would expect less planar polarization in adherens junctions and actomyosin , for example larger spread compared to wild type in their orientation quantified relative to the tube axis ( Nishimura et al . , 2012 ) . Alternatively , the balance between AB polarity and PCP can be altered by weakening AB polarity , for example mutating tight junction proteins should result in longer tubes . A similar phenotype has already been reported for Drosophila tracheal tube ( Laprise et al . , 2010 ) . With the recent advancements in in vitro systems of tubulogenesis , allowing for easy genetic manipulations and more amenable for quantitative imaging , it may in principle be possible to relate the extent of planar anisotropy in PCP mutants and strength of AB adhesion in tight junction mutants with tube length and diameter . The existing coupling between PCP and AB polarity may , however , make it challenging to tweak one polarity at a time . Our third , and probably most challenging to test , prediction is on the conditions differentiating between tubes forming perpendicular ( e . g . sea urchin gastrulation ) or parallel ( as in Drosophila gastrulation or neurulation ) to the plane of epithelium . We predict that the outcome will be defined by the orientation of PCP in the invaginating region and the geometry of the boundaries ( circular for budding and axial for wrapping ) set by for example WNT organizing signals . Recent development in imaging localization of PCP complexes in single cells ( Wu et al . , 2013; Chu and Sokol , 2016; Minegishi et al . , 2017; Habib et al . , 2013 ) allows monitoring localization of PCP complexes , and thus PCP orientation , in individual cells . By placing WNT-soaked ( Habib et al . , 2013 ) beads or WNT-secreting cells ( Chu and Sokol , 2016 ) one can vary PCP orientation in the cells at the epithelial boundary facing WNT and test for the direction of the epithelial bending and possibly tube formation . Our approach is by purpose phenomenological and by its nature cannot make predictions about specific molecular details . In all cases , we do not see our simulations as finalized predictions , but rather as pointing in the most promising direction for further exploration of these complex developmental processes . Our setup easily allows for changes as we learn more . The proposed tool should be used in close collaboration with gained experimental knowledge on initial conditions , cell generation times and differentiation processes where polarities play a central role . Our results open for a series of biological generalizations both in development and diseases . On one hand , we now may be able to explain and unify the apparently very distinct morphological transitions during gastrulation in flies , frogs , fish , mice , and humans by accounting for different initial and boundary conditions . Our model suggests how a moderate change in expression of polarities during some critical evolutionary stages could lead to widely different final morphologies . Thereby , development driven by cell–cell polarity interactions could provide major morphological transitions from local and transient modulations in polarity . On the other hand , it becomes possible to think of gastrulation , neurulation , tubulogenesis , and organogenesis as the same class of phenomena , where the orientation of the tube is guided by local organizers , and lengths/widths of the tubes are determined by the relative strength of AB polarity and PCP . At the same time , there is an emerging view that wound healing and cancer are local perturbations – for example local loss of cells , dysregulation of cell polarities ( Martin-Belmonte and Perez-Moreno , 2012 ) , proliferation , or autonomously induced organizing signals – of otherwise conceptually the same developmental processes ( Humphries and Mlodzik , 2018 ) . The power of our model is that it allows to address these hypotheses through predictive models for the dynamics of many cells that interact through combinations of AB polarity and PCP . In our model , we use the following potential to describe the pairwise interaction between cells ( 12 ) Vij=e-rij-S e-rij/β , where rij is the center–center distance between cell i and cell j , and S is the polarity factor ( 13 ) S=λ1S1+λ2S2+λ3S3 Here , λ1 , λ2 , and λ3 are the strengths of the respective polarity terms which are given as ( 14 ) S1= ( p^i× r^ij ) ⋅ ( p^j× r^ij ) , ( 15 ) S2= ( p^i× q^i ) ⋅ ( p^j× q^j ) , ( 16 ) S3= ( q^i× r^ij ) ⋅ ( q^j× r^ij ) . The unit vectors p^i , p^j and q^i , q^j represent the apical-basal polarity and planar cell polarity of cell i and j . Throughout the paper , β is a constant which we set to 5 . In order to use the Euler method , we need the gradient of Vij differentiated with respect to position , r¯i , and the two polarities , p¯i and q¯i: ( 17 ) dVijdr¯i=e−rij/β{ γ r^ij−λ1rij[ ( r^ij⋅p^j ) p^i+ ( r^ij⋅p^i ) p^j]−λ3rij[ ( r^ij⋅q^j ) q^i+ ( r^ij⋅q^i ) q^j]} , ( 18 ) dVijdp¯i=e−rij/β{ λ1[S1p^i−p^j+ ( r^ij⋅p^j ) r^ij]+λ2[S2p^i− ( q^i⋅q^j ) p^j+ ( q^i⋅p^j ) q^j]} , ( 19 ) dVijdq¯i=e−rij/β{ λ2[S2q^i− ( p^i⋅p^j ) q^j+ ( p^i⋅q^j ) p^j]+λ3[S3q^i−q^j+ ( r^ij⋅q^j ) r^ij]} . In order to derive Equations 17 , 18 , and 19 , we have used the following: ( 20 ) ddr¯ie−rij/β=1β e−rij/β r^ij , ( 21 ) γ=e−rij ( β−1 ) /β−Sβ+2rij[λ1 ( r^ij⋅p^i ) ( r^ij⋅p^j ) +λ3 ( r^ij⋅q^i ) ( r^ij⋅q^j ) ] , ( 22 ) dS1dr¯i=1rij[ ( r^ij⋅p^i ) p^j+ ( r^ij⋅p^j ) p^i]−2[ ( r^ij⋅p^i ) ( r^ij⋅p^j ) r^ij] , ( 23 ) dS1dp¯i=1pi[p^j− ( r^ij⋅p^j ) r^ij−S1p^i] , where pi is the length of the polarity of cell i which is equal to one at all times . In Figure 5 , the number of cells , N , at a given time , t , is N = 200 exp ( ln ( 2 ) t/tG ) where tG is the generation time . In these simulations , the AB potential ( Equation 6 ) between cells is set to zero when the angle between their polarity is larger than π/2 . In Figure 5 , we model resistance from the matrigel by imposing a surface force pointing toward the center of mass . The potential of the pressure in the growth medium is given by ( 24 ) VM=−Pr22rmaxwhere P is the stiffness of the medium , r is distance from the center of mass , and rmax is the distance to the cell that is the furthest away from the center of mass . The resulting force will be constant in time at the periphery . Thus , all cells on a growing sphere will be exposed to a force of equal size . However , cells that end up deep inside a folded morphology will experience weaker resistance . In Figure 5 , the number of local minima is defined as the number of cells that do not have any neighbor cells that are closer to the center of mass than themselves , and at the same time have an average angle between their AB polarity and their neighbor cells displacement vector that is less than π/2 . In Figure 6 , the semi-minor and semi-major axes correspond to the half-width and half-length of the tubes , respectively . As cells on opposite sides of a tube have AB polarity pointing in opposite directions , we approximate the semi-major and semi-minor axes , by finding the half of the maximum and minimum distance between two cells with AB polarity pointing in opposite directions . In the gastrulation simulation ( Figure 7 ) , each cell is assigned a specific value of polarity strengths ( λ1 , i , λ2 , i , and λ3 , i ) . We define the mutual interaction strength between a pair , i and j , of cells in Equation 4 with different polarity strengths by setting λ1 = mean ( λ1 , i , λ1 , j ) , and λ2 = mean ( λ2 , i , λ2 , j ) as well as λ3 = mean ( λ3 , i , λ3 , j ) . This choice makes sure that two neighbor cells interact with a force with equal magnitude but opposite sign . Furthermore , it makes sure that λ1 + λ2 + λ3 = 1 holds for all cells . In the model described by Equations 6–9 , AB polarity and PCP can influence each other’s orientation . To constrain PCP to the apical plane and thus disable its influence on AB polarity , we set λ2 = 0 when updating AB polarity in time ( during the numerical integration of Equation 10 ) . The invagination in gastrulation is implemented by adding an external force , F , that act on the AB polarity in addition to our usual intrinsic forces from Equation 10 ( 25 ) F=-k r^ e ( -x2-y2 ) /σ2 , The two parameters are k which is the strength of the force ( in Figure 7 , k = 0 . 02 ) , and σ which defines the decline of the gaussian force ( in Figure 7 , σ = 10 ) . r^is the unit vector pointing from x = y = 0 to the cells’ position , and x and y are the respective coordinates of the cells . This force is applied on the AB polarity and bends the orientation of the polarity away from a z-axis ( the anterior-posterior axis ) . MATLAB script to generate and visualize data . MATLAB R2016b or newer is required together with the Statistics and Machine Learning Toolbox . In addition , the Parallel Computing Toolbox is required if the PAR parameter in the basic script is set to 1 . The input folder contains initial conditions for three standardized systems . Bulk systems have neither apical-basal ( AB ) polarity nor planar cell polarity . Plane and shell systems have only AB polarity . ‘N’ in the file names gives the number of cells in the system . The initial polarity directions can be modified on line 4–5 in the basic . m file . Inside this file , it is also possible to set the degree of noise ( η ) , the size of the time steps ( dt ) , and the relation between the polarity strengths ( λ1 , λ2 , and λ3 ) . The parameter inc is used to speed up the simulations by only applying the neighborhood function to the nearest 100 neighbors . Generated data is saved in the output folder , and the visualization script is in a separate folder .	Cells have the power to organise themselves to form complex and stable structures , whether it is to create a fully shaped baby from a single egg , or to allow adult salamanders to grow a new limb after losing a leg . This ability has been scrutinised at many different levels . For example , researchers have looked at the chemical messages exchanged by cells , or they have recorded the different shapes an embryo goes through during development . However , it is still difficult to reconcile the information from these approaches into a description that makes sense at multiple scales . When an embryo develops , sheets of cells fold and unfold to create complex 3D shapes , like the tubes that make our lungs . Moulding sheets into tubes relies on interactions between cells that are not the same in all directions . In fact , two types of asymmetry ( or polarity ) guide these interactions . Apical-basal polarity runs across a sheet of cells , which means that the top surface of the sheet differs from the bottom . Planar cell polarity runs along the sheet and distinguishes one end from the other . For instance , apical-basal polarity marks the inner and outer surfaces of our skin , while planar cell polarity controls the direction in which our hair grows . Nissen et al . set out to investigate how these polarities help cells in an embryo organise themselves to form complicated folds and tubes . To do this , simple mathematical representations of both apical-basal and planar cell polarities were designed . The representations were then combined to create computer simulations of groups of cells as these divide and interact with each other . Simulations of ‘cells’ with only apical-basal polarity were able to generate different shapes in the ‘tissues’ produced , including many found in living organisms . External conditions , such as how cells were arranged to start with , determined the resulting shape . With both apical-basal and planar cell polarities , the simulations reproduced an important change that occurs during early development . They also replicated how the tubes that transport nutrients and oxygen form . These results show that simple properties of individual cells , such as polarities , can produce different shapes in developing tissues and organs , without the need for a complicated overarching program . Abnormal changes in cell polarity are also associated with diseases such as cancer . The mathematical model developed by Nissen et al . could therefore be a useful tool to study these events .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "physics", "of", "living", "systems" ]	2018	Theoretical tool bridging cell polarities with development of robust morphologies
Regulated exocytosis establishes a narrow fusion pore as initial aqueous connection to the extracellular space , through which small transmitter molecules such as ATP can exit . Co-release of polypeptides and hormones like insulin requires further expansion of the pore . There is evidence that pore expansion is regulated and can fail in diabetes and neurodegenerative disease . Here , we report that the cAMP-sensor Epac2 ( Rap-GEF4 ) controls fusion pore behavior by acutely recruiting two pore-restricting proteins , amisyn and dynamin-1 , to the exocytosis site in insulin-secreting beta-cells . cAMP elevation restricts and slows fusion pore expansion and peptide release , but not when Epac2 is inactivated pharmacologically or in Epac2-/- ( Rapgef4-/- ) mice . Consistently , overexpression of Epac2 impedes pore expansion . Widely used antidiabetic drugs ( GLP-1 receptor agonists and sulfonylureas ) activate this pathway and thereby paradoxically restrict hormone release . We conclude that Epac2/cAMP controls fusion pore expansion and thus the balance of hormone and transmitter release during insulin granule exocytosis . Insulin is secreted from pancreatic β-cells and acts on target tissues such as muscle and liver to regulate blood glucose . Secretion of insulin occurs by regulated exocytosis , whereby secretory granules containing the hormone and other bioactive peptides and small molecules fuse with the plasma membrane . The first aqueous contact between granule lumen and the extracellular space is a narrow fusion pore ( upper limit 3 nm; Albillos et al . , 1997 ) that is thought to consist of both lipids and proteins ( Bao et al . , 2016; Sharma and Lindau , 2018 ) . At this stage , the pore acts as a molecular sieve that allows release of small transmitter molecules such as nucleotides and catecholamines , but traps larger cargo ( Obermüller et al . , 2005; Barg et al . , 2002; MacDonald et al . , 2006; Alvarez de Toledo et al . , 1993 ) . Electrophysiological experiments have shown that the fusion pore is short-lived and flickers between closed and open states , suggesting that mechanisms exist that stabilize this channel-like structure and restrict pore expansion ( MacDonald et al . , 2006; Hanna et al . , 2009; Breckenridge and Almers , 1987; Lollike et al . , 1995 ) . The pore can then expand irreversibly ( termed full fusion ) , which leads to mixing of granule and plasma membrane and release of the bulkier hormone content ( Obermüller et al . , 2005; Barg et al . , 2002; Anantharam et al . , 2010 ) . Alternatively , the pore can close indefinitely to allow the granule to be retrieved , apparently intact , into the cell interior ( termed kiss-and-run or cavicapture ) ( Obermüller et al . , 2005; MacDonald et al . , 2006; Taraska et al . , 2003; Tsuboi and Rutter , 2003; Shin et al . , 2018 ) . Estimates in β-cells suggest that 20–50% of all exocytosis in β-cells are transient kiss-and-run events that do not lead to insulin release ( Obermüller et al . , 2005; MacDonald et al . , 2006 ) . However , kiss-and-run exocytosis contributes to local signaling within the islet because smaller granule constituents , such as nucleotides , glutamate or GABA , are released even when the fusion pore does not expand . Within the islet , ATP synchronizes β-cells ( Hellman et al . , 2004 ) , and has both inhibitory ( Salehi et al . , 2005; Poulsen et al . , 1999 ) and stimulatory ( Richards-Williams et al . , 2008 ) effects on insulin secretion . ATP suppresses glucagon release from α-cells ( Tudurí et al . , 2008 ) , and activates macrophages ( Weitz et al . , 2018 ) . Interstitial GABA leads to tonic GABA-A receptor activation and α-cell proliferation ( Jin et al . , 2013; Ben-Othman et al . , 2017 ) , and glutamate stimulates glucagon secretion ( Cabrera et al . , 2008 ) . Regulation of fusion pore behavior is not understood mechanistically , but several cellular signaling events affect both lifetime and flicker behavior . Pore behavior has been shown to be regulated by cytosolic Ca2+ , cAMP , PI ( 4 , 5 ) P2 , and activation of protein kinase C ( PKC ) ( MacDonald et al . , 2006; Hanna et al . , 2009; Alés et al . , 1999; Calejo et al . , 2013; Scepek et al . , 1998 ) and recent superresolution imaging indicates that elevated Ca2+ and dynamin promote pore closure ( Shin et al . , 2018; Chiang et al . , 2014 ) . Both myosin and the small GTPase dynamin are involved in fusion pore restriction ( Jackson et al . , 2015; Tsuboi et al . , 2004; Graham et al . , 2002; Artalejo et al . , 1995; Aoki et al . , 2010 ) , and assembly of filamentous actin promotes fusion pore expansion ( Wen et al . , 2016 ) , suggesting a link to endocytosis and the cytoskeleton . In β-cells of type-2 diabetics , upregulation of amysin leads to decreased insulin secretion because fusion pore expansion is impaired ( Collins et al . , 2016 ) , and the Parkinson’s related protein α-synuclein promotes fusion pore dilation in chromaffin cells and neurons ( Logan et al . , 2017 ) , thus providing evidence for altered fusion pore behavior in human disease . Inadequate insulin secretion in type-2 diabetes ( T2D ) is treated clinically by two main strategies . First , sulfonylureas ( e . g . tolbutamide and glibenclamide ) close the KATP channel by binding to its regulatory subunit SUR1 , which leads to increased electrical activity and Ca2+-influx that triggers insulin secretion ( Henquin , 2000 ) . Sulfonylureas are given orally and are first-line treatment for type-2 diabetes in many countries . Second , activation of the receptor for the incretin hormone glucagon-like peptide 1 ( GLP-1 ) raises cytosolic [cAMP] and thereby increases the propensity of insulin granules to undergo exocytosis . Both peptide agonists of the GLP-1 receptor ( e . g . exendin-4 ) and inhibitors of DPP-4 are used clinically for this purpose . The effect of cAMP on exocytosis is mediated by a protein-kinase A ( PKA ) -dependent pathway , and by Epac2 , a guanine nucleotide exchange factor for the Ras-like small GTPase Rap ( Kawasaki et al . , 1998 ) that is a direct target for cAMP ( Ozaki et al . , 2000 ) and is recruited to insulin granule docking sites ( Alenkvist et al . , 2017 ) . Epac2 has also been suggested to be activated by sulfonylureas ( Zhang et al . , 2009 ) , which may underlie some of their effects on insulin secretion . Here , we have studied fusion pore regulation in pancreatic β-cells , using high-resolution live-cell imaging . We report that activation of Epac2 , either through GLP1-R/cAMP signaling or via sulfonylurea , restricts expansion of the insulin granule fusion pore by recruiting dynamin and amisyn to the exocytosis site . Activation of this pathway by two classes of antidiabetic drugs therefore hinders full fusion and insulin release , which is expected to reduce their effectiveness as insulin secretagogues . To monitor single granule exocytosis , human pancreatic β-cells were infected with adenovirus encoding the granule marker NPY-Venus and imaged by TIRF microscopy . Exocytosis was evoked by local application of a solution containing 75 mM K+ , which leads to rapid depolarization and Ca2+ influx . Visually , two phenotypes of granule exocytosis were observed . In the first , termed full fusion , fluorescence of a granule that was stably situated at the plasma membrane suddenly vanished during the stimulation ( in most cases within <100 ms; Figure 1a–c , left panels ) . Since the EGFP label is relatively large ( 3 . 7 nm vs 3 nm for insulin monomers ) , this is interpreted as rapid pore widening that allowed general release of granule cargo . The sudden release of material may suggest that this release coincided with the collapse of the granule into the plasma membrane , but we cannot exclude that at least some granules remained intact ( Taraska et al . , 2003; Shin et al . , 2018; Tsuboi et al . , 2004; Huang et al . , 2018 ) . In the second type , the rapid loss of the granule marker was preceded by an increase in its fluorescence that could last for several seconds ( flash events , Figure 1a–c , right panels ) . We and others have previously shown ( Taraska et al . , 2003; Ferraro et al . , 2005; Gandasi and Barg , 2014 ) that this reflects neutralization of the acidic granule lumen and dequenching of the EGFP-label , before the labeled cargo is released . Since this neutralization occurs as the result of proton flux through the fusion pore , the fluorescence timecourse of these events can be used to quantitatively study fusion pore behavior . In the following , we will report two parameters that reflect fusion pore behavior , the fraction of exocytosis events with flash phenotype ( indicating restricted pores , about 40% in control conditions; Figure 1d ) , and the duration of the flash , referred to as ‘NPY release times’ . The latter was estimated by fitting a discontinuous function to the fluorescence timecourse ( see Figure 1c , green lines and Figure 1e ) , which limits the analysis to granules that eventually released their peptide content . The distribution of the NPY release times followed a mono-exponential function and was on average 0 . 87 ± 0 . 12 s ( 186 granules in 26 cells ) in control conditions ( Figure 1e ) . Such events are increased by elevated cAMP ( MacDonald et al . , 2006; Hanna et al . , 2009 ) and likely other conditions that stabilize the fusion pore . Indeed , when forskolin ( 2 µM; fsk ) was added to the bath solution we observed a twofold increase of exocytosis rate ( Figure 1f ) , a threefold increase of NPY release times ( Figure 1e ) , and a nearly doubled fraction of events with restricted fusion pores ( Figure 1d , f ) . The GLP-1 agonist exendin-4 ( 10 nM; Ex4 ) had comparable effects ( Figure 1d–f ) . Effects similar to those observed for human β-cells ( Figure 1 ) were observed in the insulin secreting cell line INS-1 ( Figure 1—figure supplement 1 ) . The effect of fsk on fusion pore behavior was mimicked by the specific Epac2 agonist S223 ( Schwede et al . , 2015 ) . Incubation with S223-acetomethoxyester ( 5 µM ) increased the fraction of flash events by 60% ( Figure 1d ) , doubled average NPY release times ( Figure 1e ) and doubled the event frequency ( Figure 1f ) ; the effects of fsk and S223 were not additive . In contrast , the Epac-inhibitor ESI-09 decreased the exocytosis rate in the presence of fsk by 80% ( Figure 1f ) , and the average NPY release time and the fraction of flash events were both reduced by 60% ( Figure 1d–e ) . PKA inhibition with Rp8-Br-cAMPS ( Gjertsen et al . , 1995 ) decreased neither the fraction of flash events , nor average NPY release times ( Figure 1e ) . The results indicate that Epac rather than PKA is responsible for cAMP-dependent fusion pore regulation . Paradoxically , Epac activation increases the rate of exocytosis but slows the rate of peptide release from individual granules . We studied the effect of Epac2 overexpression on fusion pore regulation . INS-1 cells were co-transfected with EGFP-Epac2 and NPY-tdmOrange2 and fluorescence was recorded simultaneously in both color channels . Epac2 overexpression had no effect on the overall exocytosis rate in either absence or presence of fsk ( Figure 2a ) , but increased the rate of flash events ( Figure 2b–c ) , supporting our finding , based on manipulation of the endogenous Epac2 activity , that Epac2 is involved in fusion pore regulation ( Figure 1d ) . NPY release times in cells overexpressing Epac2 increased threefold in the absence of fsk , and were similar to controls in presence of fsk ( Figure 2d ) . This indicates that a high Epac concentration can achieve sufficient activity to affect insulin secretion even at basal cAMP level , likely because cAMP acts in part by increasing the Epac concentration at the plasma membrane ( Alenkvist et al . , 2017 ) . To test if cAMP-dependent fusion pore restriction affects release of small transmitter molecules , we quantified nucleotide release kinetics from individual granules using patch clamp electrophysiology . The purinergic receptor cation channel P2X2 , tagged with RFP ( P2X2-RFP ) , was expressed in INS-1 cells as an autaptic nucleotide sensor ( Obermüller et al . , 2005 ) ( Figure 3a ) . The cells were voltage-clamped in whole-cell mode and exocytosis was elicited by including a solution with elevated free Ca2+ ( calculated 600 nM ) in the patch electrode . In this configuration , every exocytosis event that co-releases nucleotides causes an inward current spike , similar to those observed by carbon fiber amperometry ( Figure 3a–b ) . Including cAMP in the pipette solution increased the frequency of current spikes by 50% , consistent with accelerated exocytosis . This effect of cAMP was blocked if the Epac inhibitor ESI-09 was present ( Figure 3b–c ) . The current spikes ( see Figure 3a , right ) reflect nucleotide release kinetics during individual exocytosis events . In the presence of cAMP , but not cAMP + ESI-09 , they were markedly widened as indicated by on average 20% longer half-widths ( Figure 3d ) , 30% longer decay constants ( τ , Figure 3e ) , and 40% slower rising phases ( 25–75% slope , Figure 3f ) , compared with control . This indicates that nucleotide release is slowed by cAMP , likely because of changed fusion pore kinetics . Since the effect is blocked by ESI-09 , we conclude that the cAMP effect probably is mediated by Epac . Since ESI-09 blocks all Epac isoforms ( Zhu et al . , 2015 ) , we characterized fusion pore behavior in isolated β-cells from Epac2-/- ( Rapgef4-/- ) mice that lack all splice variants of Epac2 ( Kopperud et al . , 2017 ) . Cells from WT or Epac2-/- mice were infected with adenovirus encoding the granule marker NPY-tdmOrange2 and challenged with 75 mM K+ ( Figure 4a–b ) . In the absence of forskolin , exocytosis was significantly slower in Epac2-/- cells than WT cells , and the fraction of flash-associated exocytosis events was five-fold lower ( Figure 4c–e ) . This was paralleled by strikingly shorter fusion pore life-times in Epac2-/- cells compared with WT ( Figure 4f ) . The data suggest that Epac2 is partially activated in these conditions , consistent with elevated cAMP levels in mouse β-cells in hyperglycemic conditions ( Dyachok et al . , 2008 ) . As expected , forskolin increased both exocytosis ( Figure 4e ) and the fraction of flash events ( Figure 4c ) of WT cells . In contrast , forskolin failed to accelerate exocytosis in Epac2-/- cells , and the fraction of flash events was similar with or without forskolin ( Figure 4c , f–g ) . We conclude therefore that the effects of cAMP on fusion pore behavior are mediated specifically by Epac2 . Sulfonylureas have been reported to activate Epac ( Zhang et al . , 2009 ) , in addition to their classical role that involves the sulfonylurea receptor ( SUR ) . We therefore tested the effect of sulfonylureas on fusion pore behavior . INS-1 cells expressing NPY-tdmOrange2 were tested with three types of sulfonylureas , with different relative membrane permeability ( tolbutamide < glibenclamide < gliclazide ) . In addition , diazoxide ( 200 µM ) was present to prevent electrical activity . Exocytosis was not observed under these conditions , but could be triggered by local application of elevated K+ ( 75 mM ) . In the absence of fsk , the sulfonylureas accelerated K+-stimulated exocytosis about twofold over that observed in control ( Figure 5b , left ) , which is consistent with earlier findings that sulfonylureas augment insulin secretion via intracellular targets ( Barg et al . , 1999 ) . This effect was entirely due to an increase in flash-associated exocytosis events ( Figure 5b–c ) and the average NPY release time increased accordingly in the presence of sulfonylurea ( Figure 5d ) . Fsk strongly stimulated both flash-associated and full fusion exocytosis in absence of sulfonylurea ( Figure 5b–c , middle ) ; under these conditions , sulfonylureas tended to decrease full-fusion exocytosis without effect on the frequency of flash-associated events ( Figure 5b , middle ) . Accordingly , NPY release times were elevated compared with control ( no fsk ) , and only marginally longer than with fsk alone ( Figure 5d , right ) . Similar results were obtained in human β-cells , where glibenclamide increased exocytosis in the absence of fsk ( p=0 . 01 , n = 13 cells from four donors ) but not in its presence ( p=0 . 80 , n = 7 cells from four donors; data not shown ) . The data indicate that sulfonylureas restrict fusion pore expansion through the same intracellular pathway as cAMP , which may counteract their stimulating effect on exocytosis by preventing or delaying peptide release . Sulfonylureas also bind to SUR1 in the plasma membrane , which leads to rapid closure of KATP channels , depolarization and exocytosis . We tested the involvement of SUR1 by applying sulfonylureas acutely , which is expected to activate SUR1 in the plasma membrane but not Epac the cytosol ( Figure 5e ) . Reduced diazoxide ( 50 µM ) prevented glucose-dependent exocytosis but still allowed acute stimulation of exocytosis by sulfonylureas . Under these conditions , the fraction of flash-associated exocytosis events ( Figure 5f–g ) and the NPY release times ( Figure 5h ) were similar to control ( stimulation with elevated K+ ) for all three sulfonylureas . Taken together , the data suggest that sulfonylureas must enter the cytosol to affect fusion pore behavior , and that this effect is not mediated by the plasma membrane SUR . We excluded the possibility that sulfonylureas affect the fluorescence signal indirectly , by altering granule pH ( Figure 5—figure supplement 1 ) . Moreover , an EGFP-tagged SUR1 ( EGFP-SUR1 ) expressed in INS-1 cells did not localize to exocytosis sites or affect fusion pore behavior ( Figure 5—figure supplement 2 ) . We therefore conclude that sulfonylureas affect fusion pore behavior through Epac2 . The proteins dynamin and amisyn have previously been implicated in fusion pore regulation in β-cells ( Tsuboi et al . , 2004; Collins et al . , 2016 ) . To understand how these proteins behave around the release site , we expressed EGFP-tagged dynamin1 ( Figure 6a ) or mCherry-tagged amisyn ( Figure 6b ) together with a granule marker in INS-1 cells , and stimulated exocytosis with elevated K+ . In the presence of fsk , both of the two fluorescent proteins were recruited to the granule site during membrane fusion ( Figure 6c , f , and Figure 6—figure supplement 1 ) . Expression of both proteins was about 2–4 fold compared with endogenous levels ( Figure 6—figure supplement 2 ) , and markedly increased the NPY release times ( Figure 6d , g ) and flash-associated exocytosis events ( Figure 6e , h ) . Addition of the Epac inhibitor ESI09 prevented recruitment of both dynamin1 and amisyn during flash events and reduced flash events and NPY release times below control ( Figure 6c–h ) . In the absence of fsk , expression of the two proteins had no effect on fusion pore behavior , and only amisyn ( but not dynamin1 ) was recruited to the exocytosis site ( Figure 6i–n ) . When Epac was activated with S223 ( no fsk ) , dynamin1 and amisyn were recruited during flash events , and NPY release times and flash events were increased for both proteins ( Figure 6i–n ) . The data suggest that dynamin1 and amisyn are acutely recruited to the exocytosis site , where they participate in cAMP-dependent fusion pore restriction . cAMP-dependent signaling restricts fusion pore expansion and promotes kiss-and-run exocytosis in β-cells ( Hanna et al . , 2009 ) and neuroendocrine cells ( Calejo et al . , 2013; Machado et al . , 2001 ) ( but see Hatakeyama et al . , 2006 ) . We show here that the cAMP-mediator Epac2 orchestrates these effects by engaging dynamin and perhaps other endocytosis-related proteins at the release site ( Figure 7 ) . Since the fusion pore acts as a molecular sieve , the consequence is that insulin and other peptides remain trapped within the granule , while smaller transmitter molecules with para- or autocrine function are released ( Obermüller et al . , 2005; MacDonald et al . , 2006; Taraska et al . , 2003; Leclerc et al . , 2004 ) . Incretin signaling and Epac activation therefore delays , or altogether prevents insulin secretion from individual granules , while promoting paracrine intra-islet communication that is based mostly on release of small transmitter molecules . Paradoxically , two clinically important classes of antidiabetic drugs , GLP-1 analogs and sulfonylureas , activate Epac in β-cells and caused restriction of the fusion pore . Sulfonylureas have long been known to stimulate insulin secretion by binding to SUR1 , which results in closure of KATP channels and depolarization ( Henquin , 2000 ) . The drugs also accelerate PKA-independent granule priming in β-cells , which may involve activation of intracellularly localized SUR1 ( Eliasson et al . , 2003 ) . Our data indicate that sulfonylureas exert a third mode of action that leads to the restriction of the fusion pore and therefore limits insulin release . Two pieces of evidence suggest that SUR1 is not involved in the latter . First , acute exposure to sulfonylureas had no effect on fusion pore behavior , although it blocks KATP channels ( indicating SUR1 activation ) . Only long-term exposure to sulfonylurea resulted in restricted fusion pores , likely because it allowed the drugs to enter the cytoplasm . Second , we could not detect enrichment of SUR1 at the granule release site , which precludes any direct role of the protein in fusion pore regulation . Sulfonylurea compounds have been shown to allosterically stabilize the cAMP-dependent activation of Epac ( Takahashi et al . , 2013; Herbst et al . , 2011 ) . Our finding that sulfonylurea caused fusion pore restriction in the absence of forskolin indicates that basal cAMP concentrations are sufficient for this effect . Since gliclizide binds the CNB1 domain without activating it ( Takahashi et al . , 2013 ) and still restricts the fusion pore , Epac localization at the granule site ( Alenkvist et al . , 2017 ) may be enough to regulate the downstream proteins ( e . g . dynamin and amisyn ) . It can further be speculated that the competing stimulatory ( via exocytosis ) and inhibitor effects ( via the fusion pore ) of sulfonylureas on insulin secretion , contribute to the reduction in sulfonylurea effectiveness with time of treatment . Long-term treatment with GLP-1 analogs disturbs glucose homeostasis ( Abdulreda et al . , 2016 ) , and combination therapy of sulfonylurea and DPP4 inhibitors ( that elevate cAMP ) has been shown to lead to severe hypoglycemia ( Yabe and Seino , 2014 ) , an effect that likely depends on Epac ( Takahashi et al . , 2015 ) . Epac mediates the PKA-independent stimulation of exocytosis by cAMP ( Seino et al . , 2009 ) and our data suggests it may affect both priming and fusion pore restriction . This effect is rapid ( Eliasson et al . , 2003 ) , suggesting that Epac is preassembled at the site of the secretory machinery . Indeed , Epac concentrates at sites of docked insulin granules ( Alenkvist et al . , 2017 ) , and forms functionally relevant complexes with the tethering proteins Rim2 and Piccolo ( Fujimoto et al . , 2002 ) . However , the amount of Epac2 present at individual release sites did not correlate with fusion pore behavior , which may indicate that the protein acts indirectly by activating or recruiting other proteins . Indeed , we show here that recruitment of two other proteins , dynamin and amisyn , depends on cAMP and Epac . Other known targets of Epac are the small GTPases Rap1 and R-Ras , for which Epac is a guanine nucleotide exchange factor ( GEF ) . Rap1 is expressed on insulin granules and affects insulin secretion both directly ( Shibasaki et al . , 2007 ) , and by promoting intracellular Ca2+-release following phospholipase-C activation ( Dzhura et al . , 2011 ) . R-Ras is an activator of phosphoinositide 3-kinase ( Marte et al . , 1997 ) . By altering local phosphoinositide levels , Epac could therefore indirectly affect exocytosis via recruitment of C2-domain proteins such as Munc13 ( Kang et al . , 2006 ) , and fusion pore behavior by recruitment of the PH-domain containing proteins dynamin and amisyn ( Ramachandran and Schmid , 2008; Abbineni et al . , 2018 ) . An unresolved question is whether pore behavior is controlled by mechanisms that promote pore dilation , or that instead prevent it . Dynamin causes vesicle fission during clathrin-dependent endocytosis ( Marks et al . , 2001 ) , and since dynamin is present at the exocytosis site and required for the kiss-and-run mode ( Jackson et al . , 2015; Tsuboi et al . , 2004; Trexler et al . , 2016 ) , it may have a similar role during transient exocytosis . An active scission mechanism is also suggested by the finding that granules loose some of their membrane proteins during transient exocytosis ( Tsuboi et al . , 2004; Perrais et al . , 2004 ) . Capacitance measurements have shown that fusion pores initially flicker with conductances similar to those of large ion channels , before expanding irreversibly ( Lollike et al . , 1995 ) . This could result from pores that are initially stabilized through unknown protein interactions and that eventually give way to uncontrolled expansion . However , scission mechanisms involving dynamin can act even when the pore has dilated considerably beyond limit of reversible flicker behavior ( Shin et al . , 2018; Taraska and Almers , 2004; Zhao et al . , 2016; Anantharam et al . , 2011 ) , and even relatively large granules retain their size during fusion-fission cycles ( MacDonald et al . , 2006; Lollike et al . , 1995 ) . Separate mechanisms may therefore operate , one that prevents pore dilation by actively causing scission , similar to the role of dynamins in endocytosis , and another by shifting the equilibrium between the open and closed states of the initial fusion pore . Curvature-sensitive proteins are particularly attractive for such roles since they could accumulate at the neck of the fused granule; such ring-like assemblies that have indeed been observed for the Ca2+-sensor synaptotagmin ( Wang et al . , 2001 ) . Active pore dilation has also been proposed to be driven by crowding of SNARE proteins ( Wu et al . , 2017 ) and α-synuclein ( Logan et al . , 2017 ) . β-cell granules contain a variety of polypeptides ( insulin , IAPP , chromogranins ) and small molecule transmitter molecules ( GABA , nucleotides , 5HT ) that have important para- and autocrine functions within the islet ( Braun et al . , 2012; Caicedo , 2013 ) . Insulin modulates its own release by activating β-cell insulin receptors ( Leibiger et al . , 2008 ) , stimulates somatostatin release ( Vergari et al . , 2019 ) , and inhibits glucagon secretion ( Ravier and Rutter , 2005 ) . Insulin secretion is also inhibited by IAPP/amylin and chromogranin cleavage products such as pancreastatin ( Braun et al . , 2012 ) . Of the small transmitters , GABA inhibits glucagon secretion from α-cells ( Rorsman et al . , 1989 ) and enhances insulin secretion ( Soltani et al . , 2011 ) , and tonic GABA signaling is important for the maintenance of β-cell mass ( Soltani et al . , 2011 ) . Adenine nucleotides cause β-cell depolarization , intracellular Ca2+-release and enhanced insulin secretion ( Khan et al . , 2014; Jacques-Silva et al . , 2010 ) , but also negative effects have been reported ( Salehi et al . , 2005; Poulsen et al . , 1999 ) . Paracrine purinergic effects also coordinate Ca2+ signaling among β-cells ( Hellman et al . , 2004 ) , stimulate secretion of somatostatin from δ-cells ( Bertrand et al . , 1990 ) , and target islet vasculature and macrophages as part of the immune system ( Weitz et al . , 2018 ) . By selectively allowing small molecule release , Epac/cAMP-dependent fusion pore restriction is expected to alter both the timing and the relative volume of peptidergic vs . transmitter signaling . Given that granule priming and islet electrical activity are regulated on a second time scale , even small delays between these signals can be envisioned to affect the ratio of insulin to glucagon secretion . As illustrated by the recent finding of altered fusion pore behavior in type-2 diabetes ( Collins et al . , 2016 ) , Epac-dependent fusion pore regulation may have profound consequences for islet physiology and glucose metabolism in vivo . Human islets were obtained from the Nordic Network for Clinical Islet Transplantation Uppsala ( Goto et al . , 2004 ) under full ethical clearance ( Uppsala Regional Ethics Board 2006/348 ) and with written informed consent . Isolated islets were cultured free-floating in sterile dishes in CMRL 1066 culture medium containing 5 . 5 mM glucose , 10% fetal calf serum , 2 mM L-glutamine , streptomycin ( 100 U/ml ) , and penicillin ( 100 U/ml ) at 37°C in an atmosphere of 5% CO2 up to 2 weeks . Prior to imaging , islets were dispersed into single cells by gentle agitation using Ca2+-free cell dissociation buffer ( Thermo Fisher Scientific ) supplemented with 10% ( v/v ) trypsin ( 0 . 05% Thermo Fisher Scientific ) . INS1-cells clone 832/13 ( Hohmeier et al . , 2000 ) were maintained in RPMI 1640 ( Invitrogen ) with 10 mM glucose , 10% fetal bovine serum , streptomycin ( 100 U/ml ) , penicillin ( 100 U/ml ) , Sodium pyruvate ( 1 mM ) , and 2-mercaptoethanol ( 50 μM ) . The ins1 832/13 cells were screened by PCR and found negative for mycoplasma . Mouse islets were isolated from 5 to 12 months old WT and Epac2-/- ( Kopperud et al . , 2017 ) ( Rapgef4-/- ) animals . The Epac2 deletion involves exons 12–13 , which include the high-affinity cAMP binding domain present in all Epac2 isoforms , in contrast to previously reported knockout strain ( Shibasaki et al . , 2007 ) , which only lacks the Epac2A isoform . The mice were anesthetized and the pancreas dissected out and cleared from fat and connective tissue in ice-cold Ca5 solution ( in mM 125 NaCl , 5KCl , 1 . 2 MgCl2 , 1 . 28 CaCl2 , 10 HEPES; pH 7 . 4 with NaOH ) . Pancreas was injected with Collagenase P ( 1 mg/ml ) and cut into small pieces before mechanical dissociation ( 7 min at 37°C ) . BSA was added immediately and islets were washed 3X with ice cold Ca5 buffer with BSA . Islets were dispersed into single cells using Ca2+-free cell dissociation buffer ( supplemented with 10% ( v/v ) trypsin ) and gentle agitation . Dispersed cells were sedimented by centrifugation , resuspended in RPMI 1640 medium ( containing 5 . 5 mM glucose , 10% fetal calf serum , 100 U/ml penicillin and 100 U/ml streptomycin ) . The cells were plated onto 22 mm polylysine-coated coverslips and were transduced the next day using adenovirus ( human and mouse cells ) or transfected the same day with plasmids ( INS1 cells , using Lipofectamine2000 , Invitrogen ) encoding the granule markers NPY-Venus , NPY-EGFP or NPY-tdOrange . Imaging proceeded 24–36 hr later . The open-reading frame of human amisyn ( NM_001351940 . 1 ) was obtained as a synthetic DNA fragment ( Eurofins , Germany ) and was cloned into pCherry2 C1 ( Addgene , plasmid nr 54563 ) by seamless PCR cloning . The linker between Cherry2 and amisyn translates into the peptide SGLRSRAQASNSAV . The plasmid N1 NPY-EGFP-mCherry coding for NPY-linker ( TVPRARDPPVAT ) -EGFP-linker ( KRSGGSGGSGGS ) -mCherry was made by seamless PCR cloning . The correct open-reading frame of both Cherry2-linker-amisyn and NPY-EGFP-mCherry was confirmed by Sanger sequencing ( Eurofins , Germany ) . The NPY-tdOrange2 adeno virus was made using the RAPAd vector system ( Cell Biolabs , San Diego ) . NPY-tdOrange2 ( Gandasi et al . , 2015 ) was cloned into the pacAd5 CMVK-NpA Shuttle plasmid ( Cell Biolabs ) . Virus was produced in HEK293 cells and isolated according to the instructions of the manufacturer ( Cell Biolabs ) . Cells were imaged in ( mM ) 138 NaCl , 5 . 6 KCl , 1 . 2 MgCl2 , 2 . 6 CaCl2 , 10 D-glucose 5 HEPES ( pH 7 . 4 with NaOH ) at 32–34°C . Exocytosis was evoked with high 75 mM K+ ( equimolarly replacing Na+ ) , applied by computer-timed local pressure ejection through a pulled glass capillary . For K+-induced exocytosis , spontaneous depolarizations were prevented with 200 µM diazoxide ( 50 µM for Figure 5e–h ) . In Figure 5e–h , exocytosis was evoked by sulfonylureas ( 500 µM tolbutamide , 200 µM glibenclamide or 200 µM gliclizide ) . For electrophysiology , glucose was reduced to 3 mM , and the electrodes were filled with ( mM ) 125 CsCl , 10 NaCl , 1 . 2 MgCl2 , 5 EGTA , 4 CaCl2 , 3 Mg-ATP , 0 . 1 cAMP , 10 HEPES ( pH 7 . 15 using CsOH ) . To quantify the overexpression , INS-1 cell were transfected with either Cherry2-amisyn or Dynamin1-GFP , fixed 24 hr later in 3 . 8% formaldehyde in phosphate-buffered saline ( PBS ) for 30 min at 25°C and washed in PBS . The cells were permeabilized in 0 . 2% Triton X-100 in PBS for 5 min and washed in PBS . Blocking was done using 5% FBS in PBS for 1–2 hr at 25°C . Cells were then incubated with a primary antibody ( anti-Dynamin1 , ab52852 abcam or anti-Amisyn , ab153974 abcam ) both diluted 1/50 in 5% FCS in PBS over night at 4°C and washed again in PBS . Incubation with secondary antibody ( Alexa Fluor 488 anti-rabbit or Alexa Fluor 555 anti-rabbit , Invitrogen ) diluted 1/1000 in 5% FCS in PBS was performed for 1 hr at 25°C and subsequently the cells were washed in PBS . Human cells were imaged using a lens-type total internal reflection ( TIRF ) microscope , based on an AxioObserver Z1 with a 100x/1 . 45 objective ( Carl Zeiss ) . TIRF illumination with a calculated decay constant of ~100 nm was created using two DPSS lasers at 491 and 561 nm ( Cobolt , Stockholm , Sweden ) that passed through a cleanup filter ( zet405/488/561/640x , Chroma ) and was controlled with an acousto-optical tunable filter ( AA-Opto , France ) . Excitation and emission light were separated using a beamsplitter ( ZT405/488/561/640rpc , Chroma ) and the emission light chromatically separated ( QuadView , Roper ) onto separate areas of an EMCCD camera ( QuantEM 512SC , Roper ) with a cutoff at 565 nm ( 565dcxr , Chroma ) and emission filters ( ET525/50 m and 600/50 m , Chroma ) . Scaling was 160 nm per pixel . INS1 and mouse cells were imaged using a custom-built lens-type TIRF microscope based on an AxioObserver D1 microscope and a 100x/1 . 45 NA objective ( Carl Zeiss ) . Excitation was from two DPSS lasers at 473 nm and 561 nm ( Cobolt ) , controlled with an acousto-optical tunable filter ( AOTF , AA-Opto ) and using dichroic Di01-R488/561 ( Semrock ) . The emission light was separated onto the two halves of a 16-bit EMCCD camera ( Roper Cascade 512B , gain setting at 3800 a . u . throughout ) using an image splitter ( DualView , Photometrics ) with ET525/50 m and 600/50 m emission filters ( Chroma ) . Scaling was 100 nm per pixel for INS-1 experiments and 160 nm for mouse cells . The frame rate was 10 framess−1 , with 100 ms exposures . Exocytosis events were identified manually based on the characteristic rapid loss of the granule marker fluorescence ( most fluorescence lost within 1–2 frames ) in cells which exhibited minimum of 1 event/cell ( except mouse cells , where all cells were included ) . Events were classified as flash events if they exhibited an increase in the fluorescence signal before the rapid loss of the granule fluorescence . The NPY release times were obtained for both types of events by non-linear fitting with a discontinuous function in Origin as described previously ( Gandasi et al . , 2015 ) . Protein binding to the release site ( ΔF/S ) was measured as described previously ( Gandasi and Barg , 2014 ) . ATP release was measured in INS1 cells expressing RFP-tagged P2X2 receptor ( Obermüller et al . , 2005 ) . Cells were voltage-clamped in whole-cell mode using an EPC-9 amplifier and PatchMaster software ( Heka Elektronik , Lambrecht , Germany ) with patch-clamp electrodes pulled from borosilicate glass capillaries that were coated with Sylgard close to the tips , and fire-polished ( resistance 2–4 MΩ ) . The free [Ca2+] was calculated to be 600 nM ( WEBMAXC standard ) and elicited exocytosis that was detected as P2X2-dependent inward current spikes . Currents were filtered at 1 kHz and sampled at 5 kHz . Spike analysis was performed using automated program for amperometric recordings in IGOR Pro ( Segura et al . , 2000 ) , with the threshold set at eight times the RMS noise during event-free section of recording . Data are presented as mean ± SEM unless otherwise stated . Statistical significance was tested ( unless otherwise stated ) and is indicated by asterisks ( p<0 . 05 , p<0 . 01 , *p<0 . 001 ) . The not normally distributed exocytosis rates and ratios of flash events were tested with Kruskal Wallis with post hoc Dunn test and NPY release times were tested with Kolmogorov-Smirnov test .	Insulin is the hormone that signals to the body to take up sugar from the blood . Specialized cells in the pancreas – known as β-cells – release insulin after a meal . Before that , insulin molecules are stored in tiny granules inside the β-cells; these granules must fuse with the cells’ surface membranes to release their contents . The first step in this process creates a narrow pore that allows small molecules , but not the larger insulin molecules , to seep out . The pore then widens to release the insulin . Since the small molecules are known to act locally in the pancreas , it is possible that this “molecular sieve” is biologically important . Yet it is not clear how the pore widens . One of the problems for people with type 2 diabetes is that they release less insulin into the bloodstream . Two kinds of drugs used to treat these patients work by stimulating β-cells to release their insulin . One way to achieve this is by raising the levels of a small molecule called cAMP , which is well known to help prepare insulin granules for release . The cAMP molecule also seems to slow the widening of the pore , and Gucek et al . have now investigated how this happens at a molecular level . By observing individual granules of human β-cells using a special microscope , Gucek et al . could watch how different drugs affect pore widening and content release . They also saw that cAMP activated a protein called Epac2 , which then recruited two other proteins – amisyn and dynamin – to the small pores . These two proteins together then closed the pore , rather than expanding it to let insulin out . Type 2 diabetes patients sometimes have high levels of amisyn in their β-cells , which could explain why they do not release enough insulin . The microscopy experiments also revealed that two common anti-diabetic drugs activate Epac2 and prevent the pores from widening , thereby counteracting their positive effect on insulin release . The combined effect is likely a shift in the balance between insulin and the locally acting small molecules . These findings suggest that two common anti-diabetic drugs activate a common mechanism that may lead to unexpected outcomes , possibly even reducing how much insulin the β-cells can release . Future studies in mice and humans will have to investigate these effects in whole organisms .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2019	Fusion pore regulation by cAMP/Epac2 controls cargo release during insulin exocytosis
Animals mimicking other organisms or using camouflage to deceive predators are vital survival strategies . Modern and fossil insects can simulate diverse objects . Lichens are an ancient symbiosis between a fungus and an alga or a cyanobacterium that sometimes have a plant-like appearance and occasionally are mimicked by modern animals . Nevertheless , lichen models are almost absent in fossil record of mimicry . Here , we provide the earliest fossil evidence of a mimetic relationship between the moth lacewing mimic Lichenipolystoechotes gen . nov . and its co-occurring fossil lichen model Daohugouthallus ciliiferus . We corroborate the lichen affinity of D . ciliiferus and document this mimetic relationship by providing structural similarities and detailed measurements of the mimic’s wing and correspondingly the model’s thallus . Our discovery of lichen mimesis predates modern lichen-insect associations by 165 million years , indicating that during the mid-Mesozoic , the lichen-insect mimesis system was well established and provided lacewings with highly honed survival strategies . Modern insects have dramatic morphological specializations that match various objects of the environment . For instance , the specializations occurring in katydids and butterflies that mimic leaves , stick insects and inchworms that resemble twigs , and orchid mantids that duplicate orchid flowers , provide ecological insights for understanding mimetic associations between insect mimics and their plant models ( Stevens , 2011; Gullan and Cranston , 2014; Maran , 2017 ) . These and other fascinating cases reveal that mimesis or camouflage is highly effective when cryptic insects resemble closely the appropriate self-similar background , indicating the complexity of ecological relationships between insect mimics and their imitating models . When and how insects first evolved such an ingenious survival strategy is unclear . A Permian katydid exhibiting a mimicking pattern of wings similar to the modern relatives was considered the oldest case of insect mimicry ( Garrouste et al . , 2016 ) . However , evidence for a contemporaneous mimetic relationship in this Permian deposit was scarce , and there was no quantitative or other useful data to track the mimetic interactions among the insect , model and predator . More recent cases of insect mimicry have been recorded from the Mesozoic , indicating the existence of several such effective survival strategies . As in morphological specializations involving masterly deceit found in modern insects , several Mesozoic insect taxa developed remarkable structural adaptations resulting in highly accurate resemblances to co-existing models ( Wang et al . , 2012; Wang et al . , 2014; Yang et al . , 2020 ) . Prominent among these mimetic insects are Neuroptera ( lacewings , antlions and relatives ) , a nonspeciose relic order consisting of ca . 6000 extant species that engaged in several , impressive instances of mimicry that reveal novel and specialized strategies of deception , of which many are absent today . Striking examples are the Jurassic lacewing Bellinympha ( Saucrosmylidae ) , a compression fossil , mimicking cycadophyte leaves ( Wang et al . , 2010b ) , and larvae of the green lacewing Phyllochrysa ( Chrysopidae ) from amber , modified to resemble co-occurring liverworts ( Liu et al . , 2018 ) . Besides mimicry , other deceptive modes of appearance have been documented among Mesozoic lacewings , such as camouflaged larvae of chrysopid ( green lacewing ) and myrmeleontoid ( antlion relative ) neuropterans in amber , which evolved distinctive debris-carrying behaviors to enhance their predatory effectiveness ( Pérez-de la Fuente et al . , 2012; Pérez-de la Fuente et al . , 2018; Wang et al . , 2016; Badano et al . , 2018 ) . These cases collectively have promoted understanding of the early evolution of insect mimicry , but also have revealed that the currently species poor Neuroptera had evolved a significant repertoire of specializations involving morphologies and behaviors that adapted to a variety of Mesozoic settings . In this report , we found an exceptional system of the first lichen mimesis by a fossil lacewing . These occurrences are from the Daohugou 1 locality of Inner Mongolia in northeastern China . The new lichen-like-mimicking insects represent a new genus with two new species and exhibit remarkable wing patterns that accurately resemble the contemporaneous lichen species Daohugouthallus ciliiferus Wang , Krings et Taylor , 2010 ( Wang et al . , 2010a ) . The lichen affinity of the D . ciliiferus model previously was doubted due to the absence of evidence for fungal and algal connections that would indicate the presence of lichenization and thus the presence of a mutualistic symbiosis ( Honegger et al . , 2013; Lücking and Nelsen , 2018 ) . Our SEM results corroborate the actual presence of hyphae connected to algal cells on the D . ciliiferus specimens , indicating the foliose and subfruticose lichen growth forms were in existence during the Middle Jurassic . Present-day lichen-mimicking insects are widely recorded among several diverse orders , especially Coleoptera ( beetles ) , Lepidoptera ( moths and butterflies ) and Orthoptera ( grasshoppers , katydids and crickets ) , which have evolved unusual specializations of morphology and behavior consistent with co-occurring lichens and other habitation- or appearance similar organisms such as liverworts ( Gerson , 1973; Lücking , 2001; Capinera , 2008; Cannon , 2010; Lücking et al . , 2010 ) . An extraordinary orthopteran , the lichen dragon katydid from the modern Ecuadorian Andes , provides an excellent disguise of lichens ( Braun , 2011 ) . Other predatory and extant chrysopid larvae have a body mask adorned with affixed lichen fragments , an example of aggressive mimicry or the ‘wolf in sheep’s clothing’ syndrome ( Skorepa and Sharp , 1971; Slocum and Lawrey , 1976; Wilson and Methven , 1997; Tauber et al . , 2014 ) . Importantly , lichen-mimetic or -camouflaged insects have established a specialized lichen-association for feeding or sheltering to obtain survival advantage ( Gerson , 1973 ) . Our finding documents the earliest lichen-mimicking insect and reveals that this strategy of mimicry among insects has been in existence for minimally 165 Mya . This ancient association also will provide new insight for exploring the predator–prey relationships among insects and lichens , and the role of habitat during mid-Mesozoic time . We have studied five fossil lichen specimens , PB23120 , B0474 , B0476P/C , CNU-LICHEN-NN2019001 and CNU-LICHEN-NN2019002P/C , all of which were collected from the Daohugou 1 locality of Inner Mongolia . The newly collected specimens were identified to be Daohugouthallus ciliiferus based on careful observations of its distinctive morphology . The lichen-mimicking insects represent a new genus and two new species affiliated to Ithonidae of the order Neuroptera . The terminology of venation follows Breitkreuz et al . , 2017 . The two new species of Lichenipolystoechotes exhibit a very similar appearance , but they easily can be separated by the distinct differences of branches of the MA and CuA veins . Lichenipolystoechotes species are conspicuous based on their highly prominent , homologous , pigmentation pattern of their forewings , which implies that these insects evolved a similar defensive strategy . The closest extant relatives of Lichenipolystoechotes are Ithonidae ( moth lacewings ) , of which their ecological and biological features are poorly documented ( New , 1989 ) . The forewings of the two new species demonstrate a high similarity in their overall appearance , such as the forewing branching pattern ( Figures 3A , E and 4A ) that matches the thallus branches of the co-occurring foliose to subfruticose lichen Daohugouthallus ciliiferus ( Figure 4B–D; Wang et al . , 2010a ) . The entire forewing forms an irregular branching pattern amid rounded , diaphanous fenestrae ( windows ) that are distributed along the wing center and as U-shaped extensions occurring around the wing border . The pigmented branch pattern of the wings has uneven widths and is angulated outwardly . The variation in width of each forewing vein branch conforms well to the variation in width of the lichen’s branches , indicating a morphological similarity between the wing markings and lichen thallus ( Figure 4F; Figure 4—figure supplement 1; Supplementary file 1: Table S1 ) . Lichens often have punctiform pycnidia ( asexual reproductive structures ) with black spots appearing on their thallus , especially in extant foliose lichen families such as Parmeliaceae ( Thell et al . , 2012 ) . In Daohugouthallus ciliiferus specimens , punctiform black spots occur , but whether they are pycnidia is uncertain . It is noteworthy that a specimen of L . ramimaculatus displays similar , scattered spots on its wings that resemble the dark spots on the lichen thallus of D . ciliiferus ( Figure 4G–I ) , potentially strengthening the similarity between L . ramimaculatus and D . ciliiferus . Collectively , these details of insect morphology likely enhanced the similarity of the insect with a co-occurring lichen , providing a reasonable inference that the forewing is mimetic with the lichen thallus . It is generally known that lichens are stable , symbiotic associations of fungi and algae ( Lücking and Nelsen , 2018 ) . At the same time , lichens are regarded as pioneers in the colonization of novel surfaces such as bark , rock and soil , which dominate about 7% of the earth's terrestrial surface ( Larson , 1987 ) , and have a distribution from the polar regions to the tropics ( Lumbsch and Rikkinen , 2017 ) . They are prominent in arctic-alpine vegetation types in wet and higher montane forests ( Lumbsch and Rikkinen , 2017 ) . Many extant foliose or fruticose lichens such as taxa of Parmeliaceae are known to be epiphytic or corticolous , and bark surfaces are one of the most common substrates ( Lumbsch and Rikkinen , 2017 ) . Daohugouthallus ciliiferus is considered an epiphytic foliose to subfruticose lichen , and often is found entangled with gymnosperm seed cones ( Figures 1C and 4D; Wang et al . , 2010a ) . When Lichenipolystoechotes moth lacewings reposed in a habitat rich in D . ciliiferus , a near perfect match of their appearances would assist their concealment . Among extant Neuroptera , similar appearances of lichen-camouflage or related cases have been recorded in some larvae of green lacewings that carry packets of lichen material on their backs to hide themselves ( Slocum and Lawrey , 1976; Wilson and Methven , 1997 ) . Although Lichenipolystoechotes probably lacked the same life-habit as modern lichen-carrying chrysopoid larvae , the Jurassic taxa likely acquired a similar survival advantage when they occupied a lichen-rich habitat . Some extant Thyridosmylus species of Osmylidae , another archaic lineage of Neuroptera , possess similar complex wing markings and often occur on moss-laden surfaces of rocks , tree bark and indurated soil surfaces ( Winterton et al . , 2017; Figure 2B ) , which exhibit an impressive consistency with their surroundings ( pers . observ . by Yongjie Wang ) . Although Lichenipolystoechotes is a member of Ithonidae , phylogenetically distant to Osmylidae , we infer that their concealment strategy of mimicking cryptogam plants in certain habitats has a deep geochronologic history among ancient lacewing lineages . Unlike the models of other , co-occurring , plant-mimicking insects , lichen-mimesis of Lichenipolystoechotes appears highly specialized ( Figure 5 ) . Modern lichens can produce a variety of lichenic acids ( Gerson , 1973 ) that are unpalatable to many insects and enhance the protective sheltering for animals . Consequently , lichens and lichen-tolerant animals , such as lichen feeding insects and mites , constitute a unique micro-ecosystem . We hypothesize that such a micro-ecosystem existed 165 million-years-ago in Northeastern China that accommodated these trophic , sheltering , defensive and mimetic interactions . Although lichen mimesis is not well documented among extant insects , the most iconic such case of lichen and insect resemblance is the industrial melanism of the peppered moth Biston betularia in nineteenth century Britain ( Gerson , 1973; Stevens , 2011 ) . The Industrial Revolution caused elevated levels of soot laden air pollution that resulted in disappearance of lichen shelters for the light-colored morph of B . betularia , as their corresponding habitation sites were changed from lightly tinged to dark-hued lichen surfaces that led to their greater vulnerability to predation . This change resulted in the abrupt increase of the dark colored morph of B . betularia . When lightly hued lichens returned after aerial pollution was thwarted , B . betularia again became dominant as the lightly colored morph . The industrial melanism of B . betularia was believed as a textbook example of Darwinian evolution in action , though it was questioned by some authors ( Sargent , 1968; Sargent , 1969; Coyne , 1998; Cook and Saccheri , 2013 ) . Nevertheless , other studies demonstrated that selection pressures such as predation by birds genuinely affected the differential survival of the pale and dark colored morphs of B . betularia under differently hued backgrounds ( Howlett and Majerus , 1987; Liebert and Brakefield , 1987; Majerus , 2009; Walton and Stevens , 2018 ) . It is possible that the Jurassic Lichenipolystoechotes could have gained survival advantage from mimesis of a lichen similar to that of modern B . betularia–lichen mimesis . Specifically , if lichen models were present in the habitat occupied by Lichenipolystoechotes , survival of the mimic would be assured . It is noteworthy that the winged adults of Lichenipolystoechotes would not have been always in the shelter of a lichen model; however , when they were , the conditions of mating , laying of eggs and dispersal would be paramount for survival . If so , high-contrast lichen-like markings could contribute to concealment of the insects . Alternatively , such high-contrast markings of Lichenipolystoechotes species also can be interpreted as disruptive coloration , which would confuse the boundaries of moth lacewing and lichen to prevent the detection of a body part essential for survival ( Stevens , 2011 ) . Consequently , the lichen-like markings of Lichenipolystoechotes could likely bring the double protections to the insects-background mimicry and disruptive coloration . Was there possible benefit to D . ciliiferus from its mimetic association with Lichenipolystoechotes ? This is an open question that could raise multiple alternative explanations . Some modern insects such as ants , dipterans and larva of green lacewings are considered to potentially contribute to dispersal of lichens by transporting lichen propagules to new sites of colonization ( Gerson , 1973; Keller and Scheidegger , 2016; Ronnås et al . , 2017 ) . In a comparison with such relatively small , lichen-carrying insects , Lichenipolystoechotes possessed a considerably larger body size that likely was convenient for dispersal of lichen propagules . Notably , sexual reproductive organs such as apothecia have not been found on the D . ciliiferus thallus based on light-microscopic morphological and SEM anatomical observations; neither were vegetative propagules such as soredia or isidia seen except along marginal lobules that occasionally were present . This hypothesis of zoochory requires additional evidence for support . However , our alternative hypothesis of benefiting D . ciliiferus is based on trophic interactions . As predaceous insects , Lichenipolystoechotes inhabited a lichen-rich environment to evade their predators , but they also could have predated and consumed smaller lichen-feeding animals while simultaneously decreasing herbivore damage to the D . ciliiferus thallus . This latter hypothesis would require additional verification from evidence of a small ecological web of predator , shelter , defensive and mimetic interactions associated with Daohugouthallus and Lichenipolystoechotes in the same deposit . The accepted oldest lichen fossil was reported from the Early Devonian and lichens have existed minimally for 410 million years ( Taylor et al . , 1995; Honegger et al . , 2013; Lücking and Nelsen , 2018 ) , as have the apterygote insects ( Misof et al . , 2014 ) . Both archaic Devonian lineages have evolved more derived , diverse clades of lichens and pterygote insects resulting in a myriad of associations among their modern lineages ( Figure 6 ) . Although there is virtually no evidence to suggest when and how such association began; in this report , we describe the oldest examples of lichen mimesis that involved two lacewing species resembling a contemporaneous lichen from the same , latest Middle Jurassic deposit . These insect lineages have acquired mimicry association with lichens in less than half of the time ( 40% ) of the duration of both major lineages since the early Devonian ( Figure 6 ) . This new finding documents a unique survival strategy among mid-Mesozoic Neuroptera , and others await discovery . Specimens were collected from the Daohugou 1 locality of the Jiulongshan Formation , near Daohugou Village , Ningcheng County , approximately 80 km south of Chifeng City , in the Inner Mongolia Autonomous Region , China ( 119°14 . 318′E , 41°18 . 979′N ) . The age of this formation is 168–152 Ma based on 40Ar/39Ar and 206Pb/238U isotopic analyses ( Hy et al . , 2004; Liu , 2006; Ren , 2019 ) . CNU-NEU-NN2016040P/C and CNU-NEU-NN2016041 of Lichenipolystoechotes angustimaculatus sp . nov . , and CNU-NEU-NN2019004P/C of Lichenipolystoechotes ramimaculatus sp . nov . are housed in the College of Life Sciences and Academy for Multidisciplinary Studies , Capital Normal University ( CNU ) , Beijing , China . Lichen specimens of Daohugouthallus ciliiferus Wang , Krings et Taylor , 2010: PB23120 is housed in the paleobotanical collection of the Nanjing Institute of Geology and Palaeontology , Chinese Academy of Sciences , in Nanjing , China; B0474 and B0476P/C are housed in the Institute of Vertebrate Paleontology and Paleoanthropology , Chinese Academy of Sciences , in Beijing , China; CNU-LICHEN-NN2019001 and CNU-LICHEN-NN2019002P/C are housed in the Key Lab of Insect Evolution and Environmental Changes , College of Life Sciences and Academy for Multidisciplinary Studies , Capital Normal University , in Beijing , China . The insect and lichen fossils were examined and photographed using a Nikon SMZ25 microscope attached to a Nikon DS-Ri2 digital camera system at the Key Lab of Insect Evolution and Environmental Changes at Capital Normal University in Beijing , China . Lichen compression specimens from the Daohugou one locality were soaked in water for several seconds , dried on filter paper , and then a fragment was lifted up by the edge of a razor blade . One isolated , dried slice was examined and photographed using a Zeiss Axioscope2 compound microscope attached to a Nikon D5100 digital camera system at the State Key Laboratory of Mycology , Institute of Microbiology , at the Chinese Academy of Sciences in Beijing . That piece of lichen fossil then was sputter-coated with gold particles using an Ion Sputter E-1045 ( HITACHI ) , and SEM images were recorded using a scanning electron microscope ( Hitachi SU8010 ) with a secondary electron detector operated at 5 . 0 kV . Overlay drawings were prepared by Corel DRAW . Box plots were made with Origin 2018 software , which is used to display the distribution of the data of branch width of L . ramimaculatus’s forewing pattern and lichen thallus of D . ciliiferus . The box plots are formed by two quartiles showing the high frequency of values , and the upper and lower points of the boxes are the maximum and minimum values . All figures were composited in Adobe Photoshop .	Many insects mimic other organisms or use camouflage to hide from predators . For example , some modern animals mimic the organism lichens , which are formed from algae and fungus , and grow almost everywhere on Earth , from the Arctic to the desert . The most iconic example of an insect mimicking a species of lichen is the peppered moth . During the industrial revolution , darker colored moths were better at surviving . But when the revolution ended and pollution levels declined , species of lichen began to re-emerge and increase the survival of paler moths . Yet , it is unclear how and when insects first evolved this ingenious survival strategy , as distinctive examples of insects mimicking lichens are missing from fossil records . To answer this question , Fang et al . set out to find fossils of lichen-mimicking insects that occurred at the same time as fossils of lichens . This approach led to the discovery of two new species of lacewing insects and their related species of foliose lichen . Previous work suggested that the foliose lichen , which has a lobe like shape , did not exist more than 65 million years ago . However , the findings of Fang et al . indicate that the foliose lichen existed 165 million years ago during the age of dinosaurs , and therefore arose much earlier than previously thought . The two new species found in north-eastern China , form a new subgroup within the moth lacewing family that Fang et al . have named ‘Lichenipolystoechotes’ . Close examination of both species of lacewing and the lichen under the microscopy revealed a near perfect match in their appearance . The branching patterns of the insects’ wing markings fit the branching patterns of the lichen . Taken together , these findings suggest that , not only did lichen mimics exist in the age of the dinosaurs , but that this strategy of using lichen mimicry as a form of survival was already very effective during this time period . This discovery suggests that , 165 million years ago , a micro-ecosystem of lichens and insects existed in north-eastern China . It invites new questions about how that ecosystem worked . For example , how did the lichen benefit from its relationship with lacewing insects ? Further investigations could reveal the answers and uncover more interesting insects hidden in the fossil record .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "ecology", "evolutionary", "biology" ]	2020	Lichen mimesis in mid-Mesozoic lacewings
Cell-intrinsic and extrinsic signals regulate the state and fate of stem and progenitor cells . Recent advances in metabolomics illustrate that various metabolic pathways are also important in regulating stem cell fate . However , our understanding of the metabolic control of the state and fate of progenitor cells is in its infancy . Using Drosophila hematopoietic organ: lymph gland , we demonstrate that Fatty Acid Oxidation ( FAO ) is essential for the differentiation of blood cell progenitors . In the absence of FAO , the progenitors are unable to differentiate and exhibit altered histone acetylation . Interestingly , acetate supplementation rescues both histone acetylation and the differentiation defects . We further show that the CPT1/whd ( withered ) , the rate-limiting enzyme of FAO , is transcriptionally regulated by Jun-Kinase ( JNK ) , which has been previously implicated in progenitor differentiation . Our study thus reveals how the cellular signaling machinery integrates with the metabolic cue to facilitate the differentiation program . Recent studies have highlighted how metabolism regulates the state and fate of stem cells ( Ito and Suda , 2014; Shyh-Chang et al . , 2013; Shyh-Chang and Ng , 2017 ) . Besides catering to the bioenergetic demands of a cell , metabolic intermediates can also alter the fate of stem cells via epigenetic mechanisms like histone modifications ( Atlasi and Stunnenberg , 2017; Saraiva et al . , 2010 ) . Studies on diverse stem cell scenarios , primarily in Hematopoietic Stem Cells ( HSCs ) , have established that at various developmental stages , stem cells have different metabolic requirements ( Kohli and Passegué , 2014; Suda et al . , 2011 ) . Nevertheless , the metabolic demand of progenitors , the immediate descendants of stem cells , is yet to be fully elucidated . Studies to date have evidenced that glucose metabolism impacts the onset and magnitude of HSC induction ( Harris et al . , 2013; Shyh-Chang and Ng , 2017 ) , as well as HSC specification ( Oburoglu et al . , 2014 ) . Another major metabolic state that is active in stem and progenitor cells is fatty acid oxidation ( FAO ) ( Ito et al . , 2012; Knobloch et al . , 2017; Lin et al . , 2018; Wong et al . , 2017; Ito et al . , 2012 ) . Enzymatic activities of the members of FAO lead to the shortening of fatty acids and the production of acetyl-CoA in mitochondria . The acetyl-CoA thus produced can not only generate NADH and FADH2 through TCA cycle but also can be utilized in acetylation of various proteins including histones ( Fan et al . , 2015; Houten and Wanders , 2010; McDonnell et al . , 2016; Wong et al . , 2017 ) . The primary goal of this study was to ascertain whether FAO regulates any aspect of the hemocyte progenitors in the Drosophila larval hematopoietic organ , lymph gland . The lymph gland is a multilobed structure consisting of a well-characterized anterior lobe ( primary lobe ) and uncharacterized posterior lobes ( Figure 1A , Banerjee et al . , 2019 ) . The core of the primary lobe houses the progenitor populations and is referred to as the medullary zone ( MZ ) , while the differentiated cells define the outer cortical zone ( CZ , Figure 1A' ) . In between these two zones , lies a rim of differentiating progenitors or intermediate progenitors ( IPs ) . The blood progenitors of late larval lymph gland are arrested in G2-M phase of cell cycle ( Sharma et al . , 2019 ) , have high levels of ROS ( Owusu-Ansah and Banerjee , 2009 ) , lack differentiation markers , are multipotent ( Jung et al . , 2005 ) and are maintained by the hematopoietic niche/posterior signaling center , PSC ( Krzemień et al . , 2007; Lebestky et al . , 2003; Mandal et al . , 2007 ) . The primary lobe has been extensively used to understand intercellular communication relevant to progenitor maintenance ( Gao et al . , 2013; Giordani et al . , 2016; Gold and Brückner , 2014; Hao and Jin , 2017; Krzemień et al . , 2007; Krzemien et al . , 2010; Lebestky et al . , 2003; Mandal et al . , 2007; Mondal et al . , 2011; Morin-Poulard et al . , 2016; Sinenko et al . , 2009; Small et al . , 2014; Yu et al . , 2018 ) . Although these studies have contributed significantly toward our understanding of cellular signaling relevant for progenitor homeostasis , the role of cellular metabolism in regulating the state and fate of blood progenitors remains to be addressed . Here , we show that the G2-M arrested hemocyte progenitors of the Drosophila larval lymph gland rely on FAO for their differentiation . While the loss of FAO prevents their differentiation , upregulation of FAO in hemocyte progenitors by either genetic or pharmacological means leads to precocious differentiation . More importantly , acetate supplementation restores the histone acetylation and differentiation defects of the progenitor cells observed upon loss of FAO . Our genetic and molecular analyses reveal that FAO acts downstream to the Reactive Oxygen Species ( ROS ) and c-Jun N-terminal Kinase ( JNK ) axis , which is essential for triggering the differentiation of these progenitors ( Owusu-Ansah and Banerjee , 2009 ) . In this study , we , therefore , provide the unknown link that connects cellular signaling and metabolic circuitry essential for differentiation of the blood progenitors . Drosophila hemocyte progenitors in the lymph gland proliferate in the early larval stages ( Jung et al . , 2005; Mondal et al . , 2011 ) . Eventually , they undergo a G2-M arrest in late third instar ( Sharma et al . , 2019 ) . Studies have identified that lymph gland hemocyte progenitors of the primary lobe can be grouped into three subpopulations: Dome- pre-progenitors , Dome+ progenitors , and Dome+ Pxn+ Hml+ Intermediate progenitors ( Banerjee et al . , 2019 ) . These progenitor subpopulations will be henceforth referred to as pre-progenitors , progenitors , and IPs , respectively . The pre-progenitors can also be visualized by Pvf2 expression in first , second , and early third instar larval lymph gland ( Ferguson and Martinez-Agosto , 2017 , Figure 1—figure supplement 1A–B ) . Since in the late third instar lymph gland , only progenitors and IPs are present ( Figure 1A' ) , we analyzed the early third instar lymph gland to characterize the pre-progenitors . To ascertain the involvement of FAO , if any , in hemocyte progenitors of larval lymph gland , we monitored the expression of Hepatocyte Nuclear Factor 4 ( Hnf4 ) ( Palanker et al . , 2009 ) , an essential gene of larval FAO and lipid mobilization . As evident from Figure 1B–D , G2–M arrested progenitors ( visualized by dome-MESO-EBFP2+ ) express high levels of Hnf4 > GFP compared to pre-progenitors ( visualized by Pvf2 expression ) and IPs ( Dome+ Hml+ ) . Interestingly , neutral lipids ( visualized by Nile red staining: Figure 1E–G ) are conspicuous in late third instar blood progenitors ( dome-MESO-EBFP2+ ) , compared to the pre-progenitors ( Figure 1E–E'' and G ) and IPs ( Figure 1F–F'' and G ) . The lipid enrichment in the late progenitors is further evident upon LipidTOX ( validated marker for neutral lipids ) labeling ( Figure 1—figure supplement 2A–C ) . Based on the presence of relatively high levels of lipid droplets in a non-lipid storage tissue and the expression of Hnf4 in the progenitor cells , we speculated a developmental role of FAO in these cells . This prompted us to check for the expression of other genes involved in FAO ( Palanker et al . , 2009 , Figure 1H ) . Figure 1I shows the expression of the rate-limiting enzyme of FΑO , withered ( whd , Drosophila homolog of CPT1: Carnitine palmitoyltransferase 1 ) along with Mcad ( medium-chain acyl-CoA dehydrogenase ) , Mtpα ( mitochondrial trifunctional protein α subunit: Long-chain-3-hydroxyacyl-CoA dehydrogenase ) , scully ( 3-hydroxyacyl-CoA dehydrogenase ) , Mtpβ ( mitochondrial trifunctional protein β subunit: Long-chain-3-hydroxyacyl-CoA dehydrogenase ) and yip2 ( yippee interacting protein 2: acetyl-CoA acyltransferase ) in the late lymph gland . Elevated levels of expression of acyl-Coenzyme A dehydrogenase ( CG3902 ) is also seen in the progenitors as compared to the pre-progenitors and IPs ( Figure 1—figure supplement 2D–F ) . Major aspects of FAO takes place in the mitochondria where fat moiety is broken down to generate acetyl-CoA , NADH , and FADH2 ( Bartlett and Eaton , 2004 ) , we , therefore , looked at the status of mitochondria in the progenitors as well as the differentiated hemocytes . The presence of an abundant reticular network of mitochondria is evident in the progenitors ( dome-GAL4 >UAS-mito-HA-GFP , Figure 1—figure supplement 1C–C' , and Video 1 ) . However , for reasons unknown to us HmlΔ-GAL4 is unable to drive UAS-mito-HA-GFP , therefore , we used streptavidin labeling to visualize the mitochondrial status in the differentiated hemocytes . Interestingly , differentiated hemocytes ( HmlΔ-GAL4 > UAS-GFP ) show less reticular mitochondria ( labeled by Strepatvidin-Cy3 , Chowdhary et al . , 2017; Hollinshead et al . , 1997 , Figure 1—figure supplement 1D–D' ) in comparison to the progenitors ( Hml>GFP negative ) thereby indicating a preference for FAO in the hemocyte progenitors . Put together; the above results implicate FAO as the metabolic state of the Dome+ cells of the primary lobe of the lymph gland . These observations , in turn , encouraged us to investigate the importance of FAO in maintaining the state and fate of these progenitors during development . Drosophila ortholog of CPT1 , withered ( whd , Figure 2A ) , is a rate-limiting enzyme for FAO ( Strub et al . , 2008 ) . The loss of function of CPT1/whd , therefore , blocks mitochondrial FAO ( Schreurs et al . , 2010 ) . To investigate the role of FAO in late hemocyte progenitors , we employed a null allele of withered , whd1 ( Strub et al . , 2008 ) . The primary lobes of whd1 homozygous lymph gland have abundant Dome+ progenitors but drastically reduced number of differentiated hemocytes ( P1: plasmatocytes Figure 2B–C and D; proPO: Crystal cells , Figure 2E–F and G ) , and Intermediate progenitors ( Dome+ Pxn+ , Figure 2H–I'' and Figure 2—figure supplement 1A ) compared to control . Detailed temporal analysis of the dynamics of progenitor subpopulation during normal development , as well as upon loss of whd1 ( Figure 2J–S ) , was next carried out . In sync with an earlier report ( Ferguson and Martinez-Agosto , 2017 ) , our analysis reveals that Dome- pre-progenitors ( Figure 2K ) are present in the developing lymph gland until the early third instar ( Figure 2L and Figure 2—figure supplement 1B ) . Beyond this timeline , the subsets that populates the lymph gland are Dome+ progenitors ( Figure 2M–N and Figure 2—figure supplement 1B ) , and Dome+ Pxn+ IP cells ( Figure 2M–N and Figure 2—figure supplement 1B ) . Interestingly , in whd1 mutant lymph glands , while the pre-progenitors are present till the early third instar stage ( Figure 2P and Figure 2—figure supplement 1C ) , there is an abundance of progenitors ( Figure 2Q–S ) with a small number of IP cells ( Figure 2R–S ) . Quantification of the above results reflects that in whd1 mutant the progenitors are rather stalled instead of undergoing a natural transition to IP cells with time ( Figure 2—figure supplement 1C ) . During normal development , the first sign of differentiation ( as evidenced by Pxn expression ) occurs around 36 hr AEH , and by 96 hr AEH , there is a prominent cortical zone defined by the differentiating cells . In contrast , the cortical zone is drastically reduced due to lack of differentiation in the whd1 mutant lymph gland . The timed analysis also revealed that the defect seen in differentiation in this mutant has an early onset ( 36 hr AEH , Compare Figure 2O with Figure 2J ) . These observations implicate that the lack of FAO dampens the differentiation process of Dome+ progenitors . Since the differentiation of progenitors is affected , we next performed an RNAi-mediated downregulation of whd by the TARGET system ( McGuire et al . , 2004; Figure 2—figure supplement 1D ) . Progenitor-specific downregulation ( dome-GAL4 , UAS-GFP; tubGAL80ts20; UAS-whd RNAi ) results in a halt in differentiation , as evidenced by an increase in the area of dome > GFP and a concomitant decline in adjoining CZ ( visualized by differentiated plasmatocyte ( Nimrod: P1 , Figure 2U ) compared to control ( Figure 2T ) . Upon activation of whd RNAi from a different source ( VDRC ) by another independent progenitor specific driver TepIV-GAL4 ( Figure 2—figure supplement 1E–G ) , a similar result is obtained . Additionally , progenitor-specific knockout of whd by CRISPR/Cas9 system ( Hsu et al . , 2014 ) supports the above results ( Figure 2V and X ) . Likewise , knockdown of the key player in lipid metabolism Hnf4 results in a decline in the differentiation of the progenitors , endorsing the role of FAO in progenitor differentiation ( Figure 2W and X , Figure 2—figure supplement 2 ) . Lymph glands from a hetero-allelic combination of dHNF4 ( dHNF4Δ17/dHNF4Δ33: null allele of Hnf4 ) ( Palanker et al . , 2009 ) , exhibits an abundant progenitor pool ( Shg: DE-Cadherin ) coupled with the reduction in differentiated cells ( Pxn , compare Figure 2—figure supplement 1I with 1H , and Figure 2—figure supplement 1M ) , further denoting that FAO disruption indeed leads to compromised differentiation of progenitors . To further verify our observations , we next analyzed the homozygous mutant of two essential enzymes of β−oxidation: Mtpα ( mitochondrial trifunctional protein α subunit: Long-chain-3-hydroxyacyl-CoA dehydrogenase ) , and Mtpβ ( mitochondrial trifunctional protein β subunit: Long-chain-3-hydroxyacyl-CoA dehydrogenase ) ( Kishita et al . , 2012 ) . Primary lobes from homozygous Mtpα[KO] and Mtpβ[KO] loss of function has a large progenitor pool ( Shg: DE-Cadherin ) at the expense of differentiated cells ( Pxn ) ( compare Figure 2—figure supplement 1J–K with H ) , a phenotype similar to the whd1 ( Figure 2—figure supplement 1L and M ) . DE-cadherin enrichment in the FAO loss of function progenitors is also indicative of high-maintenance signals ( Gao et al . , 2014 ) . In addition to genetic knockdown and classical loss-of-function analyses , we performed pharmacological inhibition of FAO in Drosophila larvae ( Figure 2—figure supplement 1N–Q ) . Larvae grown in food supplemented with FAO inhibitors , Etomoxir ( Lopaschuk et al . , 1988 ) , and Mildronate also demonstrate a more than two-fold reduction in the differentiation of the progenitors of the primary lobe . Collectively , our genetic and pharmacological studies illustrate the cell-autonomous role of FAO in the differentiation of blood progenitors of the lymph gland . Previous study from our laboratory demonstrated that G2-M arrest is a hallmark of the otherwise proliferating progenitors prior to differentiation ( Sharma et al . , 2019 ) . In contrast , we found that whd1 homozygous progenitors exhibit a stark increase in EdU incorporation implicating their highly proliferating status upon loss of FAO when compared to age-isogenised controls ( Figure 3A–C ) . To have a further insight into the cell cycle status , we expressed Fly-FUCCI-fluorescent ubiquitination-based cell cycle indicator , specifically in the progenitors . This indicator employs two fluorescent probes: the first probe is an E2F moiety fused to GFP , which is degraded by Cdt2 during the S-phase . This construct allows the visualization of G1 , G2 , and M-phase cells by GFP expression . The second probe is a fusion of Red Fluorescent Protein ( mRFP . nls ) and CycB moiety . It is degraded by the anaphase-promoting complex/cyclosome ( APC/C ) during midmitosis , thereby reporting the cell in S , G2 , and M-phase . Together this system allows the visualization of cells in G1 , S , and G2/early mitosis by green , red , and yellow signals , respectively ( Zielke et al . , 2014 ) . We found that , instead of being in a G2-M arrest , the hemocyte progenitors in whd mutant are actively proliferating , as evidenced by more cells in S-phase ( red ) compared to the control ( Figure 3D–D' with E-E' ) . Quantitative analyses reveal more than three-fold increase in the number of cells in the S phase with a concomitant drop in G2-M arrested progenitors ( compare Figure 3D' with E' and F ) . Together these results assert that FAO disruption in proliferating progenitors doesn’t allow them to halt at G2-M and subsequently differentiate . The perturbation of differentiation prompted us to look at the status of both differentiation and maintenance factors , per se in these hemocyte progenitors . Although the progenitor pool in the larval lymph gland is heterogenous ( Baldeosingh et al . , 2018; Banerjee et al . , 2019; Sharma et al . , 2019 ) , our timed analysis indicates that the majority of the progenitors populating the late third instar lymph gland expresses dome . Hedgehog signaling has been implicated in the maintenance of the dome expressing progenitors ( Baldeosingh et al . , 2018; Mandal et al . , 2007; Sharma et al . , 2019; Tokusumi et al . , 2010 ) . Hematopoietic niche/PSC releases Hh , which leads to the expression of the Hh signal transducer Cubitus interruptus ( Ci155[Alexandre et al . , 1996] ) in the progenitors . The homozygous whd1 lymph gland progenitors express a higher level of Ci155 compared to control ( compare Figure 3—figure supplement 1B-B'' with A-A'' and quantitated in Figure 3—figure supplement 1C ) correlating with higher proliferation and less differentiation ( Figure 2B–D ) . This observation , along with enrichment of DE-cadherin in FAO mutants ( Figure 2—figure supplement 1H–M ) , endorses high-maintenance signal in the progenitors . Reactive oxygen species ( ROS ) is the major signal attributed to the differentiation of the hemocyte progenitors of the lymph gland ( Owusu-Ansah and Banerjee , 2009 ) . High levels of developmentally generated ROS trigger Jun Kinase ( JNK ) signaling , which sets these progenitors toward the differentiation program ( Owusu-Ansah and Banerjee , 2009 ) . Since homozygous whd1 progenitors fail to differentiate , we rationalized that this might be due to the drop in the differentiation signal ROS . To probe this possibility , we analyzed the levels of ROS in whd1 lymph glands by dihydroxy ethidium ( DHE ) staining . Quite strikingly , whd1 homozygous progenitors exhibited elevated levels of ROS ( Compare Figure 3—figure supplement 1D–D'' with 3E-E'' and quantified in Figure 3—figure supplement 1F ) . A similar observation of increased ROS in CPT1 knockdown endothelial cells has been reported in another study ( Kalucka et al . , 2018 ) . Therefore , we can infer that the halt in differentiation observed in the Ci155 enriched progenitors of homozygous whd1 is not an outcome of compromised ROS level . Collectively , these results indicate that despite achieving a high ROS level than the control , the progenitors are unable to move into differentiation , indicating that FAO might act downstream to ROS . Due to the central role of carnitine in fat metabolism , it is often used as a supplement for enhancing fat oxidation ( Pekala et al . , 2011; Wall et al . , 2011 ) . Several studies have concluded that FAO can be upregulated in the cells by L-carnitine supplementation ( Pekala et al . , 2011; Sahlin , 2011; Wall et al . , 2011 ) . Our loss-of-function genetic analyses illustrated that FAO is essential for the differentiation of hemocyte progenitors . Whether it is sufficient for the progenitor differentiation was addressed by feeding L-carnitine supplemented food ( 100 mM concentration for 48 hr ) to wandering third instar larvae . Compared to control ( Figure 4A ) , the lymph gland from L-carnitine fed larvae ( 96 hr AEH ) exhibits a drastic reduction in progenitor zone ( visualized by dome > GFP ) with a concomitant increase in differentiation ( visualized by P1: Nimrod; Figure 4B–C ) . The increase in differentiation by L-carnitine supplementation is also apparent in whd1 heterozygous mutants . Feeding L-carnitine could rescue the differentitaion defects associated with whd1 heterozygous mutants ( Compare Figure 4—figure supplement 1A–E ) . As a genetic correlate , whd was overexpressed by a Cas9-based transcriptional activator ( BDSC68139 ) ( Ewen-Campen et al . , 2017 ) in the hemocyte progenitors following the scheme in Figure 2—figure supplement 1D . Overexpression of whd indeed results in an increase in the differentiation of hemocyte progenitors ( compare Figure 4D with Figure 4E–F ) . It is interesting to note that the otherwise undifferentiated reserve progenitors of the secondary lobes also differentiate upon whd overexpression ( marked by a dotted white line in Figure 4G–H ) . We next employed dual fly-FUCCI construct , and EdU labeling to assay the cell cycle cell status of the FAO upregulated progenitors . The early third instar lymph glands from L-carnitine fed larvae exhibit a radical decline in EdU incorporation compared to the proliferating progenitors of the control larvae of similar age ( compare Figure 4I–I' with Figure 4J-J' and quantified in Figure 4K ) . As a genetic correlate , we overexpressed whd in the progenitors , which also led to a decline in EdU incorporation ( Figure 4L–N ) . Our FUCCI analysis reveals an abundance of G2-M progenitors in the early third instar larvae reared in L-carnitine supplemented food compared to control samples . Thus , less EdU incorporation due to upregulated FAO resulted in an early onset of G2-M arrest in the progenitors ( compare Figure 4O–O' with Figure 4P–P' and quantified in Figure 4Q ) . Put together; our results reveal that upon FAO upregulation , the progenitor experiences a precocious G2-M halt in their cell cycle . During normal development , the late progenitors also undergo a G2-M halt before they differentiate . We , thus , infer that FAO is imperative for the differentiation of lymph gland progenitors . Next , we investigated whether a compromise in the intracellular energy source ( ATP ) is the reason behind the differentiation defect seen in the FAO mutants . Quite intriguingly , ATP levels of homozygous whd1 larvae are comparable to similarly aged control ( Figure 5A ) . This unaltered ATP level is in sync with our observation of higher proliferation observed in homozygous whd1 hemocyte progenitors ( Figure 3B–B' ) . Since higher proliferation is driven by elevated glycolysis in different scenarios ( Lunt and Vander Heiden , 2011 ) , we hypothesized that loss of FAO might push the progenitors towards a higher glycolytic index . We performed an in vivo glucose uptake assay employing fluorescent derivative of glucose , 2-NBDG ( Zou et al . , 2005 ) in the late third instar lymph gland of both control and FAO mutant . In control , late third instar Dome+ progenitors exhibit low glucose uptake ( Figure 5B–B'' and Figure 5D ) compared to the higher uptake detectable in the peripheral hemocyte population marked by Hml ( Figure 5—figure supplement 1A–A''' and quantified in Figure 5—figure supplement 1B ) . In sharp contrast , higher glucose uptake is evident in the FAO mutant progenitors ( Figure 5C–C'' and quantified in Figure 5D ) . In concordance with the above result , in vivo lactate dehydrogenase assay ( Abu-Shumays and Fristrom , 1997 ) also revealed a high glycolytic index prevalent in the lymph glands of homozygous whd1 ( Figure 5E–F ) . Moreover , in the whd1 lymph gland , the transcript levels of HexA ( Hexokinase A ) and Pfk ( Phosphofructokinase ) , the enzymes involved in the two irreversible steps of glycolysis: exhibit a 1 . 6 fold and 1 . 7 fold increase in their expression , respectively ( Figure 5G–H ) . Together these results reveal that upon FAO disruption , lymph gland progenitors adopt high-glucose utilization/metabolism . Based on the above observations , we inferred that higher proliferation and differentiation defects observed in hemocyte progenitors with compromised FAO might be due to the surge in glucose uptake/metabolism . Upon rearing whd1 homozygous larvae in food supplemented with glycolysis inhibitor 2-Deoxy-D-glucose ( 2-DG ) , the otherwise hyper-proliferating hemocyte progenitors ( EdU , Figure 5J ) demonstrate a significant drop in EdU incorporation to a level that is comparable to control ( compare Figure 5K with Figure 5I , and quantified in Figure 5M ) . Although the glycolytic block by feeding 2-DG rescues the cell cycle status , the abrogated differentiation observed in homozygous whd1 hemocyte progenitors is not restored ( compare Figure 5P with Figure 5O and N and quantified in Figure 5R ) . Inhibition of glucose uptake by genetic perturbation of glucose transporter Glut1 in progenitor specific manner in the FAO mutant also endorses the above result ( Figure 5L–L' and M , and Figure 5Q–R ) . From these observations , it is evident that the surge in glucose metabolism encountered by the hemocyte progenitors upon FAO loss is responsible for their altered cell cycle . However , the glycolytic surge upon FAO disruption is unable to initiate progenitor differentiation . Collectively , the above results indicate that FAO plays a critical role in regulating the differentiation of hemocyte progenitors . Acetyl-CoA generated from FAO , apart from serving as a substrate for the Krebs cycle , is essential for the acetylation of various proteins , including histones . We , therefore , wondered that disruption of FAO in whd1 hemocyte progenitors might also result in altered histone acetylations , which may , in turn , result in cell cycle and differentiation defects . Histone acetylation mediated by Histone Acetyl Transferases ( HATs ) directly controls the expression of differentiation factor , thereby regulating germline stem cell differentiation ( McCarthy et al . , 2018; Xin et al . , 2013 ) . To ascertain whether HATs play a similar role in hemocyte progenitor differentiation , progenitor-specific RNAi-mediated knockdown of HATs function was done following the timeline , as shown in Figure 2—figure supplement 1D . Quite strikingly , loss of Histone Acetyl Transferase ( HAT ) genes , Gcn5 ( Carré et al . , 2005 ) and chm ( chameau ) ( Grienenberger et al . , 2002; Miotto et al . , 2006 ) phenocopy the differentiation defect seen in the hemocyte progenitors of FAO loss of function ( Figure 6A–C ) . Additionally , downregulation of Acetyl Coenzyme A synthase/AcCoAS ( the Drosophila orthologue of ACSS2 ) ( Mews et al . , 2017 ) results in a phenotype identical to HAT or FAO loss ( Figure 6D ) confirming the essential role of acetylation in hemocyte progenitor differentiation . Next , we performed an epistatic interaction of whd1 allele with other major acetyl-CoA related genes . Citrate transporter SLC25A1 or scheggia/sea in Drosophila ( Carrisi et al . , 2008 ) exports Krebs cycle metabolite citrate from the mitochondria to the cytoplasm to generate acetyl-CoA . In the cytosol , citrate gets converted to acetyl-CoA by ATP citrate lyase ( ACLY encoded by Drosophila orthologue , ATPCL ) . Trans-heterozygous loss-of-function allelic combinations of whd1 with either ATPCL01466 ( Figure 6F ) or seaEP3364 ( Figure 6G ) phenocopy differentiation defects seen in whd1 homozygous lymph gland ( Figure 6E ) . Above set of genetic correlations reveal that alteration in histone acetylation affects hemocyte progenitor differentiation ( quantified in Figure 6H ) . Knockdown of the rate-limiting enzyme of FAO:CPT1 , leads to a reduced level of H3K9 acetylation in lymphatic endothelial cells ( Wong et al . , 2017 ) . We wondered whether similar acetylation defect occurs in whd1 homozygous mutant larvae . Immunoblot analysis of extracted histones with antibody against acetylated anti-H3K9 reveals that whd1 homozygous larvae have low levels of H3K9 acetylation ( Figure 6I–J ) when compared to control . However , the level of histone H3 is comparable to that of control . Whether the tissue of our interest also reflects this decline in acetylation level , H3K9 acetylation labeling was performed in mosaic clones using hsFlp/Ay-GAL4 mediating RNAi knockdown of Drosophila CPT1 orthologue whd . The clonal patches positively marked with GFP ( where whd has been downregulated ) show a significant drop in H3K9 acetylation levels ( Figure 6M–N ) compared to surrounding hemocyte progenitors . However , histone H3 labeling in both mutant and control clonal patches are comparable ( Figure 6K–L ) and serve as a control . This observation , along with the western blot analyses , reveals the occurrence of H3K9 acetylation defects in FAO loss of function . Likewise , histone H4 acetylation visualized by pan anti-H4 acetylation antibody reveals a drastic drop in whd knockdown clonal patches ( Figure 6O–P ) . Both the expression of H3K9 and pan H4 acetylation remains unaltered in mock/wild type clones ( Figure 6—figure supplement 1A–D ) . Further , upon progenitor specific downregulation of whd function , a decline in the level of both H3K9 acetylation ( compare Figure 6—figure supplement 1H–H' with Figure 6—figure supplement 1I–I' and J ) and pan H4 acetylation ( Figure 6—figure supplement 1K–K’ with Figure 6—figure supplement 1L–L' and M ) is evident . In all these scenarios , histone H3 labeling remains unaffected ( Figure 6—figure supplement 1E–G ) . Above molecular and genetic analyses demonstrate that the downregulation of FAO in the hemocyte progenitors leads to a decline in histone acetylation . The next step was to correlate whether the differentiation defects of hemocyte progenitors in FAO loss of function is a consequence of altered histone acetylation levels . Histone acetylation in eukaryotes relies on acetyl-Coenzyme A ( acetyl-CoA ) . It has been established earlier that compromised histone acetylation levels can be restored by supplementing acetate ( Carrisi et al . , 2008; Wellen et al . , 2009; Wong et al . , 2017 ) . The supplemented acetate is converted to acetyl-CoA , which restores the endogenous histone acetylation in a cell . We wondered whether replenishing the H3K9 acetylation levels in whd1 by acetate supplementation ( 50 mM , supplemented fly food post first instar ) can rescue the differentiation defect seen in those lymph glands . Intriguingly , acetate feeding does not affect progenitor differentiation in control , whereas it rescues the differentiation defects seen in homozygous whd1 hemocyte progenitors ( Figure 7A–E ) . At the molecular level , we observed that acetate supplementation to whd1 mutant larvae leads to a restoration of the H3K9 acetylation level ( Figure 7F–G ) , which might lead to the rescue of the differentiation defect ( Figure 7A–E ) . In order to probe this possibility , the lymph gland from acetate fed larvae were dissected and assayed for the status of H3K9 acetylation level . Figure 7H–L reveals that acetate supplementation indeed restores the compromised acetylation status in the whd1 lymph gland ( compare Figure 7J–J' with Figure 7K–K' ) . Conversely , upregulation of FAO by L-carnitine feeding leads to elevated H3K9 acetylation in the lymph gland ( compare Figure 7M–M' with Figure 7N–N' and quantified in Figure 7O ) . Likewise , the level of H3K9 acetylation in L-carnitine fed larvae demonstrates a significant upregulation when compared to age-isogenised non-fed control larvae ( Figure 7P–Q ) . These results establish that the hemocyte progenitors require FAO mediated histone acetylation for their differentiation . Next , we attempted to understand how the FAO-mediated metabolic circuitry collaborates with the known differentiation signals of hemocyte progenitors . Jun-Kinase and dFOXO ( Forkhead box O ) mediated signal has been previously implicated for hemocyte progenitor differentiation ( Owusu-Ansah and Banerjee , 2009 ) . Analogous to whd1 lymph glands , expression of a dominant-negative allele of basket ( bsk , Drosophila orthologue of Jun-Kinase ) in the progenitors results in stalled differentiation ( Figure 8A–C ) . On the other hand , overexpression of FOXO , pushes the progenitor fate towards precocious differentiation ( Owusu-Ansah and Banerjee , 2009; Figure 8D–E and Figure 8H ) . However , genetic removal of one copy of whd is sufficient enough to prevent the precocious differentiation as observed in progenitor-specific overexpression of FOXO ( Figure 8F–H ) . These results illustrate an unappreciated link between the differentiation signals and FAO in the lymph gland progenitors . Next , we addressed whether JNK signaling regulates the expression of genes of FAO . The transcription of CPT1/whd , Mcad , Mtpα , scully , Mtpβ , and yip2 was assayed upon down-regulation of bsk from hemocyte progenitors . The transcript level of CPT1/whd indicates a ~ 41% drop ( Figure 8I ) while the expression of rest of the enzymes either exhibited a mild drop ( ~18% in yip2 ) or no significant alteration upon loss of bsk from the progenitors . This observation established that JNK controls the transcription of CPT1/whd , thereby regulating FAO . Interestingly , clonal analyses of bsk/JNK knockdown in hemocyte progenitors reveals a drop in the levels of H3K9 acetylation ( Figure 8L–M and Figure 8—figure supplement 1A–A''' ) and H4 pan acetylation ( Figure 8N–O and Figure 8—figure supplement 1B–B''' ) similar to whd downregulated clonal patches ( Figure 6M–P ) . Additionally , expression of bskDN in progenitor-specific manner brings about a conspicuous downregulation of both H3K9 acetylation ( Figure 8—figure supplement 1F–H ) and H4 pan acetylation ( Figure 8—figure supplement 1I–K ) . However , in the above experiments , histone H3 labeling remains unaffected ( Figure 8J–K and Figure 8—figure supplement 1C–E ) . Since compromised acetylation leads to differentiation defect in whd1 , which can be rescued upon acetate feeding , we wondered whether the block in differentiation upon JNK loss could be rescued similarly . Indeed , the hemocyte progenitors that lacked JNK ( dome >bskDN ) when reared in acetate supplemented food , demonstrate differentiation levels comparable to the similarly-aged control ( Figure 8P–T ) . Therefore , lack of histone acetylation encountered in JNK loss leads to the block in hemocyte progenitor differentiation . Moreover , whd transcription is under the regulation of JNK , which further endorses that JNK mediated regulation of FAO is crucial for differentiation . If this is true , then the upregulation of FAO in JNK loss either by L-carnitine supplementation or by overexpression of whd should facilitate differentiation . Our results demonstrate that the upregulation of FAO in hemocyte progenitors that lacked JNK indeed elicits differentiation ( Figure 8U–Y and Figure 8—figure supplement 1L–P ) . Collectively , these results are in agreement with the fact that JNK regulates the differentiation of hemocyte progenitors by FAO-mediated histone acetylation . Along with cellular signaling network , stem/progenitor cell fate and state are directly governed by their metabolism in normal development and during pathophysiological conditions ( Ito and Ito , 2016; Ito and Suda , 2014; Oginuma et al . , 2017; Shyh-Chang et al . , 2013; Shyh-Chang and Ng , 2017 ) . How the metabolic circuitry works in sync with the cell signaling machinery to achieve cellular homeostasis is yet to be fully understood . Here , we show the developmental requirement of FAO in regulating the differentiation of hemocyte progenitors in Drosophila . Our molecular genetic analyses reveal a signaling cascade that links ROS-JNK-FAO and histone modification essential for the differentiation of hemocyte progenitors ( Figure 9 ) . High ROS levels in the progenitors evoke differentiation program by triggering JNK and FOXO mediated signals ( Owusu-Ansah and Banerjee , 2009 ) . We show that activated JNK , in turn , leads to transcriptional induction of whd to facilitate the import of fat moiety into mitochondria for β-oxidation . Optimal level of acetyl-CoA , the end product of FAO , is critical for the acetylation of several proteins , including histones . Altering this pathway either at the level of JNK or FAO affects histone acetylation in a HAT dependent manner . On the other hand , it is quite possible that precocious differentiation of blood progenitors in the lymph gland of starved larvae ( Shim et al . , 2012 ) might be an outcome of starvation induced fat mobilization and increased FAO . JNK signaling has been associated with histone acetylation in different biological processes ( Miotto et al . , 2006; Wu et al . , 2008 ) . In Drosophila , Fos , a transcriptional activator of JNK , interacts with Chm ( HAT ) and causes modification of histones . Our investigation reveals that indeed upon downregulation of JNK ( Figure 8A–C ) and Chm ( Figure 6A–B ) , differentiation of hemocyte progenitors is halted . Further , we show that halt in differentiation upon JNK loss is associated with alteration of the acetylation profile of H3K9 and H4 . Given the fact that JNK signaling regulates FAO , which in turn provides the acetyl moiety for histone acetylation , our study provides a new dimension in JNK’s role for histone acetylation in an FAO dependent manner . As a result , despite having high ROS levels , the hemocytes fail to differentiate if FAO is attenuated . Thus , the current work provides a metabolic link between JNK and epigenetic regulation of gene expression . Our results show that upon disruption of FAO , hematopoietic progenitors adopt glycolysis to overcome the G2-M arrest but fail to initiate differentiation . Pharmacological and genetic inhibition of glycolysis in the FAO mutant restores their cell cycle defect but fails to facilitate their differentiation . The glycolytic surge in FAO mutants is not capable to take them through the differentiation process . This indicates that for the process of differentiation , the acetyl moiety derived from FAO plays a key role to facilitate hemocyte progenitor differentiation . A recent study has demonstrated that alteration in acetyl-CoA levels can affect proteome and cellular metabolism by modulating intracellular crosstalk ( Dieterich et al . , 2019 ) . It is intriguing to see that restoring acetylation level by the acetate supplementation is capable of rescuing hematopoietic defects in the lymph gland progenitors of FAO mutants . The acetate supplemented is converted into the end product of FAO: acetyl-CoA , the metabolite that is essential for histone acetylation . The involvement of acetyl-CoA in facilitating differentiation is further evidenced when on genetically downregulating AcCoAs ( the major enzyme in acetyl-CoA generation ) from the progenitor leads to a halt in their differentiation . Our study thus establishes that for hemocyte progenitor differentiation , the metabolic process FAO involves its metabolite acetyl-CoA for epigenetic modification . Earlier studies in diverse model systems have demonstrated that compromised in vivo histone acetylation defects can be rescued by acetate supplementation ( Gao et al . , 2016; Soliman et al . , 2012 ) . A similar finding in Drosophila hematopoiesis signifies the relevance of acetate supplementation across taxa . In light of this study , it would be interesting to see whether metabolite supplementation of FAO can modulate pathophysiological scenarios like certain forms of cancer which rely on fat oxidation . FAO has been implicated in HSCs maintenance downstream of the PML-Peroxisome proliferator-activated receptor delta ( PPARδ ) pathway ( Ito et al . , 2012 ) . Mechanistically , the PML-PPARδ-FAO pathway regulates HSC maintenance by controling asymmetric division . In FAO inhibition , HCSs undergo symmetric divisions , which lead to exhaustion and depletion of the stem cell pool resulting in their differentiation ( Ito et al . , 2012 ) . Another study in mice shows that upon short term starvation , there is a decline in the number of HSC ( Takakuwa et al . , 2019 ) . Since the HSC maintenance is FAO dependent ( Ito et al . , 2012 ) , a loss in number might be attributed to heightened fat oxidation during starvation . Interestingly , metabolic dependence on FAO has been reported in mammalian neural stem cell ( Knobloch et al . , 2017 ) , muscle stem cells ( Ryall et al . , 2015 ) and intestinal stem cells ( Chen et al . , 2020 ) . Although endothelial precursors ( Wong et al . , 2017 ) is also known to be dependent on FAO , it remains to be seen whether FAO is a preferred metabolic requirement for progenitor differentiation . The entire blood cell repertoire in Drosophila is engaged in innate immunity , maintenance of tissue integrity , wound healing , and heterogeneous stress responses , and is therefore functionally considered to be similar to myeloid cells in mammals ( Banerjee et al . , 2019; Gold and Brückner , 2014 ) . Interestingly , several molecular mechanisms that regulate Drosophila lymph gland hematopoiesis are essential players in progenitor-based hematopoiesis in vertebrates ( Banerjee et al . , 2019; Gold and Brückner , 2014; Krzemien et al . , 2010 ) . Based on the above conservations , it is reasonable to propose the requirement of FAO in progenitor differentiation described here will help us in understanding mammalian myeloid progenitor differentiation . The fly stocks used were dome-GAL4 , dome-MESO-EBFP2 , Hml-DsRed ( K . Bruckner ) , HmlΔ-GAL4 ( S . Sinenko ) Pvf2-LacZ ( M . A . Yoo ) , TepIV-GAL4 , CG3902-YFP , Mtpα[KO] , Mtpβ[KO] ( DGRC , Kyoto ) , UAS-whd RNAiKK ( VDRC , Vienna ) , OreR , w1118 , UAS-Hnf4 . miRNA ( Lin et al . , 2009 ) , UAS-whd RNAiHMS00040 ( Manzo et al . , 2018 ) , UAS-FOXO . P , Hnf4-GAL4GMR50A12 ( Tokusumi et al . , 2017 ) , UAS-FUCCI , UAS-mito-HA-GFP , UAS-chm-RNAiJF02348 ( Dietz et al . , 2015 ) , UAS-Gcn5-RNAiHMS00941 ( Janssens et al . , 2017 ) , UAS-AcCoAS RNAiHMS02314 ( Eisenberg et al . , 2014 ) , UAS-Glut1 RNAiJF03060 ( Charlton-Perkins et al . , 2017 ) , ATPCL01466 , seaEP3364 , UAS-bskDN , UAS-mCD8::GFP , U-6;sgRNA-whdTKO . GS00854 , U-6;sgRNA-whdTOE . GS00536 , whd1 , Hnf4Δ33 , Hnf4Δ17 , tub-GAL80ts20 , hsFlp and Ay-GAL4 , UAS-GFP ( BDSC , Bloomington Drosophila Stock Center ) . Following genotypes were recombined for the current study: All Stocks and crosses were maintained at 25°C , except for those used in RNAi based and GAL4-UAS expression experiments . In those cases , crosses were maintained at 29°C . For GAL80ts experiments , crosses were initially maintained at 18°C for 5 days ( equivalent to 60 hr at 25°C ) , and then shifted to 29°C till dissection ( Figure 2—figure supplement 1D ) . For synchronization of larvae , flies were allowed to lay eggs for 2 hr and newly hatched larvae within 1 hr interval were collected and transferred onto fresh food plates and aged for specified time periods at 25°C . Fatty acid β-oxidation inhibitors: Etomoxir ( Cayman Chemicals , Cay11969 , inhibitor of CPT1 ) and Mildronate ( Cayman Chemicals , Cay15997 , inhibitor of carnitine biosynthesis and transport ) were used at a concentration of 5 µM and 100 µM respectively mixed in fly food and fed to larvae from 48 hr AEH and analysis of lymph gland was done in late third instar stages . L-carnitine hydrochloride ( Sigma-Aldrich , C0283 ) at a concentration of 100 mM has been used to augment FAO by allowing the entry of palmitic acid into the mitochondria . L-carnitine was used at 100 mM concentrations in fly food and fed to larvae for 48 hr in third instar analysis and for 24 hr in second instar analysis . Glycolytic inhibitor: 2-DG ( 2-Deoxy-D-Glucose ( 2-Deoxyglucose ) ( Sigma-Aldrich , D8375 ) used at a concentration of 100 mM mixed in fly food and fed to larvae for 48 hr in third instar analysis and for 24 hr in second instar analysis . Sodium acetate ( Sigma-Aldrich , 71196 ) supplement was used at a concentration of 50 mM and fed to larvae from second instar 36 hr AEH onwards and analysis was done in late third instar stages . Similar aged larvae fed on vehicle controls served as control larvae . For all feeding experiments control larvae had same vehicle control level mixed in fly food . Fly food mixed with permissible food dye was fed to the control and experimental larvae and larvae with abundant food intake were picked for the experimental analysis . Generation of clones was done by the Ay-GAL4 system that combines the technique of Flippase ( Flp ) /FRT system and the GAL4/UAS system ( Ito et al . , 1997 ) . In this system , the Act5C promoter GAL4 fusion gene is interrupted by a FRT cassette containing yellow ( y+ ) gene . Heat shock treatment activates the Flp gene which in-turn excises the FRT cassette between the Act5C promoter and GAL4 sequence . This activates the expression of Act5C-GAL4 in cells . To induce UAS-whd RNAi and UAS-bskDN clones , mid second instar larvae of genotypes: hsFlp; Ay-GAL4 , UAS-GFP; UAS-whd RNAi and hsFlp/UAS-bskDN; Ay-GAL4 , UAS-GFP were subjected to heat shock for 90 min at 37°C , respectively . Post heat shock , larvae were transferred to 25°C to recover for 2 hr , then to express the respective knockdown constructs , larvae were reared at 29°C till dissection . The primary antibodies used in this study includes mouse anti-P1 ( Kurucz et al . , 2007 ) , rabbit anti-Pxn ( J . Fessler ) , rabbit anti-proPO ( M . Kanost ) , rat anti-DE Cadherin ( Cat# DE-cad ( DE-cadherin ) , RRID:AB_2314298 , 1:50 , DSHB ) , rat anti-Ci155 ( Cat# 2A1 , RRID:AB_2109711 , 1:2 , DSHB ) , rabbit anti-GFP ( Cat# A-11122 , RRID:AB_221569 , 1:100 , Invitrogen ) , rabbit anti-H3 ( Cat# 9927 , RRID:AB_330200 , 1:400 , Cell Signaling Technologies ) , rabbit anti-H3K9 acetylation ( Cat# 9927 , RRID:AB_330200 , 1:300 , Cell Signaling Technologies ) , rabbit anti-H4 pan acetylation ( Cat# 06–598 , RRID:AB_2295074 , 1:500 , Merck-Millipore ) . The following secondary antibodies mouse Cy3 ( Cat# 115-165-166 , RRID:AB_2338692 ) , mouse FITC ( Cat# 715-096-151 , RRID:AB_2340796 ) , rabbit Cy3 ( Cat# 711-165-152 , RRID:AB_2307443 ) , rabbit FITC ( Cat# 111-095-003 , RRID:AB_2337972 ) , rat Cy3 ( Cat# 712-165-153 , RRID:AB_2340667 ) from Jackson Immuno-research Laboratories were used at 1:400 . Lymph gland from synchronized larvae of required developmental age was dissected in cold PBS ( 1X Phosphate Buffer Saline , pH-7 . 2 ) and fixed in 4% Paraformaldehyde ( PFA ) for 45 min ( Mandal et al . , 2007 ) at room temperature ( RT ) on a shaker . Tissues were then permeabilized by 0 . 3% PBT ( 0 . 3% triton-X in 1X PBS ) for 45 min ( 3 × 15 min washes ) at RT . Blocking was then done in 10% NGS , for 30–45 min at RT . Tissues were next incubated in the respective primary antibody with appropriate dilution in 10% NGS overnight at 4°C . Post incubation in primary antibody , tissues were washed thrice in 0 . 3% PBT for 15 min each . This was followed by incubation of tissues in secondary antibody overnight at 4°C . The tissues were then subjected to four washes in 0 . 3% PBT for 15 min each , followed by incubation in DAPI solution ( Invitrogen ) for 1 hr at RT . Excess DAPI was washed off from the tissues by 1X PBS before mounting in Vectashield ( Vector Laboratories ) . Immunostaining for specific histone acetylations were performed with a slight modification of the above protocol . Lymph gland from synchronized larvae was dissected in ice cold PBS with deacetylate inhibitors ( Sodium butyrate ( 10 mM , EMD Millipore , 19–137 ) and Nicotinamide ( 10 mM , Sigma-Aldrich , 72345 ) and fixed in 4% PFA prepared in ice-cold 1X PBS ( pH 7 . 2 ) for 5 hr at 4°C . Tissue were then permeabilized by 0 . 3% PBT for 45 min . Blocking was done with 5% BSA made in 1X PBS . Primary antibody and secondary antibody incubation solutions were made in 5% BSA in 1X PBS and subsequent washings were done with 0 . 1% PBT . To detect the DE-cadherin expression , lymph glands were incubated in DE-Cadherin antibody ( 1:50 in PBS ) before fixation ( Langevin et al . , 2005 ) for 1 hr at 4°C . Tissues were then fixed in 4% PFA prepared in ice cold 1X PBS ( pH 7 . 2 ) for 5 hr at 4°C . Then , tissues were washed thrice with 0 . 3% PBT for 30 min . Secondary antibody incubation , washes , and mounting were performed following the standard protocol ( Sharma et al . , 2019 ) . Larvae were dissected in cold PBS followed by fixation in 4% PFA overnight at 4°C . This was followed by permeabilization with 0 . 1% PBT ( 0 . 1% triton-X in 1x PBS ) for 45 min at RT and incubation in Streptavidin-Cy3 in 1:200 dilution ( Molecular Probes , 434315 ) in 1XPBS for 1 hr at RT in dark . Post incubation samples were , washed thrice in PBS for 30 min . Lymph glands were then mounted in Vectashield and imaged in Zeiss LSM 780 confocal microscope . Click-iT EdU plus ( 5- ethynyl-2’- deoxyuridine , a thymidine analog ) kit ( Invitrogen , C10639 ) plus was used to perform DNA replication assay . Lymph glands were dissected and incubated in EdU solution ( 1:1000 in PBS ) for 40 min at RT for EdU incorporation . Next fixation was done in 4% PFA prepared in 1X PBS ( pH 7 . 2 ) for 45 min at RT . Tissue were then permeabilized by 0 . 3% PBT ( 0 . 3% triton-X in 1X PBS ) for 45 min at RT . Blocking was then done in 10% NGS , for 30–45 min at RT . To detect the incorporated EdU in cells , azide-based fluorophore were used as described in manufacturer protocol . EdU-labeled cell counting was done using spot detection function in Imaris Software . The protocol was slightly modified after ( Zou et al . , 2005 ) . Larvae were dissected in ice-cold PBS and incubated in PBS with 0 . 25 mM 2-NBDG ( Invitrogen , N13195 ) for 45 min at RT , washed twice in PBS for 5 min , fixed 45 min in 4% PFA and washed twice for 10 min in PBS . All washes and the fixation were done with ice-cold PBS ( 4°C ) . Lymph glands were speedily dissected and mounted in Vectashield and were imaged immediately with a Zeiss LSM 780 confocal microscope . Lactate dehydrogenase in vivo staining was modified from Abu-Shumays and Fristrom , 1997 . Lymph glands of wandering third instar larvae were dissected in cold 1X PBS ( pH8 ) . Samples were fixed for 25 min in 0 . 5% glutaraldehyde in 1X PBS at room temperature , followed by four washes in 1X PBS for 15 min each . Staining was performed at 37°C in a solution of 0 . 1M NaPO4 ( pH 7 . 4 ) , 0 . 5 mM lithium lactate , 2 . 75 mM NAD+ , 0 . 5 mg/ml NBT ( Nitro blue tetrazolium ) with 0 . 025 mg/ml PMS ( Phenazine methosulfate ) . Reaction was stopped by washing in cold 1X PBS having pH 7 . 5 . The samples were washed in four 1X PBS washes of 5 min each and immediately mounted and imaged . Larvae were dissected in cold PBS followed by fixation in 4% PFA for 1 hr at RT , permeabilized by 0 . 1% PBT ( 0 . 1% triton-X in 1X PBS ) for 45 min at RT . It was then incubated in 1X LipidTOX ( diluted from 1000X stock provided by the manufacturer; Molecular Probes , H34477 ) in PBS for 1 hr at RT in dark , washed thrice in PBS for 30 min . Lymph glands were then mounted in Vectashield and imaged in Leica SP8 confocal microscope . Larvae were dissected in cold PBS followed by fixation in 4% PFA for 1 hr at RT , permeabilized by 0 . 1% PBT ( 0 . 1% triton-X in 1X PBS ) for 45 min at RT and incubated in 0 . 5 ug/mL Nile red ( Molecular Probes , N1142 ) in PBS for 1 hr at RT in dark , washed thrice in PBS for 30 min . Lymph glands were mounted in Vectashield and imaged in Leica SP8 confocal microscope . Larvae were dissected in Schneider’s medium ( Gibco , 21720001 ) followed by incubation in 0 . 3 µM DHE ( Molecular Probes , D11347 ) in Schneider’s medium for 8 min at room temperature in dark . This was followed by two washes in 1X PBS for 5 min each; a brief fixation was done with 4% PFA for 10 min followed by two quick 1X PBS washes . Tissues were then mounted in Vectashield and imaged in Zeiss LSM 780 confocal microscope . Images were captured as confocal Z-stacks in Zeiss LSM 780 , Leica SP8 confocal , and Olympus Flouview FV10i microscopes . Same confocal imaging settings were employed for image acquisition of control and experimental samples related to an experiment . Each experiment was repeated with appropriate controls at least three times to ensure reproducibility of the results . Data expressed as mean+/-Standard Deviation ( SD ) of values from three sets of independent experiments in GraphPad . Each dot in GraphPad represents a data point . Graphs plotted in EXCEL have Error Bars representing the Standard Deviation while graphs plotted in GraphPad employs Error Bars as mean+/-Standard Deviation . At least 10 images were analyzed per genotype , and statistical analyses performed employed two-tailed Student’s t-test . Raw data related to statistical analysis are attached in the source file of each figure along with graphs plotted in excel . p-Value of <0 . 05;<0 . 01 and<0 . 001 , mentioned as * , , * respectively are considered as statistically significant while n . s . = not significant . To measure the differentiation index of the primary lobe of lymph glands , middle confocal Z-stacks of a lymph gland image covering the Medullary Zone ( MZ ) were merged into a single section using ImageJ/Fiji ( NIH ) software as previously described ( Shim et al . , 2012 ) . The merged section reflects the differentiated cell and hemocyte progenitor area clearly . For images with more than one fluorophore channel , each channel was separately analyzed . To measure differentiation index , P1-positive area was recalibrated into an identical threshold by using the Binary tool ( Process–Binary–Make binary , Image J ) . Wand tool was used to capture the area with identical threshold whereas the size was measured using the Measure tool ( Analyse–Measure ) . To measure the total area of one primary lobe of lymph gland , recalibration of the total area was then done by the Threshold tool until it was overlaid with identical threshold colour . Wand tool was used for selecting the total area for measurement . Differentiation index/fraction was estimated by dividing the size of the P1/Pxn positive area by the total size of the primary lobe . At least 10 lymph glands were analyzed per genotype , and two-tailed Student’s t-test was done to evaluate the statistical significance . Counting the number of EdU+ and FUCCI+ progenitors in lymph glands was done as described earlier ( Sharma et al . , 2019 ) , using spot detection and surface tool in Imaris software and normalized by total number of nuclei per primary lobe of lymph gland . Different progenitor sub-populations in lymph glands were counted using the surface and spot detection function in Imaris as illustrated in detail ( Sharma et al . , 2019 ) and normalized by total number of cells ( nuclei ) in primary lobe and represented as percentage of progenitors in each primary lobe . Using surface tool , surface is created over Dome+ progenitors and nuclear label channel is masked in Dome+ surface . By utilizing the spot detection tool , the number of nuclei is counted in Dome+ surface . Similarly , number of nuclei ( DAPI/Hoechst ) is counted in another surface created over Pxn+ cells . Next , in Dome+ surface , Pxn+ channel is masked . A surface is created over Dome+ Pxn+ region and nuclear channel is masked in this surface . Using spot detection tool , number of Dome+ Pxn+ IP nuclei are counted and its percentage can be calculated from the total number of nuclei in the primary lobe of lymph gland . ATP assay was performed with three biological replicates from late third instar larvae . Whole larvae were homogenized in ATP assay Lysis buffer ( Costa et al . , 2013 ) . The samples were boiled at 95°C for 5 min and diluted 1:100 in dilution buffer provided in ATP bioluminescence kit HSII ( Sigma , 11699709001 ) . Further assay was performed as per manufacturer protocol in Glomax 96 microwell Luminometer ( Promega ) . Standard curve was generated and ATP concentrations were calculated . The ATP concentration was normalized with protein concentration and expressed in percentage to plot the graph in EXCEL . Histone from late third instar larvae were extracted using Histone extraction kit ( Abcam , ab113476 ) following manufacturer protocol and quantitated by Bradford reagent ( Biorad , 5000006 ) . Equal amount of protein of each genotype was run on 4–12% SDS-PAGE and transferred to PVDF membrane ( Millipore , IPVH00010 ) . Blots were developed using Luminata Crescendo Western HRP substrate ( Millipore , WBLUR0500 ) in LAS2000 blot imaging instrument . Primary antibodies used rabbit anti-H3 ( Cat# 9927 , RRID:AB_330200 , 1:1000 , Cell Signaling Technologies ) , rabbit anti-H3K9 acetylation ( Cat# 9927 , RRID:AB_330200 , 1:1000 , Cell Signaling Technologies ) . Secondary antibody rabbit anti-IgG-HRP ( Cat# A00098-1 mg , RRID:AB_1968815 , 1:5000 , GenScript ) was used . The band intensity was measured in Image J and normalized with histone H3 as loading controls . Analysis was done using three biological replicates . Extraction of RNA from lymph gland was performed from late third instar larvae of each genotype using TRIzol ( Invitrogen , 15596018 ) followed by RNAeasy Mini Kit ( Qiagen , 74104 ) according to the manufacturer’s instructions . cDNA was prepared using the Verso cDNA Synthesis Kit ( Thermo Scientific , AB1453B ) . To quantitate transcripts , qPCR was done using iTaq Universal SYBR Green Supermix ( Biorad , 1725124 ) on a CFX96 Real-Time system/C1000 thermal Cycler ( Biorad ) . Drosophila Actin5C was used as internal control . Analysis was done using at least three biological replicates .	Stem cells are special precursor cells , found in all animals from flies to humans , that can give rise to all the mature cell types in the body . Their job is to generate supplies of new cells wherever these are needed . This is important because it allows damaged or worn-out tissues to be repaired and replaced by fresh , healthy cells . As part of this renewal process , stem cells generate pools of more specialized cells , called progenitor cells . These can be thought of as half-way to maturation and can only develop in a more restricted number of ways . For example , so-called myeloid progenitor cells from humans can only develop into a specific group of blood cell types , collectively termed the myeloid lineage . Fruit flies , like many other animals , also have several different types of blood cells . The fly’s repertoire of blood cells is very similar to the human myeloid lineage , and these cells also develop from the fly equivalent of myeloid progenitor cells . These progenitors are found in a specialized organ in fruit fly larvae called the lymph gland , where the blood forms . These similarities between fruit flies and humans mean that flies are a good model to study how myeloid progenitor cells mature . A lot is already known about the molecules that signal to progenitor cells how and when to mature . However , the role of metabolism – the chemical reactions that process nutrients and provide energy inside cells – is still poorly understood . Tiwari et al . set out to identify which metabolic reactions myeloid progenitor cells require and how these reactions might shape the progenitors’ development into mature blood cells . The experiments in this study used fruit fly larvae that had been genetically altered so that they could no longer perform key chemical reactions needed for the breakdown of fats . In these mutant larvae , the progenitors within the lymph gland could not give rise to mature blood cells . This showed that myeloid progenitor cells need to be able to break down fats in order to develop properly . These results highlight a previously unappreciated role for metabolism in controlling the development of progenitor cells . If this effect also occurs in humans , this knowledge could one day help medical researchers engineer replacement tissues in the lab , or even increase our own bodies’ ability to regenerate blood , and potentially other organs .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2020	Fatty acid β-oxidation is required for the differentiation of larval hematopoietic progenitors in Drosophila
T-cell recognition of self and foreign peptide antigens presented in major histocompatibility complex molecules ( pMHC ) is essential for life-long immunity . How the ability of the CD4+ T-cell compartment to bind self- and foreign-pMHC changes over the lifespan remains a fundamental aspect of T-cell biology that is largely unexplored . We report that , while old mice ( 18–22 months ) contain fewer CD4+ T-cells compared with adults ( 8–12 weeks ) , those that remain have a higher intrinsic affinity for self-pMHC , as measured by CD5 expression . Old mice also have more cells that bind individual or multiple distinct foreign-pMHCs , and the fold increase in pMHC-binding populations is directly related to their CD5 levels . These data demonstrate that the CD4+ T-cell compartment preferentially accumulates promiscuous constituents with age as a consequence of higher affinity T-cell receptor interactions with self-pMHC . Each T-cell expresses a T-cell receptor ( TCR ) encoded by rearranged gene segments and non-germline nucleotides . Estimates of TCR diversity imply a repertoire that can bind a universe of self and foreign peptides embedded within self-major histocompatibility complex molecules ( pMHC ) ( Davis and Bjorkman , 1988 ) . Yet , this potential cannot be realized . Thymic development limits clonal representation to T-cells bearing TCRs within an affinity window for self-pMHC ( Savage and Davis , 2001; Yin et al . , 2012; Klein et al . , 2014 ) , while peripheral space physically constrains the number of T-cells present to recognize foreign-pMHC ( Mason , 1998; Vrisekoop et al . , 2014 ) . Finally , time—with its age-associated changes in thymic expression of tissue-restricted antigens ( TRAs ) , thymic architecture , antigen experience , and homeostasis—imposes an overarching pressure that limits the binding capacity of a repertoire for self- and foreign-pMHC to each constituent's prior history of TCR–pMHC interactions ( Nikolich-Zugich , 2008; Surh and Sprent , 2008; Chinn et al . , 2012; Griffith et al . , 2012 ) . How these pressures shape the capacity of the CD4+ T-cell compartment to bind pMHC over the lifespan remains largely unexplored . Aging is associated with increased susceptibility to infections and decreased responsiveness to vaccines , suggesting that individual repertoires converge on a point where their diversity is insufficient to bind and/or mount a protective response to foreign-pMHC ( Vazquez-Boland et al . , 2001; Nichol , 2008; Nikolich-Zugich , 2008 ) . Consistent with this idea , TCR diversity within both the CD4+ and CD8+ T-cell compartments contract from adult to old mice in parallel with thymic involution ( Ahmed et al . , 2009; Rudd et al . , 2011; Britanova et al . , 2014 ) , and the number of CD8+ T-cells that bind distinct foreign class I pMHC in unprimed mice decreases over the lifespan ( Yager et al . , 2008; Rudd et al . , 2011; Decman et al . , 2012; Smithey et al . , 2012 ) . Here , we explored how aging impacts the number of naive and memory phenotype CD4+ T-cells available to bind pMHC , their relative affinity for self-pMHC , and their capacity to bind foreign-pMHC . We report that , while the absolute number of CD4+ T-cells decreases over time , those that remain have an increased affinity for self-pMHC and an increased capacity to bind foreign-pMHC . Unprimed old ( 18–22 months ) C57BL/6 mice were found to have fewer CD4+ T-cells in their secondary lymphoid organs than adults ( 8–12 weeks ) due to a loss of naive ( CD44lo ) T-cells , as expected given thymic involution ( Figure 1A , B ) ( den Braber et al . , 2012 ) . The number of memory phenotype ( CD44hi ) CD4+ T-cells increased with aging ( Figure 1C ) . This could be due to prior antigen experience and/or homeostatic proliferation ( Nikolich-Zugich , 2008; Surh and Sprent , 2008 ) . 10 . 7554/eLife . 05949 . 003Figure 1 . The CD4+ T-cell compartment contracts but accumulates CD44hiCD5hi cells with aging . The absolute numbers of T-cells in unprimed adult ( 8–12 weeks ) and old ( 18–22 months ) mice are shown as ( A ) total CD4+ T-cells in secondary lymphoid organs , ( B ) CD4+ CD44lo ( naïve ) T-cells and ( C ) CD4+ CD44hi ( memory phenotype ) T-cells . Data are concatenated from three experiments , 4 mice/group . Horizontal bar indicates median ( p < 0 . 05 , p < 0 . 0001; Mann–Whitney ) . ( D ) Relative fluorescent intensity ( RFI ) of CD5 expression on adult and old CD44hi and CD44lo CD4+ T-cells relative to CD5 expression on adult CD44lo CD4+ T-cells ( dotted line ) . Data represent four experiments with 4 mice/group ( p < 0 . 0001 , p < 0 . 05; Mann–Whitney ) . ( E ) RFI of CD3 expression on adult and old CD44hi and CD44lo CD4+ T-cells relative to CD3 expression on adult CD44lo CD4+ T-cells ( dotted line ) ( **p < 0 . 0001; Mann–Whitney ) . Results represent seven experiments with 4 mice/group . ( F ) Concatenated contour plots ( 4 mice ) showing CD5 vs BrdU incorporation in unprimed adult and old total CD4+ T-cells . Percent BrdU+ of total CD4+ T-cells ± SEM is shown in the inset ( p < 0 . 05 Mann–Whitney adult compared to old ) . ( G ) Absolute numbers of CD4+ BrdU+ T-cells . Results are representative of two experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 003 To assess steady-state TCR engagement , we measured CD5 expression , as a surrogate for the strength of tonic TCR–pMHC interactions ( Azzam et al . , 1998; Smith et al . , 2001; Mandl et al . , 2012 , 2013; Persaud et al . , 2014; Vrisekoop et al . , 2014; Fulton et al . , 2015 ) ; CD3 levels , which decrease upon TCR engagement ( Valitutti et al . , 1995 ) ; and BrdU incorporation to assess proliferation in unprimed mice . CD5 was higher on memory CD4+ T-cells in adult mice relative to adult naive T-cells , as expected ( Mandl et al . , 2013 ) , while both naive and memory CD4+ T-cells in old mice had higher CD5 expression relative to adult naive cells ( Figure 1D ) . An inverse relationship was observed between CD5 and CD3 levels , consistent with CD5 reflecting tonic TCR engagement ( Figure 1E ) . Finally , cells with high CD5 expression incorporated the most BrdU in adult and old mice , consistent with tonic TCR interactions driving homeostatic proliferation ( Figure 1F ) . A higher frequency of BrdU+ cells was observed in old mice compared with adults . However , since the total number of CD4+ T-cell drops in old mice this did not result in significantly more BrdU+CD4+ T-cells ( Figure 1F , G ) . Altogether , the data indicate that the CD4+ T-cell compartment increases in clonal representation of constituents with higher intrinsic affinity for self-pMHC . Age-related changes in the capacity of the CD4+ T-cell compartment to bind foreign-pMHC were evaluated via tetramer enrichment ( all class II pMHC tetramer validation is shown in Figure 2—figure supplements 1 , 2 ) . I-Ab tetramers presenting an immunodominant peptide ( aa 641–655 ) from West Nile Virus ( WNV ) envelope protein ( E641:I-Ab ) were used because WNV lethality increases over the lifespan of mice and humans , making it a useful model for investigating age-related defects in susceptibility to viral infection and vaccine efficacy ( Brien et al . , 2008 , 2009; Uhrlaub et al . , 2011; Suthar et al . , 2013 ) . Two-color tetramer enrichment ( Nelson et al . , 2015 ) revealed more cells binding E641:I-Ab in old mice than adults ( Figure 2—figure supplement 3 ) . To determine if this is unique to E641:I-Ab , we also enumerated CD4+ T-cells with distinct recognition properties by using a tetramer made with a subdominant ovalbumin peptide ( 326–338 ) in I-Ab ( OVA:I-Ab ) , and an allogeneic tetramer made with the moth cytochrome c peptide ( 88–103 ) bound to I-Ek ( MCC:I-Ek ) ( Savage et al . , 1999; Malherbe et al . , 2004; Moon et al . , 2007; Brien et al . , 2008 ) . OVA:I-Ab was considered to be subdominant because immunization with OVA elicited a smaller response than E641 in isolation and failed to mount a response upon co-immunization with E641 ( Figure 2—figure supplements 2 , 4 ) . OVA:I-Ab monomer is also less SDS-stable than E641:I-Ab at room temperature ( not shown ) , and pMHC stability is directly related to immunodominance ( Lazarski et al . , 2005 ) . Alloreactive cells were enumerated because they are likely to be selected on a broader range of self-pMHC and represent a broader subset of the CD4+ T-cell compartment ( Felix and Allen , 2007; Chu et al . , 2009 ) . CD4+ T-cells bound to E641:I-Ab , OVA:I-Ab , and MCC:I-Ek were simultaneously enriched from individual animals using anti-His beads against the 6× His-tag on the alpha and beta subunits of each pMHC ( Figure 2A–F and Figure 2—figure supplement 5 ) . This yielded more E641-bound adult cells than the anti-PE/APC beads ( Figure 2G and Figure 2—figure supplement 3E ) . Since tetramers cannot detect all CD4+ T-cells that respond to a given class II pMHC via weak TCR–pMHC interactions ( Sabatino et al . , 2011 ) , the more avid His-tag enrichment is likely to detect T-cells that bind tetramers with lower avidity . 10 . 7554/eLife . 05949 . 004Figure 2 . CD44lo and CD44hi CD4+ T-cells binding immunodominant , subdominant , and allogeneic pMHC increase with time . Representative plots of CD4+ T-cells bound to ( A and B ) E641:I-Ab , ( C and D ) OVA:I-Ab , and ( E and F ) MCC:I-Ek tetramers in adult ( top ) and old ( bottom ) mice . Absolute number of CD4+ CD44lo ( left Y-axis ) and CD4+ CD44hi ( right Y-axis ) T-cells bound to ( G ) E641:I-Ab , ( H ) OVA:I-Ab , or ( I ) MCC:I-Ek tetramers only enumerated after dump tetramer analysis ( ‘Materials and methods’ ) , or those binding ( J ) E641:I-Ab + OVA:I-Ab , ( K ) OVA:I-Ab + MCC:I-Ek , or ( L ) E64:I-Ab + MCC:I-Ek tetramers in combination enumerated after both dump and two-color tetramer analysis ( ‘Materials and methods’ ) . Bars indicate median ( p < 0 . 05 , p < 0 . 005 , *p < 0 . 0001 , ns = non-significant; Mann–Whitney ) . Fold change ( Δ ) in means between adult and old is shown . Results are from three experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00410 . 7554/eLife . 05949 . 005Figure 2—figure supplement 1 . Tetramer validation on T-cell hybridomas . Dot plots showing E641:I-Ab ( left panel ) , OVA: I-Ab ( middle panel ) , and MCC:I-Ek ( right panel ) tetramer bound to ( A ) parental 58α−β− TCR-negative hybridomas or those transduced to express the ( B ) OT-II TCR specific for OVA: I-Ab , or the ( C ) 5c . c7 or ( D ) 2B4 TCRs specific for MCC:I-Ek . Each tetramer was made with streptavidin-PE and streptavidin-APC for two-color analysis as labeled on axis . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00510 . 7554/eLife . 05949 . 006Figure 2—figure supplement 2 . Tetramer validation for in vivo primed CD4+ T-cells . C57BL/6 mice were immunized with 50 µg of the indicated peptide in 50 µl CFA on each side of the base of the tail or left unprimed . Spleen and lymph nodes cells were harvested on day 7-post immunization and incubated with E641:I-Ab , OVA:I-Ab , and MCC:I-Ek tetramers . No tetramer enrichment was performed . Contour plots show tetramer vs CD44 for ( A–C ) unprimed ( D–F ) E641+CFA immunized or ( G–I ) OVA+CFA immunized mice . Numbers indicate percent of CD44hi tetramer+ CD4+ T-cells . Results represent a single experiment with one mouse per condition . ( J ) Amino acid sequences of peptides contained in I-Ab and I-Ek tetramers used in this study . The bold sequences indicate the core nonamer peptide epitopes as predicted by Zhu et al . ( 2003 ) ; Birnbaum et al . ( 2012 ) ; Nelson et al . ( 2015 ) . OVA 326–338 is from chicken ovalbumin , Env 641–655 is from West Nile Virus , and MCC 88–103 is from moth cytochrome c . The putative TCR contact residues are highlighted in red . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00610 . 7554/eLife . 05949 . 007Figure 2—figure supplement 3 . WNV-specific CD4+ T-cells increase over the lifespan . Representative gating strategy for CD4+ tetramer+ T-cells via ( A ) forward and side scatter , ( B ) CD3+ CD8− T-cells , ( C ) CD3+ CD4+ T-cells , and ( D ) E641:I-Ab+ CD4+ T-cells in the tetramer enrichment bound fraction from adult and old mice . ( E ) Absolute number of E641:I-Ab+ CD4+ T-cells derived from pooled spleen and lymph node cells from unprimed adult or old mice after E641:I-Ab tetramer enrichment using anti-FP magnetic beads . Horizontal bar indicates median ( p < 0 . 05; Mann–Whitney ) . Results are shown from one of two similar experiments ( 4 mice/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00710 . 7554/eLife . 05949 . 008Figure 2—figure supplement 4 . E641 is immunodominant to OVA . C57BL/6 mouse was co-immunized with 25 µg E641 + 25 µg OVA + 50 µl CFA on each side of the base of the tail . On day 7 post immunization , pooled spleen and lymph node cells were harvested and stained with E641:I-Ab and OVA:I-Ab tetramers . No tetramer enrichment was performed . ( A ) Contour plot showing E641:I-Ab vs OVA:I-Ab tetramer+ CD4+ T-cells . ( B ) CD44 vs E641:I-Ab tetramer+ CD4+ T-cells . ( C ) CD44 vs OVA:I-Ab tetramer+ CD4+ T-cells . Number indicates percent CD44hi tetramer+ CD4+ T-cells of total CD4+ T-cells . Results are from a single experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00810 . 7554/eLife . 05949 . 009Figure 2—figure supplement 5 . Gating scheme for identification of tetramer+ cells . Representative gating scheme for identifying tetramer+ cells for Figures 2 , 4 . ( A ) Forward vs side scatter . ( B ) CD3+ CD8− T-cells . ( C ) CD4+ T-cells . After anti-His magnetic enrichment , the tetramer+ gate was set on the unbound fraction to identify ( D ) E641:I-Ab+ , ( E ) OVA:I-Ab+ , and ( F ) MCC:I-Ek+ CD4+ T-cells . ( G ) Additional gating on E641:I-Ab+ CD4+ T-cells in the bound fraction to further identify ( H–J ) tetramer single+ , double+ , and triple+ CD4+ T-cells ( E = E641:I-Ab+ , O = OVA:I-Ab+ , and M = MCC:I-Ek+ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 00910 . 7554/eLife . 05949 . 010Figure 2—figure supplement 6 . Poly-specific cells form a very small fraction of a particular total tetramer+ population . Percent tetramer single+ , double+ and triple+ CD4+ T-cells of ( A ) Total E641:I-Ab+ , ( B ) Total OVA:I-Ab+ , and ( C ) Total MCC:I-Ek+ CD4+ T-cells in adult ( left ) and old ( right ) mice . Horizontal bar indicates median . Data are from three experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 01010 . 7554/eLife . 05949 . 011Figure 2—figure supplement 7 . Two-color analysis of E641+OVA polyspecific cells ± dump tetramer exclusion . CD4+ T-cells from pooled spleen and lymph nodes from unprimed adult and old mice after tetramer enrichment using anti-His magnetic beads . Shown are representative plots of the CD4+ T-cells binding E641:I-Ab + OVA:I-Ab tetramers pre ( left ) and post ( right ) gating out the dump , MCC:I-Ek , tetramer+ cells in ( A ) adult and ( B ) old mice . Absolute number of CD44lo ( left Y-axis ) and CD44hi ( right Y-axis ) CD4+ T-cells binding E641:I-Ab + OVA:I-Ab tetramers ( C ) pre and ( D ) post gating out the dump MCC:I-Ek tetramer+ cells . Horizontal bar indicates median ( p < 0 . 005 , p < 0 . 0001 , ns = non-significant; Mann–Whitney ) . Δ: Fold change in average numbers from adult to old mice . Data are from three experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 01110 . 7554/eLife . 05949 . 012Figure 2—figure supplement 8 . Two-color analysis of OVA+MCC polyspecific cells ± dump tetramer exclusion . CD4+ T-cells from pooled spleen and lymph node from unprimed adult and old mice after tetramer enrichment using anti-His magnetic beads . Shown are representative plots of the CD4+ T-cells binding OVA:I-Ab + MCC:I-Ek tetramers pre ( left ) and post ( right ) gating out the dump , E641:I-Ab , tetramer+ cells in ( A ) adult and ( B ) old mice . Absolute number of CD44lo ( left Y-axis ) and CD44hi ( right Y-axis ) CD4+ T-cells binding OVA:I-Ab + MCC:I-Ek tetramers ( C ) pre and ( D ) post gating out the dump , E641:I-Ab , tetramer+ cells . Horizontal bar indicates median ( p < 0 . 005 , p < 0 . 0001 , ns = non-significant; Mann–Whitney ) . Δ: Fold change in average numbers from adult to old mice . Data are from three experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 01210 . 7554/eLife . 05949 . 013Figure 2—figure supplement 9 . Two-color analysis of E641+MCC polyspecific cells ± dump tetramer exclusion . CD4+ T-cells from pooled spleen and lymph node from unprimed adult and old mice after tetramer enrichment using anti-His magnetic beads . Shown are representative plots of the CD4+ T-cells binding E641:I-Ab + MCC:I-Ek tetramers pre ( left ) and post ( right ) gating out the dump , OVA:I-Ab , tetramer+ cells in ( A ) adult and ( B ) old mice . Absolute number of CD44lo ( left Y-axis ) and CD44hi ( right Y-axis ) CD4+ T-cells binding E641:I-Ab + MCC:I-Ek tetramers ( C ) pre and ( D ) post gating out the dump , OVA:I-Ab , tetramer+ cells . Horizontal bar indicates median ( p < 0 . 005 ns = non-significant; Mann–Whitney ) . Δ: Fold change in average numbers from adult to old mice . Data are from three experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 013 More naive and memory cells were observed to bind a single pMHC specificity in old mice compared with adults when using dump tetramer gating ( Figure 2G–I and Figure 2—figure supplements 5 , 6 ) ( Savage et al . , 1999 ) . This indicates that the increase in CD4+ T-cells binding E641:I-Ab is not unique . Rather , since CD4+ T-cells decline with aging , those that are left appear to bind foreign pMHC more promiscuously . Consistent with this interpretation , the number of naive cells binding OVA+MCC was higher in old mice compared with adults , as were the number of memory cells binding E641+OVA or OVA+MCC ( Figure 2J–L and Figure 2—figure supplements 7–9 ) . Altogether , these data provide evidence that the CD4+ T-cell compartment becomes polyspecific over time . Such results could reflect age-related changes in thymic selection , homeostatic signals , or both . To evaluate the former , we enriched thymocytes from adult and old mice with E641:I-Ab , OVA:I-Ab , and MCC:I-Ek tetramers . The frequency of E641-bound CD4 single positive ( SP ) cells was higher for old thymocytes compared with the adults , while the frequency of OVA and MCC-bound CD4SPs did not differ ( Figure 3 and Figure 3—figure supplement 1 ) . CD4SPs binding two distinct tetramers were not detected amongst the small number of tetramer-enriched samples . This is not surprising given that dual binders average <10% of a peripheral population ( Figure 2—figure supplement 6 ) . Since thymic output remains constant as a function of size over time ( Hale et al . , 2006 ) , the higher frequency of E641-bound CD4SP thymocytes in old mice suggests that more E641-binders leave the thymus of old mice than adults on a daily basis . However , mature CD4+ T-cells re-entering the thymus increase from ∼10% in adult mice to ∼20% in old mice ( Hale et al . , 2006 ) . Our analysis cannot resolve CD4SPs from mature CD4+ T-cells , so the impact of recirculation on our analysis is unclear . Nevertheless , the data suggest that age-related changes in thymic selection impact the clonal representation and binding capacity of the CD4+ T-cell compartment . 10 . 7554/eLife . 05949 . 014Figure 3 . Evidence for changes in selection of E641-binding CD4SP thymocytes with aging . Frequencies of ( A ) E641:I-Ab+ , ( B ) OVA:I-Ab+ , and ( C ) MCC:I-Ek+ CD4 single positive ( SP ) thymocytes per 107 CD4SP thymocytes are shown . Horizontal bar indicates median ( p < 0 . 05 and ns = non-significant; Mann–Whitney ) . Each dot represents the results from 4–5 mice pooled/group as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 01410 . 7554/eLife . 05949 . 015Figure 3—figure supplement 1 . Dump tetramer plus two-color gating scheme for identification of tetramer+ CD4SP thymocytes . Representative gating scheme for identifying tetramer+ cells for Figure 3 . ( A ) Forward vs side scatter . ( B ) CD3+ Thymocytes . ( C ) CD4SP T-cells . After anti-His magnetic enrichment , ( D–G ) E641:I-A+ CD4SP thymocytes cells were identified by dumping MCC:I-Ek+ and OVA:I-Ab+ bound cells . Similar gating strategy was employed to identify MCC:I-Ek+ and OVA:I-Ab+ CD4+ T-cells . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 015 Finally , we investigated how tonic TCR engagement relates to the capacity of the CD4+ T-cell compartment to bind foreign-pMHC ( Mandl et al . , 2013 ) . CD5 levels on the tetramer-bound adult populations , relative to those on the total adult CD4+ T-cell population , directly correlated with the fold increase in the absolute number of these populations over time ( Figure 4A , B ) . Steady-state BrdU incorporation for adult and old tetramer-bound CD4+ T-cells also mirrored the rank order ( OVA>E641>MCC ) of CD5 expression seen in both the naive and memory populations ( Figure 4C , D ) . Thus , CD5 levels are predictive of the fold-increase in pMHC-specific CD4+ T-cell subsets with aging , suggesting a link between affinity for self-pMHC , homeostatic proliferation , and expansion over time . 10 . 7554/eLife . 05949 . 016Figure 4 . CD5 levels on adult CD4+ T-cells correlate with expansion over time . Correlation between CD5 RFI for adult CD4+ tetramer+ T-cells and fold change in tetramer+ cells between adult and old populations of ( A ) CD44lo and ( B ) CD44hi CD4+ T-cells are shown as labeled . Linear regression was calculated using GraphPad Prism 5 . Steady-state in vivo proliferation was assessed by measuring percent BrdU incorporation in tetramer single+ ( C ) adult or ( D ) old CD4+ T-cells derived from unprimed mice after 6 days of BrdU exposure ( p < 0 . 05; ANOVA followed by Dunn's post-test comparison ) . Results represent two experiments with 4 mice/group . DOI: http://dx . doi . org/10 . 7554/eLife . 05949 . 016 Advances in the analysis of clonal representation , pMHC-binding capacity , and functionality within the T-cell repertoire are contributing to a broader understanding of the rules that govern its composition and function . While most studies focus on adult mouse or human T-cells , when immunity is at its peak , there is a growing appreciation that the pressures imposed by time on thymic selection and peripheral space result in a repertoire that continuously evolves in each individual . Here , we contribute to our basic understanding of T-cell biology by reporting that the size of the CD4+ T-cell compartment contracts with aging but , unlike CD8+ T-cells , the capacity of CD4+ T-cells to bind foreign-pMHC increases over the lifespan . Thymic involution could contribute to these changes in multiple ways . A decrease in cortical thymic epithelial cells and changes in antigen processing could increase competition for positively selecting pMHC ( Chinn et al . , 2012; Klein et al . , 2014 ) , favoring higher TCR affinity for self-pMHC . In addition , decreased expression of TRAs on fewer medullary TECs ( Chinn et al . , 2012; Griffith et al . , 2012 ) could lead to competition for negatively selecting pMHC with aging . Experimentally limiting thymic selection differentially impacts the CD4+ and CD8+ T-cell compartments , with CD4+ T-cells becoming more polyspecific and CD8+ T-cells becoming more pMHC focused ( Huseby et al . , 2005; Chu et al . , 2009 , 2010; Wang et al . , 2009; Yin et al . , 2012 ) . Thus , age-related changes in the thymus would be expected to restrict negative selection and result in a CD4+ T-cell compartment with a broader binding capacity , as observed here . It is also noteworthy that T-cells can productively rearrange two TCRα subunits and express two TCRs that increase reactivity to self- and allo-pMHC ( Ni et al . , 2014 ) . Whether T-cells expressing two TCRs increase over time remains unexplored . Changes in peripheral space are also likely to contribute to the results reported here . A link between higher affinity for self-pMHC and residence within the CD4+ T-cell memory pool of adult mice was previously reported ( Mandl et al . , 2013 ) . Here , we extended this observation to naive and memory CD4+ T-cells in old mice , indicating that affinity for self-pMHC influences clonal fitness over time . This would be akin to the affinity of TCR–pMHC interactions influencing clonal fitness within a polyclonal response to cognate antigens ( Lanzavecchia and Sallusto , 2002; Gett et al . , 2003; Malherbe et al . , 2004 ) . Indeed , CD5 levels on adult tetramer-binding memory subsets directly correlated with their fold expansion over the lifespan showing that CD5 levels have a clear predictive value when identifying populations with a long-term advantage for clonal representation within the CD4+ T-cell compartment . Altogether , the data presented here suggest a more complex relationship between CD4+ T cells and immune senescence than has been reported for the CD8+ T cells . While an increase in binding capacity may compensate for a decrease in total CD4+ T cell numbers , the consequences of this increase remain unclear . Certainly , a population with a higher affinity for self-pMHC and broader binding capacity poses obvious risks that could explain the increase in age-related autoimmune diseases , such as rheumatoid arthritis and giant cell arteritis ( Weyand et al . , 2003; Mohan et al . , 2011 ) . Coupling functional analysis with the results presented here will be important to gain a better understanding of the functionality of the CD4+ T cell compartment over the lifespan . Old ( 18–22 months ) male C57BL/6 mice were obtained from the National Institute of Aging breeding colony Bethesda , MD . Adult ( 8–12 weeks ) male C57BL/6 mice were purchased from the Jackson Laboratory Bar Harbor , Maine . Mice were maintained under specific pathogen-free conditions in the animal facility at The University of Arizona . Experiments were conducted under guidelines and approval of the Institutional Animal Care and Use Committee of The University of Arizona . Synthetic peptides Env 641–655 ( E641: PVGRLVTVNPFVSVA ) and OVA 323–339 ( OVA: ISQAVHAAHAEINEAGR ) were purchased at >95% purity from 21st Century Biochemicals Marlborough , MA . Complete Freund's adjuvant ( CFA ) was purchased from Sigma–Aldrich St . Louis , MO . Mice were immunized with 50 µg peptide in 50 µl CFA on each side of the base of the tail . Class II pMHC monomers were generated with baculovirus expression vectors , based on pAcGP67A ( BD Pharmingen San Jose , CA ) , encoding acidic or basic leucine zippers ( generous gift of KC Garcia ) according to the approach of Teyton and colleagues ( Scott et al . , 1996 ) . The full extracellular domains of I-Ek alpha and I-Ab alpha were expressed as fusions with the acidic leucine zipper , a BirA acceptor peptide , and a 6× his tag . The full I-Ab beta extracellular domain was expressed as fusions with the WNV Env 641–655 or OVA 326–338 peptides at the N-terminus , via a short linker similarly to Kappler and colleagues ( Crawford et al . , 1998 ) , and at the C-terminus with the basic leucine zipper and a 6× his tag . I-Ek beta fused to moth MCC 88–103 was otherwise the same . Baculovirus stocks were made in Sf9 cells and large-scale protein production was performed in Hi5 cells as previously described ( Dukkipati et al . , 2006 ) . pMHC complexes were purified from media by affinity chromatography using Ni-NTA affinity resin ( Qiagen Valencia , CA ) followed by biotinylation with BirA ( Avidity , Aurora , CO . ) and size exclusion chromatography with a Superdex-200 column ( GE Healthcare Life Sciences Pittsburgh , PA ) . Tetramers were created by mixing biotinylated peptide:I-Ab or I-Ek monomers with PE ( Biolegend San Diego , CA ) - , APC ( Biolegend ) - , or PerCPCy5 . 5 ( eBiosciences San Diego , CA ) -conjugated streptavidin at a molar ratio of 4:1 ( Tetramer Concentration: 25 nM ) . Tetramer enrichment and analysis was performed as described previously ( Moon et al . , 2007 ) with slight modifications . Inguinal , cervical , axillary , popliteal , mesenteric , and lumbar lymph nodes were harvested along with the spleen from individual mice . Single-cell suspensions of lymph node and spleens were depleted of red blood cells with ACK lysis buffer ( Gibco Life Technologies Grand Island , NY ) and Fc blocked ( mAb 2 . 4G2 hybridoma supernatant + 2% mouse serum [Caltag Laboratories Burlingame , CA] , 2% rat serum [Jackson Immuno Research Laboratories , INC West Grove , PA] ) on ice for 20 min . Each tetramer was added at a final concentration of 25 nM and incubated at room temperature in the dark for 1 hr . Cells were washed in FACS buffer ( PBS + 2% FBS , 0 . 1% NaN3 ) and resuspended in a final volume of 200 µl containing 25 µl of anti-PE and 25 µl of anti-APC microbeads ( Miltenyi Biotec San Diego , CA ) for two-color analysis of cells binding a single pMHC tetramers ( Stetson et al . , 2002; Obar et al . , 2008; Nelson et al . , 2015 ) or 50 µl of anti-His microbeads for simultaneous enrichment of cells binding three independent pMHC tetramers . After 30-min incubation at 4°C , cells were washed , resuspended in 3 ml FACS buffer and passed over a LS magnetic column at 4°C ( Miltenyi Biotec ) according to the manufacturer's instruction . The columns were removed from the magnetic field and bound cells were eluted by allowing 4 ml of FACS buffer to pass through the column by gravity at 4°C . A second elution was performed by pushing 4 ml of FACS buffer through the column with a plunger at 4°C . The tetramer-enriched ‘bound’ fraction and an aliquot of flow-thru , or ‘unbound’ fraction , were stained with a cocktail of flourochrome-labeled antibodies for 30 min at 4°C ( anti-CD19 [eBiosciences] , anti-CD8α [eBiosciences] , anti-CD11c [eBiosciences] , anti-F4/80 [Biolegend] , anti-CD3 [eBiosciences] , anti-CD4 [eBiosciences] , anti-CD44 [eBiosciences] , anti-CD5 [BD Pharmingen] ) . Cells were washed , and the samples were analyzed with a LSRII cytometer ( Beckton Dickinson Franklin Lakes , NJ ) . Analysis was performed using FlowJo software ( Treestar Ashland , OR ) . Gating was performed as shown in figure supplements . Tetramers are composed of pMHC monomers and streptavidin ( SA ) conjugated to a fluorescent protein ( FP ) . Two-color tetramer enrichment and gating for a single pMHC specificity was performed as a method for reducing false-positives in tetramer analysis ( Stetson et al . , 2002; Obar et al . , 2008; Nelson et al . , 2015 ) . The operating principle followed here is that cells which bind to tetramers via TCR–pMHC interactions will fall on a diagonal , since binding should be proportional for each , while those that bind to the FP in a TCR-independent manner should fall off the diagonal ( Figure 2—figure supplements 1 , 3 ) . The dump tetramer approach was used for samples in which E641:I-Ab , OVA:I-Ab , and MCC:I-Ek tetramers were used in a single sample for enrichment with anti-His beads ( Savage et al . , 1999; Newell et al . , 2009 ) . Here , cells bound to a tetramer of interest ( e . g . , E641:I-Ab ) were gated and then those that also bound the other two tetramers ( e . g . , OVA:I-Ab and MCC:I-Ek ) were excluded as a ‘dump’ to enumerate cells bound to one tetramer species only ( Figure 2—figure supplement 5 ) . The dumped cells could be false-positives binding SA , MHC , or the dump tetramer-associated FP nonspecifically; however , they could be bound via TCR–pMHC interactions . Importantly , the operating principles of both the dump and two-color methods were employed to enumerate cells binding two tetramers at once . Specifically , to enumerate cells bound to a specific combination ( e . g . , E641:I-Ab + OVA:I-Ab ) , those also binding the third tetramer ( e . g . , MCC:I-Ek ) were dumped prior to enumerating cells bound to both tetramers of interest by two-color analysis . Combining the two approaches should enumerate cells binding tetramers via TCR–pMHC interactions and exclude dumped cells that bind SA , MHC , or the dump-associated FP and those excluded by two-color analysis that bind the FP associated with the tetramers of interest in a non-specific manner . Combining both approaches by sequential gating ( Figure 2—figure supplement 5 ) yielded the numbers shown in Figure 2J–L . The same overall results were achieved if quadrant gating was used for two-color analysis after applying the dump gate ( Figure 2—figure supplements 7–9 ) . Thymi from old and adult mice were harvested in 1 ml of un-supplemented RPMI . Thymi from 4–5 old mice were pooled in order to take into account the drop in the total number of thymocytes in old mice due to thymic involution and to increase the total number of cells for the tetramer enrichment processing . Thymi were incubated with 3 ml of Accutase ( eBiosciences ) at 37°C for 30 min to achieve optimal cell detachment . Single cell suspension of thymocytes was depleted of red blood cells with ACK lysis buffer . The total number of thymocytes in old mice was 10-fold lower ( ∼2 × 107 ) than adults ( ∼2 × 108 ) due to thymic involution . The adult samples were then normalized for comparison by pooling 2 × 107 thymocytes from 4 to 5 adult mice . Thymocytes were Fc blocked on ice for 20 min . Cells were stained with E641:I-Ab–PerCP-Cy5 . 5 , OVA:I-Ab–PE-Cy7 , MCC:I-Ek–PE and each of these tetramers in a common FP ( APC ) . Each tetramer was added at a final concentration of 25 nM . Tetramer enrichment was carried out as described above . The tetramer enriched ‘bound’ fraction and an aliquot of flow-thru , or ‘unbound’ fraction were stained with cocktail of flourochrome-labeled antibodies for 30 min at 4°C ( anti-CD19 , anti-CD8α , anti-CD11c , anti-F4/80 , anti-CD3 , anti-CD4 , anti-CD5 ) . Cells were washed and the samples were analyzed with a LSRII cytometer ( Beckton Dickinson ) . Analysis was performed using FlowJo software ( Treestar ) as shown in Figure 3—figure supplement 1 . The single color specificities of two of the tetramers ( e . g . , OVA and MCC ) were used as dump tetramers prior to two-color analysis of the third tetramer ( e . g . , E641 ) . TCR negative 58α−β− hybridomas cells were transduced with retroviral vectors encoding the OT II , 5c . c7 or 2B4 TCR , full-length CD3 subunits , and CD4 according to previously described protocols ( Kuhns and Davis , 2007 ) . BrdU was administered to mice through drinking water at the concentration of 1 mg/ml + 1% glucose . Spleen and lymph nodes were harvested on day 6 . Post-tetramer enrichment , cells were stained with cell surface antibodies ( anti-CD3 , anti-CD4 , anti-CD19 , anti-CD8α , anti-CD11c , anti-F4/80 , anti-CD44 , and anti-CD5 ) followed by intracellular anti-BrdU ( BD Pharmingen ) antibody according to BrdU flow kit protocol ( BD Biosciences ) . Mean fluorescent intensity of cell surface antibodies and intra-cellular antibodies were obtained from FlowJo software ( Treestar ) . Statistical analyses were performed using the Mann–Whitney t-test for non-parametric data , ANOVA followed by Dunn's post-test for multiple comparisons of non-parametric data or linear regression for analyzing correlation . All statistical analysis was performed using GraphPad Prism software .	The immune system's T cells help the body to recognize and destroy harmful pathogens , such as viruses and bacteria . T cells ‘remember’ immunity-inducing fragments , called antigens , from the pathogens they have encountered . This memory then allows the immune system to quickly fend off infections if those pathogens , or even related pathogens , invade again . Vaccines exploit the ability to form immunological memory by exposing the body to harmless forms of the pathogen , or even just particular antigens from it . This allows the T cells to learn how to identify the pathogen without any risk of illness . Vaccines have been extremely successful and have helped to virtually eliminate some diseases . However , for reasons that are unclear , the immune systems of older adults become less functional , so vaccines often lose their effectiveness . Paradoxically , as people age T cells become more likely to attack the body's cells , causing autoimmune diseases like arthritis . Understanding what happens to aging T cells to cause these immune changes may help scientists design vaccines that remain effective as people age . Little is known about what happens to a particular type of T cell—the CD4+ T cells—as people age , even though this population plays a critical role in providing other immune cells with detailed instructions on when and how to fight a pathogen . Now , Deshpande et al . show that CD4+ T cells undergo a remarkable set of changes in aging mice . Mice that are nearing the end of their natural lifespan have fewer CD4+ T cells than younger mice . However , those CD4+ T cells that remain are more likely than CD4+ T cells from younger mice to be able to recognize multiple antigens . This increase in the proportion of multitasking CD4+ T cells corresponds with an increased tendency of these cells to bind to the body's own cells . If similar changes occur in older people , this may help explain some age-related autoimmune diseases . Yet , the relationship between the increase in multitasking CD4+ T cells and the decrease in immune function with aging remains to be fully explored . The challenge for scientists now is to determine how these age-related changes in CD4+ T cells affect immune responses to vaccines or pathogens in older individuals . One implication of this work is that CD4+ T cell responses may be too robust and out of balance with other arms of the immune system . This could even lead to conditions such as autoimmunity . Alternatively , while there may be more CD4+ T cells that can multitask by recognizing multiple antigens , their ability to respond appropriately to infections or vaccinations may be diminished . What is clear from the work of Deshpande et al . is that the rules that have been defined for immunity in adults change with aging . The rules that govern immunity in the elderly must be more clearly defined to realize the goal of designing immunotherapies , such as vaccines , that provide protection throughout the lifespan .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "immunology", "and", "inflammation" ]	2015	Self-recognition drives the preferential accumulation of promiscuous CD4+ T-cells in aged mice
Loss of function of the X-linked gene encoding methyl-CpG binding protein 2 ( MeCP2 ) causes the progressive neurological disorder Rett syndrome ( RTT ) . Conversely , duplication or triplication of Xq28 causes an equally wide-ranging progressive neurological disorder , MECP2 duplication syndrome , whose features overlap somewhat with RTT . To understand which MeCP2 functions cause toxicity in the duplication syndrome , we generated mouse models expressing endogenous Mecp2 along with a RTT-causing mutation in either the methyl-CpG binding domain ( MBD ) or the transcriptional repression domain ( TRD ) . We determined that both the MBD and TRD must function for doubling MeCP2 to be toxic . Mutating the MBD reproduces the null phenotype and expressing the TRD mutant produces milder RTT phenotypes , yet both mutations are harmless when expressed with endogenous Mecp2 . Surprisingly , mutating the TRD is more detrimental than deleting the entire C-terminus , indicating a dominant-negative effect on MeCP2 function , likely due to the disruption of a basic cluster . Rett syndrome ( RTT ) , the most common monogenic cause of intellectual disability in females , is a debilitating , progressive neurological disorder that is caused by mutations in the X-linked gene encoding methyl-CpG binding protein 2 ( MeCP2 ) ( Amir et al . , 1999 ) . After a 6- to 18-month period of normal development , head growth slows and affected girls lose acquired speech , dexterity , and social skills; they then develop characteristic hand stereotypies , respiratory dysrhythmias , seizures , and autistic-like features ( Hagberg et al . , 1983; Lam et al . , 2000; Klauck et al . , 2002; Carney et al . , 2003 ) . Interestingly , duplications and triplications spanning the MECP2 region on Xq28 cause a similarly progressive neurological disorder called MECP2 duplication syndrome , which has some features that overlap with RTT . Children with the duplication syndrome present with infantile hypotonia and develop severe intellectual disability , autistic-like features , recurrent respiratory infections , spasticity , seizures , and premature lethality ( del Gaudio et al . , 2006; Friez et al . , 2006; Lugtenberg et al . , 2006; Meins et al . , 2005; Van Esch et al . , 2005 ) . MeCP2 was first identified over 20 years ago as a transcriptional repressor that binds to methylated CpG dinucleotides ( Lewis et al . , 1992; Wakefield et al . , 1999; Free et al . , 2001 ) . It binds DNA directly through its N-terminal methyl-CpG binding domain ( MBD ) , whereas its C-terminal transcriptional repression domain ( TRD ) allows it to interact with corepressors such as Sin3a , HDAC1 , and HDAC2 ( Nan et al . , 1998 ) . More recent work has revealed that MeCP2 is expressed at higher levels than expected for classical site-specific transcriptional repressors: it binds as abundantly and widely throughout the genome as histone H1 , which suggests MeCP2 might have additional functions in chromatin biology ( Nan et al . , 1997; Skene et al . , 2010 ) . Further complicating the picture of MeCP2 function are transcriptional studies in mouse brains as well as human embryonic stem cell-derived neurons , which have shown that most genes are actually downregulated in RTT models that lack MeCP2 ( Chahrour et al . , 2008; Ben-Shachar et al . , 2009; Li et al . , 2013 ) . One proposal to explain this is that MeCP2 acts as a ‘transcriptional noise dampener’ , such that loss of MeCP2 function results in the diversion of basal transcriptional machinery to repetitive elements , indirectly leading to global transcriptional downregulation ( Skene et al . , 2010 ) . Additional transcriptional studies present a challenge for this hypothesis , however , as the same genes that are downregulated in the RTT models are upregulated in MECP2 duplication syndrome mouse models with double the MeCP2 expression ( Chahrour et al . , 2008 ) . How might doubling MeCP2 levels lead to the upregulation of the very same genes ? It may be that , given the abundance of MeCP2 in the normal state , doubling the protein levels , as in the duplication syndrome , exhausts the supply of normal binding partners , thus diverting them from their normal functions , but this possibility has not been formally tested . Because most RTT-causing mutations disrupt either the MBD or TRD , we decided to use a genetic approach to determine how these regions might be involved in mediating the toxicity observed in MECP2 duplication syndrome models . Studying transgenic models with known RTT-causing missense mutations on a Mecp2 null background would allow us to observe the effect of the mutation when it is the only allele present , as in RTT , whereas studying the effects of the transgene on a wild type ( WT ) background would enable us to study the importance of each domain to the neurotoxicity caused by MECP2 duplication syndrome . To explore the role of the MBD in RTT and MECP2 duplication syndrome , we chose a RTT-causing point mutation that replaces an arginine with a glycine at residue 111 ( R111G ) ( Laccone et al . , 2001 ) . R111 is critical for methyl-CpG binding in vitro: when it is mutated ( R111G ) , it completely abolishes binding to methyl-CpGs in vitro without affecting the structure of the MBD or the rest of the protein ( Free et al . , 2001; Kudo et al . , 2003 ) . We selected this mutation to determine whether MeCP2 retains any functions without a functional MBD , whether the MBD is required for the toxicity observed in duplication models , and whether doubling MeCP2 mediates toxicity by over-titrating its binding partners . Much less is known about the TRD , but the importance of the C-terminus is underscored by the clustering of common RTT-causing mutations at the very end of the TRD ( Christodoulou et al . , 2003 ) . In fact , the mutation of arginine to cysteine at residue 306 ( R306C ) is the second most common RTT-causing missense mutation . Although recent studies have shown that the R306C mutation abolishes interaction with the NCoR corepressor complex , a mouse model bearing a truncated form of MeCP2 that lacks the NCoR binding site has a milder phenotype than the published mice carrying the R306C mutation , indicating that R306C has other effects on the function of MeCP2 ( Baker et al . , 2013; Ebert et al . , 2013; Lyst et al . , 2013 ) . To study the functional consequences of these two mutations in vivo , we generated independent transgenic mouse models , each bearing one of these point mutations in the MECP2 locus . We characterized the behavioral and molecular phenotypes of mice carrying only the mutant allele ( and thus modeling RTT ) and mice carrying the mutant allele in addition to a WT allele ( modeling MECP2 duplication syndrome ) in an effort to determine how each mutation affects the function of the protein in the two diseases . To generate a modified MECP2 allele , we used a PAC ( P1-derived artificial chromosome , PAC671D9 ) containing the entire human MECP2 locus and its essential regulatory elements , but no other genes . This PAC clone was used to create the MECP2 duplication mouse model and predict the human disease , and on a Mecp2 null background it rescues loss-of-function phenotypes ( Collins et al . , 2004 ) . We used recombineering to modify the locus to carry either the R111G or the R306C mutation ( Warming et al . , 2005 ) . This PAC has reliably produced founders with protein levels and expression pattern similar to that of endogenous MeCP2 ( Collins et al . , 2004 ) . We further modified the PAC to tag MeCP2 with enhanced green fluorescent protein ( EGFP ) at the C-terminus to better identify interactors through immunoprecipitation ( IP ) and visualize their localization using immunofluorescence ( IF ) without disrupting MeCP2 function . Arginine 111 ( R111 ) is located in the middle of the MBD , whereas arginine 306 ( R306 ) is in a cluster of basic amino acids at the end of the TRD ( Figure 1A ) . After establishing transgenic lines for each mutation , we characterized the protein levels of two independent lines for each mutation by western blot and found that they expressed transgenic MeCP2 at levels similar to those of endogenous MeCP2 , thus doubling the total MeCP2 level ( Figure 1B ) . Because doubling MeCP2 consistently causes phenotypes similar to the human syndrome , we chose one line ( Line 1 ) of each mutant for further characterization ( Collins et al . , 2004 ) . Immunofluorescence using anti-MeCP2 and anti-GFP antibodies on midsagittal sections of the whole brain shows that the distribution pattern of transgenic MeCP2 parallels that of endogenous MeCP2 ( Figure 1C , upper panel ) . This colocalization also holds true upon closer inspection of the cortex and cerebellum ( Figure 1C , lower panel ) . 10 . 7554/eLife . 02676 . 003Figure 1 . Generation of transgenic lines . ( A ) Schematic representation of MeCP2-EGFP fusion protein , showing the methyl-CpG binding domain ( MBD , blue ) , transcriptional repression domain ( TRD , red ) , a cluster of basic amino acids ( yellow ) , and enhanced green fluorescent protein ( EGFP , green ) . The location of the residues arginine 111 ( R111 ) and arginine 306 ( R306 ) are shown . ( B ) Western blot analyses of whole brain lysates show that two independent lines of MeCP2-R111G and MeCP2-R306C express the transgene at levels similar to endogenous MeCP2 as judged by the similar intensities of WT and transgenic bands using an anti-MeCP2 antibody . Quantification of total MeCP2 is shown to the right for each mutant . An anti-GAPDH antibody was used to detect GAPDH as a loading control . ( C ) Immunofluorescence ( IF ) of midsagittal brain sections ( 2 months , upper panel ) shows that the expression pattern of the transgenes throughout the whole brain ( anti-GFP antibody , green ) parallels that of endogenous MeCP2 ( anti-MeCP2 antibody , red ) . Colocalization is visible as yellow in the merged image , with 4' , 6-diamidino-2-phenylindole ( DAPI , blue ) as a general marker for the nucleus . Closer examination of the cortex and cerebellum ( lower panel ) mirrors this . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 003 We crossed the transgenic males ( MECP2R111G-EGFPTg or MECP2R306C-EGFPTg ) to Mecp2+/− females ( Guy et al . , 2001 ) . These crosses resulted in males that were wild type , null , modeled RTT , or modeled the duplication syndrome , specifically: ( 1 ) Mecp2+/y ( hereafter WT ) , ( 2 ) Mecp2−/y ( hereafter null ) , ( 3 ) MECP2R111G-EGFPTg; Mecp2−/y or MECP2R306C-EGFPTg; Mecp2−/y ( hereafter R111G or R306C , respectively ) , and ( 4 ) MECP2R111G-EGFPTg; Mecp2+/y or MECP2R306C-EGFPTg; Mecp2+/y ( hereafter R111G Tg or R306C Tg , respectively ) . Note that the R111G and R306C males express only the mutated version of MeCP2 , whereas the R111G Tg and R306C Tg males express both endogenous Mecp2 as well as the mutated MECP2 . Mecp2 null mice started displaying symptoms around 4–6 weeks of age and had a median lifespan of 11 weeks ( Guy et al . , 2001 ) ( Figure 2A , red line ) , and the R111G mice died as prematurely as the null mice ( Figure 2A , green line ) . This indicates that the R111G mutation disrupts a critical function of MeCP2 , such that it becomes effectively a null allele . 10 . 7554/eLife . 02676 . 004Figure 2 . R111G mice phenocopy the null mice whereas R111G Tg mice are indistinguishable from WT mice . ( A ) Using a Kaplan–Meier survival curve , both WT ( black , n = 11 ) and R111G Tg ( blue , n = 10 ) have a normal lifespan , whereas R111G mice ( green , n = 10 ) phenocopy the premature lethality of the null mice ( red , n = 11 ) , with a median lifespan of 11 weeks . ( B–I ) Behavioral assays performed on WT ( n = 7 ) and R111G Tg ( n = 7 ) mice at 9 months of age show that R111G Tg mice are indistinguishable from WT mice in a variety of assays . The open field assay reveals that both lines travel the same distance ( B ) and have the same percentage of time traveled in the center ( C ) , indicating that they are not anxious , which is confirmed by the percentage of time spent in the dark in the light/dark assay ( D ) . Motor function and coordination were unchanged as measured by time spent on the rod ( for four trials a day for four days ) in rotarod ( E ) and footslips per centimeter traveled in parallel rod footslip ( F ) . Purposeful paw movement is also normal , as assayed by nest building after 24 hr ( G ) . R111G Tg mice also have no learning and memory deficits , as assayed by their freezing in the contextual ( H ) and cued ( I ) conditioned fear test . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± SEM . n . s . not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 004 Although the R111G Tg had a lifespan like the WT and appeared normal , we wanted to determine if they had any subtle behavioral deficits . Older mice expressing two functional copies of MeCP2 exhibit mild phenotypes , such as hypoactivity , reduced anxiety , and improved learning ( Collins et al . , 2004; Na et al . , 2012 ) , so we aged the WT and R111G Tg mice to 9 months before starting our battery of behavioral tests . We examined overall activity and anxiety using distance traveled in the open field assay ( Figure 2B ) , the center-to-total ratio in the open field assay ( Figure 2C ) , and time spent in the dark in the light/dark assay ( Figure 2D ) . We also assayed motor coordination by their ability to stay on a rotating rod in the rotarod assay ( Figure 2E ) , footslips per distance traveled in parallel rod footslip ( Figure 2F ) , and nest building after 24 hr ( Figure 2G ) . We assessed learning and memory with the contextual and cued conditioned fear test ( Figure 2H–I ) . We found no difference between WT and R111G Tg mice in any of these behaviors ( Figure 2B–I ) , indicating that an intact MBD is necessary to elicit the toxicity seen in the duplication syndrome model . These data also show that the phenotypes of the R111G mice ( when the only MeCP2 allele is mutated ) are exclusively caused by the mutation , and not insertional effects of the transgene , as the presence of a WT MeCP2 allele ( in the R111G Tg mice ) displays no phenotypes . In vitro , MeCP2-R111G abolishes binding to methyl-CpGs ( Free et al . , 2001; Kudo et al . , 2003 ) . This had never been shown in vivo , however , so we used immunofluorescence ( IF ) and DAPI staining to visualize the localization of MeCP2 to heterochromatic foci , which are regions of highly compacted chromatin . The localization of a protein to these sites can serve as a proxy for methyl-CpG binding , and WT MeCP2 usually localizes there ( Figure 3A , upper panel ) . We found that MeCP2-R111G was unable to localize to heterochromatic foci ( Figure 3A , lower panel ) , suggesting it is unable to bind methyl-CpGs in vivo . To verify that abolishing methyl-CpG binding has no effect on the ability of MeCP2 to interact with binding partners , we performed in vivo immunoprecipitation ( IP ) to determine whether MeCP2-R111G can interact with the Sin3a , HDAC1 , and HDAC2 of the Sin3a corepressor complex and HDAC3 and Tbl1 of the NCoR corepressor complex ( Nan et al . , 1998; Stancheva et al . , 2003 ) . As expected , MeCP2-R111G retained the ability to interact with both corepressor complexes to the same degree as MeCP2-EGFP ( Figure 3B ) . This demonstrates that the R111G mutation does not affect the ability of MeCP2 to bind corepressors , despite the fact that it disrupts binding to methyl-CpGs . 10 . 7554/eLife . 02676 . 005Figure 3 . R111G abolishes MeCP2 binding to methyl-CpGs but does not affect its ability to interact with known binding partners . ( A ) MeCP2-R111G does not bind to methyl-CpGs in vivo . Immunofluorescence for endogenous MeCP2 ( anti-MeCP2 antibody , red ) , the transgene ( anti-GFP antibody , green ) , and heterochromatic foci ( puncta within the nucleus visualized by DAPI , blue ) , shows that endogenous MeCP2 localizes to the heterochromatic foci and can therefore bind methyl-CpGs , whereas MeCP2-R111G is diffuse within the nucleus . ( B ) MeCP2-R111G retains the ability to bind known interactors in vivo . Immunoprecipitation ( IP ) with anti-GFP antibody was used to determine if MeCP2-R111G binds known interactors . Using WT mice lacking GFP as a negative control and MeCP2-EGFP as a positive control , we show that MeCP2-R111G retains binding to Sin3a , HDAC1 , and HDAC2 of the Sin3a complex and HDAC3 and Tbl1 of the NCoR complex . I = input , E = eluate . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 005 We next assessed the RTT-causing R306C mutation , located in the TRD . As with the R111G Tg mice , the R306C Tg mice were indistinguishable from their wild-type littermates by weight , lifespan , and brain size ( Figure 4A–C ) . The R306C mice , however , displayed a milder phenotype than the Mecp2 null mice . As early as 2 months of age , visual appraisal was sufficient to distinguish them: WT ( and R306C Tg ) mice had smooth coats , but the null mice appeared disheveled and the R306C mice were in-between . This milder phenotype was also reflected in the survival ( Figure 4A ) and weight ( Figure 4B ) curves: the null mice died early , with a median survival of 11 weeks ( Figure 4A , red line ) and were severely overweight ( Figure 4B , red line ) , but the R306C mice had a median survival of 18 weeks ( Figure 4A , green line ) and were mildly overweight ( Figure 4B , green line ) . Both R306C and null mice had smaller brains , mirroring the microcephaly seen in RTT patients ( Figure 4C ) . 10 . 7554/eLife . 02676 . 006Figure 4 . R306C mice have a milder phenotype than null mice . ( A ) A Kaplan–Meier survival curve shows that WT ( black , n = 28 ) and R306C Tg ( blue , n = 22 ) mice have a normal lifespan , while R306C ( green , n = 21 ) mice have a shortened lifespan with 50% lethality at 18 weeks , but not as severely as the null mice ( red , n = 26 ) , with 50% survival at 11 weeks . ( B ) A weight curve shows that WT and R306C Tg mice have normal weights whereas R306C mice have a milder obesity phenotype . R306C mice are more overweight than WT and R306C Tg mice , but not as severely affected as the null mice . ( C ) R306C mice have smaller brains . At 7 weeks of age , WT and R306C Tg mice have similar brain sizes , whereas R306C and null mice have brains that are about 85% the normal weight . n = 7 per genotype . ( D ) ATRX localization to heterochromatic foci is used as a marker of disease severity . At 7 weeks of age , the number of ATRX-positive foci per cell ( quantification to the upper right ) and focus intensity ( quantification to the lower right ) is indistinguishable in WT and R306C Tg mice , while they are decreased in R306C , and even more decreased in null mice . The circled cell ( white ) in the merge of the left panel is shown at increased magnification in the center panel . Immunofluorescence was performed using DAPI ( blue ) and an anti-ATRX antibody ( red ) . n = 3 per genotype . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± SEM . p<0 . 05 , p<0 . 01 , *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 006 Localization of Alpha Thalassemia/Mental Retardation Syndrome X-Linked ( ATRX ) to heterochromatic foci has recently been demonstrated to be a cell-autonomous marker of disease progression in other RTT mouse models ( Baker et al . , 2013 ) , so we evaluated ATRX localization using IF on comparable sections of the CA1 , CA2 , and CA3 regions of the hippocampus from 7-week-old mice ( Figure 4D ) . The number of ATRX-positive foci per cell and focus intensity appeared to be the same in WT and R306C Tg ( Figure 4D , upper two panels and quantification to the right ) . Null mice had the fewest ATRX-positive foci per cell and the lowest focus intensity , and R306C mice had an intermediate number of ATRX-positive foci per cell and focus intensity ( Figure 4D , lower two panels and quantification to the right ) . We next evaluated the animals for behavioral phenotypes . To fully characterize the R306C mice , we performed several behavioral tests at both 5 and 11 weeks of age . Most null mice die by 11 weeks , but they served as a point of comparison at the earlier timepoint; we were also able to compare later-manifesting phenotypes against the duplication model mice ( R306C Tg ) . At both 5 and 11 weeks , the R306C Tg mice were indistinguishable from WT mice ( Figure 5 , comparing blue to black ) . 10 . 7554/eLife . 02676 . 007Figure 5 . R306C mice recapitulate many RTT phenotypes . ( A–E ) Behavioral tests were performed at 5 weeks of age to examine the R306C phenotype . At this age , WT ( black , n = 18 ) and R306C Tg ( blue , n = 17 ) mice are indistinguishable . Like the null mice ( red , n = 14 ) , R306C mice ( green , n = 16 ) have increased anxiety , as measured by increased time spent in the dark in the light/dark assay ( A ) , purposeful paw movement deficits as measured by their ability to build nests ( B ) , motor dysfunction as measured by increased footslips per cm traveled in parallel rod footslip ( C ) , and learning and memory deficits in contextual ( D ) and cued ( E ) fear conditioning . ( F–J ) Behavioral tests were performed on a new cohort of mice at 11 weeks of age , when most of the null mice have succumbed to disease , to examine any changes in the observed phenotypes . WT ( n = 17 ) and R306C Tg ( n = 15 ) mice remain indistinguishable at this age . R306C ( n = 15 ) mice were hypoactive in the open field assay ( F ) and continue to exhibit increased anxiety as measured by decreased time spent in the center in the open field assay ( G ) . Additionally , their motor dysfunction and learning and memory deficits have both worsened ( H–J ) . The motor coordination of older R306C Tg mice was tested with a new cohort of 5 month old WT ( n = 12 ) and R306C Tg ( n = 11 ) mice subjected to rotarod ( four trials a day for 4 days ) , and no difference was observed in their time spent on the rod ( K ) . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± SEM . p<0 . 05 , p<0 . 01 , p<0 . 001 , **p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 007 In contrast , at 5 weeks of age , R306C and null mice exhibited increased anxiety , as measured by increased time spent in the dark in the light/dark assay ( Figure 5A ) . They displayed poor nest building ( Figure 5B ) and increased footslips per distance traveled in parallel rod footslip ( Figure 5C ) , indicating deficits in purposeful paw movements and motor dysfunction . They also showed learning and memory deficits as assayed by freezing in the contextual and cued conditioned fear assay ( Figure 5D–E ) . Notably , in all tests , R306C mice had a less severe phenotype than the null mice . By 11 weeks of age , R306C mice had become hypoactive as measured by distance traveled in the open field assay ( Figure 5F ) . They continued to exhibit anxiety as demonstrated by their decreased time spent in the center in the open field assay ( Figure 5G ) . Their motor dysfunction worsened ( Figure 5H ) , as did their learning and memory deficits ( Figure 5I–J ) . Additionally , a new 5-month old cohort showed that older R306C Tg mice have no motor coordination phenotype , as demonstrated by their ability to stay on a rotating rod in the rotarod assay similarly to WT ( Figure 5K ) . In summary , the R306C Tg mice displayed no phenotypes , indicating that the TRD and C-terminus must be intact to observe the toxic effects of doubling MeCP2 . Mice carrying the RTT-causing R306C mutation recapitulated many of the phenotypes seen in RTT , including learning and memory deficits and motor dysfunction , indicating that this mouse model is a reliable model of RTT that can be used for future studies . As with the R111G Tg mice , the absence of phenotypes in the R306C Tg mice demonstrates that the phenotypes observed in the R306C mice are caused by the mutation and not insertional effects of the transgene . Because R306 is not located in or near the MBD , we did not predict that mutating it would affect binding to methyl-CpGs in vivo . We performed IF to visualize heterochromatic foci as before and confirmed that MeCP2-R306C retains the ability to bind methyl-CpGs in vivo in a fashion similar to endogenous MeCP2 ( Figure 6A ) . 10 . 7554/eLife . 02676 . 008Figure 6 . R306C does not affect methyl-CpG binding , but alters binding to a subset of known interactors . ( A ) MeCP2-R306C binds methyl-CpGs in vivo . Immunofluorescence for endogenous MeCP2 ( anti-MeCP2 antibody , red ) , the transgene ( anti-GFP antibody , green ) , and heterochromatic foci ( puncta within the nucleus visualized by DAPI , blue ) shows that both endogenous MeCP2 and MeCP2-R036C localize to the heterochromatic foci and can therefore bind methyl-CpGs . ( B ) MeCP2-R306C has altered binding to a subset of known interactors . Immunoprecipitation ( IP ) with anti-GFP antibody was used to determine if MeCP2-R306C binds known interactors . Using WT mice lacking GFP as a negative control and MeCP2-EGFP as a positive control , we show that MeCP2-R306C retains binding to Sin3a , HDAC1 , and HDAC2 of the Sin3a complex , but not to HDAC3 and Tbl1 of the NCoR complex . I = input , E = eluate . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 008 Since early reports suggested that the function of the C-terminus is to bind the corepressors Sin3a , HDAC1 , and HDAC2 ( Nan et al . , 1998 ) , we performed in vivo IP to assess the ability of MeCP2-R306C to interact with these corepressors . We found , to our surprise , that MeCP2-R306C bound Sin3a , HDAC1 , and HDAC2 similarly to WT EGFP-tagged MeCP2 ( Figure 6B , upper half ) . We next tested components of the NCoR complex ( Stancheva et al . , 2003 ) . In agreement with a recent study ( Lyst et al . , 2013 ) , we found that R306C greatly reduces interaction with HDAC3 and Tbl1 of the NCoR corepressor complex ( Figure 6B , lower half ) . Although the loss of NCoR binding could play a role in the phenotypes we observe , we noticed a striking result when we compared the lifespans of previously published RTT models and the new R306C mice ( Figure 7A ) . Mice with a truncating mutation , G273X , lack the entire region mapped to interact with NCoR , so MeCP2-G273X , like MeCP2-R306C , is unable to bind NCoR . However , the median lifespan of R306C mice ( 18 weeks , this study , Lyst et al . , 2013 ) is much shorter than that of G273X mice ( 28 weeks , Baker et al . , 2013 ) . This led us to consider the potential differences between the two alleles . Because MeCP2 shows affinity for specific sequences , we tested whether these mutants differed in their ability to bind these sequences by performing in vivo chromatin immunoprecipitation ( ChIP ) experiments on brain tissue ( Chahrour et al . , 2008; Skene et al . , 2010; Baker et al . , 2013 ) . We found that MeCP2-R306C localized to the promoters of major satellite , L1 retrotransposons , somatostatin ( Sst ) , afamin ( Afm ) , corticotropin releasing hormone ( Crh ) , and Gapdh , but bound to a lesser extent than MeCP2-G273X ( Figure 7B ) . 10 . 7554/eLife . 02676 . 009Figure 7 . R306C decreases the affinity of the C-terminus of MeCP2 to DNA and reduces its DNA occupancy . ( A ) An overview of the lifespans of RTT mouse models emphasizes the fact that the R306C missense mutation is more detrimental than an earlier truncation at G273X . Full-length MeCP2 ( gray ) with the MBD ( blue ) , TRD ( red ) , and basic cluster ( yellow ) is depicted . The top shows truncating mutations , while the bottom shows the missense mutations we generated . The median lifespans of mice carrying the mutations are shown in weeks . ( B ) Chromatin immunoprecipitation ( ChIP ) using an anti-GFP antibody followed by quantitative PCR ( qPCR ) to determine enrichment of the following amplicons: Major Satellite repetitive sequences , L1 retrotransposons ( L1 ) , somatostatin ( Sst ) , afamin ( Afm ) , corticotropin releasing hormone ( Crh ) , and Gapdh shows that MeCP2-R306C has decreased binding to sequences known to be bound by MeCP2 in vivo . Comparing MeCP2-R306C ( green , n = 7 ) to MeCP2-G273X ( orange , n = 7 ) shows that the differences observed are due to an additional effect of the C-terminus that is absent in the full-length MeCP2-EGFP ( black , n = 4 ) . WT mice with no GFP ( gray , n = 3 ) were used as a negative control . ( C ) An electrophoretic mobility shift assay ( EMSA ) using increasing amounts of N-terminally GST-tagged C-terminal human MeCP2 recombinant protein ( amino acids 274–340 , from left to right: WT , R306C , R306H , K304E ) shows that the WT C-terminal fragment of MeCP2 can bind DNA and that RTT-causing mutations altering the charge of basic residues in the fragment abolish this binding . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± SEM . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02676 . 009 Since the R306C mutation reduces binding to DNA in vivo , we aimed to determine the mechanism by which this occurs . Thus , we focused on the C-terminus of the protein and generated N-terminally GST-tagged C-terminal fragments of MeCP2 ( amino acids 274–340 ) with and without the R306C mutation , which allowed us to uncover more subtle differences without the overwhelming DNA binding of the MBD . We performed electrophoretic mobility shift assays ( EMSAs ) to determine whether the C-terminal basic cluster flanking R306 binds DNA and further test whether the R306C mutation alters this binding . The WT MeCP2 fragment binds to the double-stranded DNA probe with increasing amount of protein , but the R306C mutation completely abolished the ability of the fragment to bind DNA ( Figure 7C , two left panels ) . We further confirmed the importance of the basic cluster flanking R306 by performing EMSAs with two other RTT-causing point mutations: arginine to histidine at residue 306 ( R306H ) and lysine to glutamic acid at residue 304 ( K304E ) , both of which remove a basic charge from the cluster ( Christodoulou et al . , 2003 ) . These two mutations also abolish the ability of the fragment to bind DNA ( Figure 7C , two right panels ) , thus emphasizing the importance of the basic cluster in C-terminal DNA binding . While these experiments were performed using a truncated MeCP2 peptide lacking the MBD ( amino acids 274–340 ) , we believe that the loss of this DNA binding is critical . In vivo ChIP analysis supports this notion , as despite the presence of the MBD and two functional AT-hooks in both MeCP2-G273X and MeCP2-R306C , the disruption of the basic cluster in MeCP2-R306C reduces binding to DNA even more than the complete absence of the cluster in MeCP2-G273X ( Figure 7B ) . R306 therefore has an influence on DNA binding itself in addition to its role in NCoR complex binding , and having a defective C-terminus ( R306C ) is worse than having no C-terminus at all . To gain insight into both RTT and MECP2 duplication syndrome , we generated transgenic mouse models of MeCP2 carrying the R111G or the R306C missense mutation and compared them to the existing Mecp2 null mice . Using the transgene on a Mecp2 null background , we explored how the mutations disrupt normal function in RTT; using the transgene on a wild type background allowed us to explore whether the remaining functions of the mutant alleles can cause MECP2 duplication syndrome . Our investigation of the R111G mutation led to two important conclusions about the function of MeCP2 . Since previous NMR studies demonstrated that this mutation abolishes binding to methyl-CpGs without affecting the structure of the MBD or the rest of the protein ( Free et al . , 2001 ) , we used this mutation to assess the necessity of methyl-CpG binding in both RTT and MECP2 duplication syndrome . If excessive MeCP2 disrupts neuronal function by diverting corepressors away from loci they normally repress , then abolishing methyl-CpG binding in an extra copy of MeCP2 that is still capable of interacting with corepressors ( as we show in this study ) would still produce the duplication phenotype in mice . Instead , the R111G Tg mice were indistinguishable from wild-type mice . Therefore , overabundant MeCP2 does not exert toxic effects because it diverts corepressors away from their normal binding sites: rather , trouble arises when more abundant MeCP2 occupies more methyl-CpGs . The centrality of methyl-CpG binding to MeCP2 function was underscored by the R111G mice , which phenocopied the early lethality of the Mecp2 null mice . All of this is consistent with studies asserting that lack of methyl-CpG binding leads to RTT ( Kudo et al . , 2003 ) . Like the R111G Tg mice , the R306C Tg mice were also indistinguishable from WT mice , demonstrating that a functional TRD is as necessary for the toxic effects of MeCP2 duplication as a functional MBD . The R306C mice , though displaying milder phenotypes than Mecp2 null mice , recapitulated many phenotypes seen in RTT patients , such as increased anxiety , motor dysfunction , and learning and memory deficits ( Hagberg et al . , 1983; Chahrour and Zoghbi , 2007 ) . This line is therefore a reliable mouse model for studying RTT and should be particularly useful since this is one of the most common RTT-causing mutations ( Christodoulou et al . , 2003 ) . The R306C mutation proved most useful for understanding the function of the less well-studied MeCP2 C-terminus . Because the G273X mouse model , which lacks the entire MeCP2 C-terminus , has a median survival of 28 weeks ( Baker et al . , 2013 ) —longer than both our R306C transgenic and the recently published R306C knock-in ( median survival 18 weeks , this study , Lyst et al . , 2013 ) —we hypothesized that the R306C allele must have an additional compromised function beyond its inability to bind NCoR through its TRD . Indeed , we found that MeCP2-R306C has reduced affinity for MeCP2 binding sequences in vivo , likely due to the disruption of a highly conserved basic cluster of amino acids ( 304-309 ) flanking R306 , rendering it unable to bind DNA . This study has added to our understanding of MeCP2 function by discovering that the basic cluster flanking R306 has an additional function in binding DNA and that this binding is abolished when basic residues in the cluster are mutated to neutral or acidic residues . Moreover , the R306C mutation is even more detrimental than lacking the entire basic cluster . This is reminiscent of a study documenting loss-of-function mutations in SOB3 , a gene encoding an AT-hook-containing protein . The authors showed that an arginine to histidine disruption of the AT-hook produced a more severely impaired phenotype than a truncating mutation , indicating that the missense mutation has a dominant-negative effect on the function of the protein ( Street et al . , 2008 ) . Mutation of R306 to either a cysteine or a histidine could have the same dominant-negative effect . Uncovering a role for this basic cluster adds the region around R306 to the MBD , TRD , and AT-hooks as domains known to be essential for MeCP2 function . In sum , comparisons of mouse lines carrying individual disease-causing mutations provide great insights into the function of MeCP2 and the pathogenesis of RTT . In future studies , these and additional mouse models will help us to further investigate the role of MeCP2 and its DNA binding affinity in vivo . The galK recombineering system ( Warming et al . , 2005 ) was used to modify a previously described PAC ( P1-derived artificial chromosome , PAC671D9 ) containing the entire MECP2 locus . In short , galK , flanked by 50 base pair ( bp ) homology arms to facilitate homologous recombination , was electroporated into induced SW102 cells containing the PAC to replace the residue of interest ( R111 or R306 ) . The presence of galK was selected for by growth at 30°C on minimal media plates containing galactose . Individual colonies were confirmed by streaking onto MacConkey plates followed by DNA sequencing . A single colony was chosen for the second round of electroporation to insert the desired mutation ( AGG to GGG for R111G , CGC to TGC for R306C ) . The replacement of galK with the residue of interest was selected for by plating on minimal media plates containing 2-Deoxy-D-galactose . Individual colonies were screened for the correct insert size by PCR . DNA from a single colony was sequence verified , purified , and digested with NotI followed by pulsed-field gel electrophoresis to isolate the ∼99 kb MECP2-continaing DNA from the backbone . To generate the transgenic mice , 1 ng/µl DNA was used for pronuclear injections into FVB single cell zygotes using standard procedures . Protein lysates were prepared from whole brains that were dounce homogenized in lysis buffer ( 20 mM Tris–HCl pH 8 . 0 , 180 mM NaCl , 0 . 5% NP-40 , 1 mM EDTA , Roche Complete Protease Inhibitor , Basel , Switzerland ) followed by sonication and rotation for 30 min at 4°C . After centrifugation at 4°C to remove the insoluble fraction , the supernatant was mixed with 2X NuPAGE Sample Buffer and run on a NuPAGE 4–12% Bis-Tris gradient gel with MES Running Buffer ( NuPAGE , Carlsbad , CA ) . Separated proteins were transferred to nitrocellulose membranes using the semi-dry method for 1 hr at 4°C . The membranes were blocked with 10% milk in tris buffered saline with 2% Tween-20 ( TBST ) , and incubated with primary antibody overnight at 4°C . After washing with TBST , the membranes were incubated with secondary antibody for 1 hr at room temperature followed by washing . HRP was detected using ECL detection kit ( Pierce , Rockford , IL ) . Antibodies used were: rabbit antiserum raised against the N-terminus of MeCP2 ( 1:5 , 000; Zoghbi Lab , #0535 ) , mouse anti-GAPDH 6C5 ( 1:20 , 000; Advanced Immunochemicals , 2-RGM2 , Long Beach , CA ) , rabbit anti-Sin3a ( 1:2 , 000; Abcam , ab3479 , Cambridge , England ) , rabbit anti-HDAC1 ( 1:2 , 000; Abcam , ab7028 ) , rabbit anti-HDAC2 ( 1:2 , 000; Abcam , ab7029 ) , rabbit anti-HDAC3 ( 1:500; Abcam , ab16047 ) , and rabbit anti-Tbl1 ( 1:1000; Abcam , ab24548 ) , donkey anti-rabbit HRP ( 1:20 , 000; GE Healthcare , NA934 , Little Chalfont , United Kingdom ) , and donkey anti-mouse HRP ( 1:20 , 000; Jackson ImmunoResearch Labs , 715-035-150 , West Grove , PA ) . Immunofluorescence was performed as previously described with some modifications ( Chao et al . , 2010 ) . Brains were dissected and cut midsagittally before drop-fixing in 4% paraformaldehyde ( PFA , Sigma-Aldrich , St . Louis , MO ) in phosphate buffered saline ( PBS , Sigma–Aldrich ) overnight at 4°C . Brains were cryoprotected in 30% sucrose for 1 day at 4°C and embedded in Optimum Cutting Temperature ( O . C . T . , Tissue-Tek , VWR , Radnor , PA ) . Free floating 30-µm midsagittal sections were cut on a Leica CM3050 S cryostat and stained for 24 hr with primary antibody , washed with TBST , and incubated for 24 hr with secondary antibody before mounting and imaging on a Leica TCS SP5 laser-scanning confocal microscope . Antibodies used were: rabbit anti-MeCP2 ( 1:1 , 000; Cell Signaling Technology , D4F3 , Danvers , MA ) , rabbit anti-ATRX ( 1:100; Abcam , ab97508 ) , and chicken anti-GFP ( 1:1 , 000; Abcam , ab13970 ) , goat anti-rabbit ( 1:500; Alexa Fluor 555 , Invitrogen , A-21428 , Carlsbad , CA ) and goat anti-chicken ( 1:500; Alexa Fluor 488 , Invitrogen , A-11039 ) . 4' , 6-diamidino-2-phenylindole ( DAPI , Invitrogen ) was used to stain DNA/nuclei . Transgenic R111G Tg and R306C Tg mice were maintained on a pure FVB background and genotyped using the following primers detecting GFP ( Forward 5′-CAGCAGGACCATGTGATCGC-3′ , Reverse 5′-GTGAAGTTCGAGGTGCGACAC-3′ ) . Mecp2 null mice were backcrossed and maintained on a pure 129SvEvTac background and genotyped as previously described ( Guy et al . , 2001 ) . Male transgenic mice were crossed to Mecp2 heterozygous females ( Mecp2+/− ) resulting in the 129SvEvTacxFVB F1 males that were used for all experiments . All mouse studies were approved by the Institutional Animal Care and Use Committee for Baylor College of Medicine ( IACUC Animal Welfare Assurance Number A3823-01 ) , and animal housing , husbandry , and euthanasia were conducted under the guidelines of the Center for Comparative Medicine , Baylor College of Medicine ( Protocol Number AN-1013 ) . Mice were assayed at multiple time points by the following tests in the specified sequence so as not to interfere with the other tests: ( 1 ) open field assay , ( 2 ) light/dark assay , ( 3 ) nesting , ( 4 ) parallel rod footslip , and ( 5 ) contextual and cued conditioned fear , as previously described ( Chao et al . , 2010 ) . Rotarod was performed on a new cohort of animals . Briefly , in the open field assay the movements of the mice in a 40 cm × 40 cm box , with 25 cm × 25 cm defined as the center , were tracked for 30 min and analyzed for total distance traveled and duration and distance traveled in the center using AccuScan Fusion software ( Omnitech , Columbus , OH ) . In the light/dark assay , the movements of the mice in a 40 cm × 20 cm box ( 1/3 dark , 2/3 light ) were tracked for 10 min and the time spent in the dark was analyzed using the AccuScan Fusion software . For nesting , mice were singly housed with a nestlet , and their ability to build a nest was assessed after 24 hr on a scale of 0–3 ( 0 = completely untouched , 3 = completely built nest ) . For parallel rod footslip , mice were placed in a 20 cm × 20 cm box and tracked for 10 min as they attempted to walk on a series of parallel rods . Footslips were detected by ANY-maze system ( Stoelting , Wood Dale , IL ) , and footslips per centimeter traveled were analyzed . Conditioned fear was performed over 2 days using the Actimetrics chamber and FreezeFrame 3 system ( Med Associates , St . Albans , VT ) : the mice received two 1 mA shocks accompanied by ∼85 dB white noise on the first training day , and contextual fear was tested on the second day by placement in the same box with no white noise or shock . After a rest period of at least an hour , cued fear was tested by white noise with no shock in the same box altered by different visual and olfactory cues . FreezeFrame 3 was used to analyze percent freezing during the white noise . For rotarod , mice were placed on an accelerating rotating rod ( Ugo Basile , Varese , Italy ) for four trials a day for four consecutive days for 10 min per trial , during which the rod accelerated from 4 to 40 rpm in the first 5 min . The time until the mouse fell ( or rode around the rod twice in a row ) was recorded and averaged for each day . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± the standard error of the mean ( SEM ) . Whole brains were dissected and dounce homogenized in lysis buffer ( 20 mM Tris-HCl pH 8 . 0 , 180 mM NaCl , 0 . 5% NP-40 , 1 mM EDTA , Roche Complete Protease Inhibitor ) followed by sonication and rotation for 30 min at 4°C for lysis . The insoluble fraction was removed by centrifugation , and the supernatant was rotated with camel anti-GFP beads ( Chromotek-GFP-Trap beads , Allele Biotechnology , San Diego , CA ) for 1 hr at 4°C . Beads were washed with lysis buffer before elution with 1X NuPAGE Sample Buffer by incubation at 70°C for 10 min . Inputs and eluates were subjected to western blot analysis . The chromatin immunoprecipitation ( ChIP ) was performed as previously described with some modifications ( Chahrour et al . , 2008 ) . In short , whole brain tissue was extracted and minced before being crosslinked in 1% paraformaldehyde ( PFA , Sigma-Aldrich ) for 15 min at room temperature . Incubation with 125 mM glycine in PBS for 5 min at room temperature was used to quench the crosslinking reaction . The tissue was then dounce homogenized and lysed for 10 min on ice in cold lysis buffer ( 10 mM Tris-HCl pH 7 . 5 , 10 mM NaCl , 3 mM MgCl2 , 0 . 5% NP-40 , 1 mM PMSF , Roche Complete Protease Inhibitor , Ambion RNase cocktail , Carlsbad , CA ) , followed by washing with cold lysis buffer to isolate nuclei . Nuclei were collected by centrifugation and resuspended in micrococcal nuclease buffer ( 10 mM Tris-HCl pH 7 . 5 , 10 mM NaCl , 3 mM MgCl2 , 1 mM CaCl2 , 4% NP-40 , 1 mM PMSF , Roche Complete Protease Inhibitor ) and sonicated . 75U MNase ( Worthington Biochemical Corporation , Lakewood , NJ ) was added to each sample and incubated at 37°C for exactly 5 min to digest chromatin to 100–300 bp fragments . The reaction was stopped by adding 2 mM EDTA , 1% SDS , and 200 mM NaCl . After an additional nuclear lysis by a second round of sonication followed by rotation at 4°C for 10 min , the lysate was cleared by centrifugation . The chromatin was confirmed to be sheared to 100–300 bp fragments using agarose gel electrophoresis before proceeding . For the immunoprecipitation ( IP ) , the chromatin was diluted 1:10 in ChIP dilution buffer ( 16 . 7 mM Tris–HCl pH 8 . 1 , 167 mM NaCl , 0 . 01% SDS , 1 . 1% Triton X-100 , 1 . 2 mM EDTA , 1 mg/ml BSA , 1 mM PMSF , Roche Complete Protease Inhibitor ) and precleared with Protein A Dynabeads ( Invitrogen ) before incubation with rabbit anti-GFP antibody ( 3 µg , Abcam , ab6556 ) overnight at 4°C . Protein A Dynabeads were then added and incubated with the samples for 3 hr at 4°C . The beads were washed in low salt wash buffer ( 20 mM Tris–HCl pH 8 . 1 , 150 mM NaCl , 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA ) , high salt wash buffer ( 20 mM Tris-HCl pH 8 . 1 , 500 mM NaCl , 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA ) , LiCl wash buffer ( 10 mM Tris-HCl pH 8 . 1 , 0 . 25 M LiCl , 1% NP-40 , 1% deoxycholic acid , 1 mM EDTA ) , and twice in TE ( 10 mM Tris-HCl pH 7 . 4 , 1 mM EDTA ) . The complexes were eluted twice in elution buffer ( 1% SDS , 100 mM NaHCO3 ) for 15 min at room temperature . After reversal of the crosslinking by incubation at 65°C overnight , the DNA was treated with proteinase K ( 100 µg/ml , Bioline , London , United Kingdom ) and Ambion RNase cocktail ( 2 µl per sample ) . DNA was isolated using phenol chloroform extraction followed by ethanol precipitation . The DNA pellet was resuspended in 30 µl TE buffer , and 1 µl was used for each individual quantitative PCR reaction . Quantitative PCR ( qPCR ) reactions were performed in triplicate using iTaq SYBR Green Supermix ( BioRad , Hercules , CA ) on a CFX96 Real-Time System ( BioRad ) using a fast 2-step cycling program according to the manufacturer's instructions . The following primers were used to detect the promoters of Major Satellite: Forward 5′-CATCCACTTGACGACTTGAAAA-3′ Reverse 5′-GAGGTCCTTCAGTGTGCATTT-3′ , L1 retrotransposon: Forward 5′-AGAAGAAACGGGAGACAGCA-3′ Reverse 5′-CTGCCGTCTACTCCTCTTGG-3′ , somatostatin ( Sst ) : Forward 5′-CATTGACAGGTACCCAACTGA-3′ Reverse 5′-CAGCCACATAGGAGCACACTT-3′ , afamin ( Afm ) : Forward 5′-AGACAGGCTGGCCTGAGAGTCA-3′ Reverse 5′-TTCCAATGCACGCGTCTCACCC-3′ , corticotropin releasing hormone ( Crh ) : Forward 5′-GTCACCAAGGAGGCGATACCTA-3′ Reverse 5′-TAAATAATAGGGCCCTGCCAAG-3′ , Gapdh: Forward 5′-CCAGCTACTCGCGGCTTTACGG-3′ Reverse 5′-CCTCCCGCCCTGCTTATCCAGT-3′ . The threshold cycle ( Ct ) values were averaged for each sample and normalized to input . Data were analyzed by an ordinary one-way ANOVA followed by Tukey's multiple comparisons test . Results were plotted as the mean ± SEM . WT human MeCP2 and MeCP2-R306C ( amino acids 274–340 ) were cloned in-frame to an N-terminal GST tag ( GST-MeCP2 pGEX-5x3 ) using the In-Fusion EcoDry Cloning system ( Clontech , Mountain View , CA ) , and C-terminal point mutations were generated using the QuikChange Multi Site-Directed Mutagenesis Kit ( Agilent , Santa Clara , CA ) . Recombinant proteins were expressed in BL21 ( DE3 ) E . coli cells following 1 mM IPTG induction and purified using Glutathione Sepharose 4B beads ( GE Healthcare ) before elution into glutathione buffer ( 50 mM Tris–HCl pH 8 . 0 , 10 mM glutathione ) and dialyzed into PBS overnight at 4°C . Proteins were stored at −80°C . The electrophoretic mobility shift assay ( EMSA ) was performed as previously described with some modifications ( Baker et al . , 2013 ) . In short , increasing amounts of protein were combined with 0 . 1 mg/ml BSA , 66 . 7 nM DNA probe , and EMSA buffer ( 10 mM Tris–HCl pH 7 . 4 , 50 mM KCl , 0 . 5 mM MgCl2 , 0 . 1 mM EDTA , 5% glycerol ) and incubated at room temperature for 30 min before running on a 1% agarose gel in 1X TAE at 4°C . The strands of the 64-bp probe were annealed by boiling both strands for 5 min and cooling down to room temperature over the course of 3 hr . The sequence was as follows: Forward 5′-GGACTCCAGGTCCAGGACCGCGTTTTTCGCGCGCACGGCGCGGGAGGTCCAGCTGTCCACCTCC-3′ , Reverse 5′-GGAGGTGGACAGCTGGACCTCCCGCGCCGTGCGCGCGAAAAACGCGGTCCTGGACCTGGAGTCC-3′ . Gels were post-stained with ethidium bromide and imaged under ultraviolet light to detect DNA .	Rett syndrome is a disorder that affects the development of the brain after birth . Infants with this condition develop as normal until they are 6–18 months old , when the development of their language and motor skills stops , or even regresses . Most cases of Rett syndrome are caused by mutations in a gene called MECP2 . If an individual mistakenly inherits an extra copy of the MECP2 gene , it can cause another developmental disorder called MECP2 duplication syndrome . This condition , which also affects the brain , gets worse over time and shares many features with Rett syndrome . The extra copy of the MECP2 gene leads to the production of too much MeCP2 protein . However , how doubling the level of this protein causes the syndrome and , in particular , which parts of the protein are involved are unknown . Previously , researchers engineered mice that expressed a copy of the human MECP2 gene alongside their own version of the gene . These mice developed a condition similar to MECP2 duplication syndrome and many of these mice suffered from seizures and died within their first year . Heckman et al . have now engineered mice that also have an extra human MECP2 gene but with one of two mutations that cause Rett syndrome in humans . Some mice had a mutation in a part of the MeCP2 protein that binds to DNA that is marked with small chemical tags called methyl groups . Other mice had a mutation in a domain of the protein that works to switch off genes . Heckman et al . found that mice with extra MeCP2 protein with either of these two mutations were as healthy as normal mice and showed none of the signs of MECP2 duplication syndrome . This indicates that both of these domains must be intact for doubling the levels of the MeCP2 protein to be harmful . Furthermore , Heckman et al . discovered that the mutation in the part of MeCP2 that works to switch genes off also reduces the protein's ability to bind to DNA . The next challenge is to understand the mechanism by which doubling the levels of this protein causes harm to the brain . Further work is also needed to uncover why having too much MeCP2 protein or none at all cause syndromes that share many features .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Rett-causing mutations reveal two domains critical for MeCP2 function and for toxicity in MECP2 duplication syndrome mice
John Maynard Smith compared protein evolution to the game where one word is converted into another a single letter at a time , with the constraint that all intermediates are words: WORD→WORE→GORE→GONE→GENE . In this analogy , epistasis constrains evolution , with some mutations tolerated only after the occurrence of others . To test whether epistasis similarly constrains actual protein evolution , we created all intermediates along a 39-mutation evolutionary trajectory of influenza nucleoprotein , and also introduced each mutation individually into the parent . Several mutations were deleterious to the parent despite becoming fixed during evolution without negative impact . These mutations were destabilizing , and were preceded or accompanied by stabilizing mutations that alleviated their adverse effects . The constrained mutations occurred at sites enriched in T-cell epitopes , suggesting they promote viral immune escape . Our results paint a coherent portrait of epistasis during nucleoprotein evolution , with stabilizing mutations permitting otherwise inaccessible destabilizing mutations which are sometimes of adaptive value . Epistasis can play a key role in evolution , such as by constraining accessible evolutionary pathways ( Weinreich et al . , 2005; Kryazhimskiy et al . , 2011 ) and increasing the role of contingency in adaptation ( Blount et al . , 2008; Bridgham et al . , 2009 ) . One of the simplest types of epistasis is that which occurs between mutations within a single protein . That such epistasis is common has long been considered self-evident–for example , in their seminal 1965 analysis of protein evolution , Emile Zuckerkandl and Linus Pauling wrote , “Of course … the functional effect of a given single substitution will frequently depend on the presence or absence of a number of other substitutions ( Zuckerkandl and Pauling , 1965 ) . ” But although numerous laboratory evolution and site-directed mutagenesis experiments have demonstrated that mutations can in principle exhibit strong epistatic interactions ( Bershtein et al . , 2006; Bloom et al . , 2006; Lunzer et al . , 2010; Salverda et al . , 2011 ) , surprisingly little is known about the actual role of epistasis in natural protein evolution . A few studies have reconstructed naturally occurring mutations involved in antibiotic resistance or steroid-receptor ligand specificity ( Wang et al . , 2002; Weinreich et al . , 2006; Ortlund et al . , 2007; Bridgham et al . , 2009 ) and found strong epistatic interactions . However , these studies have focused on small numbers of mutations pre-selected for analysis due to their putative adaptive role , and in most cases the actual temporal order of mutations is unknown . As a result , many basic questions remain without clear answers: What is the prevalence of epistasis during protein evolution ? How does epistasis arise from an evolutionary process that is conceived as proceeding through the incremental accumulation of mutations ? And is it possible to coherently understand epistasis in terms of the underlying protein biophysics ? An experimental approach to address these questions is suggested by John Maynard Smith’s classic analogy between protein evolution and the game where the goal is to convert one word into another a single letter at a time passing only through intermediates that are also words ( Maynard Smith , 1970 ) :WORD→WORE→GORE→GONE→GENE . Implicit in this analogy is the idea that epistasis constrains evolution—for example , the original parent sequence does not tolerate three of the four eventual changes , as GORD , WERD and WOND are not words . We sought to similarly test for epistasis in actual protein evolution by reconstructing an extended natural evolutionary trajectory , and then also introducing each mutation individually into the original parent ( Figure 1 ) . While this experimental strategy is not guaranteed to find every possible epistatic interaction , it will systematically identify all mutations that have different effects in the original parent and the evolutionary intermediates in which they actually occurred . The experimental strategy in Figure 1 also offers the possibility of determining how epistatically interacting mutations were actually fixed—for example through sequential functional intermediates as posited by Maynard Smith , or by the simultaneous or closely coupled fixation of several individually deleterious mutations ( Kimura , 1985; Meer et al . , 2010 ) . 10 . 7554/eLife . 00631 . 003Figure 1 . Outline of experiment designed to parallel Maynard Smith’s analogy . The actual evolutionary trajectory involves the accumulation of mutations , and consists of a series of evolutionary intermediates . We recreate and experimentally assay each of these evolutionary intermediates . We also introduce each mutation individually into the original parent sequence , and experimentally assay these single mutants . If Maynard Smith is correct , each of the naturally occurring evolutionary intermediates should be a functional protein . However , some of the single mutants could exhibit impaired function if there is significant epistasis among mutations along the evolutionary trajectory . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 003 The experiment outlined in Figure 1 requires a protein for which it is possible both to reconstruct the natural evolution and to assay for the functions that contribute to biological fitness . Human H3N2 influenza A virus is exceptionally suited to the first requirement , as the extensive availability of contemporary and historical sequences enables the detailed reconstruction of evolutionary trajectories . We focused on the 498-residue nucleoprotein ( NP ) . Although NP’s evolution is less rapid and medically infamous than that of its surface counterparts hemagglutinin and neuraminidase , NP still accumulates roughly one amino-acid substitution per year ( Rambaut et al . , 2008 ) . Crucially for our experiment , NP’s primary function—serving as a scaffold for viral RNA during transcription and genome packaging ( Portela and Digard , 2002; Ye et al . , 2006 ) —occurs within the interior of infected cells , and so is probably fairly authentically represented in tissue-culture assays . NP is also a target of cytotoxic T lymphocytes ( CTLs ) , and so is under continual pressure for change in CTL epitopes ( Berkhoff et al . , 2004 , 2007; Valkenburg et al . , 2011 ) —a pressure partially countered by the fact that some of these epitopes are under functional constraint ( Rimmelzwaan , et al . , 2004a; Berkhoff et al . , 2005 , 2006 ) . CTL selection in influenza is thought to be weaker than antibody selection on the viral surface proteins , and so much of NP’s evolution is shaped by stochastic forces such as population bottlenecks and hitchhiking with antigenic mutations in the surface proteins ( Rambaut et al . , 2008; Bhatt et al . , 2011 ) —stochastic forces that in some cases can also accelerate the fixation CTL escape mutations ( Gog et al . , 2003 ) . Our experiments do not measure these complexities of immune pressure as they do not include CTL selection , but as described later in this paper , existing data enable us to identify adaptive CTL-escape mutations . In the work reported here , we use the strategy in Figure 1 to synthesize information about influenza’s natural evolution with our own experiments to examine epistasis in NP evolution . We find that epistasis constrains both the sequence evolution and ongoing adaptation of NP , and that the mechanistic basis for most of this epistasis can be understood in remarkably simple terms . We focused on the evolutionary trajectory separating NPs from two human H3N2 strains isolated 39 years apart , A/Aichi/2/1968 and A/Brisbane/10/2007 ( Figure 2 ) . To map this trajectory , we developed a probabilistic technique to estimate the posterior distribution of mutational events ( Minin and Suchard , 2008; O’Brien et al . , 2009 ) and , original to this work , their time-orderings along an unknown phylogenetic tree within the BEAST software package ( Drummond et al . , 2012 ) . Each sample from this posterior distribution represents a mutational path from Aichi/1968 to Brisbane/2007 , which in turn can be represented as a directed graph through protein sequence space ( Figure 2—figure supplement 1 ) . Summarizing these graph samples effectively integrates over uncertainty in the tree and substitution process , and yields the marginal posterior distribution of the evolutionary trajectory from Aichi/1968 to Brisbane/2007 ( Figure 2 ) . The most probable trajectory consists of 39 mutational steps at 33 sites ( 5 mutations revert; 1 site mutates to two identities ) . The fact that NP sequences are available for every year since 1968 allows us to reconstruct the trajectory with remarkable precision: there are >1046 possible orderings of 39 mutations , yet we can confidently identify the sequences of 25 of the evolutionary intermediates; the remainder fall along regions of the trajectory where two or more mutations occurred in an unknown order ( Figure 2 ) . 10 . 7554/eLife . 00631 . 004Figure 2 . Inferred evolutionary trajectory . ( A ) Evolutionary trajectory through protein sequence space from Aichi/1968 to Brisbane/2007 NP . Each circle represents a unique inferred sequence , with areas and intensities proportional to the probability that sequence was part of the evolutionary trajectory ( Figure 2—figure supplement 1 ) . Mutations for which the parent and descendent are clearly resolved are in black; mutations that occurred in an unknown order are in red . High-confidence evolutionary intermediates have numeric labels . The estimated date of occurrence of each mutation is shown in Figure 2—figure supplement 2 . ( B ) Phylogenetic tree of the human H3N2 NPs from 1968 to 2011 that were used to infer the evolutionary trajectory . The lines of descent connecting Aichi/1968 and Brisbane/2007 to their common ancestor are in black . Data and computer code are provided in Figure 2—source code 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 00410 . 7554/eLife . 00631 . 005Figure 2—source code 1 . The sequence data and source code used to generate the evolutionary trajectory and phylogenetic tree . This code is in the form of a ZIP file with BEAST input files and Python code . A README file explains the contents . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 00510 . 7554/eLife . 00631 . 006Figure 2—figure supplement 1 . Construction of the evolutionary trajectory . We estimated the joint posterior distribution of phylogenetic trees with mutations mapped on the branches in BEAST . Portions of two example trees with mutations are shown at the top . Tracing along the line of descent ( shown in red ) gives a specific sequence of mutations; we can use these to reconstruct the corresponding path through sequence space ( a directed graph through the space of all possible protein sequences ) , as shown at the bottom of the figure . We summarize samples of individual directed graphs drawn from the posterior distribution , with the posterior probabilities of specific points and connections equaling the fraction of the samples that correspond to directed graphs containing these points and connections . We represent visually the probabilities of the points and connections by varying the areas of the circles/widths of the lines , and their intensities . At bottom right is the result of integrating just the two partial paths shown in this figure . The full evolutionary trajectory in Figure 2 represents the integration of many full paths . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 00610 . 7554/eLife . 00631 . 007Figure 2—figure supplement 2 . Dates at which mutations fixed along the evolutionary trajectory in Figure 2A . This plot shows the posterior median and 80% Bayesian credible intervals for the dates . In many cases the date intervals for two mutations overlap even though in the evolutionary trajectory one mutation can be placed before another . This is because the dates are correlated in the individual tree-mutation samples drawn from the posterior , so that the difference in times between two mutations tend to be more precisely defined than the absolute date ranges . Two mutations ( red ) tend to arise on branches going from the common ancestor to Aichi/1968 , while the remaining mutations ( blue ) tend to arise on branches from the common ancestor to Brisbane/2007 . The code used to generate this figure is in Figure 2—source code 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 007 We created plasmids encoding each of the high-confidence protein intermediates along the trajectory , using the same codons found in the natural sequences but not introducing any of the synonymous mutations that occurred during this timeframe . We used a mini-replicon system to assess the transcriptional activity of each NP in combination with polymerase proteins ( PB2 , PB1 , PA ) from the human H3N2 strain A/Nanchang/933/1995 ( Figure 3—figure supplement 1 ) . All evolutionary intermediates exhibited high activity ( Figure 3A ) , supporting Maynard Smith’s notion that evolution proceeds through functional sequences . 10 . 7554/eLife . 00631 . 008Figure 3 . Three mutations are strongly deleterious when introduced individually into the parent Aichi/1968 NP , despite eventually becoming fixed along the evolutionary trajectory without apparent negative effect . ( A ) The transcriptional activity for the high-confidence evolutionary intermediates , quantified using a GFP reporter ( Figure 3—figure supplement 1 ) . Activity is scaled so that the parent Aichi/1968 NP has an activity of one . The numeric labels of the evolutionary intermediates match those used in Figure 2A . ( B ) The change in activity caused by introducing each mutation individually into the parent Aichi/1968 NP . The deleterious effects on activity caused by L259S , R348G , and V280A are not caused by the genetic background of influenza polymerase proteins , as these three mutants are impaired regardless of whether the polymerase proteins are derived from Nanchang/1995 , Aichi/1968 , or Brisbane/2007 ( Figure 3—figure supplement 3 ) . ( C ) All three of the individual mutations that reduce activity also impair growth , yet there is no defect in the growth of viruses carrying the NPs of the first high-confidence evolutionary intermediates in which these individually deleterious mutations were actually fixed . Viral growth is quantified as described in Figure 3—figure supplement 2 . Numerical data are in Figure 3—source data 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 00810 . 7554/eLife . 00631 . 009Figure 3—source data 1 . Summary of transcriptional activity data ( mean and standard error ) for all variants from this study . The means and standard errors are computed from at least triplicate assays , and are reported standardized so the value for Aichi/1968 NP is at 1 . The data are provided in a CSV file . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 00910 . 7554/eLife . 00631 . 010Figure 3—source data 2 . Summary of viral growth data ( mean and standard error ) for all variants for which this property was measured in this study . The means and standard errors are computed from at least triplicate assays , and are reported standardized so the value for Aichi/1968 NP is at 1 . The data are provided in a CSV file . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01010 . 7554/eLife . 00631 . 011Figure 3—source data 3 . Transcriptional activity data in the alternative polymerase genetic backgrounds shown in Figure 3—figure supplement 3 . The data are provided in a CSV file . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01110 . 7554/eLife . 00631 . 012Figure 3—figure supplement 1 . Schematic of NP transcriptional activity assay . ( A ) 293T cells are co-transfected with plasmids encoding PB2 , PB1 , and PA from Nanchang/1995 , an NP variant , and a reporter plasmid expressing a negative-sense influenza viral-RNA encoding GFP . The four influenza proteins ( PB2 , PB1 , PA , and NP ) collaborate to transcribe the reporter GFP vRNA into mRNA , which is translated into protein , causing the cells to fluoresce . Activity is quantified by flow cytometry as the mean fluorescence intensity ( MFI ) above the background of control cells that receive no NP , relative to the activity of three wild-type Aichi/1968 NP replicates . ( B ) Example flow cytometry plots . ( C ) Total transcriptional activity as a function of the amount of transfected NP plasmid . The PB2 , PB1 , PA , and GFP reporter plasmids are transfected at 200 ng each . The NP plasmid was varied from 0 to 300 ng , and the fluorescence was quantified . At low levels of NP plasmid , the signal increases with increasing plasmid concentration , but at high levels the signal is saturated . We performed our assays with 50 ng of NP plasmid , which is near the midpoint of the curve—with this amount of plasmid , the signal changes in a sensitive manner with amount of NP . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01210 . 7554/eLife . 00631 . 013Figure 3—figure supplement 2 . Schematic of viral growth assay . ( A ) Co-cultures of 293T-CMV-Nan95-PB1 and MDCK-SIAT1-CMV-Nan95-PB1 cells were co-transfected with reverse-genetics plasmids encoding PB2/PB1/PA from Nanchang/1995 , an NP variant , HA/NA/M/NS from A/WSN/1933 , and a reporter plasmid expressing a negative-sense influenza viral-RNA encoding GFP with flanking sequences from PB1 . A media change to low-serum media was performed after 20–24 hr , and at 66 hr post-transfection the viral supernatant was harvested . Dilutions of the supernatant were used to infect fresh MDCK-SIAT1-CMV-Nan95-PB1 cells , and 16 hr post-infection , the viral titer was determined by quantifying the fraction of GFP positive cells by flow cytometry using the Poisson formula . The titers are standardized to the average of three Aichi/1968 NP controls . ( B ) Example flow cytometry plots . The cells are at 105 per well , so the computed titers for these example plots are −105 × ln ( 1 − 0 . 0277 ) = 2809 infectious particles per μl for Aichi/1968 , and −105 × ln ( 1 − 0 . 0144 ) /10 = 145 infectious particles per μl for V280A . ( C ) Viral titers in the supernatant at the indicated times post-transfection for the Aichi/1968 NP . Based on this timecourse , we chose to determine titers for all of the tested variants at 66 hr , when the virus is near but not yet at peak titers . Typically at this time point we would observe titers for the Aichi/1968 NP of around 103 infectious particles per μl; however , titers varied somewhat from day-to-day , which is why the values for each assay were standardized to three Aichi/1968 NP controls run on that same day . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01310 . 7554/eLife . 00631 . 014Figure 3—figure supplement 3 . Effects of key NP mutations on transcriptional activity in different polymerase genetic background of the viral polymerase genes . The six indicated NP variants were tested in the transcriptional activity assay with ( A ) the PB2 , PB1 , and PA genes derived from Nanchang/1995 as in all of the assays shown in Figure 3; ( B ) with these polymerase genes derived from Aichi/1968; or ( C ) with these polymerase genes derived from Brisbane/2007 . In the last case ( the Brisbane/2007 polymerase genes ) , the magnitude of the activity reduction is somewhat less , but all three mutants are still clearly impaired . In each plot , the activity is normalized relative to that observed for the parental Aichi/1968 NP in that polymerase background , and so the different panels have y-axis units that cannot be compared across panels . The NP plasmid amount is the same ( 50 ng ) in all three backgrounds , but note that although this makes the readings at the middle of the dose-response curve for the Nanchang/1995 polymerase background ( Figure 3—figure supplement 1 ) , this may not be true for the other polymerase backgrounds . As can be seen from the plots , L259S , R384G , and V280A are impaired in all three backgrounds , suggesting that the mutational effects are intrinsic to NP and not due to interactions with polymerase genetic background . Data are in Figure 3—source data 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 014 We then introduced each mutation individually into the parent Aichi/1968 NP and measured its effect on activity ( Figure 3B ) . Most single mutants were highly functional , but three ( L259S , R384G , V280A ) exhibited large decreases in activity . These three mutations are also deleterious in the background of polymerase proteins from Aichi/1968 and Brisbane/2007 ( Figure 3—figure supplement 3 ) , suggesting that the deleterious effect is intrinsic to NP itself . Because RNA transcription is essential for influenza replication , impaired activity should be devastating to viral fitness . To confirm this , we used reverse genetics ( Hoffmann et al . , 2000 ) to generate GFP-carrying viruses ( Bloom et al . , 2010 ) with the polymerase genes from Nanchang/1995 and the remaining genes from the lab-adapted A/WSN/1933 ( H1N1 ) strain ( Figure 3—figure supplement 2 ) . Viruses with the parent ( Aichi/1968 ) or final ( Brisbane/2007 ) NP grew to comparably high titers in tissue culture ( Figure 3C ) , but growth of the three transcriptionally impaired mutants was dramatically lower ( >1000-fold lower for L259S and R384G , and >20-fold lower for V280A ) . However , we observed good growth of the first high-confidence evolutionary intermediates in which these mutations were actually fixed ( Figure 3C ) . To understand how the three individually deleterious mutations fixed along the line of descent without a substantial fitness cost , we examined the evolutionary trajectory . L259S fixed in an unknown order with another mutation ( N334H ) in an evolutionary intermediate that we have labeled Step 10 ( Figure 4A ) . L259S is deleterious to activity and growth in both Aichi/1968 and Step 10 , while N334H has no major effect in either background ( Figure 4A ) . But N334H rescues the deleterious effect of L259S in both Aichi/1968 and Step 10 , indicating that fixation of L259S was enabled by N334H . We cannot determine the order of these two mutations during the evolution of the virus: they could have occurred simultaneously , or one could have preceded the other by a sufficiently small amount of time that no influenza isolates were sequenced in the intervening time period . 10 . 7554/eLife . 00631 . 015Figure 4 . Effects of the individually deleterious mutations in the evolutionary intermediates in which they occurred . ( A ) L259S impairs the transcriptional activity and viral growth of both the parent Aichi/1968 and the evolutionary intermediate Step 10 , but is rescued by N334H in both backgrounds . N334H alone has little effect on activity or growth in either background . The actual evolutionary trajectory involved the fixation of L259S and N334H in an unknown order . ( B ) R384G impairs activity and ablates growth of Aichi/1968 , but has no effect on activity and a reduced adverse effect on growth in the high-confidence evolutionary intermediate ( Step 21 ) in which it and several other mutations occurred in an unknown order . Addition of E375G to Step 21 with R384G fully rescues viral growth , but E375G alone worsens the impact of R384G . The reversion of L259S that preceded Step 21 plays an important role in enabling R384G , as the evolutionary intermediate without this reversion ( Step 20 ) is more negatively affected by R384G . ( C ) V280A is deleterious in Aichi/1968 but not in the Step 35 evolutionary intermediate in which it actually occurred . M136I , which precedes V280A , largely rescues its effect . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 015 The next individually deleterious mutation , R384G , fixed in Step 21 in an unknown order with several other mutations ( Figure 4B ) . Immediately prior to this , the individually deleterious mutation L259S reverted in the transition from Step 20 to Step 21 . R384G is deleterious to activity and growth in both Aichi/1968 and Step 20 ( Figure 4B ) . However in Step 21 , the deleterious effect on activity disappears while that on growth diminishes , indicating that reversion of L259S alleviates the impact of R384G . The further addition of E375G to Step 21 containing R384G eliminates the remaining growth impairment—but E375G alone fails to rescue R384G in the background of Aichi/1968 ( Figure 4B ) . Note that E375G has also been previously reported to partially compensate for R384G in slightly different genetic backgrounds ( Rimmelzwaan et al . , 2004a ) . Fixation of R384G was therefore partially enabled by the preceding reversion of L259S , with further assistance from E375G . S259L clearly preceded R384G , but we are unable to resolve whether the second enabling mutation ( E375G ) occurred before , after , or simultaneously with R384G . The final individually deleterious mutation , V280A , fixed in Step 35 . Although V280A is deleterious in Aichi/1968 , it has no negative impact in Step 35 ( Figure 4C ) . M136I , which immediately preceded V280A in the natural evolution , mostly rescues its deleterious effect in the background of Aichi/1968 ( Figure 4C ) . Therefore , V280A was enabled by mutations ( including M136I ) that occurred prior to its own fixation . What is the mechanistic explanation for these mutational effects ? None of the identified epistatically interacting residues are in contact in the monomeric or known oligomeric NP crystal structures ( Ye et al . , 2006 , 2012; Ng et al . , 2008 ) , nor are any of them in the protein’s RNA-binding groove ( Figure 5 and Figure 5—figure supplement 1 ) . We therefore hypothesized that the individually deleterious mutations might destabilize NP , and that the epistasis might be mediated by counterbalancing stabilizing mutations . This hypothesis is consistent with the fact that N334H rescues L259S , and reversion of L259S in turn partially rescues R384G , despite the lack of physical contact among these residues . 10 . 7554/eLife . 00631 . 016Figure 5 . There is no obvious structural basis for the observed epistasis , as none of the epistatically interacting mutations are in contact in the solved crystal structures of NP . Shown above is one monomer from PDB structure 2IQH; the mutations are also not in contact in any of the known oligomeric structures ( Figure 5—figure supplement 1 ) . The sites of the three individually deleterious mutations are in orange , those of the rescuing mutations are in green , and the site of E375G ( which can rescue R384G depending on genetic background ) is in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01610 . 7554/eLife . 00631 . 017Figure 5—figure supplement 1 . None of the epistatically interacting mutations are in contact in the known oligomeric structures of NP . ( A ) Sites of the mutations ( spheres ) in the trimeric structure of NP from PDB 2IQH . ( B ) Sites of the mutations in the dimeric structure from PDB 2Q06 . ( C ) Sites of the mutations in the dimeric structure from PDB 3MIR . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 017 We began by pairing N334H with each of the individually deleterious mutations in the background of the parent Aichi/1968 NP . N334H rescues activity for each of these mutations ( Figure 6A ) . N334H also rescues growth for L259S and V280A , and largely rescues growth for R384G ( Figure 6B ) . To test if this rescue was related to in vivo protein levels , we quantified NP in transfected human cells ( Figure 6C , D ) . The parent Aichi/1968 NP and its N334H mutant were present at comparably high levels , but levels were markedly reduced for variants carrying each of the three individually deleterious mutations . Addition of N334H to these mutants restored wild-type protein levels , indicating that N334H can counteract the decrease in protein levels associated with the individually deleterious mutations . 10 . 7554/eLife . 00631 . 018Figure 6 . The epistasis correlates with mutational effects on NP stability . ( A ) and ( B ) N334H rescues activity and mostly rescues viral growth of all three individually deleterious mutations . ( C ) Western blot showing that the three individually deleterious mutations all reduce NP levels in transfected cells relative to an mCherry control expressed from an IRES in the same plasmid; this effect is counteracted by N334H . ( D ) Quantification of NP levels from triplicate Western blots ( Figure 6—figure supplement 1 ) . ( E ) The deleterious mutations decrease and N334H increases the stability of NP , as measured by thermal denaturation of purified protein monitored by circular dichroism ( Figure 6—figure supplements 2–6 and Figure 6—source data 1 ) . Figure 6—figure supplement 7 shows that M136I , which precedes V280A in the natural evolution , and is modestly stabilizing ( Figure 6—figure supplements 4–7 ) , also partially rescues the levels of V280A NP in transfected cells and the activity of all three individually deleterious mutations . Together , the two stabilizing mutations N334H and M136I can rescue the activity of combinations of the individually deleterious mutations ( Figure 6—figure supplement 8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01810 . 7554/eLife . 00631 . 019Figure 6—source data 1 . A table of all of the melting temperatures and the changes in stability relative to Aichi/1968 , in CSV format . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 01910 . 7554/eLife . 00631 . 020Figure 6—figure supplement 1 . The full set of triplicate Western blots used to quantify the NP levels in transfected cells . NP levels were quantified by transfecting 293T cells with a FLAG-tagged NP on a plasmid also containing a FLAG-tagged mCherry under an IRES element . The mCherry therefore serves as a transfection/loading control . Three triplicate blots each from independent transfections were performed to quantify NP levels . For each blot , the ratio of NP to mCherry was quantified , and is shown below the band . These ratios were then standardized relative to that for the Aichi/1968 NP , and the averages of the triplicates computed . Note that all three blots confirm that L259S , V280A , and R384G are present at reduced levels—and that N334H can rescue this defect . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02010 . 7554/eLife . 00631 . 021Figure 6—figure supplement 2 . Purification of Aichi/1968 NP with deletion of residues 2–7 , mutation R416A , and a C-terminal 6-His tag ( expected molecular weight 56 . 6 kDa ) . These modifications were necessary to obtain monomeric RNA-free in NP in a CD-compatible buffer . The other NP variants behaved similarly by the measures in this figure . ( A ) NP was at high purity after elution off a cobalt column , as judged by SDS-PAGE . ( B ) NP was further purified over a Superdex 200 size-exclusion column . It eluted at roughly the expected size . ( C ) After size-exclusion , NP had a 260/280 nm ratio much less than 1 , indicating protein largely free of RNA . ( D ) NP at 20°C exhibits a CD spectrum characteristic of an alpha-helical protein . This spectrum disappears after heating , and does not reappear after cooling back to 20°C ( indicating irreversible denaturation ) . We monitored unfolding at 209 nm ( dashed vertical line ) . All CD was performed at a NP concentration of 5 μM with a cuvette path length of 0 . 1 cm . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02110 . 7554/eLife . 00631 . 022Figure 6—figure supplement 3 . Circular dichroism wavelength scans for all variants that were tested ( those with thermal melts shown in Figure 6—figure supplements 4–6 ) . All variants exhibit similar CD spectra characteristic of an alpha-helical protein . The magnitude of the spectra are also similar , indicating all protein variants are indeed at roughly the same concentration ( which was measured at 5 μM as judged by absorbance at 280 nm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02210 . 7554/eLife . 00631 . 023Figure 6—figure supplement 4 . The first 15 thermal denaturation curves . NP variants were thermally denatured with unfolding monitored by ellipticity at 209 nm , at a scan rate of 2°C per minute . Melting temperatures were obtained from sigmoidal fits over the range 20–60°C . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02310 . 7554/eLife . 00631 . 024Figure 6—figure supplement 5 . The second 15 thermal denaturation curves . NP variants were thermally denatured with unfolding monitored by ellipticity at 209 nm , at a scan rate of 2°C per minute . Melting temperatures were obtained from sigmoidal fits over the range 20–60°C . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02410 . 7554/eLife . 00631 . 025Figure 6—figure supplement 6 . The last 13 thermal denaturation curves . NP variants were thermally denatured with unfolding monitored by ellipticity at 209 nm , at a scan rate of 2°C per minute . Melting temperatures were obtained from sigmoidal fits over the range 20–60°C . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02510 . 7554/eLife . 00631 . 026Figure 6—figure supplement 7 . M136I partially rescues the activity of the three individually deleterious mutations , and mostly rescues protein levels for V280A . M136I preceded V280A in the natural evolution of NP . Note that Figure 6—figure supplements 4–6 shows that M136I is modestly stabilizing . ( A ) In the activity assay , M136I alone has no effect , but it can partially rescue the activity of all three individually deleterious mutations . ( B ) Western blots showing the levels of NP protein relative to the mCherry control . M136I is present at wild-type levels , whereas levels of V280A are reduced . Adding M136I to V280A largely rescues the NP levels . Three triplicate Western blots from independent transfections are shown . The numbers below the bands represent the ratio of the NP band intensity to the mCherry band intensity . ( C ) Quantification of the triplicate Western blots . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02610 . 7554/eLife . 00631 . 027Figure 6—figure supplement 8 . Activities for NPs with combinations of the stabilizing and destabilizing mutations . Various combinations of the three epistatically constrained destabilizing mutations ( L259S , V280A , R384G ) and the stabilizing mutations ( N334H , M136I ) were introduced into the Aichi/1968 NP and the total transcriptional activity was quantified . The final NP variant at the end of the trajectory contained V280A , R384G , N334H , and M136I—but not L259S , as this mutation reverts before the occurrence of R384G . An NP with just those four mutations exhibits wild-type levels of activity . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 027 To see if these changes in in vivo protein levels correlated with global protein stability , we purified the NP variants with an additional mutation in the tail loop ( Ye et al . , 2006 ) that enabled us to obtain monomeric RNA-free protein that exhibited the expected alpha-helical circular-dichroism spectrum ( Figure 6—figure supplement 2 ) . All NP variants exhibited similar circular-dichroism spectra ( Figure 6—figure supplement 3 ) and unfolded with a single cooperative transition ( Figure 6—figure supplements 4–6 ) , allowing us to determine melting temperatures ( Tm ) for irreversible thermal denaturation . The three individually deleterious mutations were all destabilizing , with changes in melting temperatures ( ΔTm ) relative to the parent Aichi/1968 NP that ranged from −3 . 6°C to −4 . 9°C . N334H was stabilizing ( ΔTm of 4 . 4°C ) , and adding N334H to each of the individual destabilized mutants restored their stability to roughly wild-type values ( Figure 6E ) . M136I is also modestly stabilizing , and partially rescues all three individually deleterious mutations ( Figure 6—figure supplements 4–7 ) . The final Brisbane/2007 NP contains two of the three identified destabilizing mutations ( L259S reverts ) and both of the identified stabilizing mutations—combining all four mutations in the parental Aichi/1968 background gives wild-type levels of activity ( Figure 6—figure supplement 8 ) . These results suggest that most of the epistasis that we identified during NP’s evolution is due to counterbalancing stabilizing and destabilizing mutations . To obtain a more complete portrait , we measured the stabilities of all of the resolved evolutionary intermediates ( Figure 6—figure supplements 4–6 ) . Figure 7A shows the correlation between activity and stability for all NP variants for which both properties were measured . For variants with melting temperatures exceeding 43°C , activity is independent of stability—changes in stability above this threshold are neutral with respect to activity . But once stability begins to fall below 43°C , there is a rapid decline in activity . A similar pattern is observed in the correlation between stability and viral growth , with growth declining precipitously once stability drops below 43°C ( Figure 7B ) . The exception to this relationship is that variants containing R384G without E375G exhibit reduced growth even when they possess adequate stability and activity ( Figures 4B , 7A , B ) . E375G is modestly destabilizing ( ΔTm = −1 . 0°C ) , and so epistatically interacts with R384G by a mechanism other than protein stability . E375G and R384G induce opposite charge changes and occur on the same surface of NP ( Figure 5 and Figure 7—supplement figure 1 ) —we hypothesize that maintenance of the electrostatic charge on this surface might be important for NP’s interaction with some partner late in the viral life cycle after RNA transcription is complete . 10 . 7554/eLife . 00631 . 028Figure 7 . Most of the epistasis that we identified in NP’s evolution can be explained by counterbalancing stabilizing and destabilizing mutations . ( A ) The relationship between NP stability and transcriptional activity for all variants for which both properties were measured . As long as the stability is greater than a threshold around 42°C , changes in stability are neutral with respect to activity . Below this threshold , activity declines sharply with decreasing stability . ( B ) The relationship between viral growth and NP stability exhibits a similar behavior . The exception is that growth of variants with R384G is fully rescued only by combining a stabilizing mutation with E375G ( Figure 4B ) , an effect that we hypothesize is related to the electrostatic charge on one of NP’s surfaces ( Figure 7—figure supplement 1 ) . ( C ) The dynamics of protein stability during NP evolution . Shown are the measured stabilities for evolutionary intermediates from the trajectory in Figure 2A . The lines along the y-axis at the far left show the stabilities of the five indicated individual point-mutants of the Aichi/1968 NP . Although the destabilizing mutations L259S , R384G , and V280A are deleterious to the Aichi/1968 parent , during evolution they are counterbalanced by stabilizing mutations . In the top panels , selected points are labeled with the NP variant name; the full data plotted in this figure are in Figure 7—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02810 . 7554/eLife . 00631 . 029Figure 7—source data 1 . The activity , viral growth , and stability data shown in Figure 7 . The data are provided in CSV format . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 02910 . 7554/eLife . 00631 . 030Figure 7—figure supplement 1 . We hypothesize that E375G helps counteract R348G by maintaining the electrostatic charge on one of NP’s surfaces . As shown in Figure 4B , the activity of the destabilizing mutation R384G is fully rescued by a stabilizing mutation , such as N334H in Aichi/1968 or reversion of L259S in Step 21 . However , full rescue of viral growth also requires E375G . E375G is destabilizing , and alone does not rescue growth or activity of R384G . We hypothesize that rescue of growth of R384G requires two events . The first is counteraction of NP destabilization . The second is some effect specific to E375G that is apparent in growth but not transcriptional activity . E375G ( which causes loss of a negative charge ) counterbalances the loss of a positive charge due to R384G . The image shows an electrostatic surface ( PyMol , PDB 2IQH ) . R384 and E375 are on the same surface , contributing positive and negative charges . Mutating both to neutral glycine maintains the net charge . Perhaps the charge of this surface is important for interaction with other protein partners late in the life cycle ( after RNA transcription ) , explaining why E375G is important for growth but not activity . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 030 This caveat about R384G and E375G notwithstanding , it is striking that we can explain all of the other observed mutational effects simply in terms of protein stability . Figure 7C shows the trajectory of stability during NP’s evolution . The parent Aichi/1968 is only marginally more stable than the minimal threshold where activity and viral growth begin to suffer . For this reason , the three destabilizing mutations L259S , R384G , and V280A are highly deleterious to Aichi/1968 . During NP’s evolution , stability fluctuates as the protein fixes stabilizing and destabilizing mutations . Each of the three destabilizing mutations that we identified as being under epistatic constraint is closely associated with a stabilizing mutation . The stabilizing M136I preceded the destabilizing V280A , and provided a stability cushion to eliminate V280A’s otherwise deleterious effect ( Figures 4C , 7C ) . The stabilizing S259L preceded the destabilizing R384G , and was necessary ( in conjunction with E375G ) to alleviate R384G’s otherwise deleterious effect ( Figures 4B , 7C ) . The stabilizing N334H occurred in close temporal proximity to the destabilizing L259S and fully counteracts L259S’s otherwise deleterious effect ( Figures 4B , 7C ) —although in this case there is insufficient natural sequence data to determine which of these mutations occurred first ( it is also possible that they occurred simultaneously ) . The aforementioned results illuminate the evolutionary steps that gave rise to the fixation of the individually deleterious destabilizing mutations , but they do not provide any indication of what forces may have driven this fixation . The destabilizing mutations could have been fixed stochastically by genetic drift or hitchhiking , or they could have been directly favored by selection for viral immune escape . As discussed in the Introduction , NP is a target of CTLs , and mutations in CTL epitopes benefit influenza by helping it evade immune memory that accumulates in the human population ( Berkhoff et al . , 2007; Valkenburg et al . , 2011 ) . We began by searching the literature for experimentally validated human CTL epitopes in NP . All three destabilizing mutations occur in characterized epitopes ( DiBrino et al . , 1995; Voeten et al . , 2000; Rimmelzwaan et al . , 2004b; Berkhoff et al . , 2007; Assarsson et al . , 2008; Alexander et al . , 2010; Cheung et al . , 2012 ) , and mutations at two of these sites have been shown to reduce CTL recognition ( Voeten et al . , 2000; Berkhoff et al . , 2004; Rimmelzwaan et al . , 2004b; Berkhoff et al . , 2007; Figure 8—source data 1 ) . To test if the epistatically constrained mutations are in more epitopes than expected by chance , we considered two approaches to comprehensively identify epitopes in NP: mining of a database of literature-characterized epitopes ( Vita et al . , 2010 ) , and computational prediction of epitopes from protein sequence ( Stranzl et al . , 2010 ) . The first approach has the advantage of only identifying experimentally validated epitopes , but the disadvantage that this set of epitopes is subject to unknown biases due to experimental choices about HLA types and viral strains . Computational prediction has the advantage of being unbiased with respect to HLA types and viral strains , but the disadvantage that the predictions may not be accurate . As it turns out , both approaches give the same result—the three epistatically constrained mutations are significantly enriched in CTL epitopes relative to all sites in NP and to the set of sites that experienced substitutions along the evolutionary trajectory ( Figure 8 and Figure 8—figure supplement 2 ) . These three destabilizing mutations are thus disproportionately important for viral immune escape , and may have been favored by selection for this property ( a dN/dS test [Murrell et al . , 2013] is inconclusive , probably due to lack of sequence data; Figure 8—figure supplement 3 and Figure 8—source code 2 ) . Stability-mediated epistasis therefore constrains the adaptive process of CTL escape as well as the sequence evolution of NP . The destabilizing CTL-escape mutations L259S , R384G , and V280A were inaccessible to NP during much of its evolutionary trajectory , but were fixed after stabilizing mutations made the protein permissive to their occurrence . 10 . 7554/eLife . 00631 . 031Figure 8 . The three epistatically constrained destabilizing mutations occur at sites significantly enriched in human CTL epitopes . Distributions of the numbers of experimentally characterized epitopes per residue for all sites or sites that experienced mutations along the evolutionary trajectory are in blue and red , respectively . Sites 259 , 280 , and 384 are in significantly more epitopes than three random positions from all sites ( p=0 . 001 ) or the mutated sites ( p=0 . 004 ) ; however , three random positions from the mutated sites are not in significantly more epitopes than three random positions from all sites ( p=0 . 157 ) . Epitopes with experimentally characterized CTL responses were mined from the Immune Epitope Database ( Figure 8—source code 1 and Figure 8—figure supplement 1 ) . The primary citation and summary information for epitopes involving sites 259 , 280 , and 384 are in Figure 8—source data 1 . Similar results are obtained if CTL epitopes are instead predicted computationally ( Figure 8—figure supplements 1 , 2 ) . A dN/dS analysis is inconclusive about whether the three sites are under positive or negative selection , probably due to lack of sequence data ( Figure 8—figure supplement 3 and Figure 8—source code 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03110 . 7554/eLife . 00631 . 032Figure 8—source data 1 . Literature-characterized NP human CTL epitopes that include residues 259 , 280 , or 384 and contain sequences conserved or nearly conserved in Aichi/1968 . The epitopes are listed in a TXT file . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03210 . 7554/eLife . 00631 . 033Figure 8—source code 1 . The input data files and the custom Python scripts used to identify human CTL epitopes in NP . The code and data are provided as a ZIP file , which contains a README text file that describes the contents in more detail . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03310 . 7554/eLife . 00631 . 034Figure 8—source code 2 . The data and source code used for the dN/dS analysis . The code and data are provided in a ZIP file; a README file explains the contents . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03410 . 7554/eLife . 00631 . 035Figure 8—figure supplement 1 . Distribution of human CTL epitopes along the NP primary sequence . This plot shows the number of CTL epitopes for each residue in NP . Sites of specific mutations along the evolutionary trajectory are indicated . The top blue bars show the number of experimentally characterized CTL epitopes as mined from the Immune Epitope Database . The bottom red bars show the number of CTL epitopes predicted by NetCTLpan1 . 1 , a computational epitope prediction method , applied to all HLA supertypes . In both cases , the epitope density is scaled to have a minimum of 0 and a maximum of 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03510 . 7554/eLife . 00631 . 036Figure 8—figure supplement 2 . The three epistatically constrained mutations are at sites significantly enriched in CTL epitopes as predicted by a computational epitope prediction program NetCTLpan1 . 1 . The blue curve shows the distribution of epitopes per site for all sites in NP , while the red curve shows the distribution of epitopes per site for the sites in NP that substituted along the evolutionary trajectory . The sites of the three epistatically constrained mutations ( 280 , 259 , and 384 ) are indicated . These three sites contain significantly more epitopes than would be expected from three randomly chosen positions from all sites ( p=0 . 008 ) or from the mutated sites ( p=0 . 028 ) . A random selection of three positions from all mutated sites contains significantly more epitopes than a random selection of three positions from all sites ( p=0 . 011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 03610 . 7554/eLife . 00631 . 037Figure 8—figure supplement 3 . It is largely inconclusive whether the sites of the three epistatically constrained mutations are under positive or negative selection as quantified by dN/dS values . Human H3N2 influenza NP sequences from 1968 to 2011 were analyzed for positive selection using the dN/dS analysis implemented in FUBAR ( Figure 8—source code 2 ) . The plots show the posterior probability that sites are under ( A ) negative or ( B ) positive selection . At reasonable levels of significance , we can conclude that site 384 is not under negative selection and is probably under positive selection . For sites 259 and 280 , there is not strong evidence for either positive or negative selection . We note that the lack of decisive results may be due largely to insufficient sequence data—although we used hundreds of NP sequences , they are closely related with a trunk-like phylogeny as shown in Figure 4B , which will reduce the power of the test . In addition , we note that mutations which are beneficial to viral CTL escape may still fix slowly ( reducing their sites’ dN/dS values ) even if they are under positive selection due to the types of epistatic constraint described in this paper . DOI: http://dx . doi . org/10 . 7554/eLife . 00631 . 037 Our results paint a remarkably coherent picture of epistasis in NP evolution . We identified three mutations that are strongly deleterious to the original parent , yet were eventually fixed without adverse effect . All three mutations decrease NP’s thermal stability . This decreased stability reduces in vivo protein levels , in turn reducing total transcriptional activity and viral growth . On the other hand , stabilizing mutations have little effect on protein levels , activity , or growth in the background of the parent NP—presumably because this parent is already sufficiently stable for its cellular environment . However , these stabilizing mutations play a crucial evolutionary role by counteracting the destabilizing mutations and enabling them to fix during evolution . Of course , we do not wish to caricature protein evolution by suggesting that all epistasis is mediated by stability . In principle , mutations can affect a multitude of properties of NP , including its homo-oligomerization , association with RNA and other proteins , and cellular transport . We do not suggest that these properties are unimportant . In fact , we observe an epistatic interaction between R384G and E375G that likely relates to the electrostatic charge on one of NP’s surfaces . Our assays are also of finite sensitivity , and so may miss small effects that are still significant to natural selection . There is also the potential for epistasis between NP and other viral proteins , although we see no evidence for such epistasis here , since all NP evolutionary intermediates that we tested are functional in a fixed background of other proteins . But doubtless some of these other mechanisms of epistasis would become apparent if we examined even longer evolutionary trajectories . However , the overriding message from our experiments is that stability-mediated epistasis is the dominant constraint on NP evolution . Epistatically interacting mutations can be fixed in several ways . The mutations can accumulate sequentially without ever passing through a low-fitness intermediate ( as in Maynard Smith’s analogy ) , an initial deleterious mutation can be compensated by a subsequent mutation , or multiple mutations can occur simultaneously . We have identified two instances that clearly conform to Maynard Smith’s paradigm: the stabilizing M136I preceded V280A , and the stabilizing S259L preceded R384G ( Figure 4B , C ) . We also identified two instances where the actual evolutionary path is unclear due to a lack of natural sequence data from the relevant timeframe: N334H/L259S , and E375G/R384G ( Figure 4A , B ) . However , in both cases it is at least possible that evolution conformed to Maynard Smith’s paradigm: no simultaneous mutations or deleterious intermediates need have occurred if N334H preceded L259S , and if E375G preceded R384G ( Figure 4A , B ) . Previous experimental studies have identified stabilizing mutations as contributing to the evolution of enzyme specificities ( Bloom et al . , 2006 ) , bacterial ( Wang et al . , 2002; Bershtein et al . , 2006; Weinreich et al . , 2006 ) and viral ( Chang and Torbett , 2011 ) drug resistance , and H5N1 transmissibility ( Imai et al . , 2012 ) . At a broader level , analyses of large datasets have shown that it is common for mutations to be deleterious to one protein homolog but benign to another ( Kondrashov et al . , 2002; Baresic et al . , 2010 ) . Our work illustrates how the dynamics of stability during evolution might explain these findings . As shown in Figure 7C , most of the intermediates during NP’s evolution are only marginally more stable than the minimal threshold where function begins to suffer . This marginal stability of natural proteins has been noted previously , and been given two distinct explanations . The first explanation holds that evolution actively selects for marginal stability because both insufficient and excess stability are deleterious ( DePristo et al . , 2005; Tokuriki and Tawfik , 2009 ) . The second explanation holds that evolution only selects against insufficient stability , but that proteins typically are marginally stable because most mutations are destabilizing and so extra stability is rapidly eroded by functionally neutral but destabilizing mutations ( Taverna and Goldstein , 2002; Bloom et al . , 2007 ) . Our results decisively favor the second explanation for NP , since the evolving protein is usually marginally stable despite the fact that higher stability has no deleterious effect ( Figure 7A , B ) . We therefore suggest the following: functionally neutral but stabilizing mutations occasionally fix by stochastic forces such as genetic drift , population bottlenecks , or hitchhiking . These stabilizing mutations enable NP to tolerate otherwise deleterious destabilizing mutations . Although these destabilizing mutations could in principle also fix by stochastic forces , the three that we have identified are actually adaptive since they contribute to viral immune escape . Stability-mediated epistasis therefore constrains NP’s adaptation as well as its sequence evolution , since the accessibility of immune-escape mutations is dependent on the acquisition of enabling mutations . It is intriguing to speculate whether similar forms of epistasis might constrain the evolution of other proteins . For example , the antigenic evolution of influenza hemagglutinin is punctuated , with a fairly constant rate of sequence change nonetheless leading to periodic jumps in antigenicity that require reformulation of the annual influenza vaccine ( Smith et al . , 2004 ) . One explanation that has been posited for this punctuated pattern is that adaptive antigenic change is limited not by the overall rate of substitution , but rather by the waiting time for the protein to accumulate antigenically neutral mutations that can be productively combined with mutations causing large antigenic changes ( Koelle et al . , 2006; van Nimwegen , 2006 ) . Stability-mediated epistasis of the type that we have observed for NP provides at least one plausible mechanistic explanation for this and other cases of constrained molecular evolution . Via Markov chain Monte Carlo implemented in BEAST ( Drummond et al . , 2012 ) , we estimated the joint posterior distribution of phylogenetic trees and mutations along the branches of these trees given 431 date-stamped human H3N2 NP protein sequences from the Influenza Virus Resource . We assumed a Jones—Taylor—Thornton model ( Jones et al . , 1992 ) of protein substitution , a strict molecular clock and a relatively uninformative coalescent-based prior on the tree . We inferred the unobserved mutations via a data augmentation procedure that exploits uniformization and is robust to model misspecification ( Minin and Suchard , 2008; O’Brien et al . , 2009 ) . Figure 2B reports the maximum clade credible tree from the posterior distribution . Novel to this work , for each posterior sample , we converted the order of mutations along the line of descent from Aichi/1968 and Brisbane/2007 into a directed graph through sequence space ( Figure 2—figure supplement 1 ) . Summarizing these graph samples effectively integrates over uncertainty in the tree and substitution process , returning the marginal posterior distribution of the evolutionary trajectory of interest . In our GraphViz visualization , each circle represents a unique inferred sequence . Areas and intensities of circles are proportional to the posterior probability that the true trajectory visited that sequence . Lines correspond to mutations , with thickness and intensity proportional to the posterior probability that specific mutation connected those two sequences . We labeled connections with posterior probability ≥60% in black in Figure 2; mutations lacking high-confidence connections are in red . Likewise , we considered sequences with posterior probability ≥60% as high confidence and assigned them numeric labels . Figure 2—source code 1 contains the relevant computer code . We reverse-transcribed the Aichi/1968 NP and the Nanchang/1995 PB2 , PB1 , and PA from viral RNA ( BEI Resources NR-9534 and NR-3222 ) , and cloned them into pHW2000 ( Hoffmann et al . , 2000 ) ( a gift from Y Kawaoka ) to create pHWAichi68-NP , pHWNan95-PB2 , pHWNan95-PB1 , and pHWNan95-PA ( Supplementary file 1 ) . Similar plasmids were constructed for the PB2 , PB1 , and PA of Aichi/1968 and Brisbane/2007 and named pHWAichi68-PB2 , pHWAichi68-PB1 , pHWAichi68-PA , pHWBR07-PB2 , pHWBR07-PB1 , and pHWBR07-PA ( Supplementary file 1 ) . We used site-directed mutagenesis to create the other NP variants ( Supplementary file 2 ) . Note that the final sequence at the end of our trajectory ( Step 39 ) matches the Brisbane/2007 protein sequence , but does not match the nucleotide sequence of this strain as we did not introduce any of the synonymous mutations . We measured activity using a previously described ( Bloom et al . , 2010 ) reporter plasmid encoding a GFP vRNA with termini from the A/WSN/1933 PB1 . We co-transfected this reporter into 12-well dishes of 293T cells along with 50 ng of NP plasmid and 200 ng each of pHWNan95-PB2 , pHWNan95-PB1 , and pHWNan95-PA ( Figure 3—figure supplement 1 ) . This amount of NP plasmid is near the midpoint of the dose-response curve ( Figure 3—figure supplement 1 ) . After 20 hr , we quantified the GFP mean-fluorescence intensity ( MFI ) by flow cytometry . We seeded 293T cells at 2 × 105 per well 20–24 hr pre-transfection in D10 ( DMEM with 10% heat-inactivated fetal bovine serum , 2 mM L-glutamine , 100 U/ml penicillin , and 100 μg/ml streptomycin ) . We quantified the activity relative to the average for three replicates of wild-type pHWAichi68-NP . We performed at least three biological replicates for each variant , with each replicate performed on a different day using an independent plasmid mini-prep . Figure 3—source data 1 gives the means and standard errors of the activities for all NP variants . We grew viruses carrying GFP in the PB1 gene using a modification of a previously described system ( Bloom et al . , 2010; Figure 3—figure supplement 2 ) . We used lentiviral transduction to create the cell lines 293T-CMV-Nan95-PB1 and MDCK-SIAT1-CMV-Nan95-PB1 , which express the coding sequence of the Nanchang/1995 PB1 with the F2 peptide disrupted after its eighth codon ( Chen et al . , 2001 ) under control of a CMV promoter . We seeded co-cultures of 2 × 105 293T-CMV-Nan95-PB1 and 2 × 104 MDCK-SIAT1-CMV-Nan95-PB1 cells in D10 media in six-well dishes , and 20–24 hr later transfected with 250 ng each of pHWNan95-PB2 , pHH-PB1flank-eGFP ( Bloom et al . , 2010 ) , pHWNan95-PA , a pHWAichi68-NP variant , pHW184-HA , pHW186-NA , pHW187-M , and pHW188-NS . These last four plasmids ( Hoffmann et al . , 2000 ) encode genes from A/WSN/1933 ( gifts from Y Kawaoka ) . After 20–24 hr , we replaced the D10 with influenza growth media ( Opti-MEM I with 0 . 3% BSA , 0 . 01% heat-inactivated fetal bovine serum , 100 U/ml penicillin , 100 μg/ml streptomycin , and 100 μg/ml calcium chloride ) . After 66 hr ( shortly before peak titers , Figure 3—figure supplement 2 ) , we collected the supernatant and clarified it for 5 min at 2000×g . We infected dilutions of supernatant into 12-well dishes seeded 8 hr earlier at 105 MDCK-SIAT1-CMV-Nan95-PB1/well in influenza growth media , and 16 hr later determined the titer by flow cytometry by using the Poisson equation to estimate the viral titer based on the fraction of GFP positive cells . We performed these titering infections with various volumes of viral supernatant such that the titer could be computed from an infection with between 0 . 5% and 10% of cells green . Note that these titers reflect the number of particles that are able to productively infect cells and transcribe high levels of GFP from viral RNA—they may not reflect the same titers as would be determined using other approaches such as plaque assays or tissue-culture infectious dose 50% assays . We quantified the titer relative to the average of three replicates of the wild-type pHWAichi68-NP ( typically around 103 infectious particles per microliter ) . We performed at least three biological replicates for each variant , with each replicate performed on a different day using an independent plasmid mini-prep . Figure 3—source data 2 gives the means and standard errors . We cloned the NP coding sequence into a mammalian expression plasmid under control of a CMV promoter with a FLAG tag inserted between the N-terminal methionine and the second residue . After the NP stop codon , we added an internal ribosome entry site ( IRES ) followed by mCherry with a C-terminal FLAG tag . We seeded 293T cells at 2 × 105 per well in D10 in 12-well dishes , and 20–24 hr later transfected with 400 ng of plasmid . After 20 hr , we collected the cells and lysed them on ice in 100 μl of RIPA buffer with one protease-inhibitor tablet ( Roche , Basel , Switzerland , 05892791001 ) per 10 ml . We pelleted debris at 21 , 000×g for 10 min , and loaded 2 . 5 μl of clarified supernatant on an SDS-PAGE gel after boiling with a reducing sample-loading buffer . We transferred the protein to a PVDF membrane and stained with a 1:5000 dilution of mouse anti-FLAG ( Sigma , St . Louis , MO , F1804 ) followed by a 1:2500 dilution of Alexa Flour 680-conjugated goat anti-mouse ( Invitrogen , Grand Island , NY , A-21058 ) , using Li-Cor Odyssey ( Lincoln , NE ) blocking buffer ( Li-Cor 927-40 , 000 ) and performing washes with TBS-T ( Pierce , Rockford , IL , 28360 ) . We quantified the ratio of NP to the corresponding mCherry control using a Li-Cor Odyssey Infrared Imaging System , and normalized this ratio to that for wild-type Aichi/1968 NP ( Figure 6—figure supplements 1 , 7 ) . In order to obtain non-aggregated RNA-free NP in a CD-compatible buffer , we introduced two previously described ( Ye et al . , 2006 ) modifications: deletion of residues 2–7 and R416A . We cloned NP with these modifications and a C-terminal 6-histidine tag into pET28b ( + ) , transformed into BL21 Star DE3 ( Invitrogen , Grand Island , NY , C6010-03 ) , and grew 1 L of these bacteria to an OD600 of 0 . 3–0 . 6 at 37°C . We then chilled the cultures on ice , reduced the shaker temperature to 20°C and induced with 500 μM IPTG . After overnight growth , we pelleted the cells and lysed them on ice by sonication in 50 ml of 50 mM sodium phosphate pH 8 . 0 , 500 mM sodium chloride , 0 . 5% Triton X-100 , 10 mM imidazole , 1 mM PMSF , 0 . 1 mg/ml magnesium chloride , 1 mM lysozyme , and 1000 units of benzonase ( Sigma , St . Louis , MO , E1014 ) . We clarified the supernatant for 30 min at 10 , 000×g and 4°C , passed it through a 0 . 45-μm filter , and purified NP over a cobalt column ( Pierce , Rockford , IL , 89969 ) using the manufacturer’s protocol but eluting with 200 mM imidazole . We concentrated the protein with an Amicon Ultra 30 kDa filter , and dialyzed it against CD buffer ( 20 mM sodium phosphate pH 7 . 0 with 300 mM sodium fluoride ) in a 20-kDa dialysis device . We further purified the protein over a Superdex 200 GL size-exclusion column . All variants eluted in a single monomeric peak , and all had ratios of absorbance at 260 nm to absorbance at 280 nm of less than 0 . 65 ( Figure 6—figure supplement 2 ) . We diluted the proteins to 5 μM in CD buffer as determined by the absorbance at 280 nm in a 1 cm quartz cuvette using an extinction coefficient of 0 . 0566 μM/cm , and acquired CD spectra at 20°C with a Jasco J-815 spectropolarimeter . All variants exhibited similar spectra with typical alpha-helical characteristics ( Figure 6—figure supplement 2 , 3 ) . We performed thermal melts at a scan rate of 2°C per minute , monitoring ellipticity at 209 nm . All variants unfolded with a single cooperative transition , allowing us to obtain melting temperatures from sigmoidal curve fits ( Figure 6—figure supplements 4–6 and Figure 6—source data 1 ) . The melting was irreversible ( Figure 6—figure supplement 2 ) , preventing us from calculating equilibrium thermodynamic stabilities . We identified human CTL epitopes of ≤12 residues with at least 89% conservation in Aichi/1968 NP and a verified T-cell response from the Immune Epitope Database ( Vita et al . , 2010 ) . Figure 8—source data 1 lists primary citations for epitopes involving residues 259 , 280 , and 384 . Figure 8—figure supplement 1 shows the number of epitopes at each position . We computed p-values by randomly drawing three different residues from the set of all sites or all mutated sites , and comparing the number of epitopes for these sites to the number for sites 259 , 280 , and 384 ( Figure 8 ) . p-values represent the fraction of 105 random draws that contained at least as many epitopes as sites 259 , 280 , and 384 . We performed a similar analysis for epitopes predicted by NetCTLpan 1 . 1 ( Stranzl et al . , 2010 ) using the default settings for 9-mer peptides and the HLA supertypes ( Figure 8—figure supplement 2 ) . The data and computer code are in Figure 8—source code 1 . The dN/dS comparisons shown in Figure 8—figure supplement 3 were performed with FUBAR ( Murrell et al . , 2013 ) using the DataMonkey server . The sequence data , results , and analysis are in Figure 8—source code 2 .	During evolution , the effect of one mutation on a protein can depend on whether another mutation is also present . This phenomenon is similar to the game in which one word is converted to another word , one letter at a time , subject to the rule that all the intermediate steps are also valid words: for example , the word WORD can be converted to the word GENE as follows: WORD→WORE→GORE→GONE→GENE . In this example , the D must be changed to an E before the W is changed to a G , because GORD is not a valid word . Similarly , during the evolution of a virus , a mutation that helps the virus evade the human immune system might only be tolerated if the virus has acquired another mutation beforehand . This type of mutational interaction would constrain the evolution of the virus , since its capacity to take advantage of the second mutation depends on the first mutation having already occurred . Gong et al . examined whether such interactions have indeed constrained evolution of the influenza virus . Between 1968 and 2007 , the nucleoprotein—which acts as a scaffold for the replication of genetic material—in the human H3N2 influenza virus underwent a series of 39 mutations . To test whether all of these mutations could have been tolerated by the 1968 virus , Gong et al . introduced each one individually into the 1968 nucleoprotein . They found that several mutations greatly reduced the fitness of the 1968 virus when introduced on their own , which strongly suggests that these ‘constrained mutations’ became part of the virus’s genetic makeup as a result of interactions with ‘enabling’ mutations . The constrained mutations decreased the stability of the nucleoprotein at high temperatures , while the enabling mutations counteracted this effect . It may , therefore , be possible to identify enabling mutations based on their effects on thermal stability . Intriguingly , the constrained mutations helped the virus overcome one form of human immunity to influenza , suggesting that interactions between mutations might limit the rate at which viruses evolve to evade the immune system . Overall , these results show that interactions among mutations constrain the evolution of the influenza nucleoprotein in a fashion that can be largely understood in terms of protein stability . If the same is true for other proteins and viruses , this work could lead to a deeper understanding of the constraints that govern evolution at the molecular level .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "microbiology", "and", "infectious", "disease" ]	2013	Stability-mediated epistasis constrains the evolution of an influenza protein
A newfound signaling pathway employs a GGDEF enzyme with unique activity compared to the majority of homologs associated with bacterial cyclic di-GMP signaling . This system provides a rare opportunity to study how signaling proteins natively gain distinct function . Using genetic knockouts , riboswitch reporters , and RNA-Seq , we show that GacA , the Hypr GGDEF in Geobacter sulfurreducens , specifically regulates cyclic GMP-AMP ( 3′ , 3′-cGAMP ) levels in vivo to stimulate gene expression associated with metal reduction separate from electricity production . To reconcile these in vivo findings with prior in vitro results that showed GacA was promiscuous , we developed a full kinetic model combining experimental data and mathematical modeling to reveal mechanisms that contribute to in vivo specificity . A 1 . 4 Å-resolution crystal structure of the Geobacter Hypr GGDEF domain was determined to understand the molecular basis for those mechanisms , including key cross-dimer interactions . Together these results demonstrate that specific signaling can result from a promiscuous enzyme . A cell’s sensory system is composed of complex signaling networks that permit timely responses to changes in environmental conditions , cues from neighboring cells , and feedback from contacting surfaces ( Camilli and Bassler , 2006; Capra and Laub , 2012 ) . How new signaling pathways emerge to control distinct functions remains an important underlying question ( Rowland and Deeds , 2014 ) . While highly conserved signaling enzymes are generally easy to identify at the sequence level , it is challenging to predict their specific activity or role ( Danchin et al . , 2018; Seshasayee et al . , 2010 ) . Another complication is that signaling enzymes often exhibit promiscuous or off-target activity when studied in vitro ( Rowland and Deeds , 2014 ) . In vivo confirmation of both their product and physiological function is essential . In the domain Bacteria , the second messenger signaling pathway involving cyclic di-GMP ( cdiG ) is used almost universally to shift between a free-living and surface-attached biofilm state , which requires coordinated changes in physiology and gene regulation ( Jenal et al . , 2017; Römling et al . , 2013 ) . Multiple processes such as flagellar motility , exopolysaccharide production , quorum sensing , and pilus retraction are controlled by cdiG ( Hengge , 2009 ) . In pathogenic bacteria , cdiG regulates additional virulence factor secretion , host suppression , and defense mechanisms ( Chen et al . , 2014; Valentini and Filloux , 2016 ) . These processes are each driven by specific sets of signaling enzymes: diguanylate cyclases harboring GGDEF domains and EAL/HD-GYP domain phosphodiesterases , which respectively synthesize or degrade cdiG in response to sensory modules fused directly to the enzyme or acting upstream . A grand challenge in bacterial signaling is to understand how cdiG networks utilize only one intracellular output to control diverse adaptations . While many explanations to the ‘one signal , many phenotypes’ problem have been explored ( Hobley et al . , 2012; Sarenko et al . , 2017; Hug et al . , 2017 ) , we recently discovered an unexpected alternative that involves a new activity for GGDEF enzymes . A sub-class of GGDEFs demonstrates promiscuous activity and is capable of producing all three known bacterial cyclic dinucleotides , cdiG , cyclic di-AMP ( cdiA ) , and cyclic GMP-AMP ( 3′ , 3′-cGAMP , also called cyclic AMP-GMP ) ( Hallberg et al . , 2016 ) . The hybrid promiscuous ( Hypr ) GGDEF enzyme GSU1658 ( renamed GacA for GMP-AMP cyclase ) was hypothesized to regulate a new signaling pathway through activation of cGAMP-specific GEMM-Ib riboswitches in Geobacter sulfurreducens , an environmental bacterium known for its unique ability to perform extracellular electron transfer and accelerate bioremediation of subsurface contaminants ( Kellenberger et al . , 2015; Nelson et al . , 2015 ) . However , the physiological function of cGAMP signaling was not established , and it was unclear whether GacA regulated a single ( only cGAMP ) or multiple cyclic dinucleotide signaling pathways in vivo . Furthermore , there were general questions about the molecular mechanism of homodimeric GGDEF enzymes , the most abundant class of cyclic dinucleotide signaling enzymes in bacteria ( Seshasayee et al . , 2010 ) , such as the function of highly conserved residues ( Schirmer , 2016 ) and the identity of the general base , as well as specific questions for how Hypr variants could perform preferential synthesis of cGAMP , a heterodimeric product . In this work , we show for the first time that GacA specifically affects cGAMP levels and cGAMP riboswitch transcripts in G . sulfurreducens and is important during bacterial growth on particulate acceptors , such as mineral Fe ( III ) oxides . In contrast , GacA is not essential for biofilm growth on electrodes , a phenotype associated with cdiG signaling . These results reveal that the general physiological function for cGAMP is to establish a transiently attached lifestyle that is distinct from the permanently attached biofilm lifestyle signaled by cdiG . Furthermore , we sought to understand the molecular mechanism for GacA by obtaining a 1 . 4 Å-resolution structure of the Geobacter Hypr GGDEF domain bound to GTP . Combining this structure with kinetic analyses and mathematical modeling afforded insights into how GGDEF enzymatic activity is regulated in general as well as uncovered natural variations that give rise to cGAMP synthesis . Together , these genetic and biochemical analyses provide evidence that this cGAMP signaling pathway emerged from components of cdiG signaling to regulate a distinct surface-associated lifestyle , and gives a full picture of cGAMP signaling from the molecular to the cellular to the environmental level . Geobacter isolates produce energy via contact-dependent electron transfer to extracellular metals , which exist as insoluble precipitates at neutral pH ( Navrotsky et al . , 2008 ) . Stimulation of this biological metal reduction activity is useful for bioremediation of metal-rich sites and anaerobic oxidation of petroleum-based groundwater pollutants ( Chang et al . , 2005; Lovley et al . , 2011; Rooney-Varga et al . , 1999 ) . Geobacter also can grow via electron transfer to electrode surfaces , where their biofilms produce electricity in bioelectrochemical devices that use wastewater or contaminated groundwater ( Bond and Lovley , 2003; Logan and Rabaey , 2012; Lovley , 2012 ) . The ability to transfer electrons to extracellular substrates requires multiple extracellular structures , including pili , polysaccharides , and cytochromes localized to the outer cell surface . cGAMP-responsive riboswitches ( GEMM-Ib family ) are conserved upstream of many cytochrome , pilus assembly , and polysaccharide biosynthesis genes in most Geobacter species ( Kellenberger et al . , 2015; Nelson et al . , 2015 ) , suggesting a possible role for this cyclic dinucleotide in attachment to extracellular surfaces that serve as electron acceptors . The discovery of a GMP-AMP cyclase in Geobacter ( GacA ) ( Hallberg et al . , 2016 ) presented the hypothesis that GacA synthesizes cGAMP in vivo to alter gene expression via cGAMP-specific riboswitches . However , no Hypr GGDEF enzyme including GacA has been linked yet to intracellular cGAMP levels or to phenotypic changes in any organism . To assess the physiological role of cGAMP signaling in G . sulfurreducens , we constructed a scarless gacA deletion strain and tested its ability to respire soluble and particulate extracellular electron acceptors ( See Key Resources Table ) . The ΔgacA strain was defective in reducing Fe ( III ) oxide particles , including both akaganeite and amorphous insoluble Fe ( III ) - ( oxyhydr ) oxide , but grew normally with soluble compounds such as fumarate or Fe ( III ) -citrate ( Figure 1A and B ) . Mutants lacking gacA always demonstrated a ~5 d lag during reduction of Fe ( III ) oxides , but if left exposed for over 14 d , the ∆gacA strain eventually produced Fe ( II ) at levels similar to wild type . Re-expressing gacA as a single copy on the chromosome restored Fe ( III ) reduction , and caused it to initiate sooner than wild type . In contrast , after a ~1 day lag , the rate and extent of growth as a biofilm attached to electrodes poised at −0 . 1 V vs SHE ( a potential chosen to mimic Fe ( III ) oxides ) was unaffected by deletion of gacA ( Figure 1C ) . The ∆gacA defect with Fe ( III ) oxides was the opposite of mutants in esn genes encoding chemosensory , histidine kinase , and diguanylate cyclase response regulator proteins . Mutants in esn genes reduce Fe ( III ) oxides similar to wild type , but show poor biofilm growth on electrodes ( Chan et al . , 2017 ) . One of these esn genes , esnD ( GSU3376 ) , encodes a GGDEF diguanylate cyclase that produces only cdiG based on biosensor analysis ( Hallberg et al . , 2016 ) . We compared growth of a mutant lacking cdiG-producing EsnD under the same conditions used to study the mutant lacking cGAMP-producing GacA . The ∆esnD mutant reduced Fe ( III ) oxides more rapidly than wild type , but showed a > 3 d lag in colonizing electrode surfaces and never reached current levels observed in wild type ( Figure 1A and C ) . Re-expressing esnD as a single copy integrated into the chromosome restored biofilm growth on −0 . 1 V vs . SHE electrodes ( Figure 1C ) . These data support that cdiG contributes to biofilm growth on electrodes , while cGAMP is involved in reduction of Fe ( III ) oxide particles . In addition , the enhanced growth of ∆esnD mutants with Fe ( III ) oxides suggests an antagonistic effect of cyclic di-GMP on Fe ( III ) oxide reduction . To test if deletion of gacA altered cGAMP levels within the cell , we developed a new luciferase-based reporter by cloning the cGAMP-specific pgcA riboswitch upstream of a nano-luciferase ( NLuc ) gene , then integrating this reporter as a single copy into the Tn7 insertion site of G . sulfurreducens . For these experiments , all constructs were grown under the same conditions to stationary phase under electron acceptor limitation . Luminescence in the cGAMP reporter strain declined over 80% when gacA was deleted , and recovered to wild type luminescence levels when gacA was re-expressed from its native promoter ( Figures 1D , 2A and B ) . Over-expression of gacA from a constitutive promoter increased luminescence to 200% of wild type . This result links gacA to intracellular levels of cGAMP in Geobacter , and correlates with parallel LC/MS analysis of cell extracts , which also showed that cGAMP levels fell below the detection limit in the ∆gacA strain ( Figure 2C ) . The new cGAMP reporter assay also allowed us to interrogate roles of conserved residues in GacA that are critical for activity in diguanylate cyclases . GacA has an N-terminal CheY-like receiver domain and a C-terminal GGDEF domain . In WspR , a Pseudomonas aeruginosa diguanylate cyclase with similar domain architecture , phosphorylation of a conserved aspartate in the receiver domain promotes dimerization , leading to an active enzyme ( De et al . , 2008; Huangyutitham et al . , 2013 ) . In a G . sulfurreducens ∆gacA background , expression of a GacAD52A variant with the phosphorylation site replaced by alanine produced similar levels of cGAMP as GacAWT . Expression of gacAD52A also restored Fe ( III ) oxide reduction to a ∆gacA mutant ( Figure 1D and E ) . Thus , it appears that the conserved aspartate is not essential for GacA activation . Aspartate phosphorylation instead may deactivate GacA or the receiver domain may be activated through non-canonical mechanisms ( Lin et al . , 2009; Ocasio et al . , 2015; Trajtenberg et al . , 2014; Wang et al . , 2009 ) . A second mechanism regulating canonical GGDEF domains involves a conserved allosteric inhibitory site ( I-site ) that binds cyclic dinucleotides . We previously showed that WT GacA co-purifies predominantly with cdiG bound and has low activity when expressed in E . coli , whereas the R393A I-site mutant of GacA does not purify with bound dinucleotides and exhibits increased in vitro activity ( Hallberg et al . , 2016 ) . In extracts of cells overexpressing GacA , the major CDN present is cGAMP ( Hallberg et al . , 2016 ) , yet GacA still purifies with bound cdiG , which supports the hypothesis that the I-site is specific for cdiG . When we expressed the GacAR393A variant insensitive to allosteric inhibition in a ∆gacA background , cGAMP-dependent luciferase reporter activity increased 30-fold compared to wild type . Expressing this highly active I-site mutant in the ∆gacA strain also led to the highest observed rates of Fe ( III ) oxide reduction , nearly doubling the level of Fe ( II ) produced at all time points ( Figure 1D and E ) . These data suggest that occupancy of the I-site , rather than phosphorylation of the receiver domain , primarily regulates GacA activity . The increased Fe ( III ) reduction activity of the cGAMP-overproducing strain further supports a role for cGAMP in enhancing metal oxide reduction . The ΔesnD strain also reduced amorphous insoluble Fe ( III ) -oxides more rapidly than the wild type strain ( Figure 1A ) , leading to a hypothesis that deletion of esnD increased cGAMP levels in G . sulfurreducens . Consistent with this hypothesis , activity of the cGAMP reporter was 3-fold higher when esnD was deleted ( Figure 1D ) . Using a different reporter construct comprised of an engineered cdiG-responsive riboswitch ( Figures 1D , 2A and B ) , we confirmed that deletion of esnD caused a detectable decrease in cdiG , while cdiG levels did not change significantly in ∆gacA . These results support cGAMP synthesis by GacA being inhibited by intracellular cdiG levels . Mutation of the I-site appears to relieve this allosteric inhibition , as does lowering of cdiG levels in the ∆esnD strain , triggering corresponding increases in Fe ( III ) oxide reduction rates . Many GEMM-I riboswitches from Geobacter selectively bind cGAMP over other cyclic dinucleotides such as cdiG and are the founding members of the GEMM-Ib subclass . For example , in-line probing showed the riboswitch upstream of pgcA used for our reporter analysis was ~1200 fold more selective for cGAMP over cdiG ( Kellenberger et al . , 2015 ) , and similar results were reported in ( Nelson et al . , 2015 ) . The correlation we observed between increased Fe ( III ) oxide reduction and increased cGAMP levels , combined with the in vitro activity of cGAMP riboswitches , suggests that cGAMP could be a global effector of genes crucial to metal reduction . There are 17 GEMM-I riboswitches in G . sulfurreducens , and in several cases , two riboswitches occur in tandem upstream of a gene or operon . When gacA was deleted , RNAseq analysis showed that transcription of all genes downstream of GEMM-Ib riboswitches declined ( Figure 3A , Supplementary file 1 ) . For example , deletion of gacA decreased expression of an operon containing outer membrane cytochromes OmcAHG ( GSU2885-GSU2882 , 16-fold decrease ) , an operon of uncharacterized lipoprotein transpeptidases ( GSU0181-0183 , 16-fold decrease ) , the extracellular cytochrome PgcA ( 10-fold decrease ) , a transcriptional regulator and genes within the pilMNOP operon ( 2–3 fold decrease ) , the extracellular multiheme cytochromes OmcST , ( 2–3 fold decrease ) , and multiple hypothetical genes ( Figure 3A ) . In contrast , cells lacking esnD showed a ~2 fold increase in expression of these same genes controlled by cGAMP-responsive riboswitches ( Figure 3B ) . Every gene that showed a decrease in expression due to gacA deletion also showed an increase due to ∆esnD . The smallest effect was in the operon containing OmcST , which was previously reported to contain a riboswitch sensitive to both cGAMP and cdiG ( Kellenberger et al . , 2015 ) . Interestingly , some hypothetical genes ( GSU0919 , GSU3250 , and GSU3409 ) , and the entire operon for the multiheme cytochrome OmcZ were downregulated in both ∆gacA and ∆esnD despite a lack of known riboswitch sequences , suggesting additional modes for cyclic dinucleotide regulation . As OmcZ is known to be essential for growth in electrode biofilms ( Nevin et al . , 2009 ) , the unexpected decrease in OmcZ due to gacA deletion likely explains the lag in electrode growth seen in Figure 1C . This effect may be due to minor contribution by GacA to cdiG levels or from pleiotropic effects from cGAMP signaling or its many downstream effectors . Closer inspection of intergenic regions confirmed lower riboswitch mRNA levels and increased termination near the cGAMP recognition site in ∆gacA mutants . A table showing the inverse relationship in riboswitch RNA levels between ΔgacA and ΔesnD strains is shown in Figure 3C and a map of two different untranslated regions is shown in Figure 3D . For example , in the tandem riboswitch upstream of the OmcAHG gene cluster , deletion of gacA eliminated detectable RNA by the second riboswitch sequence , while RNA levels in each riboswitch region increased ~2 fold in the ΔesnD strain . Previously published experiments selecting for faster Fe ( III ) oxide reduction activity resulted in evolved G . sulfurreducens variants containing mutations in the pgcA GEMM-Ib riboswitch ( GSUR3008 ) ( Tremblay et al . , 2011 ) . Similarly , we identified a naturally evolved variant with a single A inserted after residue 93 of the pgcA riboswitch that had accelerated growth under laboratory conditions in which Fe ( III ) oxides were abundant . The A insertion is predicted to destabilize the riboswitch terminator , which in another terminator mutant we showed bypasses the need for cGAMP binding to turn on gene expression ( Kellenberger et al . , 2015 ) . To test if faster Fe ( III ) reduction could be explained by cells becoming insensitive to cGAMP regulation , a scarless and markerless G . sulfurreducens strain was made to reconstruct this natural variation in a clean genetic background ( Figure 2E ) . The Fe ( III ) oxide reduction rate was indeed increased in this 93+A strain . When gacA was deleted in this 93+A strain , pgcA expression and metal reduction rate remained high . Global transcriptional analysis showed that only expression of pgcA was increased , and was insensitive to cGAMP signaling , with no other cGAMP-dependent genes , genes for other electron transfer proteins , or genes for pili affected ( data not shown ) . These genetic and RNAseq experiments establish a unique phenotype controlled by 3’3’-cGAMP , and show that GacA is primarily responsible for formation of this second messenger . An opposing relationship for cGAMP and cdiG is established , with each cyclic dinucleotide enhancing extracellular electron transfer to a distinct type of surface . To our knowledge , this is the first time that these two contact-dependent electron transfer processes have been shown to be differentially regulated on a global scale . In this context , the signaling enzyme GacA presents an enigma: we previously discovered that this founding member of Hypr GGDEFs can produce mixtures of cdiG , cGAMP , and cdiA , depending on in vitro conditions ( Hallberg et al . , 2016 ) . In the next sections , we employ biochemical analysis , mathematical modeling , and structural elucidation to determine how this homodimeric enzyme ‘breaks symmetry’ in several ways to produce the heterodimeric cGAMP signal in the cell . In addition , the role of several residues ultra-conserved across all GGDEF enzymes but with previously unassigned function is revealed . To date , two other enzymes have been discovered that produce cGAMP or the related compound , mixed linkage cGAMP: DncV from Vibrio cholerae , and cGAS in metazoans ( Davies et al . , 2012; Sun et al . , 2013; Wu et al . , 2013 ) . Both enzymes have one active site per monomer and operate via a two-step mechanism , wherein a linear dinucleotide intermediate is formed , rotated in the active site , then cyclized . Importantly , DncV and cGAS differ in the order in which the nucleotide linkages are formed ( Figure 4A ) . DncV initially produces pppA[3′ , 5′]pG , utilizing ATP as the nucleophile donor and GTP as the electrophile acceptor , whereas cGAS produces pppG[2′ , 5′]pA ( Gao et al . , 2013; Kranzusch et al . , 2014 ) . Thus , these two enzymes have opposite preferences for the first phosphodiester bond formed . In contrast , GacA is a homodimeric enzyme that has one nucleotide substrate binding site per GGDEF domain , or half of the active site per monomer . We observed that GacA generates both types of linear intermediates in the presence of nonhydrolyzable analogs ( Figure 4B ) , which means that both ATP and GTP can serve as donor and acceptor . This result reveals a marked difference between GacA and the other two dinucleotide cyclases , DncV and cGAS , and is consistent with the increased promiscuity of GacA to produce homodimeric products , for example cdiG and cdiA . To gain insight into how GacA preferentially produces cGAMP in vivo , it was necessary to establish a full kinetic model for the enzyme ( Figure 5A ) . First , we measured initial rates for product formation with single substrates ( ATP or GTP ) using an enzyme-coupled assay for pyrophosphate detection ( Burns et al . , 2014 ) . In these two cases , the kinetic model is greatly simplified because only one homodimeric product is generated , and this model has been validated for canonical GGDEF enzymes ( Oliveira et al . , 2015 ) . Interestingly , kcat values are similar for production of cdiA and cdiG ( 0 . 03–0 . 04 sec−1 ) . The main difference instead appears to be substrate binding , as GTP is the preferred substrate over ATP ( Table 1 ) . Second , to obtain values for the two heterodimeric equilibrium constants ( e . g . KA\|G , binding constant for ATP given GTP is pre-bound ) , we compared computationally modeled product ratios to experimental measurements to find KA\|G and KG\|A values that optimally fit the data ( Figure 5B and C , Figure 5—figure supplement 1 ) . In these models , kcat , AG was set conservatively to equal the catalytic rate constants determined for the homodimeric products ( 0 . 03 sec−1 ) . This assumption is supported by the fact that ATP and GTP are equally competent as donor and acceptor ( Figure 4B ) . We found that the best-fit values to solve the full kinetic model were KA\|G = 71 µM and KG\|A = 10 µM ( Table 1 ) . This result shows that there is positive cooperativity ( K1 > K2 ) facilitating binding of the second substrate for all reaction pathways . We also analyzed whether binding constants for the second nucleotide are different depending on whether enzyme first bound ATP or GTP , for example comparing KG\|A to K2G and KA\|G to K2A . This effect is termed selective cooperativity , as it affects the product ratios . Comparison of KG\|A to K2G shows that there is a 2-fold enhancement of GTP binding to the A-bound vs . G-bound enzyme , which would lead to preferential cGAMP production ( Figure 5A ) . Comparison of KA\|G to K2A shows a 1 . 4-fold enhancement of ATP binding to the A-bound vs G-bound enzyme , but in this case the model fit is relatively insensitive to changes in KA\|G value ( Figure 5B ) , so this effect may or may not be significant . Taken together , the kinetic model provides support for selective cooperativity favoring cGAMP production by enhancing GTP binding to the A-bound enzyme . The A-bound form is favored under cellular conditions where ATP levels are typically 3-fold or more relative to GTP levels ( Buckstein et al . , 2008 ) . While performing the kinetic modeling , we observed that substrate depletion during in vitro reactions can skew product ratios over time . Since NTP levels are expected to be maintained at cellular homeostasis , we used the computational model to simulate in vivo product ratios with substrate concentrations remaining constant . This model demonstrates that GacA is predominantly a cGAMP synthase across the entire physiological range of substrate ratios ( Figure 5D ) , and is unlikely to switch between producing different signals in vivo , as had been an alternative proposed function ( Hallberg et al . , 2016 ) . The kinetic model also allows us to estimate the impact of receiver ( Rec ) domain activation . While our data indicate that the traditional phosphorylation site D52 is not required for GacA activity in vivo ( Figure 1D , E ) and in vitro ( Hallberg et al . , 2016 ) , alternative activation of Rec domains by S/T phosphorylation , kinase binding , or ligand binding has been shown ( Lin et al . , 2009; Ocasio et al . , 2015; Trajtenberg et al . , 2014; Wang et al . , 2009 ) . The low kcat values that we measured in vitro are in the range observed for other non-activated GGDEF enzymes ( Wassmann et al . , 2007 ) , and activation has been shown to increase canonical GGDEF activity by up to 50-fold ( Huangyutitham et al . , 2013; Paul et al . , 2007 ) . While uniform effects on kcat values would not change product ratios , we used the kinetic model to simulate product ratios if kcat , AG was increased asymmetrically by activation more than kcat , diG or kcat , diA ( Figure 5E ) . The result is that GacA can be almost fully selective ( >90% cGAMP ) if the proposed asymmetric activation leads to kcat , AG that is nine times kcat , diG or kcat , diA ( Figure 5E ) , which would result from a difference in activation energy ( ΔΔGo ) of only 1 . 3 kcal/mol . Taken together , these results show how cooperative binding , including selective cooperativity induced by the first substrate bound , and tuning substrate affinities to cellular concentrations , could make GacA predominantly produce cGAMP . Mathematical modeling also led to the hypothesis that asymmetric activation could further favor GacA behaving exclusively as a cGAMP signaling enzyme in vivo . To gain insight into the molecular basis for function and mechanism of this signaling enzyme , we pursued structural characterization of the GacA GGDEF domain from Geobacter metallireducens in the presence of GTP , and obtained a 1 . 4 Å resolution x-ray crystal structure as an N-terminal fusion with T4 lysozyme ( Table 2 , Figures 6 and 7 ) . The GacA GGDEF domain has a βααββαβαβ global topology that positions one guanosine substrate above the signature [G/A/S]G[D/E]E[F/Y] motif and can be overlaid with a canonical GGDEF domain with an RMSD value of 1 . 152 Å ( Figures 6B and 8 ) . A region behind the two alpha helices that support substrate binding is modified from a beta sheet to a helical/loop motif , and varies considerably between GGDEF structures ( Chen et al . , 2016; Dahlstrom et al . , 2015; Deepthi et al . , 2014; Yang et al . , 2011 ) . Electron density for three guanine nucleotides was found in the GacA structure , two in nucleotide-interacting regions that are conserved in other GGDEF domains ( Figure 7 ) ( Chan et al . , 2004 ) . One guanine nucleotide is bound near the canonical allosteric inhibitory site ( I-site ) ( Chan et al . , 2004; Christen et al . , 2006 ) and the second nucleotide is bound in the active site above the GGDEF motif ( Figures 6A and 7 ) . For the latter , we were only able to find partial localized electron density for the alpha phosphate ( Figure 7B ) . It is likely that GTP was hydrolyzed during crystallization but remained coordinated in the active site with density now visible for the guanosine and beta-gamma pyrophosphate ( PPi ) . For clarity , we show both the original and modeled structures with the alpha phosphate . The final guanosine nucleotide binds at the T4 lysozyme-GGDEF interface and may act to stabilize the construct in a way that ATP cannot . To our knowledge , this is the first structure of a Hypr GGDEF domain . Our structure of the Hypr GGDEF domain with bound guanosine plus PPi shows that two specific hydrogen-bonding interactions are made with the nucleobase , a Watson-Crick interaction with Ser348 and a sugar-face interaction with Asn339 ( Figure 6B ) . The phosphate backbone is recognized via several specific interactions ( Figure 8 ) . In particular , a magnesium ion coordinates to the β and γ phosphates and is held in place by the side chain of Asp374 in the GGDEF motif . A signature difference between Hypr and canonical GGDEF domains is that Hypr GGDEFs have Ser348 , whereas the canonical GGDEFs have an aspartate residue at that position . We previously hypothesized that Ser348 would form hydrogen bonds on the Watson-Crick face of either guanine or adenine ( Hallberg et al . , 2016 ) , and this can be seen in our structure . In fact , we have shown that Hypr GGDEFs can accept other purine NTPs as substrates ( Figure 8 ) . However , making the corresponding D-to-S mutant for canonical GGDEFs unexpectedly inactivates the enzyme in four out of five cases ( Hallberg et al . , 2016 ) . One enzyme , GSU3350 , remained active , but solely as a diguanylate cyclase . Overlaying Hypr and canonical GGDEF domains provides a potential explanation for these earlier results ( Figure 6B ) , although an important caveat to this analysis is that the Hypr GGDEF structure contains guanosine plus PPi bound instead of intact GTP in the case of the canonical GGDEF . To interact with serine , which is a shorter side chain than aspartate , the guanosine shifts toward the α2 helix in the Hypr GGDEF compared to GTP in the canonical structure . This shift maintains the hydrogen-bonding distance ( 2 . 6 Å in ImDGC and 2 . 7 Å in GmGacA ) , but requires a corresponding shift of other interactions that would stabilize GTP in the active site . In the Hypr GGDEF structure , the Mg2+ coordinating the phosphates also moves toward the α2 helix relative to the canonical structure , even though there is no bond between the guanosine and phosphates in our structure . We hypothesized that this key compensatory shift is due to the presence of Asp374 in the GGDEF motif of GmGacA , which is shorter than the glutamate residue found in the GGEEF motif of the canonical enzyme whose structure was overlaid . To test whether the Mg2+-GGDEF interaction is indeed critical to GacA activity , we made the D374E mutant of GsGacA , which converts the motif to GGEEF ( the indicates that the numbering used corresponds to the G . metallireducens GacA structure , because GsGacA is shorter by one amino acid ) . The D374E mutant is inactive , as expected for the Ser348/GGEEF combination causing loss of substrate binding ( Figure 6D , Figure 6—figure supplement 1 ) . The S348D/D374E double mutant , which represents the Asp348/GGEEF combination , restores activity , but solely as a diguanylate cyclase . The S348D mutant also becomes an active diguanylate cyclase . Taken together , these observations support a ‘Goldilocks’ model for the GacA active site ( Figure 6C ) , in which the residue interacting with the WC face of the nucleotide , Ser348 , and the magnesium coordinated the GGDEF motif must be the appropriate distance apart . If the two components are too far apart , as in the Ser348/GGEEF case , substrate binding cannot occur . GacA mutants that represent Asp348/GGEEF and Asp348/GGDEF combinations retain activity because they also can match the right distance , the latter through flexibility of the aspartate side chains . However , these mutants become diguanylate cyclases because the S348D mutation drives specificity for GTP . In support of this model , bioinformatics analysis shows that natural GGDEF enzymes with Ser/Thr at position 348 harbor the GGDE[F/Y] motif exclusively ( Figure 6—figure supplement 1D ) , whereas GGDEF enzymes with Asp at position 348 are almost evenly divided between D and E at the central position of the motif ( 57% and 43% , respectively ) ( Figure 6—figure supplement 1E ) . To further demonstrate that the Goldilocks model applies to canonical GGDEF enzymes , we performed mutational analysis on WspR , a diguanylate cyclase from Pseudomonas aeruginosa ( Hickman et al . , 2005 ) . The WspR mutants recapitulate the same activity trends that were shown for GacA ( Figure 6—figure supplement 1B ) . Furthermore , the model explains our prior D-to-S mutagenesis results; the four inactive enzymes have a GGEEF motif , while the enzyme that retained activity , GSU3350 , has a GGDEF motif . While the Hypr GGDEF monomer structure gave some insights into substrate binding , the enzyme functions as a homodimer , with one NTP binding site per monomer . To elucidate the function of other conserved residues , we superimposed our structure onto both GGDEF domains of the C2 symmetric enzyme dimer structure from Idiomarina sp . A28L ( Gourinchas et al . , 2017 ) . In both dimer structures , the glutamate residue that is the fourth residue in GGDEF is close to the GTP/guanosine:PPi bound to the opposite monomer , and in the case of GacA , is oriented appropriately to deprotonate the 3’ hydroxyl group from the substrate ( Figure 9A ) . This observation strongly suggests that this glutamate is the general base that activates the nucleophile donor . This glutamate was among the ultra-conserved residues in GGDEF domains that had no previously assigned function ( Schirmer , 2016 ) . Mutating this residue to glutamine knocks out catalytic activity of GacA ( Figure 9E ) , which is in line with prior experiments demonstrating that this residue is required for canonical GGDEF function ( Malone et al . , 2007 ) . Identifying this glutamate as the general base provides molecular insight into the regulatory mechanisms for GGDEF enzymes . Some GGDEF enzymes are activated by shifting oligomeric states , from monomer to dimer or even to higher order oligomers ( Huangyutitham et al . , 2013; Paul et al . , 2007 ) . Monomers are inactive because each monomer binds only one NTP . However , other GGDEF enzymes are predicted to be activated by changing the dimer conformation ( Gourinchas et al . , 2017; Zähringer et al . , 2013 ) . In these cases , the orientation of the two monomers can affect whether the newly identified general base is poised to deprotonate the 3′ hydroxyl across the dimer . We observed another cross-dimer interaction with the guanosine substrate that had different residue identities for Hypr versus canonical GGDEFs . In the diguanylate cyclase dimer , Arg537 appears to form a cation-π interaction by stacking above the nucleobase in the opposite active site ( Figure 9B ) . This conserved residue ( 94% of predicted diguanylate cyclases ) also had no prior assigned function ( Schirmer , 2016 ) . Interestingly , the modeled dimer of the G . metallireducens Hypr GGDEF has a tyrosine ( Tyr304 ) at this position , which is tucked away in the monomer structure ( Figure 9C ) . However , with side chain rotation it can form a π-π stacking interaction with either adenine or guanine ( Figure 9D , Figure 9—figure supplement 1 ) . Thus , our analysis of the structures suggests that Arg537 is a previously unappreciated determinant of substrate specificity in diguanylate cyclases , which is replaced by other residues in Hypr GGDEF enzymes . In support of this functional assignment , the corresponding Y304*R mutant was found to have a product ratio more skewed towards cdiG , which is consistent with the cation-π interaction favoring guanine over adenine ( Figure 9E , Figure 9—figure supplement 1C ) . Also , an analysis of Hypr GGDEFs from different bacteria previously showed that two enzymes harboring an arginine produce more cdiG ( Cabther_A1065 and Ddes1475 ) , whereas enzymes harboring tyrosine , serine , alanine , or glutamine produce predominantly cGAMP ( Hallberg et al . , 2016 ) . These results reveal that this cross-dimer interaction affects product distribution , leading us to propose a mechanism for the cooperative binding and putative asymmetric activation effects shown by kinetic modeling . As shown in the model , the status of one monomer , for example the identity of nucleotide substrate and/or Rec activation , can be communicated by residue ( s ) that make cross-dimer interactions to the substrate in the other monomer’s active site . For example , changes to the orientation of Tyr304 will tune the binding energy , possibly in a differential manner for guanosine or adenosine substrates . There may be other residues besides Tyr304 that are involved in this cross-dimer communication . Taken together , our analysis of the crystal structure shows that consideration of cross-dimer interactions may be key to unlocking residue functions for both Hypr and canonical GGDEFs . For bacteria , obtaining energy is key to niche survival , whether in a host or outside environment . In dynamic anaerobic environments , where oxygen is unavailable or limiting , microbes must seek alternative electron acceptors . One such strategy that profoundly impacts Earth’s biogeochemistry is the process of extracellular electron transfer to metals , surfaces , and other cells . Use of environmental metal oxides as terminal electron acceptors by Geobacter requires cell-metal contact to facilitate electron transfer , and while attachment to surfaces is typically regulated by cdiG signaling , our results demonstrate that a separate mechanism has emerged for metal particle attachment . In retrospect , permanent biofilm-like attachment as driven by cdiG signaling would be a poor choice for interacting with environmental metal oxides , as Fe ( III ) oxides are usually nanophase ( <100 nm ) , and a single metal particle cannot provide enough energy to support cell division ( Levar , 2013; Zacharoff et al . , 2017 ) . Thus , based on energetics and size , metal oxides present a conundrum for metal-reducing bacteria: a surface that requires transient , rather than permanent , contact . We hypothesize that cGAMP signaling — and thus GacA — arose as a divergent signaling system for this separate , transiently surface-associated state ( Figure 10 ) . Specifically , GacA helps coordinate electron transfer to Fe ( III ) oxides , but is not involved in permanent biofilm growth on electrodes , or planktonic growth with soluble metals . These phenotypes contrast with the involvement of a canonical cdiG-synthesizing GGDEF enzyme , EsnD , in biofilm-based electricity production on electrodes ( Chan et al . , 2017 ) . Along with growing evidence that transient-attached and permanent-attached states are distinct stages in the biofilm lifestyles of bacteria ( Lee et al . , 2018 ) , this study provides the first evidence that these modes can be signaled by two different cyclic dinucleotides . Whether this paradigm is more widespread or whether different mechanisms are present in other bacteria is the subject of future work . Interestingly , Vibrio cholerae has a completely distinct cGAMP signaling pathway from the GacA-cGAMP-riboswitch pathway analyzed in this study ( Davies et al . , 2012; Severin et al . , 2018 ) . As exemplified by GacA , Hypr GGDEFs likely arose from divergent evolution of diguanylate cyclase enzymes in ancestral deltaproteobacteria , as it is conserved across species of Geobacter , Myxobacteria , and others . In contrast , DncV is found in a pathogenicity island unique to the El Tor strain of V . cholerae that also contains a cGAMP-activated phospholipase ( Severin et al . , 2018 ) . Activation of this phospholipase changes membrane fatty acid composition and inhibits cell growth ( Severin et al . , 2018 ) , which contrasts with the riboswitch-driven transcriptional response and electron transfer phenotype in G . sulfurreducens . A challenge we faced at the outset was to reconcile our in vivo observations that showed GacA produces cGAMP-specific phenotypes with prior in vitro observations that showed GacA to be a promiscuous dinucleotide cyclase . This paradox mirrors a common problem in studying two-component signaling: histidine kinases can phosphorylate non-cognate receiver domains in vitro , whereas this crosstalk is not observed in vivo . By combining biochemical analysis with mathematical modeling , we demonstrate that under standard cellular conditions , substrate-assisted cooperative binding biases production to give predominantly cGAMP in vivo ( Figure 5A ) . The main side product , cdiG , is likely produced by GacA at sufficiently low amounts that a housekeeping phosphodiesterase can prevent cross-signaling in vivo , as shown for PdeH in E . coli ( Sarenko et al . , 2017 ) . A new hypothesis that arises from the model is that with asymmetric activation , GacA can act as a completely selective enzyme in vivo ( Figure 5E ) . One molecular mechanism that we propose here and would be very intriguing to explore in the future is that asymmetric activation could occur via modulation of a single Rec domain . These results run counter to the intuition that enzymes with homodimeric active sites can only produce symmetrical products . In fact , they lead to a newly intuitive explanation: active site symmetry is broken once the first substrate binds , and the identity of that substrate can influence the second binding event , giving rise to substrate-assisted selectivity . Our structural analysis further reveals a signature cross-dimer residue in Hypr GGDEFs that is poised to ‘read’ substrate identity and allosterically transfer that information to the other half active site ( Figure 9B–D ) . Combining structural and biochemical analyses also led to several insights into the function of GGDEF enzymes in general , in terms of substrate recognition and catalysis . The Goldilocks model explains why Ser348/GGEEF enzymes are non-functional and provides a basis of selectivity against pyrimidine substrates . The natural pyrimidine NTPs most likely are too short to interact with both the active site Mg2+ and residues that recognize the nucleobase , even if hydrogen bonds were matched . GGDEF enzymes capable of coordinating larger divalent metals or otherwise shortening the distance may be able to accommodate pyrimidine substrates . The identification of cross-dimer residues as the general base and involved in substrate recognition provides a molecular basis for activation mechanisms that involve conformational changes of the dimer . From the broader perspectives of protein evolution and engineering , GacA provides an important case study for divergent evolution to a new in vivo function . Our results reveal functional intermediates in a potential mutational pathway to evolve cGAMP cyclases from diguanylate cyclases . First , only GGDEF not GGEEF enzymes are on pathway , followed by R304Y and D348S in either order . Importantly , this implies that cGAMP signaling can arise in any bacteria harboring GGDEF-type diguanylate cyclases . Another major finding is that this ‘promiscuous’ enzyme is the functional endpoint and in fact is attuned to play a highly specific role in the cell , as shown by phenotypic data that support its key role in cGAMP production and signaling . However , in vitro analysis or biochemical screening that does not account for cellular substrate concentration and homeostasis would result in an incorrect functional assignment . In fact , this in vitro ‘blindspot’ caused Hypr GGDEF enzymes to remain undiscovered until we performed an in vivo biosensor-based screen ( Hallberg et al . , 2016 ) . Taken together , these insights are instructive for future efforts to discover and design signaling enzymes that produce other cyclic dinucleotides , besides the four currently known in bacteria and mammals . All oligonucleotides were purchased from Elim Biopharmaceuticals ( Hayward , CA ) or IDT ( Coralville , IA ) . The codon-optimized WspR gene was purchased from IDT as a gBlock ( See Key Resources Table ) . Cyclic dinucleotide standards were purchased from Axxora ( Farmingdale , NY ) or enzymatically synthesized . NTP stocks were purchased from New England Biolabs ( Boston , MA ) . All strains and plasmids used in this study are listed in the Key Resources Table . Antibiotics were used in the following concentration for E . coli; kanamycin 50 µg/mL; spectinomycin 50 µg/mL , chloramphenicol 25 µg/mL , carbenicillin 50 µg/mL , and ampicillin 100 µg/mL . For G . sulfurreducens; kanamycin 200 µg/mL and spectinomycin 50 µg/mL . G . sulfurreducens strains and mutants were grown in anoxic medium with excess acetate ( 20 mM ) and limiting fumarate ( 40 mM ) as described ( Chan et al . , 2015 ) . Agar ( 1 . 5% ) was added to the acetate-fumarate medium to culture for clonal isolates on semisolid surface in a H2:CO2:N2 ( 5:20:75 ) atmosphere in an anaerobic workstation ( Don Whitley ) . All growth analyses were initiated by picking a single colony from acetate-fumarate agar using freshly streaked , −80°C culture stocks . When electrodes were used as the electron acceptor , fumarate was replaced with 50 mM NaCl to maintain a similar ionic strength . When insoluble Fe ( III ) oxide or soluble Fe ( III ) citrate was used as the electron acceptor , a non-chelated mineral mix was used ( Chan et al . , 2015 ) . XRD amorphous insoluble Fe ( III ) - ( oxyhydr ) oxide was produced by first synthesizing Schwertmannite ( Fe8O8 ( OH ) 6 ( SO4 ) ·nH2O ) , combining 10 g of Fe ( II ) sulfate in 1 L of water with 5 . 5 mL of 30% H2O2 overnight ( Levar et al . , 2017 ) . The solids were centrifuged at 3 , 700 × g and re-suspended in dH2O three times to obtain Schwertmannite in a pH ~5 solution . This stable product could be added to basal medium and sterilized by autoclaving , resulting in ~30 mM Fe ( III ) ( based on Fe ( III ) extractable by NH3OH ) . Autoclaving in pH 7 basal medium converts Schwertmannite into a high surface area , XRD amorphous insoluble Fe ( III ) - ( oxyhydr ) oxide and is the primary form of insoluble Fe ( III ) oxides presented in this study , which we refer to as simply Fe ( III ) oxides . Media containing akaganeite ( β-FeOOH ) , another form of insoluble Fe ( III ) commonly synthesized by slow NaOH addition to FeCl3 solutions , was also used and showed similar growth trends but at a slower rate . Minimal medium containing 20 mM acetate as the electron donor and Fe ( III ) oxide or Fe ( III ) citrate as the sole electron acceptor was inoculated 1:100 from the acetate-fumarate grown culture . To monitor Fe ( III ) reduction over time , 0 . 1 mL of the Fe ( III ) medium was removed at regular intervals and dissolved in 0 . 9 mL of 0 . 5 N HCl for at least 24 hr in the dark . The acid extractable Fe ( II ) was measured using a modified FerroZine assay ( Chan et al . , 2015 ) . Three-electrode bioreactors with a working volume of 15 ml were assembled as previously described ( Marsili et al . , 2008 ) . The potential of the polished graphite working electrode with a surface area of 3 cm2 was maintained at −0 . 10 V vs . standard hydrogen electrode ( SHE ) using a VMP3 multichannel potentiostat ( Biologic ) , a platinum counter electrode and calomel reference . This potential mimics Fe ( III ) oxides used in parallel experiments . Reactors were inoculated as previously described ( Chan et al . , 2017 ) . Bioreactors were maintained at 30°C under a constant stream of humidified N2:CO2 ( 80:20 ) scrubbed free of oxygen by passage over a heated copper furnace . In prior work , the oxygen concentration in the headspace of these reactors has been shown to be ~1 ppm . The sucrose-SacB counter-selection strategy was used to generate a scarless gacA or esnD deletion strain ( Chan et al . , 2015 ) . Supplementary file 2 lists primers and restriction enzymes used to generate the ~750 bp flanking fragments of the gacA or esnD sequences to ligate into pK18mobsacB . The E . coli S17-1 donor strain mobilized plasmids into G . sulfurreducens . To integrate downstream of the glmS gene using Tn7 ( Damron et al . , 2013 ) , derivatives of E . coli MFDpir carrying pTNS3 ( encoding the Tn7 transposase TnsABCD ) and MFDpir carrying a modified suicide vector pTJ1 with the sequence of interest cloned between the n7L and n7R sites was combined with G . sulfurreducens recipient strains by centrifugation in the anoxic glovebox , then incubation of the cell mixture on top of a filter paper disk ( Millipore GPWP04700 ) placed on 1 . 5% agar with acetate-fumarate plates for 4 hr before plating on spectinomycin selective medium . Amplification and sequencing of the insertion junction revealed that TnsABCD mediated Tn7 integration is site specific in G . sulfurreducens and is 25 bp downstream of the glmS ( GSU0270 ) stop codon . Plasmids expressing gacA and gacA site variants were cloned into pRK2-Geo2 ( Chan et al . , 2015 ) with the native promoter ( replacing the acpP promoter ) or over-expressed from the acpP promoter in pRK2-Geo2 . The native gacA promoter was fused to gacA and gacA variants by extending two oligos coding for the gacA promoter with gacA PCR fragments using overlap PCR . The cGAMP selective riboswitch controlling GSU1761 was fused to the Nanoluc gene with overlap PCR and cloned into pTOPO2 . 1 . The luminescent reporter provided a strong signal even at low levels of expression compared to GFP , and allowed us to circumvent the problem of high autofluorescence in crude protein lysates due to the abundance of cytochromes in G . sulfurreducens . The reporter was made by fusing the natural cGAMP-specific GSU1761 riboswitch upstream of the Nanoluc gene ( pNL1 . 1 , Promega ) . Analysis of the riboswitch expression platform suggests that cGAMP binding stabilizes anti-terminator formation and thus turns on expression of the Nanoluc reporter gene . The cGAMP selective GSU1761 riboswitch was replaced with a mutant Gmet_0970 riboswitch using Gibson assembly to generate a cdiG selective Nanoluc fusion . The cdiG selectivity of this mutant GEMM riboswitch is confirmed using gel-shift analysis ( Figure 2 ) . Tn7 integrative plasmids expressing gacA , gacA mutants , esnD and Nanoluc fusions were sub-cloned with either the native promoter or the acpP promoter into a derivative of pTJ1 ( Damron et al . , 2013 ) from the pRK2-Geo2 backbone by sequential digest with NheI and blunted with Klenow enzyme before digesting with AscI . Genes under the native promoter or acpP promoter were then ligated into the AscI and PmeI site into the pTJ1 derivative Tn7 integration plasmid . Cyclic dinucleotide Nanoluc fusion plasmids were subcloned into pRK2-Geo2 and Tn7 integrative plasmids to report cGAMP or cdiG levels in G . sulfurreducens strains . For the crystallography construct , the T4 lysozyme sequence containing an E11Q inactivating mutation without stop codon was placed upstream of the Hypr GGDEF domain of Gmet_1914 ( residues 294–459 ) sequence . This chimeric protein coding sequence was inserted between the NdeI and XhoI restriction sites of pET24a using restriction digest-ligation techniques . For in vitro analysis of mutants , site-directed-mutagenesis with the around-the-horn mutagenesis technique [https://openwetware . org/wiki/%27Round-the-horn_site-directed_mutagenesis] was used on a previously reported plasmid for expression of MBP-tagged R393A GSU1658 ( Hallberg et al . , 2016 ) to generate GacA mutant constructs . For WspR constructs used in flow cytometry assays , codon-optimized WspR ( Key Resources Table ) was inserted between the NdeI and XhoI restriction sites of pCOLADuet-1 using restriction digest-ligation techniques . This wild-type sequence was used as the template for round-the-horn mutagenesis . All primers and restriction enzyme used are listed in Supplementary file 2 . G . sulfurreducens strains with either the cdiG- or the cGAMP-Nanoluc reporter integrated into the Tn7 site were grown to mid-log fumarate-limited medium ( 40 mM acetate and 80 mM fumarate ) , the same ratio of acetate:fumarate as in RNA-seq conditions , and lysed at room temperature for 5 min in a phosphate-buffered saline ( PBS ) solution containing 1 × BugBuster ( Novagen ) and 0 . 3 mg/ml DNase . 10 μl of the Nanoglo reagent ( Promega ) and 10 μl of the cell lysate were combined in a white-bottom , 96 well plate and luminescence was detected at 461 nm ( Molecular Devices ) . Biological replicates ( n = 3 ) were assayed . In assays where the luminescence exceeded the linear range of the spectrophotometer or deviated from steady-state , lysates were diluted in PBS before combining with the Nanoglo reagent . Total RNA was extracted from 10 mL of G . sulfurreducens electron acceptor limited culture grown to mid-log ( 0 . 25–0 . 3 OD ) . Cell pellets were washed in RNAprotect ( Qiagen ) and frozen at −80°C before RNA extraction using RNeasy with on column DNase treatment ( Qiagen ) . Ribosomal RNA was depleted using RiboZero ( Illumina ) by the University of Minnesota Genomics Center before stranded synthesis and sequenced on Illumina HiSeq 2500 , 125 bp pair-ended mode . Residual ribosomal RNA sequences ( <1% ) were removed before analysis using Rockhopper , an RNAseq analysis program specifically designed to analyze bacterial transcriptomes ( McClure et al . , 2013 ) . Duplicate biological samples were analyzed for each strain . Each replicate had between 13–14 M passing filter reads . Rockhopper aligned the rRNA depleted reads to our laboratory re-sequenced and re-annotated G . sulfurreducens genome , then normalized read counts from each experimental replicate by the upper quartile gene expression before they are compared . Raw reads and re-sequenced genome data have been deposited to the NCBI SRA database PRJNA290373 ( Chan et al . , 2015 ) . Full RNA-seq expression data are in Supplementary file 1 . Full-length proteins with N-terminal His6-MBP tags encoded in pET16-derived plasmids ( cGAS and DncV plasmids are from ( Kranzusch et al . , 2014 ) , WT GacA is from ( Hallberg et al . , 2016 ) , and mutants are from this study ) and the T4 lysozyme-Gmet_1914294-459 GGDEF chimera protein with C-terminal His6 tag encoded in pET24a were overexpressed in E . coli BL21 ( DE3 ) Star cells harboring the pRARE2 plasmid encoding human tRNAs ( Novagen ) . Briefly , an aliquot of the overnight starter culture was re-inoculated into LB with antibiotics ( LB/Carb/Chlor for pET16 , LB/Kan/Chlor for pET24a ) and grown to an OD600 ~0 . 7 , after which cultures were induced with 1 mM IPTG for 10 hr . After centrifugation to isolate the cell pellet , cells were lysed by sonication in lysis buffer ( 25 mM Tris-HCl ( pH 8 . 2 ) , 500 mM NaCl , 20 mM imidazole , and 5 mM beta-mercaptoethanol ) . Lysate was then clarified by centrifugation at 10 , 000 × g for 45 min at 4°C . Clarified lysate was bound to Ni-NTA agarose ( Qiagen ) , and resin was washed with 3 × 20 mL lysis buffer prior to elution with lysis buffer supplemented with 500 mM imidazole . Purified proteins were dialyzed overnight at 4°C into storage buffer ( 20 mM HEPES-KOH ( pH 7 . 5 ) , 250 mM KCl , 1 mM TCEP , and 5% ( v/v ) glycerol ) . Proteins purified in this way were concentrated to ~5–10 mg/mL , flash frozen in liquid nitrogen , and stored at −80°C . Protein purity was assessed by SDS-PAGE . The crystallization fusion construct , T4lysozyme-Gmet_1914294-459-His6 , was further purified by size-exclusion chromatography on a Superdex 200 16/60 column in gel-filtration buffer ( 20 mM HEPES-KOH ( pH 7 . 5 ) , 250 mM KCl , 1 mM TCEP , and 5% ( v/v ) glycerol ) , and eluted protein was concentrated to 10 mg/mL . Purified protein was used immediately for x-ray crystallography or flash frozen in liquid nitrogen and stored at −80°C for biochemical experiments . In vitro activity assays were performed as previously described ( Kranzusch et al . , 2014 ) as independent technical replicates ( n = 3 , assays used the same stock enzyme preparation in separate reaction mixtures ) , with the following modifications . Enzyme ( 10 µM ) was incubated in reaction buffer ( 50 mM Tris-HCl ( pH 7 . 5 ) , 100 mM NaCl , 10 mM MgCl2 , and 5 mM dithiothreitol ) with 100 µM each NTP substrate and/or nonhydrolyzable analog , and ~0 . 1 µCi radiolabeled [α-32P]-ATP or [α-32P]-GTP ( Perkin Elmer ) at 28°C for 1 hr . After the reaction , the mixture was treated with 20 U of Calf Intestinal Alkaline Phosphatase ( NEB ) at 28°C for 30 min to digest unincorporated NTPs , followed by heating to 95°C for 30 s to terminate the reaction . The reaction mixture ( 1 µL ) was spotted onto a PEI-cellulose F thin layer chromatography plate ( Millipore ) , and allowed to dry for 15 min at room-temperature . TLC plates were developed using 1 M KH2PO4 ( pH 3 . 6 ) as the mobile phase . Plates were dried overnight and radiolabeled products were detected using a phosphor-imager screen ( GE Healthcare ) and Typhoon Trio +scanner ( GE Healthcare ) . In vitro activity assays were performed as previously described for diguanylate cyclases ( Burns et al . , 2014 ) as independent technical replicates ( n = 3 , assays used the same stock enzyme preparation in separate reaction mixtures ) , with the following modifications . The EnzChek pyrophosphate kit ( Life Technologies ) was used according to the manufacturer’s instructions except the buffer was supplemented with KCl to a final concentration of 100 mM and MgCl2 to a final concentration of 10 mM , and the reactions were initiated with addition of ATP or GTP . Assays were performed in triplicate in Corning Costar 96 well black , clear-bottomed plates containing 1 µM protein and varying NTP concentrations ( 0–10 mM ) . Absorbance at 360 nm in each well was measured using a SpectraMax i3x plate reader ( Molecular Devices ) and SoftMax Pro 6 . 5 . 1 software . Subsequent analyses to determine enzymatic rates were performed using the Excel Solver package . Activity assays with ATP , GTP , or mixtures of ATP and GTP were performed as described previously ( Burhenne and Kaever , 2013; Hallberg et al . , 2016 ) in independent technical replicates ( n = 3 , assays used the same stock enzyme preparation in separate reaction mixtures ) . For unnatural substrates , the reactions were performed using 5 μM GacA and 200 μM of unnatural NTP at 37°C for 16–20 hr . Prior to LC-MS analysis , samples were treated with a 60°C incubation as in ref . ( Gentner et al . , 2012 ) to ensure all analyzed CDN samples were monomeric . LC-MS analysis of enzyme reactions was performed using an Agilent 1260 Quadrupole LC-MS with an Agilent 1260 Infinity HPLC equipped with a diode array detector . Sample volumes of 10 µL were separated on a Poroshell 120 EC C18 column ( 50 mm length ×4 . 6 mm internal diameter , 2 . 7 µm particle size , Agilent ) at a flow rate of 0 . 4 mL/min . For analysis of enzyme reactions , an elution program consisting of 0% B for 5 min , followed by linear elution of 0% to 10% B over 1 . 5 min , isocratic elution at 10% B for 2 min , linear elution of 10% to 30% B over 2 . 5 min , linear elution from 30% to 0% B over 10 min , and isocratic elution of 0% B for 4 min , 50 s was used . Solvent A was 10 mM ammonium acetate/0 . 1% acetic acid and solvent B was HPLC-grade methanol . Under these conditions , the retention times are 10 . 23 ± 0 . 02 min for cdiG , 10 . 56 ± 0 . 02 min for cGAMP , and 11 . 09 ± 0 . 05 min for cdiA . The assignment of cyclic dinucleotide identity was confirmed through analysis of the mass spectra in the positive ion mode using m/z range = 150 to 1000 . Product ratios were quantified using peak integrations at 254 nm by comparison to standard curves generated for each cyclic dinucleotide at known concentrations . We develop our mathematical derivation from the scheme presented in Figure 5A . This kinetic scheme assumes that the active dimer ( E ) remains at a constant total concentration throughout the reaction compared to inactive monomeric enzyme , which is not included in the model . The model also assumes that the rate-determining step is the production of the linear intermediate , which implies that the intermediate is efficiently converted to the cyclic dinucleotide . While the linear intermediate ( pppGpG ) is sometimes observed for diguanylate cyclases in vitro ( Skotnicka et al . , 2016 ) , this may be due to the characterization of non-activated enzymes in vitro . We do not see significant buildup of any linear intermediates for in vitro or in vivo reactions with GacA , which supports our model . With this assumption , our kinetic model contains 14 equations , describing the change in concentration of each relevant compound in the reaction: ( 1 ) d[E]/dt=kn1G[EnG]−k1G[E][G]+kn1G[EGn]−k1G[E][G]+kn1A[EAn]−k1A[E][A]+kn1A[EnA]−k1A[E][A]+kcat[EGG]+kcat[EGA]+kcat[EAA]+kcat[EAG] ( 2 ) d[EnG]/dt=k1G[E][G]−kn1G[EnG]−k2G[EnG][G]+kn2G[EGG]+knA\|G[EAG]−kA\|G[EnG][A] ( 3 ) d[EGn]/dt=k1G[E][G]−kn1G[EGn]+kn2G[EGG]−k2G[EGn][G]+knA\|G[EGA]−kA\|G[EGn][A] ( 4 ) d[EnA]/dt=k1A[E][A]−kn1A[EnA]+knG\|A[EGA]−kG\|A[EnA][G]+kn2A[EAA]−k2A[EnA][A] ( 5 ) d[EAn]/dt=k1A[E][A]−kn1A[EAn]+kn2A[EAA]−k2A[EAn][A]+knG\|A[EAG]−kG\|A[Ean][G] ( 6 ) d[EGG]/dt=k2G[EnG][G]−kn2G[EGG]+k2G[EGn][G]−kn2G[EGG]−kcat[EGG] ( 7 ) d[EGA]/dt=kA\|G[EGn][A]−knA\|G[EGA]+kG\|A[EnA]][G]−knG\|A[EGA]−kcat[EGA] ( 8 ) d[EAG]/dt=kA\|G[EnG][A]−knA\|G[EAG]+kG\|A[EAn][G]−knG\|A[EAG]−kcat[EAG] ( 9 ) d[EAA]/dt=k2A[EnA][A]−kn2A[EAA]+k2A[EAn][A]−kn2A[EAA]−kcat[EAA] ( 10 ) d[cdiA]/dt=kcat , cdiA[EAA] ( 11 ) d[cdiG]/dt=kcat , cdiG[EGG] ( 12 ) d[cAG]/dt=kcat , cAG[EAG]+kcat , cAG[EGA] ( 13 ) d[A]/dt=−k1A[E][A]+kn1A[EAn]−k1A[E][A]+kn1A[EnA]−kA\|G[EGn][A]+knA\|G[EGA]−k2A[EnA][A]+kn2A[EAA]−k2A[EAn][A]+kn2A[EAA]−kA\|G[EnG][A]+knA\|G[EAG] ( 14 ) d[G]/dt=−k1G[E][G]+kn1G[EnG]−k1G[E][G]+kn1G[EGn]−k2G[EnG][G]+kn2G[EGG]−k2G[EGn][G]+kn2G[EGG]−kG\|A[EnA][G]+knG\|A[EGA]−kG\|A[EAn][G]+knG\|A[EAG] Where each variable is: E , active enzyme A and G , ATP and GTP EnG , enzyme with binding pocket 1 empty and binding pocket 2 with GTP EGn , enzyme with binding pocket 1 with GTP , binding pocket 2 empty EnA , as EnG , except with ATP EAn , as EGn , except with ATP EGG , enzyme with two GTP bound EAA , enzyme with two ATP bound EAG , enzyme with ATP in binding pocket 1 , GTP in binding pocket 2 EGA , enzyme with GTP in binding pocket 1 , ATP in binding pocket 2 And where reverse rate constants of a reaction are denoted by ‘n’ ( i . e . k1G is the forward rate constant for the first GTP binding event , whereas kn1G is the rate constant for GTP dissociating ) . Thus , the equilibrium constant values shown in Figure 3A are related to on/off rates by the following equations: ( 15 ) K1A=kn1A/k1A ( 16 ) K2A=kn2A/k2A ( 17 ) K1G=kn1G/k1G ( 18 ) K2G=kn2G/k2G ( 19 ) KA\|G=knA\|G/kA\|G ( 20 ) KG\|A=knG\|A/kG\|A For the single-substrate case , we utilize the exact solution provided by Oliveira et al . to obtain all dissociation constants for the first and second binding events , as well as the catalytic rate constant . Importantly , this only gives dissociation constants , which we convert to on and off rate constants using an arbitrary assignment of kno . Because KD=koff/kon , and because the kcat values are <<1 sec−1 , we arbitrarily set kon values to 1 µM−1sec−1 . Thus , k1A , k2A , k1G , k2G , kA\|G , andkG\|A are all set to 1 µM−1sec−1 . To calculate endpoint product ratios , we performed numerical integration ( using the Python ODEint solver package in NumPy ) of the system of differential equations using the starting concentrations over an hour-long time course — equivalent in length to our experimental procedures — using 1 s intervals . As stated in the main text , the value of kcat , AG was set conservatively to the same value as kcat , diG ( 0 . 03 sec−1 ) . The values for KA\|G and KG\|A were tested between 1 and 100 µM ( corresponding to varying kno between 1 and 100 µM sec−1 ) using a step size of 1 µM . Thus , for each combination of KA\|G and KG\|A tested ( 10 , 000 possible combinations ) , we performed linear updates over 3 , 600 1 s steps of the 14 analytes , using the generic equation: ( 21 ) [Analyte]t+1=[Analyte]t+t∗d[Analyte]/dt Because we do not include any noise in these equations , the simulation gives the same result each time . The ratio of each cyclic dinucleotide , which was calculated for the 1 hr endpoint , is: ( 22 ) ratioCDN=[CDN]1h/ ( [cdiG]1h+[cAG]1h+[cdiA]1h ) We calculated the model error as the sum of the least squares difference between the experimental product ratios and modeled results ( Equation 22 ) for each starting ATP/GTP ratio . The best fit values for KA\|G and KG\|A were the combination that gave the lowest model error ( Figure 3B ) . Parameter values for the best fit kinetic model are shown below ( also see Table 1 ) : To model cellular homeostasis as shown in Figure 5D and E , the numerical integration program was run with d[A]/dt and d[G]/dt ( Equations 13 and 14 ) set to zero . BL21 ( DE3 ) Star cells containing the pRARE2 plasmid ( Novagen ) and pET24a plasmid encoding synthase enzyme constructs were inoculated into LB/kan/chlor at 37°C with shaking at 250 rpm overnight . An aliquot of the starter culture was re-inoculated into LB/kan/chlor media and grown to an OD600 ~0 . 3 , after which cultures were induced with 1 mM IPTG at 28°C for 4 hr . Cells were harvested by centrifugation at 4 , 700 rpm for 15 min at 4°C , and pellets were stored at −80°C . Cyclic dinucleotides were extracted as described previously ( Hallberg et al . , 2016 ) from two biological replicates . LC-MS analysis of E . coli cell extracts was performed as described previously ( Hallberg et al . , 2016 ) . Prior to crystallization , T4Lysozyme-Gmet_1914294–459 protein was incubated for 10 min at rt at a concentration of 6 . 5 mg ml−1 in the presence of 10 mM GTP and 10 mM MgCl2 . The T4Lysozyme-Gmet_1914294–459–Guanosine nucleotide complex was crystallized in a hanging-drop vapor diffusion format using the final optimized crystallization conditions of 30 mM HEPES-KOH ( pH 7 . 5 ) , 300 mM Na ( OAc ) , and 26% PEG-4000 . Crystals were grown in Easy-Xtal 15-well trays ( Qiagen ) in 2 μl hanging drops with a 1:1 ( protein:reservoir ) ratio over 350 μl of reservoir solution . Crystals required incubation at 18°C for 2–4 days for complete growth , and then were transferred with a nylon loop to a new drop containing reservoir solution supplemented with 10% glycerol as a cryoprotectant and incubated for 30 s before flash-freezing in liquid nitrogen . Native and anomalous data were collected under cryogenic conditions at the Lawrence Berkeley National Laboratory Advanced Light Source ( Beamline 8 . 3 . 1 ) . X-ray diffraction data were processed with XDS and AIMLESS ( Kabsch , 2010 ) in the monoclinic spacegroup C 2 . Phase information was determined with a combination of molecular replacement and sulfur single-wavelength anomalous dispersion ( SAD ) . Briefly , iterative sulfur-SAD data sets were collected at ~7 , 235 eV and merged from independent portions of a large T4Lysozyme-Gmet_1914294–459 crystal as previously described ( Lee et al . , 2016a ) . A minimal core of T4-Lysozyme ( PDB 5JWS ) ( Lee et al . , 2016b ) was used a search model for molecular replacement and sub-structure determination . The placed T4-Lysozyme fragment was then used to guide SAD identification of 17 sites with HySS in PHENIX ( Adams et al . , 2010 ) corresponding to 12 sulfur atoms in T4Lysozyme-Gmet_1914294–459 and 5 solvent ion positions . SOLVE/RESOLVE ( Terwilliger , 1999 ) was used to extend phases to the native T4Lysozyme-Gmet_1914294–459 data processed to ~1 . 35 Å and model building and refinement were completed with Coot ( Emsley and Cowtan , 2004 ) and PHENIX ( Table 2 ) . A Python-based program was developed to extract alignment data for a library of 139 , 801 putative GGDEF domain-containing proteins from the Uniprot database ( obtained through Pfam , accession PF00990 , http://pfam . xfam . org/ , accessed 06/05/2014 ) . In particular , positions critical for catalytic activity ( i . e . the GG[D/E]EF sequence ) and selectivity ( i . e . positions 347 and 303 in GsGacA ) were identified and analyzed for each sequence . Given previous results with some DGCs possessing altered signature motifs , we assigned any diguanylate cyclase with a [G/A/S]G[D/E][F/Y] motif to be active .	Microscopic organisms known as bacteria are found in virtually every environment on the planet . One reason bacteria are so successful is that they are able to form communities known as biofilms on surfaces in animals and other living things , as well as on rocks and other features in the environment . These biofilms protect the bacteria from fluctuations in the environment and toxins . For over 30 years , a class of enzymes called the GGDEF enzymes were thought to make a single signal known as cyclic di-GMP that regulates the formation of biofilms . However , in 2016 , a team of researchers reported that some GGDEF enzymes , including one from a bacterium called Geobacter sulfurreducens , were also able to produce two other signals known as cGAMP and cyclic di-AMP . The experiments involved making the enzymes and testing their activity outside the cell . Therefore , it remained unclear whether these enzymes ( dubbed ‘Hypr’ GGDEF enzymes ) actually produce all three signals inside cells and play a role in forming bacterial biofilms . G . sulfurreducens is unusual because it is able to grow on metallic minerals or electrodes to generate electrical energy . As part of a community of microorganisms , they help break down pollutants in contaminated areas and can generate electricity from wastewater . Now , Hallberg , Chan et al . – including many of the researchers involved in the 2016 work – combined several experimental and mathematical approaches to study the Hypr GGDEF enzymes in G . sulfurreducens . The experiments show that the Hypr GGDEF enzymes produced cGAMP , but not the other two signals , inside the cells . This cGAMP regulated the ability of G . sulfurreducens to grow by extracting electrical energy from the metallic minerals , which appears to be a new , biofilm-less lifestyle . Further experiments revealed how Hypr GGDEF enzymes have evolved to preferentially make cGAMP over the other two signals . Together , these findings demonstrate that enzymes with the ability to make several different signals , are capable of generating specific responses in bacterial cells . By understanding how bacteria make decisions , it may be possible to change their behaviors . The findings of Hallberg , Chan et al . help to identify the signaling pathways involved in this decision-making and provide new tools to study them in the future .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "microbiology", "and", "infectious", "disease" ]	2019	Structure and mechanism of a Hypr GGDEF enzyme that activates cGAMP signaling to control extracellular metal respiration
Morphogens function in concentration-dependent manners to instruct cell fate during tissue patterning . The cytoneme morphogen transport model posits that specialized filopodia extend between morphogen-sending and responding cells to ensure that appropriate signaling thresholds are achieved . How morphogens are transported along and deployed from cytonemes , how quickly a cytoneme-delivered , receptor-dependent signal is initiated , and whether these processes are conserved across phyla are not known . Herein , we reveal that the actin motor Myosin 10 promotes vesicular transport of Sonic Hedgehog ( SHH ) morphogen in mouse cell cytonemes , and that SHH morphogen gradient organization is altered in neural tubes of Myo10-/- mice . We demonstrate that cytoneme-mediated deposition of SHH onto receiving cells induces a rapid , receptor-dependent signal response that occurs within seconds of ligand delivery . This activity is dependent upon a novel Dispatched ( DISP ) -BOC/CDON co-receptor complex that functions in ligand-producing cells to promote cytoneme occurrence and facilitate ligand delivery for signal activation . The Hedgehog ( HH ) pathway is an evolutionarily conserved signaling relay that contributes to embryonic development through influencing cell fate , cell proliferation , cell and tissue polarity , stem cell maintenance , and tissue homeostasis ( reviewed in Briscoe and Thérond , 2013 ) . During development , HH family morphogens , which include HH in flies and Sonic ( SHH ) , Desert ( DHH ) and Indian ( IHH ) Hedgehogs in vertebrates , function in concentration-dependent manners to instruct tissue morphogenesis through Patched ( PTCH ) receptors and Cell Adhesion Molecule-Related/Down-Regulated by Oncogenes ( CDON ) , Brother of CDON ( BOC ) or Growth Arrest-Specific 1 ( GAS1 ) co-receptors ( Allen et al . , 2007; Marigo et al . , 1996; Okada et al . , 2006; Yao et al . , 2006 ) . Binding of SHH to PTCH receptor complexes allows for activation of the G-protein-coupled receptor Smoothened ( SMO ) , which can signal through both G-protein-dependent and independent effector routes to induce downstream responses ( reviewed in Arensdorf et al . , 2016 ) . G-protein-dependent noncanonical SMO signals induce transcription-independent responses including cytoskeletal rearrangement for cell migration , lipid metabolic responses , and Ca2+ release ( Arensdorf et al . , 2017; Belgacem and Borodinsky , 2011; Ho Wei et al . , 2018 ) . The canonical SMO effector route is thought to be largely G-protein-independent , and requires SMO translocation into the primary cilium for activation of GLI transcriptional effectors to induce target gene expression ( Arensdorf et al . , 2016; Corbit et al . , 2005 ) . To induce appropriate responses in target cells , HH ligands must deploy from their site of synthesis and transport to short- and long-range target cells . HH family ligands are unique in that they are lipid-modified by a covalently linked cholesterol moiety on their carboxyl termini , and by a long-chain fatty acid such as palmitate on their amino termini ( Pepinsky et al . , 1998; Porter et al . , 1996 ) . These modifications result in high affinity of the ligand for producing-cell membranes , necessitating specific deployment and transport mechanisms to assure the formation and fitness of HH morphogen gradients during tissue patterning ( reviewed in Hall et al . , 2019 ) . In vertebrates , proteins dedicated to deployment of SHH include the twelve-transmembrane spanning protein Dispatched ( DISP ) and the secreted glycoprotein SCUBE2 ( Burke et al . , 1999; Creanga et al . , 2012; Tukachinsky et al . , 2012 ) . Mechanisms by which DISP and SCUBE2 collaborate to ensure SHH deployment and influence its establishment of its morphogen gradient are not yet clear , but several models have been proposed . The cytoneme model posits that long , specialized filopodia extend from producing and receiving cells to facilitate transport and exchange of HHs across developing tissues ( Kornberg and Roy , 2014; Ramírez-Weber and Kornberg , 1999 ) . Cytonemes are thin filopodia , typically smaller than 200 nm in diameter , that can reach up to ~300 μm from the originating cell body ( Kornberg , 2014 ) . Cytonemes functioning during morphogenesis were first recognized in Drosophila wing imaginal discs ( Ramírez-Weber and Kornberg , 1999 ) . Subsequent studies found them to contain signaling molecules including HH , WNT , TGFβ , Notch ligands , and growth factors ( reviewed in Fairchild and Barna , 2014; Kornberg , 2014 ) . Interrogations of HH-containing cytonemes in fly and chick systems revealed localization of DISP , BOC ( BOI in Drosophila ) , CDON ( iHOG , Drosophila ) , and PTCH to the cellular structures along with ligand ( Bischoff et al . , 2013; Bodeen et al . , 2017; Chen et al . , 2017; Gradilla et al . , 2014; Sanders et al . , 2013 ) . The cytoneme-specific functionalities of these proteins , and how they facilitate HH ligand movement , have not yet been determined . Although specialized filopodia containing SHH have been observed in vertebrate systems ( Sanders et al . , 2013 ) , little is known about their cell biology in mammals . The full cast of proteins dedicated to initiation , growth , and maintenance of cytonemes are not yet known , nor are the mechanisms by which morphogens are loaded into and transported along the cellular structures . Herein we interrogate cytoneme-based transport of SHH in mouse cells . By using newly developed imaging protocols and developing a rapid read-out for SHH pathway induction , we reveal that Myosin 10 ( MYO10 ) is required for movement of SHH to cytoneme tips to initiate a signal response in target cells . We find that disruption of MYO10 in vivo leads to neural tube patterning defects consistent with attenuated SHH morphogen gradient function , confirming that cell-based interrogation of cytonemes can predict in vivo relevance . Our studies also reveal that a novel complex between SHH , DISP , and BOC/CDON contributes to SHH cytoneme occurrence and ligand delivery , and that BOC/CDON activity in ligand-producing cells is required for cytoneme-mediated induction of an SHH signal response in target cells . To interrogate cytoneme-based SHH transport in mammalian systems , NIH3T3 cells and mouse embryonic fibroblasts ( MEFs ) expressing SHH plus the membrane marker mCherry-Mem ( mCherry fused to the first 20 residues of neuromodulin ) were fixed using the MEM-fix technique , and then examined by confocal microscopy ( Figure 1A–B’’; Bodeen et al . , 2017; Hall and Ogden , 2018 ) . Image analysis of the mCherry-Mem signal in SHH-expressing cells revealed long extensions from both NIH3T3 cells and MEFs that reached around and over neighboring cells ( Figure 1A–B’’ ) . Depth analysis of the mCherry-Mem signal revealed that cytoneme-like projections originated from portions of the cell membranes that were not in contact with the growth surface ( Figure 1A’–A’’’ , B’’ ) . The small diameter , long lengths , and growth patterns of these extensions are consistent with the documented characteristics of cytonemes , indicating NIH3T3 cells and MEFs can be used to interrogate the specialized filopodia ( Bodeen et al . , 2017; Hall and Ogden , 2018; Kornberg and Roy , 2014; Ramírez-Weber and Kornberg , 1999 ) . The atypical actin-based motor protein MYO10 is thought to promote filopodial outgrowth through facilitating anterograde transport of protein cargo that supports growth and maintenance of the cellular outgrowths ( Bohil et al . , 2006; Hirano et al . , 2011; Tokuo and Ikebe , 2004; Wei et al . , 2011; Zhang et al . , 2004 ) . We tested for localization of MYO10 to SHH-containing projections by expressing GFP tagged MYO10 in NIH3T3 cells . Indeed , MYO10-GFP was enriched in tips of projections from SHH-expressing cells , and showed co-localization with SHH puncta in a subset of the extensions ( Figure 1C , C’ ) . Live imaging revealed that the MYO10-tip-enriched extensions were highly dynamic and capable of forming stable connections with MYO10-positive extensions from neighboring cells ( Videos 1–3 ) . Shorter extensions that maintained contact with the growth substrate were largely immobile , suggesting they were likely adhesion or retraction fibers , rather than dynamic cytonemes ( Videos 1 and 2; Kornberg and Roy , 2014; Zhang et al . , 2004 ) . Cytoneme-like structures were detected in the absence of SHH expression in both MEFs and NIH3T3 cells , indicating that the morphogen is not required for their initiation . However , SHH expression raised the cytoneme occurrence rate in both cell types from ~30% to ~60% ( Figure 1D ) , and more than doubled the average number of cytonemes per NIH3T3 cell ( Figure 1E ) . To test whether increased cytoneme occurrence rates correlated with SHH expression level , cytonemes were quantified in HEK cells stably transfected with a ponasterone A-inducible SHH expression vector ( Goetz et al . , 2006 ) . Likely due to a low level of ‘leaky’ SHH expression occurring in the absence of induction , baseline cytoneme occurrence in inducible cells was increased compared to untransfected controls ( Figure 1—figure supplement 1A–F ) . Addition of increasing concentrations of ponasterone A led to a dose-dependent increase in cytoneme occurrence that correlated with increased SHH protein production ( Figure 1—figure supplement 1B–F ) . These results suggest SHH-producing cells may tune cytoneme occurrence rates proportional to morphogen expression levels . To test whether the SHH signal transducing protein SMO contributed to increased cytoneme occurrence rates observed in cultured cells expressing ligand , NIH3T3 cells were treated with the direct SMO agonist SAG and antagonist vismodegib ( Figure 1—figure supplement 1G ) . SAG failed to significantly alter cytoneme occurrence in the absence or presence of SHH , indicating that direct induction of SMO is not sufficient to induce a cytoneme response . Vismodegib treatment raised baseline cytoneme occurrence and blunted the ability of SHH to increase occurrence rate over the elevated baseline . To clarify whether these changes resulted from loss of SMO induction , cytonemes were quantified in CRISPR generated Smo-/- NIH3T3 cells . Smo-/- cells showed an ~15% increase in occurrence rates in response to SHH expression , indicating that SMO is not required for a cytoneme response . However , the increase in occurrence rate was reduced compared to what was observed in control cells ( ~28% increase ) , suggesting active SMO may enhance the ability of SHH to increase cytonemes in cultured cells . The ability of SHH to induce cytonemes independent of its canonical signal transducing protein SMO suggests the morphogen may act in a cell autonomous manner to control its own deployment . To determine whether other morphogens or developmental signaling molecules documented to localize to cytonemes might also be able to influence cytonemes , we quantified occurrence rates in NIH3T3 cells expressing the Notch ligand Jagged , BMP2 , FGF2 , or Wnt3A ( Figure 1F ) . Like SHH , Jagged and FGF2 triggered an approximate doubling of cytoneme occurrence . Wnt3A also increased occurrence rates , albeit to a lesser extent . BMP2 expression did not alter baseline cytoneme occurrence , indicating that some , but not all developmentally relevant signaling proteins can modulate activity of the specialized filopodia in NIH3T3 cells . To determine whether SHH-containing cytonemes transmitted an activation signal to receiving cells , we developed a contact-mediated activation assay in which SHH pathway induction in receiving cells could be rapidly detected . The current temporal indicator of pathway induction tracks accumulation of the SHH signal transducing G-protein-coupled receptor Smoothened ( SMO ) into primary cilia ( Corbit et al . , 2005 ) . However , both active and inactive SMO proteins cycle through the primary cilium , making this assay sub-optimal for tracking SMO activation in real time ( Milenkovic et al . , 2015; Rohatgi et al . , 2009 ) . Active SMO signals through Gαi heterotrimeric G proteins to raise intracellular Ca2+ , which we reasoned would be a rapid and activation-specific read-out for pathway induction resulting from cytoneme-based ligand delivery ( Adachi et al . , 2019; Belgacem and Borodinsky , 2011; Klatt Shaw et al . , 2018 ) . A previous study examining filopodial-based transport of SHH used a truncated SHH-N construct , which is amenable to palmitoylation , but lacks the carboxyl-terminal cholesterol modification ( Sanders et al . , 2013 ) . Because cholesterol contributes to both SHH morphogen gradient formation and PTCH binding on receiving cells ( Gong et al . , 2018; Li et al . , 2006 ) , we wanted to ensure that we analyzed trafficking of the physiologically relevant dually lipidated molecule . We generated SHH-GFP/mCherry such that both lipid modifications are added to a signaling molecule with an internal GFP or mCherry tag ( Figure 2—figure supplement 1A–B’ ) . Transcriptional reporter assays confirmed functionality of the internally-tagged fluorescent SHH proteins ( Figure 2—figure supplement 1C , C’ ) . Once SHH expression constructs were validated , we monitored Ca2+ flux by co-culturing R-GECO-expressing Ca2+ reporter cells with NIH3T3 cells expressing palmitoylated and cholesterol-modified SHH-GFP or GFP control . R-GECO cells in contact with cytonemes , but not cell body , from control or SHH GFP-positive cells were monitored for Ca2+ reporter flux ( Figure 2A–B’ and Videos 4–5 ) . To increase the likelihood that signals would result from deposition of SHH via cytonemes , and not from SHH secreted into the media , media wash-out was performed at 15 min intervals for the duration of data acquisition . We counted any R-GECO Ca2+ flux with a minimum peak fluorescence of 50 to be a positive event ( Figure 2B–C’ ) . This flux value was determined by examining the distribution of fluorescence intensity of R-GECO cells in contact with GFP control cells ( n = 27 cells ) . A value of 50 was determined to be within the 90th and 95th percentile of flux value distributions in control cells ( Figure 2B ) , indicating any flux over 50 would likely be significantly above the control intensity range . Consistent with a SHH-induced signal , Ca2+ flux rates of R-GECO reporter cells in contact with SHH containing cytonemes were more than twice the rate of flux observed in reporter cells contacted by cytonemes from GFP-expressing control cells ( Figure 2B , B’ , D ) . Furthermore , R-GECO cells in continuous contact with SHH-containing cytonemes had a significantly higher total time spent in flux than cells in contact with control cell cytonemes ( Figure 2E ) . R-GECO reporter cells typically produced transient Ca2+ pulses within ~10–20 s of SHH-GFP release from cytonemes docked to their cell membranes ( Figure 2B’ , red dashes ) . To determine whether there was a statistically significant correlation between cytoneme-mediated ligand delivery and Ca2+ response , we documented all flux events with a mean intensity of over 50 that occurred within 20 s of SHH deposition ( n = 15 cells ) . A Wilcoxon signed rank test performed against these results confirmed that reporter cells spent a significantly greater amount of time in positive flux within a 20 s window following an SHH deposit than they did outside this response window ( p=6 . 1e-05 , Figure 2F ) . As such , a significant correlation between SHH delivery and Ca2+ release was confirmed . Importantly , SHH-stimulated Ca2+ flux was blocked in Smo-/- R-GECO cells ( Figure 2D columns 5 and 6 ) or by treatment with the inverse SMO agonist cyclopamine ( Figure 2C–D columns 7 and 8 ) , confirming specificity of the Ca2+ response to SHH pathway activation . To confirm that signal induction resulted from SHH delivered through cytonemes , and not from SHH ligand that may have been secreted into the culture medium from ligand-expressing cells , GFP and SHH-GFP expressing cells were co-cultured with R-GECO reporter cells . Flux was monitored in reporter cells in contact with cytonemes from SHH-GFP or GFP-expressing control cells ( Figure 2D ) . Reporter cells in contact with SHH-GFP-expressing cytonemes maintained higher flux rates than reporter cells in contact with GFP-expressing control cells , and exhibited a flux rate similar to what was observed in SHH-GFP/R-GECO culture conditions ( Figure 2D column 2 vs . column 4 ) . Reporter cells in contact with GFP-expressing cells in the mixed culture exhibited a flux rate similar to reporter cells cultured with GFP-expressing cells in the absence of SHH-GFP co-culture ( column 1 vs . column 3 ) . We conclude secreted SHH does not contribute to the observed Ca2+ response in cells in this assay . Thus , cytonemes can deliver SHH to induce a bona fide SMO activation signal . Our observation that cytonemes appeared to transport distinct puncta of SHH toward target cells ( Figure 2A ) prompted us to investigate the mode by which the morphogen reached the cytoneme tip . HHs are covalently modified by cholesterol at their carboxyl-termini , raising the possibility that they could travel along cytoneme membranes inside cholesterol-rich lipid rafts ( Callejo et al . , 2011; Creanga et al . , 2012; Porter et al . , 1996; Rietveld et al . , 1999 ) . To test for SHH localization to rafts along NIH3T3 cytoneme membranes , we assayed for ligand colocalization with a fluorescently-labeled cholera toxin ( CTX ) raft marker . Although rafts were evident along the length of cytonemes , they rarely contained SHH , as evidenced by a negative correlation coefficient between CTX and SHH signals ( Figure 3A–B ) . Thus , SHH is unlikely to transport along cytoneme membranes in raft-like domains . We next considered that SHH might load into vesicular structures for transport inside cytonemes because studies in both Drosophila and mouse suggest that HH ligands are released from producing cells in exosomes ( Coulter et al . , 2018; Gradilla et al . , 2014 ) . To investigate this , we tested whether SHH localized inside cytonemes , or to the outside leaflet of cytoneme membranes . Cells expressing SHH-GFP were subjected to extracellular immuno-staining with anti-SHH antibody prior to fixation . GFP fluorescence was used to track total SHH and antibody signal ( ex-SHH ) was used to monitor the ligand pool exposed to the extracellular environment . SHH-GFP and ex-SHH signals were both evident in puncta on the plasma membrane of SHH-expressing cells ( Figure 3C , arrows; Figure 3—figure supplement 1A–D for antibody controls ) . Notably , surface aggregates of ex-SHH evident along membranes of the cell body were rarely seen along cytonemes . Conversely , SHH-GFP signal was consistently detected in cytonemes , suggesting SHH is positioned inside cytoneme membranes ( Figure 3C’ ) . Consistent with this hypothesis , SHH colocalized with the tetraspanin exosomal markers CD9-mCherry and CD81-mCherry along cytonemes ( Figure 3D , E ) , suggesting ligand likely traffics inside cytonemes through a vesicular transport mechanism . Although HH family ligands have been reported to enrich in RAB18 and CD63 containing exosomes in mouse neuronal cells and Drosophila tissues , respectively , SHH failed to colocalize with these markers in cytonemes of NIH3T3 cells ( Figure 3—figure supplement 1E–G; Coulter et al . , 2018; Gradilla et al . , 2014 ) . Thus , cell-type-specific vesicular loading may occur . To evaluate whether SHH trafficked through cytonemes in vesicles , SHH-mCherry-expressing NIH3T3 cells were examined by immuno-electron microscopy using anti-mCherry antibody ( Figure 3F and Figure 3—figure supplement 2 for control experiments ) . Transmission electron microscopy was performed on 70 nm sections of cells and their cytonemes . Cytonemes of SHH-producing cells contained multiple vesicles ( Figure 3F , arrowheads ) , many of which were positive for SHH-mCherry ( Figure 3F , F’ arrows , Figure 3—figure supplement 2A ) . Consistent with what was observed with anti-exSHH ( Figure 3C ) , clusters observed along the plasma membrane of the cell body were not evident along cytoneme membrane in TEM sections ( Figure 3—figure supplement 2B ) . We hypothesized that if SHH undergoes vesicular trafficking inside cytonemes , a molecular motor would likely contribute to its movement . Because MYO10 colocalized with SHH at cytoneme tips ( Figure 1C’ ) , we tested whether inhibition of MYO10 would impact ligand movement along the specialized filopodia . MYO10-dependent effects on SHH cytoneme mobility were assayed by monitoring fluorescence recovery after photobleaching ( FRAP ) of the two proteins . Cytonemes of NIH3T3 cells expressing cytoplasmic GFP , mCherry-Mem , MYO10-GFP , or SHH-mCherry were photobleached , and recovery of each protein to cytoneme tips was calculated in the absence or presence of ionomycin , which is proposed to attenuate MYO10 motor activity through raising intracellular Ca2+ ( Homma et al . , 2001; Morgan and Jacob , 1994 ) . SHH-mCherry and MYO10-GFP cytoneme signals recovered at similar rates in vehicle-treated cells ( ~0 . 25 ± 0 . 12 and 0 . 27 ± 0 . 15 µm/s , respectively ) , but failed to recover following ionomycin treatment . Conversely , membrane diffusion rates , which were calculated by monitoring mCherry-Mem recovery to cytoneme tips , were not reduced by ionomycin treatment ( Figure 4A–B , Figure 4—figure supplement 1 , Videos 6 and 7 ) . Recovery of cytoplasmic GFP signal to cytoneme tips was so rapid we were unable to calculate accurate recovery rates in either condition . These results suggest SHH is unlikely to travel along cytonemes through cytoplasmic or membrane diffusion-based mechanisms , and is instead actively transported along the specialized filopodia , potentially by MYO10 . To test whether SHH enrichment in cytonemes required MYO10 function , we generated MYO10-null MEFs from Myo10m1J/m1J mutant mice , and assessed SHH cytoneme dynamics in this genetic background ( Heimsath et al . , 2017 ) . SHH was expressed in MYO10 mutant MEFs , and cytoneme to cell body SHH signal intensity ratios were determined ( Figure 4C ) . MYO10 mutant MEFs exhibited low SHH cytoneme to cell body signal intensity ratios , indicating inefficient cytoneme enrichment of the morphogen in the absence of MYO10 . Enrichment of SHH in cytonemes was restored by co-expression of wild type or pleckstrin homology domain-deficient ( ΔPH ) MYO10 , but not by a MYO10 mutant lacking its cargo binding domains ( MYO10-HMM ) ( Berg and Cheney , 2002 ) . Comparable effects on SHH cytoneme enrichment were seen in wild type NIH3T3 cells over-expressing these MYO10-GFP variants ( Figure 4—figure supplement 2A–F ) . Although wild type and MYO10ΔPH did not alter the ratio of SHH or SHH-mCherry in cytonemes , MYO10-HMM over-expression reduced SHH cytoneme localization and attenuated SHH-mCherry FRAP to cytoneme tips ( Figure 4D , Figure 4—figure supplement 2A–F ) . All MYO10 variants showed similar cytoneme FRAP rates , suggesting that failure of SHH to be transported by MYO10-HMM was likely due to compromised cargo binding , and not due to alteration of MYO10-HMM motor activity ( Figure 4—figure supplement 2G ) . MYO10 contributes to the formation and maintenance of filopodia , and SHH increases cytoneme occurrence rates in HEK , NIH3T3 , and wild-type MEFs ( Figure 1D , E , Figure 1—figure supplement 1F; Bohil et al . , 2006 ) . Hence , we next tested for a role for MYO10 in SHH-stimulated cytoneme occurrence in wild-type and Myo10 mutant cells . In NIH3T3 cells , expression of wild-type MYO10 modestly enhanced the ability of SHH to increase cytoneme occurrence rates over baseline . MYO10-HMM over-expression suppressed occurrence rates ( Figure 4E ) , which we speculate resulted from the mutant protein oligomerizing with endogenous MYO10 to disrupt its ability to bind and transport cargo ( Figure 4E and Figure 4—figure supplement 2G ) . In Myo10-/- MEFs , SHH failed to stimulate cytoneme occurrence . Cytoneme occurrence rate increases in the presence of SHH were rescued by reintroduction of either wild type MYO10 or MYO10-ΔPH , but not by MYO10-HMM ( Figure 4E’ ) , further supporting an essential role for the cargo domain for SHH cytoneme biology . We next tested the ability of GFP-MYO10-HMM expressing cells to deliver an SHH activation signal . Consistent with our hypothesis , co-expression of GFP-MYO10-HMM with SHH in ligand-producing NIH3T3 cells reduced SHH-induced Ca2+ flux in co-cultured R-GECO reporter cells to significantly lower rates than those observed upon co-culture with ligand-producing cells co-expressing GFP or wild type MYO10-GFP with SHH ( Figure 4F ) . Combined , these results suggest MYO10 is required for the cytoneme-promoting effects of SHH , and also for cargo domain-mediated transport of SHH to cytoneme tips for delivery to target cells . Myo10-null mice ( Myo10m1J/m1J ) are semi-lethal with ~60% of homozygous mutants exhibiting exencephaly and embryonic or perinatal lethality . Surviving animals display white belly spots , with a subset of these animals also exhibiting syndactyly ( Bachg et al . , 2019; Heimsath et al . , 2017 ) . Exencephaly and syndactyly can be attributed to reduced Bone Morphogenic Protein ( BMP ) signaling and de-repression , rather than disruption , of SHH signaling ( Nikolopoulou et al . , 2017; Patterson et al . , 2009 ) . These phenotypes are seemingly inconsistent with our in vitro observations that MYO10 promotes cytoneme stability and transport of SHH ( Figure 4A–F ) . Therefore , in an effort to understand the effects of MYO10 loss on SHH signaling in vivo , we analyzed developing neural tubes from Myo10m1J/m1J E9 . 5 embryos ( Heimsath et al . , 2017 ) . SHH expressed in the notochord signals to the adjacent floor plate of the developing neural tube to induce SHH , which then signals in a ventral to dorsal trajectory to specify neural progenitor domains . Control of these domains is exquisitely sensitive to alteration of SHH signaling , so monitoring their induction allows for robust analysis of gradient function ( Kutejova et al . , 2016; Placzek and Briscoe , 2018 ) . In order to track the zone of ventral SHH activity in MYO10 mutant animals , we introduced an Shh::GFP allele in which GFP is inserted into the endogenous Shh locus such that the mature ligand retains an internal GFP tag adjacent to the carboxyl-terminal cholesterol modification ( Chamberlain et al . , 2008 ) . Examination of neural tubes from ShhGFP/+ and ShhGFP/+; Myo10m1J/m1J neural tubes revealed altered floor plate induction , as evidenced by reduced SHH::GFP signal in the ventral-most neural tube ( Figure 5A–D ) . Consistent with attenuated SHH signaling activity in neural tubes of MYO10-null animals , the expression domain of SHH-induced Gli1 was compressed in Myo10m1J/m1J embryos with exencephaly ( Figure 5E–G ) . This correlated with attenuated SHH-controlled progenitor domain induction , as indicated by reduced Olig2 expression ( Figure 5K–M ) . Moreover , we noted a consistent delay in initiation of notochord regression in MYO10-null animals , further supporting attenuated GLI activity ( Figure 5H–J; Park et al . , 2000 ) . These results support a role for MYO10 in SHH morphogen gradient function in vivo . Having identified MYO10 as a new functional player in cytoneme occurrence and SHH transport , we next wanted to determine whether known SHH-binding partners would also impact cytoneme occurrence in mammalian cells . Experiments in Drosophila and chick model systems suggest a role for the SHH deployment protein DISP and co-receptors BOC/BOI and CDON/iHOG in cytoneme function ( Bodeen et al . , 2017; Callejo et al . , 2011; González-Méndez et al . , 2017; Gradilla et al . , 2014; Sanders et al . , 2013 ) . In flies , DISP promotes cytoneme stability of ligand producing cells , and in chick , BOC stabilizes cytonemes of SHH receiving cells ( Bodeen et al . , 2017; Sanders et al . , 2013 ) . Because iHOG has been reported to stabilize cytonemes and localize to exovesicles from ligand-producing cells in flies ( González-Méndez et al . , 2017; Gradilla et al . , 2014 ) , we hypothesized that DISP and CDON or BOC might function together to influence cytoneme occurrence and/or function in mouse cells . DISP-HA and GFP-tagged BOC or CDON were co-expressed in NIH3T3 cells in the absence and presence of SHH , and colocalization between the three proteins was assessed . Confocal microscopy revealed that all three proteins localized to cytonemes , but that DISP did not significantly colocalize with either BOC or CDON along cytoneme membranes in the absence of SHH ( Figure 6A , D , F , G ) . Co-expression of SHH increased colocalization between BOC and DISP throughout cytoneme membrane and in SHH-positive puncta ( Figure 6A–B’ , F ) . Notably , puncta containing all three proteins were evident in cells abutting SHH-containing cytonemes ( Figure 6B and zoom in B’ , arrowheads ) . As such , DISP and BOC may be released to target cells along with ligand , as has been reported for iHOG in Drosophila ( Gradilla et al . , 2014 ) . Consistent with ligand-containing endosomes being internalized by receiving cells , immunoelectron microscopy revealed early and late endosomal structures containing SHH near cytoneme contact points on the signal-receiving cell ( Figure 6C ) . To better understand the associations between cytoneme-localized DISP , BOC , and SHH , stimulated emission depletion ( STED ) microscopy was used to examine individual SHH puncta within cytonemes . STED showed DISP and SHH consistently positioned adjacent to BOC , suggesting that a trimeric complex may occur in cytonemes in the presence of ligand ( Figure 6H , I ) . To test for an interaction between the three proteins , co-immunoprecipitation experiments were performed using lysates from NIH3T3 cells expressing DISP-FLAG and BOC-EGFP in the absence and presence of SHH . BOC-EGFP was captured on anti-FLAG beads in the absence of SHH , suggesting that DISP and BOC can associate ( Figure 6J , and Figure 6—figure supplement 1A for the uncropped blot ) . Upon ligand expression , SHH incorporated into DISP-FLAG/BOC-EGFP immunocomplexes without significantly altering BOC binding ( Figure 6J ) . Thus , SHH is not required for biochemical association between DISP and BOC , but may promote enrichment of the trimeric complex in cytonemes . BOC and CDON are semi-redundant for co-receptor function in PTCH-SHH binding , but do show differential expression and functionality in temporal and tissue-specific contexts ( Allen et al . , 2011; Bergeron et al . , 2011; Cardozo et al . , 2014; Okada et al . , 2006; Tenzen et al . , 2006 ) . Likely consistent with context-dependent functionality , co-localization dynamics between DISP and CDON differed from what was observed for DISP and BOC . CDON colocalized with DISP at SHH-positive puncta in cytonemes , albeit to a slightly lesser extent than did BOC ( Figure 6D–G ) . However , unlike what was observed for BOC , ligand expression did not increase colocalization between DISP and CDON along the length of cytonemes ( Figure 6F , G ) . Biochemical interrogation of DISP-CDON binding by immunoprecipitation analysis revealed that whereas CDON-GFP was captured on FLAG beads by DISP-FLAG in the absence of ligand , its association with DISP-FLAG was reduced upon SHH-DISP binding ( Figure 6K , Figure 6—figure supplement 1B for the uncropped blot ) . Thus , distinct functional pools of CDON with differential affinity toward DISP ± SHH may exist . Drosophila iHOG/CDON and chick BOC proteins have been reported to localize to cytonemes and influence their behavior in vivo ( Callejo et al . , 2011; González-Méndez et al . , 2017; Sanders et al . , 2013 ) . To determine the effects of BOC and CDON on cytonemes of SHH-expressing murine cells , the co-receptors were expressed with SHH in MEFs , and cytoneme occurrence rates were determined . The vertebrate-specific SHH co-receptor GAS1 , which has not yet been investigated for a role in cytonemes , was also tested ( Allen et al . , 2007 ) . GAS1-expressing cells showed baseline and SHH-induced cytoneme occurrence rates similar to GFP-expressing control cells , indicating that GAS1 over-expression does not actively promote or stabilize cytonemes in the absence or presence of ligand ( Figure 6L ) . Conversely , both BOC and CDON expression elevated baseline cytoneme occurrence rates near to SHH-stimulated levels , and modestly enhanced the ability of SHH to increase occurrence ( Figure 6L ) . To determine whether BOC , CDON , GAS1 or a combination of the co-receptors was required for SHH-induced cytoneme occurrence rate increases , SHH was expressed in Boc-/- , Cdon-/- , Gas1-/- triple KO MEFs ( BCG KO ) ( Allen et al . , 2011 ) , and occurrence rates were quantified in control and co-receptor re-expressed conditions ( Figure 6L , right panel ) . Unlike Ptch-/- or Disp1-/- cells , which showed cytoneme occurrence increases in response to SHH expression , BCG KO cells failed to increase cytoneme occurrence in response to SHH expression ( Figure 6L–M ) . BCG KO cells were also compromised in their ability to induce a Ca2+ response in co-cultured R-GECO reporter cells in the absence of direct cell body contact , indicating that at least one of the co-receptors is required to facilitate SHH-induced cytoneme occurrence and delivery ( Figure 6L–N ) . Consistent with previous reports that PTCH can be activated by SHH that is tethered to neighboring cell membranes ( Caspary et al . , 2002; Tokhunts et al . , 2010 ) , BCG KO cells were able to induce a Ca2+ response in R-GECO reporter cells upon direct cell-cell body contact ( Figure 6N ) . GAS1 re-expression did not rescue the ability of SHH to promote cytoneme occurrence , further supporting that GAS1 is not a modulator of SHH cytoneme function . Re-expression of BOC rescued SHH-mediated cytoneme occurrence increases in the triple KO cells , and restored the ability of cells to deliver an activation signal from BCG KO cells stably expressing ligand to co-cultured Ca2+ reporter cells ( Figure 6L , N ) . CDON re-expression also restored cytoneme occurrence and receiving-cell signal induction , but not as effectively as BOC . Thus , we conclude that either BOC or CDON are required for SHH-induced cytoneme biogenesis or stability , and that BOC is likely the predominant co-receptor functioning in the specialized filopodia . The formation of a morphogen gradient of sufficient robustness to confer tissue patterning is a complex process that likely involves integration of multiple molecular mechanisms promoting morphogen release and transport . SHH morphogen is unique in that it harbors two essential lipid modifications , including an amino-terminal palmitate and a carboxyl-terminal cholesterol , that must be overcome to facilitate release from producing cell membranes for transport through the extracellular milieu ( Pepinsky et al . , 1998; Porter et al . , 1996 ) . Reported dissemination mechanisms proposed to neutralize the lipid modifications include sheddase-directed release , in which the lipids are cleaved , free diffusion of multimeric SHH in which the multimer configuration buries the lipids , and assisted diffusion , in which extracellular chaperones counteract hydrophobic behavior of the lipids ( Creanga et al . , 2012; Ohlig et al . , 2012; Tukachinsky et al . , 2012; Zeng et al . , 2001 ) . Proposed deployment mechanisms in which lipid-modified ligand is packaged or transported include exosome-mediated deployment , lipoprotein particle association , and transport through specialized filopodia called cytonemes ( Eugster et al . , 2007; Gradilla et al . , 2014; Ramírez-Weber and Kornberg , 1999 ) . Despite growing bodies of work supporting each of these transport mechanisms , the molecular machinery driving them , how they coordinate activity , and the cell and tissue contexts in which they occur remain unclear . Herein , we investigated cytoneme-based transport of SHH . We focused on molecular components facilitating movement and delivery of SHH ligands through the specialized filopodia for activation of signaling in target cells . High-resolution live imaging microscopy combined with our improved ability to fix cytonemes of cultured cells , allowed us to interrogate the cell biology of SHH cytonemes , identify new components of the cellular machinery contributing to cytoneme-based SHH delivery , and validate contribution of one of the components in vivo . By developing and validating an assay in which ligand-activated SMO-induced Ca2+ release can be monitored in real time , we were able to demonstrate direct , contact-dependent induction of a SHH response within seconds of cytoneme-mediated ligand delivery . Using this assay as a probe , we identified roles for the actin motor MYO10 and the adhesion co-receptors BOC and CDON during cytoneme-based SHH transport and delivery . Consistent with its reported role in filopodial outgrowth ( Bohil et al . , 2006 ) , Myo10-/- MEFs showed an approximate 10% reduction in basal cytoneme occurrence rates compared to control MEFs , and failed to increase occurrence rates following SHH expression . Thus , MYO10 is likely important for cytoneme biogenesis . Our results suggest that the molecular motor also plays a specific role in cytoneme-based transport of SHH that is facilitated through the MYO10 cargo binding domain . This hypothesis is supported by the observations that SHH and MYO10 traffic along cytonemes at similar velocities , and that cells expressing a MYO10 cargo binding mutant fail to enrich SHH in cytonemes , or to induce a pronounced SHH signal response in co-cultured R-GECO reporter cells . In vivo studies of MYO10 mutant mice reveal ventral neural tube patterning defects in animals with exencephaly that are consistent with alteration of the SHH morphogen gradient . These include a reduced zone of SHH activity at the floorplate , compressed Gli1 and Olig2 expression domains , and altered notochord regression . In addition we failed to observe SHH expression in the floor plate of a subset of embryos examined . However , floor plate specification was not consistently lost , indicating SHH signaling from the notochord to the floor plate can occur in the absence of MYO10 function . Homozygous mutants that did not show exencephaly survived to adulthood without developing SHH phenotypes . The failure of MYO10 mutants to consistently exhibit embryonic or adult SHH loss-of-function phenotypes is not unprecedented given the variable phenotypic penetrance observed in Myo10 knockout animals ( Heimsath et al . , 2017 ) . We speculate that functional compensation by cytoneme-independent mechanisms of SHH distribution likely occur to limit the impact of MYO10 loss on SHH notochord to floor plate signaling and morphogen gradient establishment ( reviewed in Hall et al . , 2019 ) . Further , functional redundancy with other actin-based motors may provide some compensation for MYO10 loss . One candidate is MYO5a , which can also localize to filopodia , bind cargo , and traffic toward filopodial tips ( Kerber and Cheney , 2011 ) . Nevertheless , the observation that the majority of severely affected Myo10 null embryos show exencephaly may suggest attenuated cytoneme function in these animals . This is because exencephaly can result from attenuated signaling by BMP , the Drosophila ortholog of which has been demonstrated to transport along cytonemes ( Nikolopoulou et al . , 2017; Patterson et al . , 2009; Roy et al . , 2014 ) . The mechanism by which MYO10 transports SHH to filopodial tips has yet to be determined . Given that SHH localized to vesicles inside cytonemes , we anticipate that MYO10-dependent movement of ligand occurs through motor-driven vesicular transport along actin filaments . The ΔPH mutant , which lacks phospholipid-binding capability , was not compromised for SHH transport or cytoneme occurrence . Thus , we hypothesize that MYO10 might connect to SHH-containing vesicles through its cargo domain via an adaptor protein localized to vesicular membranes such as a tetraspanin ( Andreu and Yáñez-Mó , 2014 ) . Consistent with this possibility , immuno-localization studies revealed co-localization of SHH with CD9 and CD81 tetraspanins in discrete puncta along cytonemes . In addition to identifying MYO10 as a functional partner in cytoneme-based SHH transport , our studies also revealed a novel role for co-receptors BOC and CDON in signal producing cell cytonemes . Although the co-receptors have previously been reported to localize to cytonemes ( Gradilla et al . , 2014; Sanders et al . , 2013 ) , our study is the first to reveal that they can associate with the SHH deployment protein DISP , and act in producing cells to promote cytoneme formation and stability for SHH delivery . Despite functional redundancy between the two proteins ( Allen et al . , 2011; Zheng et al . , 2010 ) , we found that their associations with DISP-SHH complexes differed . Whereas both BOC and CDON associated with DISP in the absence of SHH , the amount of CDON in association with DISP was reduced upon SHH expression . DISP , CDON and SHH were observed to co-localize to puncta along cytonemes , suggesting the minor fraction still in association with DISP in the presence of ligand may be specific to cytoneme function . However , BOC may be the preferred cytoneme-localized co-receptor because it showed greater colocalization with DISP and facilitated a stronger SHH signal response in signal-receiving cells . Notably , either BOC or CDON was required for SHH to promote cytoneme occurrence because cells lacking all co-receptors failed to initiate cytonemes , or to induce long range signals without re-expression of one of the adhesion co-receptors . Importantly , BOC/CDON/GAS1 mutant cells are not compromised for contact-mediated SHH release because the SHH-expressing BCG KO cells can induce a response when directly abutting a signal-receiving cell . As such , we conclude BOC and CDON contribute to SHH transport from producing cells through promoting cytoneme occurrence and/or stability in the presence of ligand . Determining the mechanism ( s ) by which SHH-containing co-receptor complexes promote cytoneme induction or stability is beyond the scope of the current study . However , the reported ability of BOC to activate the cytoskeletal regulator JNK during neuronal differentiation ( Vuong et al . , 2017 ) , suggests that BOC or CDON might connect SHH with actin remodelers for cytoneme movement . Thus , our results suggest the exciting possibility that SHH may ‘reverse-signal’ in an autocrine fashion to promote its own transport . We anticipate that this reverse signal is SMO/GLI independent , and instead , occurs through BOC or CDON co-receptor in complex with SHH . Our observation that SHH expression can increase cytoneme occurrence rates in both Ptch-/- and Disp1-/- fibroblasts indicates that trimeric co-receptor complexes are not required for cytoneme initiation or stabilization in either signal sending or receiving cells . Thus , PTCH or DISP association with cytoneme-localized co-receptors likely confers specificity for ligand release or canonical signal induction without directly influencing cytoneme behavior . In Drosophila , DISP is required for HH to promote cytoneme stability , suggesting functionality of cytoneme-localized DISP may differ between fly and vertebrate systems ( Bodeen et al . , 2017 ) . We do not know how DISP function evolved between the systems , but speculate that the additional SHH-binding proteins such as GAS1 and SCUBE2 , which are present in vertebrates and lacking in flies , may account for the discrepancy ( Allen et al . , 2007; Creanga et al . , 2012 ) . Future studies will be required to determine how SHH is loaded into cytoneme vesicles along with deployment complex components , and how ligand is transferred from signal producing cytonemes to target cells . The observed accumulation of exSHH puncta along the cell body combined with the paucity of exSHH signal along cytonemes may hint at how ligand enters the specialized filopodia . Studies in Drosophila indicate that HH ligands are initially directed to apical membranes where they are re-internalized by DISP prior to release for long range signaling ( Callejo et al . , 2011; D'Angelo et al . , 2015; reviewed in Hall et al . , 2019 ) . Thus , the exSHH signal could represent prepackaged protein that is poised for vesicular loading with BOC/CDON by DISP . It is possible that SHH may fail to load into or be released from specialized filopodia without deployment complex activity because cytoneme membranes may have a composition unique from that of bulk plasma membrane . Filopodia are documented to enrich for select phosphatidylinositol species ( Jacquemet et al . , 2019 ) , which may alter the overall membrane composition to prevent diffusion of SHH onto cytoneme membranes without assistance . Due to improved methods for preserving and imaging cytonemes of cultured cells , and for monitoring responses in signal-receiving cells , we are now poised to address these provocative questions . The proven utility of cultured cells for analyzing cytoneme biology reveals that in vitro systems can function as tractable models for interrogating morphogen transport . Furthermore , cultured cells may also allow for investigation of how cytonemes synergize with other morphogen dispersion processes to ensure gradient robustness during tissue development . Cell fixation and staining were performed using MEM-fixation protocols ( Hall and Ogden , 2018 ) . The following antibodies and dilutions were used: rabbit anti-SHH ( H-160 ) ( 1:100; Santa Cruz ) , mouse anti-GFP ( 4B10 ) ( 1:500; CST ) , rat anti-HA ( 1:250; Roche ) , mouse anti-CD63 ( E-12 ) ( 1:100; Santa Cruz ) , rabbit anti-Myc-Tag ( 2272 ) ( 1:400; CST ) . Secondary antibodies ( Jackson ImmunoResearch and Invitrogen ) were used at a 1:1000 dilution . For additional information please refer to Appendix 1—key resources table . Lipid raft staining was performed using Cholera Toxin Subunit B ( Recombinant ) ( CTX ) , Alexa Fluor 488 Conjugate ( Invitrogen ) . CTX was dissolved in chilled PBS to a final concentration of 1 . 0 mg/mL . CTX was incubated with cells for 20 min at 4°C to prevent endocytosis . Cells were rinsed three times in chilled PBS prior to MEM-fixation . Extracellular staining was performed by diluting antibodies in 4°C PBS supplemented with 5% normal goat serum . Antibody solutions were then incubated for 30 min on live cells on ice to prevent endocytosis . Cells were rinsed three times in chilled PBS prior to MEM-fixation . Microscopy images were taken with a TCS SP8 STED 3X confocal microscope ( Leica ) for fixed and live cell imaging . FRAP assays were carried out on a Bruker Opterra swept field confocal microscope , equipped with an enclosure box at 5% CO2 and imaging and objective heater at 37°C . NIH3T3 cells were imaged in phenol red-free standard growth media . For assays involving ionomycin , standard growth media was replaced immediately prior to imaging with phenol red- and serum-free media supplemented with 0 . 083% DMSO ( control ) , or DMSO with 2 . 5 µM ionomycin ( #9995 , CST ) . Image acquisition was performed with 60x/1 . 4NA/Oil objective lens ( CFI Plan Apo Lambda ) with a 30 µm pinhole array and 70 µm width slit . Fluorescence was recorded with a 5 s baseline followed by a complete photobleaching of a cytoneme with 488 nm and 561 nm lasers . Fluorescence recovery was recorded for 120 s with 100 ms exposure per channel with frames taken every 277 ms . Fluorescence recovery of individual regions of interest ( ROI ) along the cytoneme to its tip were normalized with pre-FRAP equal to 100% and post-FRAP equal to 0% . FRAP curves were corrected for any loss of fluorescence during acquisition ( Fritzsche and Charras , 2015 ) . Cells were cultured at 37°C in 5% CO2 . NIH3T3 ( CRL-1658 ) , HEK293T ( CRL-11268 ) , and LightII ( JHU-68 ) cells were obtained from ATCC . HEK ( Bosc 23 ) ponasterone A inducible SHH cells were obtained from D . Robbins ( Goetz et al . , 2006 ) . Boc/Cdon/Gas1-/- MEFs were obtained from Allen et al . , 2011 . Myo10-/- MEFs were generated from mice obtained from and cryo-recovered by The Jackson Laboratory ( stock number 024583 , B6 . Cg-Myo10m1J/GrsrJ ) . Smo-/- Flp-In-3T3 cells were generated using CRISPR-Cas9 technology . Briefly , 400 , 000 Flp-In-3T3 cells were transiently co-transfected with precomplexed ribonuclear proteins ( RNPs ) consisting of 100 pmol of chemically modified sgRNA ( mSmo . sg–NA - 5’- CAGCUACAUCGCAGCCUUCG -3’ , Synthego ) , 33 pmol of spCas9 protein ( St . Jude Protein Production Core ) , and 200 ng of pMaxGFP ( Lonza ) . The transfection was performed via nucleofection ( Lonza , 4D-Nucleofector X-unit ) using solution SG and program EN158 in a small ( 20 µl ) cuvette according to the manufacturer’s recommended protocol . Five days post-nucleofection , cells were single cell sorted for GFP+ ( transfected ) cells by FACs and clonally expanded . Clones were screened and verified for the desired out-of-frame indel modifications via targeted deep sequencing using gene specific primers with partial Illumina adapter overhangs ( mSmo . F – 5’- ttccttccccgtgtcagaacgaggt -3’ and mSmo . R – 5’- gcggccatgcagtgaagtgagggtc -3’ , overhangs not shown ) as previously described ( Sentmanat et al . , 2018 ) . In brief , clonal cell pellets were lysed and used to generate gene specific amplicons with partial Illumina adapters in PCR#1 . Amplicons were indexed in PCR#2 and pooled with other targeted amplicons for other loci to create sequence diversity . Additionally , 10% PhiX Sequencing Control V3 ( Illumina ) was added to the pooled amplicon library prior to running the sample on an MiSeq Sequencer System ( Illumina ) to generate paired 2 × 250 bp reads . Samples were demultiplexed using the index sequences , fastq files were generated , and NGS analysis was performed using CRIS . py ( Connelly and Pruett-Miller , 2019 ) . MEFs were generated as previously described ( Jozefczuk et al . , 2012 ) . Briefly , pregnant dams were harvested at E12 . 5–13 . 5 and embryos were dissected in 1X PBS , then decapitated and internal organs removed . The remaining tissue was rinsed in 1X PBS , then finely minced into pieces in a dish containing Trypsin-EDTA ( 0 . 25% ) ( Gibco ) . The dish was placed at 37°C in an incubator for 15 min , then an additional 2 mL of Trypsin-EDTA was added , tissue was vigorously pipetted , then placed back in the incubator at 37°C for an additional 10 min . The solution was transferred to a 15 mL conical tube and contents were allowed to settle for 2 min . Supernatant was removed , and then centrifuged for 5 min at 200 x g . The cell pellet was resuspended in MEF media ( see below ) and plated in a 60 mm plate and left overnight . Each line was then SV40-transformed and single cell selection was performed by serial dilution in a 96-well plate . MEF lines were derived from five different Myo10 mutant embryos and five wild type littermates . Cells were maintained in DMEM ( Life Technologies ) supplemented with 10% bovine calf serum ( Fisher Scientific ) and 1% Penicillin Streptomycin solution ( Gibco ) . HEK293T cells utilized 10% heat-inactivated fetal bovine serum ( Corning ) . Cell lines were routinely validated by functional assay and western blot as appropriate and tested monthly for mycoplasma contamination by MycoAlert ( Lonza ) . Transfection of plasmid DNA was performed with Lipofectamine 3000 and P3000 reagent ( Thermo Fisher Scientific ) , according to manufacturer’s instructions . When required , the final amount of DNA used for transfection was kept constant by the addition of control vector DNA . All cells were harvested 36 hr after transient DNA plasmid transfection for subsequent assays . SHH inducible cells were incubated in HEK media with ponasterone A at the indicated concentrations for 16 hr prior to analysis . Incubation of SMO modulators was performed with 100 nm SAG or 200 nm vismodegib ( LC laboratories ) for 16 hr prior to analysis . Wild-type , Shh::gfp ( JAX # 008466 ) , and Myo10m1J/m1J ( JAX # 024583 ) embryos in the C57BL/6 background were harvested and processed for immunohistochemistry at E9 . 5 . Pregnant dams were harvested , uterine horns removed , and embryos were dissected in 1X PBS , then rinsed three times . Embryos were fixed overnight at 4°C in 2% PFA . The following day , embryos were rinsed three times in 1X PBS and moved to 30% sucrose to cryo-protect . The following day , embryos were frozen in O . C . T . Compound ( Tissue-Tek ) on dry ice . Embryos were sectioned transverse at 10 µm thickness on a Leica Microm CM1950 cryo-stat . Sections were briefly dried , then washed in 1X TBST , then blocked with 2% BSA , 1% goat serum , 0 . 1% Triton-X-100 in 1X PBS . Antibodies were diluted in blocking buffer and incubated overnight on sections at room temperature . The following antibodies and dilutions were used: chicken anti-GFP ( 1:500; Aves ) , mouse anti-PAX6 ( 1:25; DSHB ) , and rabbit anti-OLIG2 ( 1:300; Millipore ) . Primary antibody was removed , sections were washed with 1X TBST three times , then incubated for 3 hr in secondary antibodies ( Invitrogen ) used at a 1:500 dilution . Sections were washed three times in 1X TBST , then rinsed with tap water , and cover slips were applied with ProLong Diamond mounting media . Sections were imaged on a Leica DMi8 widefield microscope and processed using LAS X . SHH::GFP neural tube domains were calculated as the mean area per section of the neural tube taken at the cardiac level ( n = 3–4 sections per mouse ) . A minimum of five embryos per genotype were analyzed . For in situ hybridization E9 . 5 embryos were harvested and fixed for 4 hr at 4°C in 4% PFA . In situ hybridization experiments were performed as described previously ( Abler et al . , 2011 ) with the following modifications . Frozen embryos were sectioned transverse at 15 µm thickness and mounted on charged glass slides . The sense and antisense digoxigenin-labeled RNA probes were made using a DIG RNA labeling kit and following the manufacturer’s instructions ( Roche ) , sense probes were used as negative controls and no positive signal was observed . Three embryos per genotype were analyzed . All Myo10m1J/m1J embryos analyzed displayed overt exencephaly . Approximately 0 . 4 × 106 NIH3T3 cells were seeded into individual wells of 6-well plates one day prior to transfection . A total of 2 µg plasmid DNA was transfected into individual wells . pCMV-R-GECO1 for ‘receiving’ sensor cells , and the appropriate construct combination ( e . g . SHH-mCherry , GFP , MYO10-GFP , etc… ) for the ‘producing’ cells . Six hours after transfection , cells were trypsinized and replated into eight well , polystyrene chambers on a 1 . 5 borosilicate coverglass , 0 . 7 cm2/well ( Nunc Lab-Tek II ) . Prior to cell addition , chamber wells were divided in half with #0 ( 0 . 08–0 . 13 mm ) thick coverslips ( Electron Microscopy Sciences ) cut to size , with vacuum grease ( Dow Corning ) added along the lateral edges to retain a liquid-tight barrier . Receiving and producing cells were seeded onto opposite sides of the barrier and allowed to recover overnight . The following day the barrier was removed 3 hr prior to imaging allowing sufficient time for cells to migrate and cytonemes to extend between producing and receiving cells . In GFP and SHH-GFP coculture experiments , cells were seeded into chambered wells at ~40% cell density ( 20% R-GECO , 10% GFP and 10% SHH-GFP ) and allowed to recover for a minimum of 5 hr before imaging . Media was removed prior to imaging , and cells were gently washed in PBS . Immediately prior to any imaging event , media was replaced to remove secreted SHH . In experiments where SMO activation was inhibited , 10 µM cyclopamine ( LC Laboratories ) was added to R-GECO cells 16 hr prior to imaging . Live imaging was performed at 37°C , 5% CO2 with resonant scanning for 15 min per area over the entire cytoneme/s depth ( ~4–6 µm ) , with Z-steps of ~0 . 6–1 . 0 µm . Maximum intensity projections were generated for subsequent analysis of the time-lapses . Time-lapses of cells were analyzed using LAS X ( Leica ) . SHH-GFP deposits onto R-GECO sensor cells were recognized if a SHH puncta was detected traversing a cytoneme from a producing cell body to accumulate at the tip , where in a successive frame fluorescence was absent and did not undergo retrograde movement . Puncta were identified by a fluorescent intensity signal >50% than background cytoneme fluorescence by single line scan along a cytoneme . R-GECO fluorescent intensity histograms of individual cells were normalized for each cell with minimum fluorescence equal to 0 and maximum to 100 . Ca2+ flux occurrence was quantified as a relative peak in R-GECO fluorescence within a ~20 s window with a minimum peak fluorescence of 50 . Maintained fluorescence over 20 s was considered a single flux . Total flux time was determined using a threshold to define an increased flux ( i . e . a peak ) using the R-GECO cells in contact with GFP-control samples . The threshold was determined to be a flux value of greater than 50 which lies between the overall 90th and 95th percentiles of the control samples . Next , the proportion of time ( seconds ) when flux values were greater than 50 was calculated for each control and SHH case sample . The proportion of time when flux values were greater than 50 was compared between cases and controls using the Wilcoxon rank sum test . For SHH cases only , the proportion of time when flux values were greater than 50 was compared by the occurrence of a SHH deposit ( yes vs . no ) using the Wilcoxon signed rank test . Twenty seconds was considered a biologically relevant time frame in which a deposit and a subsequent increased flux should occur ( Adachi et al . , 2019; Tewson et al . , 2012 ) . Statistical analyses were conducted using SAS software version 9 . 4 ( SAS Institute , Cary , NC ) and R version 3 . 6 . 0 ( R Foundation for Statistical Computing , Vienna , Austria ) . A two-sided significance level of p<0 . 05 was considered statistically significant . R-GECO cells in contact with cytonemes that did not exhibit a single flux during the time-lapse were excluded from analysis . Image analysis was performed using CellProfiler ( McQuin et al . , 2018 ) . An image analysis pipeline was constructed to mask and isolate cytonemes . Output images were then run through a secondary pipeline measuring Pearson’s correlation coefficient between pixels of different fluorophores to calculate the relative colocalization of the proteins of interest within a cytoneme . A third pipeline was used for determining protein colocalization of two proteins at SHH puncta in cytonemes . This pipeline segmented SHH pixels as a reference point within cytonemes to measure Pearson’s correlation coefficient between pixels of the other fluorophores of interest in contact with SHH . A minimum of 30 cytonemes were analyzed per condition . All analyses were performed using GraphPad Prism . One-way analysis of variance was performed for multiple comparisons , with Tukey’s multiple comparison as a posttest . Significant differences between two conditions were determined by two-tailed Student’s t tests . All quantified data are presented as mean ± SD , with p<0 . 05 considered statistically significant . Significance depicted as p<0 . 05 , p<0 . 01 , p<0 . 001 , **p<0 . 0001 , ns = not significant . The following plasmids were used in this study: pCDNA ( control vector ) ( Clontech ) , pCDNA3-EGFP ( Addgene Plasmid #13031 ) , pCMV-mCherry-Mem ( Addgene Plasmid #55779 ) , pcDNA-Wnt3A ( Addgene Plasmid #35908 ) , pIRES-Jag1-HA ( Addgene Plasmid #17336 ) , pCMV-R-GECO1 ( Addgene Plasmid #32444 ) , pCMV-mCherry-CD9 ( Addgene Plasmid #55013 ) , pCMV-mCherry-CD81 ( Addgene Plasmid #55012 ) , pEGFP-CD63-C2 ( Addgene Plasmid #62964 ) , pCMV-EGFP-Rab18 ( Addgene Plasmid #49550 ) , pEGFP-C1-hMyoX ( Addgene Plasmid #47608 ) , pCMV6-hGAS1-Myc-DDK ( Origene Cat: RC224804 ) , pCMV3-mFGF2-N-GFPSpark ( SinoBiological Cat: MG50037-ANG ) , pCMV3-mBMP2-C-GFPSpark ( SinoBiological Cat: MG51115-ACG ) , pCMV-mCherry2 , pCDNA3-mSHH-FL , pCDNA3-mSHH-N , pCDNA3-mSHH-FL-EGFP , pCDNA3-mSHH-FL-mCherry2 , pCDNA3-V5-Disp-HA , pCS2-hBOC-EGFP and pCS2-hCDON-EGFP ( a gift from A . Salic ) , pEGFP-C2-bMyo10-HMM ( Berg and Cheney , 2002 ) and pEGFP-C2-bMyo10-Δ3PH , a modified version of pEGFP-bMyo10 where the 3 PH domains were removed via deletion of aa 1168–1491 . For the generation of fluorescently tagged SHH , GFP or mCherry2 was introduced into the SHH protein immediately 3’ to amino acid Gly198 with the addition of 10 amino acids ( Alanine188 - AENSVAAKSG - Glycine197 ) downstream of GFP or mCherry2 including the intein cleavage-cholesterol attachment site , similar to what was previously descried ( Chamberlain et al . , 2008 ) . Briefly , a Bgl2 site was introduced into SHH-FL after Gly198 by Quikchange ( Agilent ) using primers ( forward 5’ GTGGCGGCCAAATCCGGCGGCAGATCTGGCTGTTTCCCGGGATCCGCC and reverse 5’ ggcggatcccgggaaacagccagatctgccgccggatttggccgccac ) . The 10 amino acid duplication 3’ to GFP or mCherry2 on SHH protein was introduced using Infusion ( Clontech ) using primers ( forward 5’TCCGGCGGCAGATCTGCAGAGAACTCCGTGGCGGCCAAATCCGGCGGCTGTTTCCCGGGA and reverse 5’ tcccgggaaacagccgccggatttggccgccacggagttctctgcagatctgccgccgga ) . GFP or mCherry2 with Bgl2 sites was generated by Phusion PCR ( NEB ) with the following primers ( forward 5’ GAATTCAGATCTATGGTGAGCAAGGGCGAG and reverse 5’ gaattcagatctcttgtacagctcgtccatg ) or ( forward 5’ GAATTCAGATCTATGGTGAGCAAGGGCGAGGAG and reverse 5’ gaattcagatctcttgtacagctcgtccatgccg ) using pEGFP ( Clontech ) or pCMV-mCherry2 ( Clontech ) as the DNA template , respectively . For western blotting , cells were washed twice in PBS , harvested in 1% NP-40 Lysis Buffer ( 50 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl , 1% NP-40 , 0 . 1% SDS , 1X Protease Inhibitor Cocktail and 0 . 5 mM DTT ) and incubated for 30 min at 4°C . Extracts were cleared by centrifugation at 14 , 000 x g at 4°C for 45 min and analyzed . The supernatant was removed , and protein concentrations were determined by bicinchoninic acid ( BCA ) assay ( Pierce ) . Equal amounts of total protein from each sample were analyzed by SDS-PAGE on Criterion gels ( Bio-Rad ) . SDS-PAGE samples were transferred onto Protran Nitrocellulose ( GE ) or Immobilon-P PVDF ( Millipore ) using Tris/Glycine/SDS Buffer ( Bio-Rad ) at 100V for one hour at 22°C . Membranes were blocked with 5% milk and 0 . 1% Tween-20 in Tris-buffered saline ( TBS ) for 1 hr at room temperature . Membranes were immunoblotted for 1 hr at 22°C using the following antibodies: rat anti-HA ( 1:3000; Roche ) , mouse anti-V5 ( 1:5000; Life Technologies ) , rabbit anti-SHH ( H-160 ) ( 1:1000; Santa Cruz ) , rabbit anti-GFP ( 1:8000 , Rockland ) , rabbit anti-Kinesin ( anti-Kif5B , 1:5000; Abcam ) , mouse anti-α-Tubulin ( DM1A ) ( 1:5000 , CST ) , followed by three 5 min washes in secondary milk ( primary milk diluted to 25% with TBS ) . Corresponding HRP-conjugated secondary antibodies ( Jackson Immuno ) were incubated for 1 hr at RT at a 1:5000 concentration . Blots were developed using an Odyssey Fc ( Li-Cor ) with ECL Prime ( GE ) . For immunoprecipitation assays , proteins of interest were expressed in NIH3T3 cells . Cell lysates were prepared ~48 hr post-transfection using a 0 . 5% NP-40 Lysis Buffer ( 30 mM Tris-HCl , pH 7 . 4 , 75 mM NaCl , 0 . 5% NP-40 , 5% glycerol , 2 mM MgCl2 , 4 mM KCl , 1 mM EDTA , and 1X Protease Inhibitor Cocktail ) and incubated for 30 min at 4°C with two units per mL of Benzonase Nuclease to degrade DNA from protein samples . Extracts were cleared by centrifugation at 14 , 000 x g at 4°C for 30 min , supernatant was collected , and protein concentration was determined by BCA assay ( Pierce ) . Equal total protein amounts for each sample were used in co-immunoprecipitation assays and analyzed by SDS-PAGE on Criterion gels ( Bio-Rad ) . Co-immunoprecipitation assays were performed as described ( Stewart et al . , 2018 ) with the following modifications . Samples were pre-cleared with A/G Plus Agarose for 30 min with gentle rotation . Samples were centrifuged at 1000 x g for 1 min and set up in new tubes with either anti-Mouse IgG1 control or EZview Red Anti-Flag Affinity Gel ( Sigma ) ( to immunoprecipitate Flag epitope-tagged proteins ) for three hours at 4°C with gentle rotation . Samples were then centrifuged at 1000 x g for 1 min and supernatant was removed . Beads were washed 3x for 5 min each with 0 . 5% NP-40 Lysis Buffer with gentle rotation at room temperature . Proteins were eluted from agarose beads with 1X SDS sample buffer ( 2% SDS , 4% v/v Glycerol , 40 mM Tris-HCl , pH 6 . 8 , 0 . 1% Bromophenol blue ) by incubating them at room temperature for 5 min . Samples were centrifuged at 2000 x g for 2 min and the eluent was transferred to a new tube . Immunoprecipitates were analyzed by western blot using the following antibodies: rabbit anti-GFP ( 1:8000; Rockland ) , rabbit anti-Flag ( DDDDK ) ( 1:3000 , Abcam ) , rabbit anti-Shh ( 1:2000; SCBT ) , and rabbit anti-Kif5B , ( 1:5000; Abcam ) . For co-culture Gli-reporter assays , HEK293T cells were seeded at a density of 1 × 106 cells per 60 mm plate . The following day , pCDNA3-GFP ( 2 µg ) , pCDNA3-SHH-FL-GFP-10aa linker ( 4 µg ) , pCMV-mCherry2 ( 2 µg ) and pCDNA3-SHH-FL-mCherry2-10aa linker ( 4 µg ) were transfected into HEK293T cells . In a six-well plate Light II reporter cells were seeded at a density of 0 . 5 × 106 cells per well in DMEM-10% FBS complete growth media and grown overnight at 37°C , 5% CO2 . The following day , transfected HEK293T cells underwent trypsinization and were seeded into the Light II wells at a density of 0 . 5 × 106 cells per well . Cells were allowed to recover for 4 hr at 37°C , 5% CO2 . Media was removed from cells , washed twice with PBS and once with DMEM serum-free complete media ( phenol red free ) . DMEM serum-free media was added back to each well and allowed to incubate for 2 hr . Washing was carried out over 6 hr repeating the above wash steps . After 6 hr , 3 mL of DMEM Serum-free Complete Media was added to each well and the cells were incubated for ~36 hr . Reporter assays were carried out according to Dual Luciferase Reporter Assay Kit instructions ( Promega ) . Experiments were repeated three times in triplicate . NIH3T3 cells expressing SHH-mCherry were seeded at 60% confluency into eight well , Permanox slide , 0 . 8 cm2/well ( Nunc Lab-Tek II ) and , for pre-embedding immunolabeling , were fixed in a 0 . 5% glutaraldehyde with 4% PFA fixative in 0 . 1 M phosphate buffer . Prior to labeling with primary antibody , samples were washed with buffer and excess aldehyde groups neutralized with glycine . Samples were blocked with 1% BSA in 10 mM PBS ( BSA/PBS ) then a blocking solution matched to the species of the secondary antibody ( Aurion , Wageningen , The Netherlands ) in PBS . Samples were incubated with chicken anti-mCherry ( 1:1000 , Abcam ) diluted in BSA/PBS overnight at 4°C . Following primary antibody incubation , samples were washed in BSA/PBS then incubated with a streptavidin conjugated secondary antibody . Samples were rinsed with PBS and incubated with biotinylated nanogold ( Nanoprobes ) then washed with PBS and fixed in 1% glutaraldehyde ( Electron Microscopy Sciences ( EMS ) ) . Following fixation , samples were successively rinsed in distilled water and 0 . 2 M citrate buffer then incubated with HQ Silver Enhancement reagent ( Nanoprobes ) prepared per manufacturer instructions . Enhancement reaction was halted by rinsing with distilled water . Samples were contrasted successively with 1% osmium tetroxide ( EMS ) and 1% uranyl acetate ( EMS ) in water with water washes between contrasting steps . Samples were then dehydrated in an ascending series of alcohols , infiltrated with EmBed 812 ( EMS ) and polymerized at 80°C overnight . Samples were sectioned on a Leica ultramicrotome ( Wetzlar , Austria ) at 70 nm and examined in a Tecnai G² F20-TWIN transmission electron microscope . Images were recorded using an AMT side mount camera system . Unless specified , all chemical and reagents were from Sigma .	During development , cells must work together and talk to each other to build the organs and tissues of the growing embryo . To communicate precisely with long-distance targets , cells can project a series of thin finger-like structures known as cytonemes . Cells use these miniature highways to exchange cargo and signals , such as the protein sonic hedgehog ( SHH for short ) . Alterations to the way SHH is exchanged during development predispose to cancer and lead to disorders of the nervous system . Yet , the mechanisms by which cytonemes work in mammals remain to be fully elucidated . In particular , it is still unclear how the structures start to form , and how the proteins are loaded and transported from one end to another . A ‘molecular motor’ called myosin 10 , which can carry cargo along the internal skeleton of cells , may be involved in these processes . To find out , Hall et al . used fluorescent probes to track both myosin 10 and SHH in mouse cells , showing that myosin 10 carries SHH from the core of the signal-producing cell to the tips of cytonemes . There , the protein is passed to the target cell upon contact , triggering a quick response . SHH also appeared to be more than just passive cargo , interacting with another group of proteins in the signal-emitting cell before reaching its target . This mechanism then encourages the signalling cells to produce more cytonemes towards their neighbours . SHH is crucial during development , but also after birth: in fact , changes to SHH transport in adulthood can also disrupt tissue balance and hinder healing . Understanding how healthy tissues send this signal may reveal why and how disease emerges .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2021	Cytoneme delivery of Sonic Hedgehog from ligand-producing cells requires Myosin 10 and a Dispatched-BOC/CDON co-receptor complex
During eukaryotic evolution , genome size has increased disproportionately to nuclear volume , necessitating greater degrees of chromatin compaction in higher eukaryotes , which have evolved several mechanisms for genome compaction . However , it is unknown whether histones themselves have evolved to regulate chromatin compaction . Analysis of histone sequences from 160 eukaryotes revealed that the H2A N-terminus has systematically acquired arginines as genomes expanded . Insertion of arginines into their evolutionarily conserved position in H2A of a small-genome organism increased linear compaction by as much as 40% , while their absence markedly diminished compaction in cells with large genomes . This effect was recapitulated in vitro with nucleosomal arrays using unmodified histones , indicating that the H2A N-terminus directly modulates the chromatin fiber likely through intra- and inter-nucleosomal arginine–DNA contacts to enable tighter nucleosomal packing . Our findings reveal a novel evolutionary mechanism for regulation of chromatin compaction and may explain the frequent mutations of the H2A N-terminus in cancer . Genome size , defined as the haploid DNA content of a cell , has increased as eukaryotes evolved from single-cell species to more complex , multicellular organisms . Within the same evolutionary timeframe , nuclear volume has also increased but at a slower rate than genome size expansion ( Maul and Deaven , 1977; Olmo , 1982 ) . While the ratio of nuclear to cell size has remained essentially constant in eukaryotes ( Cavalier-Smith , 2005 ) , the disproportional increase in genome size relative to the nuclear volume has required organisms with larger genomes to compact their chromatin to greater extents than organisms with small sized genomes . Indeed there is a positive correlation between genome size and native chromatin compaction as measured by dye incorporation into chromatin ( Vinogradov , 2005 ) . In most eukaryotes , the genome is organized into chromatin by the repeating nucleosomal structure ( Luger et al . , 1997 ) . The nucleosomes stack and fold into higher order structures , serving to systematically compact the genome ( Lieberman-Aiden et al . , 2009; Duan et al . , 2010 ) and to regulate molecular processes that are based on DNA ( Celeste et al . , 2002; Vogelauer et al . , 2002; Fischle et al . , 2005; Kouzarides , 2007; Fussner et al . , 2011 ) . The surface of the histone octamer has 14 DNA interaction sites . Each interaction is mediated by an arginine residue that intercalates into the minor groove of the DNA to stabilize the nucleosomal structure ( Luger et al . , 1997; West et al . , 2010 ) . Arginine is the most commonly used amino acid for interaction with DNA due to its positive charge and the lower energetic cost compared to lysine for displacing water when intercalating into the minor groove ( Rohs et al . , 2009 ) . Nucleosomes mediate chromatin compaction through a variety of mechanisms . For instance , nucleosomes form higher order structures through inter-nucleosomal contacts between the histone H4 N-terminal domain ( NTD ) and the acidic patch of H2A and between two H2B C-terminal domains ( CTD ) ( Luger et al . , 1997; Dorigo et al . , 2003 , 2004; Gordon et al . , 2005; Schalch et al . , 2005 ) . Histone variants , such as H2A . Z or H2A . Bbd , as well as post-translational modifications of histones , such as H4K16ac , can further regulate the degree of compaction ( Suto et al . , 2000; Bao et al . , 2004; Shogren-Knaak et al . , 2006; Zhou et al . , 2007; Chandrasekharan et al . , 2009; Kim et al . , 2009; Fierz et al . , 2011 ) . Polycomb complexes compact large domains of chromatin ( Eskeland et al . , 2010 ) and are important for proper development . Histones of the H1 family promote additional compaction by binding between nucleosomes to linker DNA near the DNA entry/exit site on nucleosomes and stabilize the intrinsic ability of nucleosomal arrays to fold in vitro ( Carruthers et al . , 1998; Robinson et al . , 2008; Szerlong and Hansen , 2010 ) . Linker histones may affect chromatin compaction globally ( Fan et al . , 2005 ) , at specific stages of the cell cycle such as mitosis ( Maresca et al . , 2005 ) or at specific regions of the genome ( Li et al . , 2012 ) . In contrast to canonical histones , the linker histones are much less conserved ( Caterino and Hayes , 2010 ) , and ectopic expression of human linker histones in the budding yeast even at low levels is lethal for the cell ( Miloshev et al . , 1994 ) . Finally , structural proteins such as condensin also contribute to chromatin condensation ( Tada et al . , 2011 ) . Many of these modulatory mechanisms are dynamic in nature ( Luger et al . , 2012 ) and may help explain why multicellular organisms can compact chromatin to different degrees in different cell types . However , despite the existence of these mechanisms for genome compaction in higher eukaryotes , it has not been known whether the canonical histones themselves have evolved sequence features that also contribute to the generally increased chromatin compaction observed in organisms with larger genomes . In this study , we provide evidence from analysis of 160 fully-sequenced eukaryotic genomes that arginine ( R ) residues at specific positions in the N-terminal tail of histone H2A—which protrudes from the nucleosome on the opposite side of DNA entry/exit site—have co-evolved with increasing genome size with a concomitant decrease in serines/threonines . Although increases in genome size are associated with phylogenetic evolution from protozoa to fungi to more complex plants and animals , we present genetic and molecular evidence from the budding yeast and human cells as well as in vitro biochemical data to demonstrate that the evolutionary changes in H2A directly regulate chromatin compaction in vivo and in vitro with consequences for the nuclear volume . The evolutionary changes in H2A regulate chromatin compaction in yeast and human cells , revealing a surprising flexibility in the dynamics of the chromatin fiber that has been conserved across distant eukaryotes . This previously unrecognized structural feature of the nucleosome has evolved to enable greater chromatin compaction when genome size is disproportionately larger than the nuclear volume . Our findings also suggest that the reported mutations in the histone H2A NTD may contribute to the altered chromatin compaction that is commonly observed in cancer cells ( Zink et al . , 2004 ) . To determine whether specific residues in the four core histones have co-evolved with increasing genome size , we performed residue composition analysis of canonical histone protein sequences from 160 fully sequenced eukaryotes with genome sizes ranging from 8 to 5600 Mbp encompassing protozoa , fungi , plants , and animals . The canonical histone proteins for each organism were defined based on at least 90% overlap and 35% identity with the histone fold domain of the corresponding human sequence ( ‘Materials and methods’ ) . Each organism was categorized as having a small ( <100 Mbp ) , medium ( 100–1000 Mbp ) , or large ( >1000 Mbp ) genome ( Figure 1—figure supplement 1A ) . Of the canonical histones , the H2A NTD showed the most statistically significant variability in amino acid residues , where the number of arginines increased with increasing genome size ( Figure 1A ) , while the number of serines ( S ) and threonines ( T ) decreased ( Figure 1B ) . Other amino acid residues in the H2A NTD , including lysines ( K ) , did not correlate with genome size ( Figure 1—figure supplement 1B ) . 10 . 7554/eLife . 02792 . 003Figure 1 . Histone H2A N-terminal sequence has co-evolved with genome size . Violin plots of the number of ( A ) arginines or ( B ) serines/threonines in the H2A NTD for species with small , medium , and large genomes . Plot widths correspond to species frequency within each group . ( C ) H2A NTD sequences for S . cerevisiae and H . sapiens . ( D ) Heat map of H2A NTD residue composition at the indicated positions ordered by genome size . Example species are shown with kingdom and genome size information . ( E ) Protein sequence motifs surrounding the four H2A NTD arginine residues . ( F ) Positioning of evolutionarily variable residues relative to the H2A N-terminus ( left ) or histone fold ( right ) . See also Figure 1—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 00310 . 7554/eLife . 02792 . 004Figure 1—source data 1 . H2A multiple sequence alignments , heat map data , and canonical H2A isoforms . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 00410 . 7554/eLife . 02792 . 005Figure 1—figure supplement 1 . Phylogenetic distribution of species analyzed in this paper . ( A ) The bar graph indicates the number and proportion of organisms in our data set that belong to the indicated phylogenetic kingdoms for each genome size category . ( B ) Violin plots of the number of lysines in the H2A NTD grouped by genome size as in Figure 1 . ( C ) Boxplot of genome sizes of the organisms which have an H2A without ( left ) or with ( right ) the indicated residue . p-values ( Mann–Whitney U test ) of the difference in means between the absence or presence of the indicated residue are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 00510 . 7554/eLife . 02792 . 006Figure 1—figure supplement 2 . H2A arginines 3 and 11 are situated adjacent to DNA within the nucleosome . Structure of a di-nucleosome obtained from the crystal lattice of a mono-nucleosome structure is shown from ( A ) the side or ( B ) close up highlighting potential intra- and inter-nucleosomal interactions between arginines and DNA backbones . Green is H2A , yellow is H2B , cyan is H3 , and salmon is H4 . The red and blue spheres are R3 and R11 , respectively , both of which are in chain C ( Davey et al . , 2002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 006 The acquisition of arginines and loss of serines/threonines in the H2A NTD with increasing genome size occur at specific positions and in sequential order . For instance , the human H2A NTD contains arginine residues at positions 3 and 11 that are absent in yeast and at position 20 which is correspondingly a lysine in yeast ( Figure 1C ) . In contrast , the human sequence lacks S10 and S15 that are observed in the yeast H2A NTD ( Figure 1C ) . Alignment of all H2A NTD sequences also revealed similar trends across all eukaryotes studied here . The heat map in Figure 1D shows the occurrence of arginines and serines/threonines in the H2A NTD as a function of genome size ( see Figure 1—source data 1 for raw data and Figure 1—figure supplement 1C for statistical analysis ) . At position 3 , an arginine ( R3 ) is predominantly present in medium and large species but is lacking in small species . At position 11 , a lysine ( K11 ) is observed in species with medium genomes that evolves to an arginine ( R11 ) mainly in organisms with large genomes . R17 is present in most organisms examined , suggesting a very conserved function for this residue ( Zheng et al . , 2010 ) . At position 20 , small genomes contain predominantly a lysine residue , which converts to arginine in medium and large genomes . In contrast , serines/threonines at positions 10 and 15 are found primarily in organisms with small genomes and much less so in organisms with medium and large genomes ( Figure 1D ) . Additionally , each of the four H2A NTD arginines is surrounded by a conserved motif ( Figure 1E ) . The residues surrounding R3 and R17 are mainly glycine and serine , respectively . At position 11 , the motif varies based on genome size . Species with medium-sized genomes contain VKG and those with large genomes contain ARA . The same is true of position 20 , where AKA is present in organisms with small genomes and ( S/T ) RA in larger genome species ( Figure 1E ) . Interestingly , except for R3 , the positions of all the other evolutionarily varying residues in the H2A NTD are strongly conserved relative to the histone fold domain and not the N-terminus ( Figure 1F ) . When counting conventionally from the N-terminus , amino acids R11 , R17 , and R20—which are numbered based on the human sequence—were not observed consistently at the same positions in other organisms . However , these residues are respectively 12 , 6 , and 3 amino acids away from the histone fold in most species ( note the vertical axes in Figure 1F ) . S10 and S15—which are numbered based on the yeast sequence—also show more uniform positioning when counted from the histone fold . Altogether , as genome size increases , arginines appear in conserved positions within the H2A NTD relative to the histone fold , and serines and threonines are lost . To determine whether arginines and serines/threonines of the H2A NTD affect chromatin compaction in vivo , we took advantage of a strain of Saccharomyces cerevisiae , that has both chromosomal copies of H2A deleted and carries a single copy of H2A on a plasmid ( TSY107 ) , to construct mutant strains containing single or multiple insertions of arginines into their conserved motifs , deletions of serines , or combinations thereof ( see Table 1 for specific amino acid changes and Supplementary file 1A for a description of the mutant strains ) . Two mutants , R3 ( ΔGS10 ) R11 and R11ΔS15 , were also designed such that the spacing between R3 and R11 or R11 and the histone fold , respectively , is the same as in the H2A NTD of organisms with large genomes ( Figure 1C , F ) . As a control for positive charge , mutant strains with lysines inserted in the same positions as arginines were also generated . 10 . 7554/eLife . 02792 . 007Table 1 . List of H2A mutations , sequence changes and their effects on chromatin compaction and nuclear volumeDOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 007FISHNuclear volumeH2A mutantH2A NTD Protein sequence% Changep-value% Changep-valueYeast WTSG–GKG–GKAGSA–AKASQSRSAKAG–1 . 0E+00–1 . 0E+00 R3SGRGKG–GKAGSA–AKASQSRSAKAG−189 . 5E−04−54 . 1E−01 R11SG–GKG–GKAGSARAKASQSRSAKAG−158 . 6E−04−205 . 9E−05 R3R11SGRGKG–GKAGSARAKASQSRSAKAG−228 . 2E−06−163 . 0E−03 R3 ( ΔGS10 ) R11SGRGKG–GKA··ARAKASQSRSAKAG−302 . 1E−06+63 . 7E−01 R11ΔS15SG–GKG–GKAGSARAKA·QSRSAKAG−413 . 9E−08−94 . 7E−04 K3SGKGKG–GKAGSA–AKASQSRSAKAG+98 . 6E−01+139 . 4E−03 K11SG–GKG–GKAGSAKAKASQSRSAKAG+163 . 1E−01+32 . 6E−01 K3K11SGKGKG–GKAGSAKAKASQSRSAKAG+68 . 3E−01+315 . 4E−08 K11ΔS15SG–GKG–GKAGSAKAKA·QSRSAKAG−79 . 2E−02+26 . 6E−01 ΔGS10SG–GKG–GKA··A–AKASQSRSAKAG−63 . 2E−02+103 . 0E−02 ΔS15SG–GKG–GKAGSA–AKA·QSRSAKAG+39 . 4E−02+99 . 7E−03 R6SG–GKGRGKAGSA–AKASQSRSAKAG−55 . 6E−02+101 . 0E−03 K20RSG–GKG–GKAGSA–AKASQSRSARAG−33 . 0E−01+71 . 5E−02 R17KSG–GKG–GKAGSA–AKASQSKSAKAG−17 . 9E−0102 . 7E−01Human—HA Tag WTSGRGKQGGKTRAKAKSRSSRAG–1 . 0E+00–1 . 0E+00 ΔR3SG·GKQGGKTRAKAKSRSSRAG+398 . 3E−03+421 . 3E−08 R11KSGRGKQGGKTKAKAKSRSSRAG+202 . 3E−02+141 . 2E−03 R11ASGRGKQGGKTAAKAKSRSSRAG+431 . 0E−05+215 . 7E−07 ΔR3R11ASG·GKQGGKTAAKAKSRSSRAG+353 . 5E−03+185 . 9E−04Human—FLAG Tag WTSGRGKQGGKARAKAKSRSSRAG–1 . 0E+00–1 . 0E+00 Δ1–12············KAKSRSSRAG+474 . 9E−03+182 . 8E−04The -marks indicate spacing for sequence alignment purposes . The inserted residues are bold typed and underlined . Deletions are indicated by · . Percent ( % ) change refers to the difference in median values relative to WT unless otherwise indicated; the statistically significant differences are bold typed . p-values were calculated using the t-test ( yeast ) and Mann–Whitney U test ( human ) . percent change was calculated relative to isogenic WT control ( ΔS15 ) . To test the effects of H2A NTD changes on chromatin compaction , the physical distance between two probes on chromosome XVI spaced 275 kb apart was assessed in each of the H2A mutants using fluorescent in situ hybridization ( FISH ) ( Figure 2A; Guacci et al . , 1997; Bystricky et al . , 2004 ) . The probes were differentially labeled and visualized by confocal microscopy . The distance between the probes was measured in a single plane in which both probes were present within each nucleus ( Bystricky et al . , 2004 ) . When compared to the isogenic wild type ( WT ) , addition of a single arginine at position 3 ( R3 ) or 11 ( R11 ) to the H2A NTD was sufficient to significantly decrease the average interprobe distance by 18% and 15% , respectively ( Figure 2B , C; Table 1 ) . The average interprobe distance was further decreased by 22% when both arginines were present ( R3R11 ) and even more so ( 30% ) in R3 ( ΔGS10 ) R11 . Deleting G9S10 ( ΔGS10 ) alone caused slightly increased compaction with low statistical significance ( Table 1 ) . The largest decrease in interprobe distance ( 41% ) was observed in the R11ΔS15 mutant , which places R11 12 amino acids from the histone fold , the same position as in organisms with large genomes . Removal of S15 ( ΔS15 ) alone had no effect . The effect of arginines was not simply due to increasing the positive charge of the H2A NTD , as insertions of lysines at positions 3 and 11 did not significantly affect the interprobe distances ( Figure 2C and Supplementary file 2 ) . Although lysines are found at these positions in certain species ( Figure 1D ) , the lack of potential compaction by lysines may be due to the absence of other evolutionary changes in yeast histones ( see Figure 1E ) . Additionally , R17K or K20R mutations did not affect compaction , nor did a randomly inserted arginine at position 6 ( R6 ) ( Figure 2C and Supplementary file 2 ) , suggesting that not every arginine in the H2A NTD contributes to chromatin compaction . 10 . 7554/eLife . 02792 . 008Figure 2 . Ectopic expression of H2A NTD arginines causes compaction in yeast . ( A ) Schematic position of probes on chromosome XVI that were used for FISH . The letters correspond to the probe sets . ( B ) FISH images and ( C ) boxplot of the distributions of interprobe distances for probe set A in the indicated strains . ( D ) The mean interprobe distances for the indicated yeast strains for probe sets A , B , C , and D are plotted as a function of genomic distance . Solid lines are best fit equations . ( E ) Boxplot of the distributions of interprobe distances for probe set A in the indicated strains . Dashed lines mark the median value for the WT strain . The boxplot whiskers contain 90% of the data . All scale bars are 1 µm . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; p<0 . 001; *p<0 . 0001 ( For exact values , see Supplementary file 2 ) . ( F ) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains including the densitometric profiles comparing the WT to each of the mutant H2A strains for a given amount of enzyme . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 00810 . 7554/eLife . 02792 . 009Figure 2—figure supplement 1 . Ectopic expression of H2A NTD arginines causes compaction in yeast . ( A–C ) . Boxplot of the distributions of interprobe distances in the indicated H2A mutant strains for probe sets B , C , and D as shown in Figure 3A . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; **p<0 . 0001 ( Supplementary file 2 ) . Cell cycle analysis of yeast strains in the TSY107 ( D ) or FY406 ( E ) background . 1C and 2C refer to G1 and G2 DNA content , respectively . Note that the WT strains ( TSY107 , FY406 ) carry one copy of the H2A gene on a plasmid with the two chromosomal copies deleted ( Schuster et al . , 1986; Hirschhorn et al . , 1995 ) . Dosage alterations of the H2A protein cause G2/M arrest ( Sopko et al . , 2006 ) which is evident in our WT strains . Nonetheless , all the mutants display highly similar cell cycle profiles . ( F ) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains . The amount of enzyme used to digest chromatin is indicated . Also shown are the densitometric profiles of the agarose gel that compares WT to each of the indicated mutant H2A strains for a given amount of enzyme . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 009 We further confirmed the effects of R11 on chromatin compaction using three additional probe sets ( Figure 2A ) . The level of compaction seen in our WT strain is similar to what has been previously reported in yeast using a different strain background ( Bystricky et al . , 2004 ) . The interprobe distances for all probe sets were significantly decreased in R11 compared to WT and even more so in R11ΔS15 but not ΔS15 alone ( Figure 2—figure supplement 1A–C and Supplementary file 2 ) . Plotting the physical vs genomic distances for all probe sets revealed uniform compaction across large genomic distances ( Figure 2D ) . The effect of R11 on chromatin compaction was not strain-specific as H2A R11 and R11ΔS15 , but not ΔS15 , caused chromatin compaction in a different strain background ( Figure 2E; Supplementary file 2 ) . We therefore conclude that H2A arginines at positions 3 and 11 , especially when R11 is placed at the evolutionarily-conserved position relative to the histone fold , increase the degree of chromatin compaction . Chromatin is differentially compacted at different cell cycle stages ( Guacci et al . , 1997 ) . Cell cycle profile analysis showed little difference between the strain harboring WT H2A and any of the mutant strains ( Figure 2—figure supplement 1D–E ) , indicating that the observed differences in chromatin compaction are not due to altered cell cycle profiles . Chromatin compaction may also be influenced by nucleosomal spacing; indeed the linker DNA length is larger in human cells than in yeast ( Grigoryev , 2012 ) . We find that there are essentially no differences in nucleosomal density in H2A arginine mutants using Micrococcal nuclease ( MNase ) digestion ( Figure 2F , Figure 2—figure supplement 1F ) , indicating that the average nucleosomal spacing is not affected by these mutations . But the more compact mutants displayed decreased accessibility to MNase as indicated by the delayed appearance of the nucleosomal digestion pattern ( Figure 2F , Figure 2—figure supplement 1F ) . Since chromatin structure may influence the volume of the nucleus ( Cavalier-Smith , 2005 ) , we asked whether nuclear volume was affected by H2A-mediated chromatin compaction . We tagged a nuclear pore protein , Nup49 , in its chromosomal locus with GFP to visualize the nuclear membrane and used confocal microscopy to capture three-dimensional images of the nucleus to quantify volumes of ≥150 cells per H2A mutant ( Figure 3A , B; Table 1; Supplementary file 3 , see ‘Materials and methods’ for volume calculations ) . As compared to WT cells , H2A mutants containing R11 or R3R11 , both of which contain more compact chromatin , displayed significantly decreased nuclear volumes . The average nuclear volume in the R3 mutant was also less than WT but did not reach statistical significance . Interestingly , H2A mutants from which serines 10 and 15 were removed displayed larger nuclear volumes . Simultaneous insertions of arginines into these strains ( R3 ( ΔGS10 ) R11 and R11ΔS15 ) decreased their nuclear volume ( R11ΔS15 p<0 . 001 compared to ΔS15 ) , restoring them to levels similar to WT . The control strains with either lysines or R6 had nuclear volumes similar to or larger than WT . Neither arginines nor serines had any effect on total cell size as measured by concanavalin A staining ( Figure 3—figure supplement 1A , B; Supplementary file 3 ) . In the FY406 strain background , ΔS15 did not cause an increase in nuclear volume; and thus both R11 and R11ΔS15 strains exhibited smaller nuclear volumes than isogenic WT ( Figure 3C; Supplementary file 3 ) . These data suggest that modulation of chromatin compaction through the H2A NTD , especially in the presence of R11 , affects the nuclear volume but this effect may be indirect ( see human data below ) . 10 . 7554/eLife . 02792 . 010Figure 3 . Ectopic expression of H2A NTD arginines decreases nuclear volume in yeast . ( A ) Images of the nuclear envelope , as visualized by Nup49p-GFP , and boxplot of the distributions of nuclear volumes in the indicated strains in the TSY107 background ( B ) or the FY406 background ( C ) . Dashed lines mark the median value for the WT strain . The boxplot whiskers contain 90% of the data . All scale bars are 1 µm . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; p<0 . 001; p<0 . 0001 ( Supplementary file 3 ) . Red dagger ( † ) indicates that mean nuclear volume of R11ΔS15 is significantly smaller than its isogenic WT strain ( ΔS15; p<0 . 001 ) . See also Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 01010 . 7554/eLife . 02792 . 011Figure 3—figure supplement 1 . H2A arginines do not affect cell size . ( A ) Images of the cell wall , as visualized by concanavalin A staining , and ( B ) boxplot of the distributions of cellular volumes in the indicated yeast strains . The scale bar is 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 011 Since the H2A NTD in large genomes contains both R3 and R11 , we expected that their removal would cause de-compaction of chromatin . To test this prediction , we ectopically expressed WT or mutant H2A in several human cell lines and measured the distances between probes 0 . 49 Mbp apart on chromosome 1 by FISH , as well as the largest nuclear cross-sectional areas ( ‘Materials and methods’ ) . The H2A gene was HA-tagged and mutated to remove R3 ( ΔR3 ) , to replace R11 with alanine ( R11A ) or lysine ( R11K ) , or to combine two mutations ( ΔR3R11A ) . The H2A constructs were overexpressed using the strong CMV promoter in the normal human IMR90 fibroblasts , the breast cancer cell line MDA-MB-453 , or the HEK293 cells . Cells overexpressing ΔR3 , R11A , or ΔR3R11A H2A mutants had increased interprobe distances , indicating de-compaction of chromatin . Expression of H2A R11K had modest effects on chromatin de-compaction with marginal statistical significance ( Figure 4A , B , Figure 4—figure supplement 1A , B; Supplementary file 4 ) . Cells expressing any of the H2A mutants displayed larger nuclear areas , suggesting that nuclear size is increased ( Figure 4C , D , Figure 4—figure supplement 1C–F; Supplementary file 5 ) . Equal degrees of overexpression were confirmed by immunofluorescence analysis with an anti-HA antibody and detection of HA-H2A by western blotting ( Figure 4D , Figure 4—figure supplement 1G ) . Ectopic expression of a C-terminally FLAG-tagged H2A mutant missing residues 1–12 ( Δ1–12 ) also caused significant de-compaction of chromatin and increased nuclear area despite being expressed at a lower level than WT ( Figure 4E , F ) . These data demonstrate that , consistent with our predictions , the H2A NTD , especially arginines 3 and 11 , function to compact chromatin in human cells . 10 . 7554/eLife . 02792 . 012Figure 4 . Loss of H2A NTD arginines decreases chromatin compaction in human cells . ( A ) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated . ( B ) Boxplot of the distributions of inter-probe distances . Note that R11K was only marginally significant at p=0 . 023 . ( C ) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated . ( D ) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells . Bottom: boxplot of the distributions of α-HA fluorescence intensities . ( E ) Left: FISH images , as in ( A ) , of IMR90 cells expressing a C-terminal FLAG-tagged WT or tailless ( Δ1–12 ) H2A . Right: boxplot of the distributions of inter-probe distances . ( F ) Top: immunofluorescence images of IMR90 cells overexpressing FLAG-tagged WT or tailless H2A . Bottom: boxplot of nuclear areas and fluorescence intensities , as indicated . Dashed lines mark the median value for the WT strain . All scale bars are 10 µm . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; p<0 . 001; p<0 . 0001 ( Supplementary files 4 and 5 ) . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 01210 . 7554/eLife . 02792 . 013Figure 4—figure supplement 1 . Loss of H2A NTD arginines decreases chromatin compaction in human cells . ( A ) FISH images of probes on chromosome 1 in MDA-MB-453 cells with either WT or mutant HA-tagged H2A overexpressed . ( B ) Boxplot of the distributions of interprobe distances . Immunofluorescence images of ( C ) MDA-MB-453 or ( E ) HEK293 cells overexpressing WT or mutant HA-tagged H2A . Boxplot of the distributions of the largest nuclear cross-sectional areas in ( D ) MDA-MB-453 or ( F ) HEK293 for the indicated H2A over-expressing cells . ( G ) Western blot of lysates from HEK293 cells overexpressing the indicated WT or mutant HA-tagged H2A . All scale bars are 10 µm . Dashed lines mark the median value for the WT strain . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; **p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 013 Because R11 compacts chromatin in vivo , we investigated whether this effect is directly on the chromatin fiber . We used step-wise salt dialysis to assemble nucleosomal arrays with a DNA template containing 12 copies of the 177 bp ‘601’ nucleosome positioning sequence ( 601-177-12 ) and recombinant Xenopus laevis histone octamers that contain either WT H2A or one with R11 deleted ( ΔR11 ) . We assembled nucleosomal arrays at different octamer-to-template ratios ( 0 . 9 , 1 , and 1 . 1 octamer to 1 template ) and monitored the quality of the arrays by MgCl2 precipitation and restriction digest analysis using ScaI . We found that a 1:1 octamer-to-template ratio gave the best results as the ScaI digest demonstrated well-assembled arrays compared to the 5% free DNA loaded as a comparison ( Figure 5A ) . We used analytical ultracentrifugation to determine the sedimentation velocity combined with van Holde–Weischet analysis ( Weischet et al . , 1978 ) to ascertain the distribution of sedimentation coefficients ( S ) for each nucleosomal array in the absence or presence of 0 . 8 mM MgCl2 , a concentration of the divalent cation that promotes intra-molecular folding of nucleosomal arrays ( Schwarz and Hansen , 1994 ) . In the absence of Mg2+ , arrays containing WT H2A sedimented with a coefficient of 33 . 1 , which is a value that has been previously shown for similar arrays ( Dorigo et al . , 2003; Shogren-Knaak et al . , 2006; Zhou et al . , 2007 ) . In contrast , arrays missing R11 adopted a more extended conformation with a smaller sedimentation coefficient of 31 . 0 ( Figure 5B ) . Addition of Mg2+ increased compaction of both arrays and shifted the sedimentation coefficients to 39 . 3 and 37 . 4 for WT and ΔR11 H2A , respectively ( Figure 5B ) . A second independent chromatin assembly and ultracentrifuge analysis confirmed these results ( Figure 5—figure supplement 1 ) . Thus , in the absence of R11 in the H2A NTD , nucleosomal arrays adopt a less compact conformation even in the presence of divalent cations , showing that R11 directly increases chromatin compaction . 10 . 7554/eLife . 02792 . 014Figure 5 . H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro . ( A ) Polyacrylamide gel electrophoresis ( PAGE ) of ScaI-digested 601-177-12 DNA template assembled with octamers containing recombinant WT or ΔR11 H2A . As a control , 5% of the 601-177-12 DNA without octamers was also digested . ( B ) The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0 . 8 mM MgCl2 as indicated . S20°C , W is the sedimentation coefficient corrected to water at 20°C . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 01410 . 7554/eLife . 02792 . 015Figure 5—figure supplement 1 . H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro . The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0 . 6 mM MgCl2 as indicated . S20°C , W is the sedimentation coefficient corrected to water at 20°C . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 015 To determine whether chromatin compaction through H2A arginines interferes with transcription regulation , we examined gene expression patterns in the H2A yeast mutants . Remarkably , there was a high level of correlation ( ≥0 . 99 ) between all strains examined ( Figure 6A ) , and no specific gene ontology was found among the genes that were differentially expressed by twofold or more . The expression levels of the histone genes were similar , indicating that altered levels of histone genes expression do not account for the changes in chromatin compaction . These data indicate that compaction of chromatin by H2A does not significantly alter global gene expression in exponentially growing cells . 10 . 7554/eLife . 02792 . 016Figure 6 . Mutations to H2A NTD decrease the fitness of yeast . ( A ) Pearson correlations between the global gene expressions of the indicated strains grown in YPD . Correlations are calculated from an average of at least two experiments . ( B ) Growth curves of the indicated H2A yeast strains over 10 hr in YPD . ( C ) Spot tests with 10-fold serial dilutions for the indicated strains in the presence of different drugs . ( D ) The proportion of yeast cells in a co-culture of WT and the indicated mutant H2A carrying Pgk1 gene fusion to GFP ( green ) or RFP ( red ) as indicated by color . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 016 All strains also showed similar growth rates in rich media ( Figure 6B ) and no significant differences in sensitivity to hydroxyurea , methyl methanesulfonate ( MMS ) , bleomycin , 4-nitroquinoline 1-oxide ( 4NQO ) , cycloheximide , and rapamycin , indicating no major defects with DNA replication or repair , protein synthesis , or the TOR signaling pathways ( Figure 6C ) . But in competition growth assays in which equal amounts of WT and H2A mutant cells harboring the PGK1 gene fused to either GFP or RFP were co-cultured , the H2A mutants regardless of any effect on chromatin compaction , were outcompeted ( Figure 6D ) . This suggests that changes in the H2A NTD sequence can affect the overall fitness of the cell . Deregulated chromatin compaction is often a pathological hallmark of cancer cells ( Edens et al . , 2012 ) , although the underlying mechanisms are not well-understood . A survey of the COSMIC database ( Forbes et al . , 2011 ) , as of the time of writing , revealed 41 documented missense mutations within the H2A NTD with 29 ( 71% ) affecting a residue within one of the four arginine motifs ( Figure 7A ) . R11 , which had the strongest effect of any single arginine residue on chromatin compaction , is the most commonly mutated residue in the H2A NTD . We tested the effects of three of these mutations , R11C , H , and P and found that ectopic expression of each in normal human fibroblasts decreases chromatin compaction significantly with R11P having the strongest effect ( Figure 7B , C ) . These cancer mutations have little effect on increasing nuclear area , however ( Figure 7D , E ) , in contrast to R11A ( Figure 4D ) . It is unclear to what extent the H2A mutants have to be expressed in cancer cells relative to the 17 canonical H2A genes in the human genome to affect chromatin compaction . But our data suggest that over-expression of an H2A mutant has the potential to disrupt chromatin compaction in cancer . 10 . 7554/eLife . 02792 . 017Figure 7 . Mutations of H2A NTD found in cancers decreases chromatin compaction in human cells . ( A ) Schematic of the H2A NTD showing only the mutations within the arginine motifs found in various cancers as indicated by the colored shapes ( Forbes et al . , 2011 ) . The letter within each shape represents the mutated amino acid . ( B ) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated . ( C ) Boxplot of the distributions of inter-probe distances . ( D ) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated . Anti-HA primary and Alexa Fluor 647-conjugated secondary antibodies were used to determine expression in FISH images and for measurement of nuclear areas . ( E ) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells . Bottom: boxplot of the distributions of α-HA fluorescence intensities . Dashed lines mark the median value for the WT strain . All scale bars are 10 µm . Boxes are colored if the mean of the indicated strain is significantly different from WT . Red stars denote level of significance: p<0 . 01; **p<0 . 0001 ( Supplementary files 4 and 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02792 . 017 In this study , we describe evolutionary adaptations of the histone H2A whereby single arginines in the NTD function to dramatically affect the degree of genome compaction . This mechanism is distinct from several other known chromatin compaction mechanisms in higher eukaryotes ( Bednar et al . , 1998; Dorigo et al . , 2003; Shogren-Knaak et al . , 2006; Zhou et al . , 2007; Fierz et al . , 2011 ) in that it involves the histone proteins themselves . The H2A-mediated chromatin compaction thus provides a novel but potentially complementary mechanism for genome compaction . Organisms with small genomes but also very small cell size may face similar physical constraints as those with larger genomes , and may therefore use arginine-containing H2A as a means for chromatin compaction . For instance , Ostreococcus tauri which possesses an R3-containing H2A , is a free-living unicellular algae that has a very small genome of 12 . 6 Mbp but a cell diameter of 0 . 8 µm ( Palenik et al . , 2007 ) . S . cerevisiae , which does not contain an H2A with R3 , has a similarly sized genome but has a cell diameter that is approximately five times larger . Furthermore , certain organisms such as Oikopleura dioica , which has one of the smallest genomes in animals , have distinctive life cycles and possess H2A genes with and without arginines , which may enable them to dynamically regulate genome compaction at different stages of their life cycles ( Moosmann et al . , 2011 ) ( for species with H2A isoforms , see attached spreadsheet ) . So , the H2A arginines may have evolved in circumstances when the genome size became disproportionately large compared to nuclear volume . Interestingly , the toad , Bufo gargarizans , which has a genome size that is twice as large as the human genome , possesses an H2A gene with not only R3 and R11 but also glutamine 6 replaced with an arginine , suggesting that additional arginines in the H2A tail may enable further compaction in organisms with even larger genomes . To better understand the three dimensional positions of the H2A NTD arginines , we examined a crystal structure of the mono-nucleosome in which R3 , R11 , R17 , and R20 were all simultaneously crystalized , and visualized interactions between nucleosomes within the crystal lattice ( Davey et al . , 2002 ) . Interestingly , while R17 and R20 are more buried within the octamer , R3 and R11 are situated close to the DNA backbone . R3 is at 2 . 87 Å from the DNA and could potentially bind the DNA gyre as the DNA wraps around the histone octamer . R11 forms close contacts with the DNA phosphate backbone of self and neighboring nucleosomes ( 4 . 09 , 2 . 90 Å , respectively ) . Although these interactions may have helped form the crystal lattice , they also suggest a possible mechanism for tighter nucleosomal stacking in vivo through shielding of the DNA negative charge ( Figure 1—figure supplement 2A , B; Davey et al . , 2002 ) . Thus , the evolutionary appearance of arginines in the H2A NTD sequence at positions 3 and 11 corresponds to strategic positioning of R3 and R11 within the nucleosome structure that may enable interactions with the DNA , leading to more compact chromatin . Our in vivo data in yeast cells demonstrate that interprobe distances shorten in the presence of the H2A arginines R3 and R11 . While the mechanism of this shortening is still unknown , the two most likely explanations are due to linear chromosomal compaction or increased chromatin looping ( Bohn and Heermann , 2010 ) . However , our data are more consistent with increased linear compaction due to several reasons . First , our analysis of multiple probes along Chr XVI in yeast demonstrates a uniform compaction between all probe pairs examined . Second , our in vivo data with human cell lines shows de-compaction of chromatin in the absence of R11 . If chromatin looping was the mechanism , loops would have to be disassembled in human nuclei independent of factors such as CTCF and condensin . Third , our in vitro data show that R11 alone affects chromatin compaction even in absence of divalent cations . Because the in vitro experiments were performed with unmodified histones in arrays with equal linker lengths , this strongly points to a direct effect of H2A NTD arginines on chromatin compaction that occurs in short arrays . Both R3 and R11 are at contact distances from the DNA , and R11 may also contact the DNA backbone of the neighboring nucleosome ( Figure 1—figure supplement 2 ) . These intra- and inter-nucleosomal interactions with the arginines and the DNA may serve to neutralize the negative charge of the DNA backbone , leading to enhanced stacking of nucleosomes and hence increased compaction . Consistent with this model , the other two arginine residues in the H2A NTD , R17 and R20 , which are more buried from the surface , do not affect compaction by themselves . However , their functions may be to modulate the effects of the surrounding residues . Although all canonical H2A genes contain R3 and R11 in humans , the cell may still be able to dynamically regulate chromatin compaction by these arginines . For instance , arginines may be subject to posttranslational modifications , such a methylation which makes the arginine residue bulkier , or citrullination which removes the positive charge ( Wang et al . , 2001; Hagiwara et al . , 2005; Di Lorenzo and Bedford , 2011; Waldmann et al . , 2011 ) . Interestingly , the H2A NTD is situated in close proximity to the H2B CTD which when ubiquitylated , disrupts chromatin compaction in vitro ( Fierz et al . , 2011 ) , lending support to the ability of this region of the nucleosome to modulate chromatin compaction . The inability of lysines , especially at position 11 , to increase chromatin compaction suggests exquisite structural constraints for H2A-mediated chromatin compaction . Although lysines and arginines both are positively charged , the positive charge of arginine is due to the presence of a guanidinium group that is structurally different from the positive charge of an amino group of a lysine residue . In this regard , it is interesting to note that only arginines contact DNA as it wraps around the nucleosome core ( Luger et al . , 1997 ) ; and arginines preferentially bind the minor groove of DNA compared to lysines ( Rohs et al . , 2009 ) . The context in which lysines appear in evolution may be important as well . We did not observe a lysine at position 3 in our list of organisms , and K11 was present in organisms with medium-sized genomes and surrounded mainly by the motif VKG ( Figure 1D , E ) . When tested in our S . cerevisiae strains , K11 was in the context of AKA . So , it is conceivable that additional amino acid changes would be required for lysines in the H2A NTD to increase genome compaction . The nucleoskeletal theory proposes that chromatin structure influences the shape of the nucleus , and thus is a major determinant of nuclear volume ( Cavalier-Smith , 2005 ) , although the amount of DNA per se does not affect nuclear volume ( Neumann and Nurse , 2007 ) . Non-chromatin components such as nuclear import factors from the cytoplasm may also modulate nuclear size ( Levy and Heald , 2010 ) . Our data suggest that in particular cases , the effects of H2A NTD mutations on chromatin compaction are linked to nuclear volume , although not in a straightforward relationship . While arginines at positions 3 and 11 increase chromatin compaction and reduce nuclear volume , lysines at the same positions have no effect on chromatin compaction yet increase nuclear volume . Removal of serines at positions 10 or 15 has little effect on compaction but also increase nuclear volume . In human cells , expression of all R11 mutants ( R11A , C , H , and P ) decreased compaction but only R11A also affected nuclear area . Although we do not observe a clear-cut relationship between nuclear volume and chromatin compaction , our data identify a region of the nucleosome that is directly or indirectly linked to nuclear volume control mechanisms . Since alterations in chromatin structure often cause changes in transcription ( Parra and Wyrick , 2007 ) , we were surprised that mutant H2A-containing yeast had very similar gene expression profiles as WT cells , grew at similar rates , did not have altered cell cycle profiles , and were not sensitive to DNA damaging drugs or environmental challenges . These data raise the possibility that H2A-mediated compaction of chromatin may have evolved as a mechanism to enable regulation of chromatin compaction without having to make compensatory changes to all other processes that are also based on DNA such as transcription . Nevertheless , it remains to be determined what molecular or cellular processes govern the optimal level of chromatin compaction and nuclear volume for an organism . The stable alterations of chromatin compaction in eukaryotic model organisms through genetic manipulation of H2A should facilitate further experiments to uncover these processes . The yeast strains used in this study are listed in Supplementary file 1A . Yeast cells were grown in YPD at 30°C unless otherwise noted . C-terminal tagging of yeast proteins was performed as described previously ( Longtine et al . , 1998 ) . Mammalian cell lines were maintained at 37°C and 5% CO2 and cultured with 10% fetal bovine serum and DMEM ( Life Technologies , Grand Island , NY ) . Sequences were initially extracted from the Entrez database using a keyword search for ‘histone’ , and removing non-histone sequences by using keyword searches such as ‘histone-like’ , ‘ubiquitin’ , and ‘acetyl’ , yielding 54 , 646 results . Blast 2 . 0 ( Camacho et al . , 2009 ) was used to align the sequences against the highly conserved histone fold region of the four core histones from Homo sapiens . Thresholds for true hits were set at >35% identity match and >90% overlap match with the histone fold globular domain region . All duplicate sequences were removed , and further sequence comparisons were made for histone H3 and H2A sequences to filter variants within them . The canonical sequence data sets comprised 672 sequences for histone H3 , 357 sequences for histone H4 , 518 sequences for histone H2B , and 435 sequences for histone H2A . To further select one canonical sequence for a species among isotypes and variants when annotation was missing , the sequences were compared to the canonical H . sapiens and S . cerevisiae sequence , and the sequence with the highest similarity was selected . Using only completely sequenced species , the final histone sequence data set included canonical sequences for 160 species from plants , fungi , protozoa , and animals , with genome sizes ranging from 8 to 5600 Mbp . Sequences for the four core histones were subsequently split into the N-terminal tail , globular domain , and C-terminal tail ( in the case of H2A and H2B ) sub-sequences . For discovery of patterns of residue changes according to genome size , each of the sub-sequences was further sub-grouped into small ( <100 Mbp ) , medium ( 100–1000 Mbp ) , and large ( >1000 Mbp ) genome sizes . The frequency of the amino acid residues in each sequence in the sub-groups was determined , and a p-value for the comparison between sub groups was obtained using a Mann–Whitney U Test . Multiple sequence alignment profiles were created using the Muscle sequence comparison tool from Embl-EBI ( Edgar , 2004a , 2004b ) . Weblogo3 ( Schneider and Stephens , 1990; Crooks et al . , 2004 ) was used for motif discovery . Heat maps for residue positions were constructed using Cluster 3 . 0 ( de Hoon et al . , 2004 ) and Java Treeview ( Saldanha , 2004 ) . Site directed mutagenesis was performed using the QuickChange Lightning kit ( Agilent Technologies , Santa Clara , CA ) on the pFL142 plasmid . Supplementary file 1B contains all the plasmids that were used and constructed in this study . The sequences of primers are listed in Supplementary file 1C . The correct mutation was verified by sequencing . Yeast strains were generated that contained a C-terminally tagged Nup49p-GFP fusion . Cells were grown in a rich medium to 0 . 6–0 . 8 × 107 cells/ml , fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature , washed twice in PBS , and mounted on a poly-L-lysine-coated slide with mounting medium ( Vector Laboratories , Burlingame , CA ) . Z-stacks were obtained as described in the microscopy imaging section , and GFP excitation was achieved at 488 nm . Resulting z-stack images were de-convolved using a constrained iterative algorithm from SlideBook 5 . 0 software and nuclear volumes were measured by masking the inside of each nucleus , which were delineated by the GFP signal . The resulting mask was used to calculate volumes through the SlideBook software . Statistical analysis was performed using the Student's t test . Yeast strains were grown in rich medium to 0 . 6–0 . 8 × 107 cells/ml , fixed in growth medium with 4% paraformaldehyde for 15 min at room temperature , washed twice in PBS , and stained with a 1:50 dilution of concanavalin A conjugated with tetramethylrhodamine ( Life Technologies ) for 15 min at room temperature . Cells were washed twice in PBS , once in water , and mounted on a poly-L-lysine-coated slide with mounting medium . Z-stacks were obtained as described in the microscopy imaging section with mRFP excitation . Cell volume was measured by masking the inside of the RFP signal as described in the measurement of yeast nuclear volume . For yeast FISH analysis , DNA templates for probes 1 , 3 , and 4 came from cosmids 71042 , 70912 , and 70982 ( American Type Culture Collection , Manassas , VA ) as described elsewhere ( Guacci et al . , 1994 ) . DNA templates for Probe 2 were obtained by PCR amplification of a 10-kb region starting at coordinate 364647 of chromosome 16 using three primer pairs ( Probe2_P1 , Probe2_P2 , Probe2_P3 , Supplementary file 1C ) . All DNA templates were digested to smaller fragments using Sau3a ( New England BioLabs , Ipswich , MA ) . Fragments were directly labeled using BioPrime labeling kit ( Life Technologies ) with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP ( Life Technologies ) . For human cell FISH analysis , DNA templates for probes came from BACS RP11-252L24 and RP11-195J4 spaced 0 . 488 Mb apart on chromosome 1 . Each BAC was digested into smaller fragments using Sau3a and fragments were directly labeled using BioPrime labeling kit with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP , as described above . Yeast strains were grown in rich medium to 0 . 6–0 . 8 × 107 cells/ml and fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature . Cells were then washed twice in the growth medium and re-suspended in 2 ml of EDTA-KOH ( 0 . 1 M , pH 8 . 0 ) and 10 mM DTT and incubated for 10 min with shaking at 30°C . Cells were spun down and re-suspended in 2 ml of YPD + 1 . 2 M sorbitol with 50 µg/ml of Zymolyase 100-T ( Sunrise Science Products , San Diego , CA ) and 400 U/ml of lyticase ( Sigma-Aldrich , St . Louis , MO ) and incubated at 30°C for 16 min with shaking . Spheroplasts were then washed twice in YPD + 1 . 2 M sorbitol and transferred to a poly-L-lysine-coated slide . After settling for 5 min , excess liquid was aspirated away and the slides were allowed to air dry for 5 additional min . The slides were washed in methanol for 10 min and then acetone for 30 s before air drying . Cells were then dehydrated in a series of cold ethanol washes ( 70% , 80% , 90% , 100% , 1 min each ) and allowed to air dry . Denaturing solution ( 70% deionized formamide , 2 × SSC ) was added to the slide , and cells were denatured at 75°C for 7–10 min . The slides were immediately put through another cold ethanol dehydration series and allowed to air dry . Hybridization solution ( 50% deionized formamide , 2 × SSC , 10% dextran sulfate , 100 ng/µl salmon sperm DNA ) containing fresh probes was added to the slide , and the probes were hybridized for 40–48 hr at 37°C . The slides were then washed in two 5 min washes in 0 . 05 × SSC at 48°C and washed twice in BT Buffer ( 0 . 15 M NaHCO3 pH 7 . 5 , 0 . 1% Tween ) for 5 min at room temperature . Mounting medium containing DAPI ( Vector Laboratories ) was added to the slides , and a coverslip was sealed with nail polish . Inter-probe distances were measured in single projections as described elsewhere ( Bystricky et al . , 2004 ) by finding the pixel distance between weighted centers of the green signal and red signal and converted to nm by the appropriate factor . A 3i Marianas SDC confocal microscope equipped with a Zeiss AxioObserver Z1 with a 100 × /1 . 45 NA objective and Yokogawa CSU-22 confocal head was used . Images were captured by a Hamamatsu EMCCD C9100-13 camera controlled by Slidebook 5 . 0/5 . 5 ( Intelligent Imaging Innovations , Denver , CO ) . DAPI , GFP , mRFP , and Far-red images were acquired by excitation at 360 nm , 488 nm , 561 nm , and 640 nm from a high-speed AOTF laser launch line . A step size of 0 . 3 ( yeast ) or 0 . 5 ( human ) µm was used for z-stack acquisition . Micrococcal nuclease ( MNase ) digestions were performed on exponentially growing yeast cells as described previously , except that the enzyme was obtained from Sigma-Aldrich ( Sigma-Aldrich ) ( Rando , 2010 ) . RNA was extracted from exponentially growing yeast as described previously ( Schmitt et al . , 1990 ) . PolyA-RNA was prepared , labeled , and hybridized to Affymetrix Gene ChIP Yeast Genome 2 . 0 array by the UCLA clinical microarray core facility and data normalized according to manufacturer's indications . The data are accessible at Gene Expression Omnibus with accession number GSE50440 . A plasmid containing 12 tandem 177 bp repeats of the high affinity 601 sequence was obtained from Craig L Peterson's laboratory ( Shogren-Knaak et al . , 2006 ) . DNA arrays were prepared as described previously ( Luger et al . , 1999 ) . After excision with EcoRV , the arrays were gel purified . QuikChange Lightning Site-Directed Mutagenesis ( Agilent Technologies ) was used to create H2A ΔR11 using primers as listed in Supplementary file 1C . Recombinant X . laevis histones were expressed in bacteria and purified as described previously ( Luger et al . , 1999 ) . Equimolar amounts of all histones were co-folded to form octamers . Intact octamers were purified from aggregates and free H2A-H2B dimers using Pharmacia Superdex 200 gel filtration column ( GE Healthcare Bio-Sciences , Pittsburgh , PA ) . Recombinant histone octamers and the 601-177-12 DNA template ( Lowary and Widom , 1998 ) were combined in stoichiometric amounts where 1 . 0 equivalent of histone octamers and 1 . 0 equivalents of DNA template were mixed in 2 . 0 M NaCl . Nucleosome arrays were assembled by step-wise salt dialysis in decreasing NaCl concentration: 1 . 6 M , 1 . 2 M , 1 . 0 M , 0 . 6 M , 0 . 4 M , 0 . 1 M , and 0 . 025 M ( in 10 mM Tris pH 8 . 0 , 0 . 25 mM EDTA ) , followed by exchanges with 2 . 5 mM NaCl and 10 mM Tris pH 8 . 0 without EDTA . Each dialysis step was performed at 4°C for 4 hr to overnight . Partially assembled chromatin was eliminated by precipitation in 4 . 0 mM MgCl2 ( Dorigo et al . , 2003 ) . The extent of array saturation was assessed by ScaI digestion ( 200 ng total DNA/chromatin , 3 units ScaI , 50 mM NaCl , 50 mM Tris pH 7 . 4 , 0 . 5 mM MgCl2 ) , performed for 16 hr at room temperature followed by 1 hr at 37°C , and subsequent analysis using a 5% native polyacrylamide gel ( Luger et al . , 1999 ) . Nucleosome arrays were allowed to equilibrate at room temperature in buffer ( 2 . 5 mM NaCl , 10 mM Tris–HCl pH 8 . 0 ) containing either 0 . 1 mM EDTA or 0 . 6 and 0 . 8 mM MgCl2 . Samples were centrifuged at 20 , 000 RPM on a Beckman Optima XL-I analytical ultracentrifuge using an An60 Ti rotor after a 1 hr equilibration at 20°C under vacuum . Time-dependent sedimentation was monitored at 260 nm . Boundaries were analyzed by the method of van Holde and Weischet ( Weischet et al . , 1978; Hansen and Turgeon , 1999 ) . N-terminally HA-tagged WT H2A of X . laevis was cloned by PCR into mammalian expression vector pCMV-HA ( Clontech Laboratories , Mountain View , CA ) between EcoRI and NotI sites . C-terminally Myc-FLAG-tagged human H2A , in a mammalian expression vector , was obtained from OriGene ( RC200688 , Origene Technologies , Rockville , MD ) . Site directed mutagenesis was performed using the QuickChange Lightning kit ( Agilent Technologies ) on these expression plasmids . Human cells ( HEK293 , IMR90 and MDA-MB-453 ) were grown on glass coverslips in 24-well plates in DMEM containing 10% fetal bovine serum and transfected with the indicated H2A expression plasmids using BioT transfection reagent ( Bioland Scientific , Paramount , CA ) or Lipofectamine LTX with Plus reagent ( Life Technologies ) . Cells were grown for 48 hr post-transfection . For immunofluorescence only , transfected cells were fixed with ice-cold methanol for 15 min at −20°C followed by washing with PBS-T . For combined immunofluorescence and FISH , transfected cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature followed by washing with PBS . Cells were then permeabilized in 0 . 5% Triton X-100 in PBS for 10 min at room temperature followed by washing with PBS . Cells were blocked in 5% BSA and incubated with anti-HA antibody ( ab9110; 1:250 dilution , Abcam , Cambridge , MA ) or anti-FLAG antibody ( F1804; 1:1000 dilution , Sigma-Aldrich ) . Cells were washed and incubated with secondary antibody ( A11008; 1:500 Alexa Fluor 488 goat anti-rabbit , A21245; 1:250 Alexa Fluor 647 goat anti-rabbit , A11001; 1:500 Alexa Fluor 488 goat anti-mouse , or A21235; 1:100 Alexa Fluor 647 goat anti-mouse , Life Technologies ) . For immunofluorescence , cells were washed and then incubated with Hoechst stain ( 0 . 001 mg/ml in PBS ) . After final washes , cover slips were mounted and imaged . Fluorescence was visualized as above except with the use of 63X magnification . For FISH , cells were washed , following secondary antibody incubation , in CSK buffer ( 100 mM NaCl , 300 mM sucrose , 3 mM MgCl2 , 10 mM PIPES pH 6 . 8 ) and permeabilized in CSKT buffer ( CSK+0 . 5% Triton X-100 ) before being fixed for 10 min in 4% paraformaldehyde in PBS at room temperature . Cells were immediately put through a cold ethanol dehydration series ( 5 min each at 85% , 95% , and 100% ) and allowed to air dry . Cells were rehydrated in 2 × SSC for 5 min and then RNase-treated for 30 min at 37°C in a humid chamber . Cells were washed with 2 × SSC and denatured at 80°C for 15–20 min with 70% deionized formamide and 2 × SSC . They were immediately cooled with cold 2 × SSC and put through another cold ethanol dehydration series . Probes were added to cells and allowed to hybridize for 48 hr . After hybridization , cells were washed with 50% formamide in 2 × SSC , 2 × SSC , and 1 × SSC containing DAPI . Slides were mounted , imaged , and analyzed as described above . Nuclear staining , in H2A-expressing cells , was used to measure lengths of the long and short orthogonal nuclear axes . Estimated nuclear cross-sectional area was calculated using the following formula: Area = ( D1/2 ) ( D2/2 ) *π , where D1 and D2 are long and short axis lengths , respectively . Two sets of yeast strains were generated in which Pgk1p was C-terminally fused with either GFP or RFP ( Supplementary file 1A ) . GFP-labeled WT H2A strains were co-cultured with RFP-labeled mutant H2A strains at a 1:1 ratio and at an optical density of ∼0 . 4 . Corresponding co-cultures with switched fluorescent labels were also made . Cultures were incubated at 30°C for 72 hr and were diluted every 6–12 hr to maintain cells in exponential growth phase . Samples were collected every 12 hr for analysis by flow cytometry . Collected cells were fixed in 70% ethanol , washed , and re-suspended in 50 mM sodium citrate , pH 7 . 0 , and mildly sonicated to disrupt aggregates . GFP- and RFP- labeled cells were counted using a Becton Dickinson FACScan cytometer , and the proportion of each in the population was calculated . Cell cycle analysis of exponentially growing cells was performed essentially as described previously ( Zou et al . , 1997 ) , except that cells were stained with 1 μM SYTOX Green ( Life Technologies ) . Approximately 1 . 0 × 107 exponentially growing yeast cells were collected and re-suspended in 100 μl of H2O and 10-fold serially diluted . Subsequently , 5 μl was spotted on agar plates containing media and drugs as indicated in the figures and incubated at 30°C for 2–6 days .	There are up to three meters of DNA in a human cell . To fit this length into the cell's nucleus in an organized manner , DNA is wrapped around proteins called histones and then tightly packaged to form a structure called chromatin . The interaction between the histones and the DNA is helped by certain amino acids on the surface of the histones fitting snugly into the DNA molecule . Plants and animals have genomes that are significantly larger than those of single-celled organisms . However , although genome size has increased gradually during the evolution of complex organisms , the size of the nucleus has not undergone a similar expansion . Large genomes are therefore packaged more tightly than small genomes . However , we do not fully understand how different species evolved the ability to do this . Now Macadangdang , Oberai , Spektor et al . have compared the histones of 160 species ranging from single-celled microorganisms to plants and animals . This revealed that the amino acids in a particular type of histone—histone 2A—vary according to genome size . Organisms with small genomes use histone 2A proteins with fewer arginine amino acids on their surface than organisms with large genomes . Further experiments showed that yeast cells engineered to contain arginine-rich histones wind their DNA more tightly; and , in some cases when the chromatin was more compacted , the nuclei were also smaller . On the other hand , removing arginines from histones in human cells cause the chromatin to be loosely packed and the nuclei to be larger than normal . Moreover , chromatin is often abnormally packed in cancer cells and Macadangdang et al . found that many of these cells contained histones with fewer arginines than normal . Plants , animals , and other eukaryotes have evolved a variety of mechanisms to control how much they compact their chromatin in addition to the way discovered by Macadangdang et al . Future work is now needed to determine how these different mechanisms work together in different species such that the chromatin is compacted to the optimal level .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "cell", "biology" ]	2014	Evolution of histone 2A for chromatin compaction in eukaryotes
The Endosomal Sorting Complexes Required for Transport III ( ESCRT-III ) proteins are critical for cellular membrane scission processes with topologies inverted relative to clathrin-mediated endocytosis . Some viruses appropriate ESCRT-IIIs for their release . By imaging single assembling viral-like particles of HIV-1 , we observed that ESCRT-IIIs and the ATPase VPS4 arrive after most of the virion membrane is bent , linger for tens of seconds , and depart ~20 s before scission . These observations suggest that ESCRT-IIIs are recruited by a combination of membrane curvature and the late domains of the HIV-1 Gag protein . ESCRT-IIIs may pull the neck into a narrower form but must leave to allow scission . If scission does not occur within minutes of ESCRT departure , ESCRT-IIIs and VPS4 are recruited again . This mechanistic insight is likely relevant for other ESCRT-dependent scission processes including cell division , endosome tubulation , multivesicular body and nuclear envelope formation , and secretion of exosomes and ectosomes . ESCRTs are categorized into groups −0 through -III and act in various cellular processes including cell division , multivesicular body formation , and wound repair ( Hurley , 2015 ) . An investigation of protein sorting between endosomes and lysosomes initially led to the discovery of ESCRT-I ( Katzmann et al . , 2001 ) , with other ESCRTs identified soon afterwards ( Babst et al . , 2002a , 2002b; Katzmann et al . , 2003 ) . Sequential recruitment of the ESCRTs enables division of membrane compartments , with ESCRT-III being critical for the membrane scission process ( Henne et al . , 2013 ) . ESCRT-IIIs have a five alpha-helix core structure and assemble in vitro into macromolecular rings or spirals ( McCullough et al . , 2015; Muzioł et al . , 2006 ) . ESCRT-IIIs are believed to polymerize in the neck of the membrane constriction to drive membrane fission ( Cashikar et al . , 2014; Dobro et al . , 2013; Fabrikant et al . , 2009; Hanson et al . , 2008; Henne et al . , 2012; Lata et al . , 2008; McCullough et al . , 2015 ) . Scission of the membrane , even in the presence of ESCRT-IIIs , stalls in the absence of the hexameric AAA+ ATPase ( Morita et al . , 2010 ) , VPS4A/B , which contains a microtubule interacting and transport ( MIT ) domain that binds to MIT interacting motifs ( MIM ) of ESCRT-IIIs ( Obita et al . , 2007; Stuchell-Brereton et al . , 2007 ) . The final scission process is believed to be associated with VPS4 working on ESCRT-IIIs , but the mechanism is still unresolved . In some models , the ESCRT-IIIs provide the motive force for scission and the VPS4 is required after scission to recycle the ESCRT-IIIs for subsequent scissions ( Lata et al . , 2008; Wollert et al . , 2009 ) . In other models , the VPS4 is actively engaging the ESCRT-IIIs prior or during scission by actively remodeling ESCRT-IIIs in order to force scission ( Saksena et al . , 2009 ) , by rearranging ESCRT-IIIs as part of the pathway toward scission ( Cashikar et al . , 2014 ) , or by binding to ESCRT dome structures in order to add rigidity necessary for scission ( Fabrikant et al . , 2009 ) . ESCRT complexes are hijacked by HIV to enable separation of the viral particle from the host cell plasma membrane . The production of enveloped HIV-1 at the plasma membrane occurs with the recruitment of the structural protein Gag at individual assembly sites . The carboxyl terminus of Gag has a motif that is essential for recruitment of ESCRTs . First , Gag recruits the ‘early’ factors like ESCRT-I/TSG101 and ALIX which contribute to subsequent recruitment of ESCRT-III proteins . The ESCRT-IIIs then polymerize into structures that are believed to constrict the neck and drive membrane fission . HIV release appears to require fewer members of the ESCRT family than other processes . Redundancy likely makes many variants , such as CHMP5 , CHMP6 and CHMP7 , only conditionally necessary ( Morita et al . , 2011 ) . ESCRT-IIIs that are essential for assembly of HIV-1 include CHMP2 ( either A or B variant ) and CHMP4B , which are recruited to site of budding with other proteins such as ESCRT-I/TSG101 and ALIX which interact with Gag ( Morita et al . , 2011 ) . The reduced number of required ESCRTs makes HIV assembly an approachable system for studying the biophysics of ESCRT-mediated membrane scission . Prior to viral particle separation from a host cell , a roughly spherical particle is formed ( Martin-Serrano et al . , 2003 ) , but the topological pathway and timing of events to reaching the Gag sphere has not previously been followed in vivo . The order of some of the events in virion assembly has been resolved by live-cell microscopy . First , the HIV- genome is recruited to the membrane , potentially with a few Gag molecules ( Jouvenet et al . , 2009 ) . Then , over a 5–30 min period , the Gag accumulates around the genome ( Ivanchenko et al . , 2009; Jouvenet et al . , 2008 ) . Once Gag reaches a steady-state , ESCRT-III and VPS4 are transiently recruited at the site of assembly ( Baumgärtel et al . , 2011; Bleck et al . , 2014; Jouvenet et al . , 2011 ) . The timing of some critical steps is not known impacting our understanding of the mechanism . It is not known whether bending occurs before , during , or after the transient recruitment of ESCRTs . Thus , is membrane bending driven by Gag multimerization , by Gag engagement with the HIV-1 genome or by the ESCRTs ? It is also not known if scission occurs before , during , or after the transient recruitment of ESCRT-IIIs or VPS4 . Do they generate the force for scission , do they prepare the membrane for scission , or does VPS4 recycle ESCRT-IIIs after scission ? Here , we investigated , during the assembly of HIV Gag , the temporal recruitment of ESCRT-III proteins and VPS4 relative to membrane scission . We also examined membrane curvature during Gag assembly to determine when a spherical particle forms relative to ESCRT-III recruitment . We find that membrane bending occurs contemporaneous with recruitment of Gag and prior to arrival ESCRT-III . The ESCRT-IIIs and the VPS4 ATPase arrive after Gag assembly has concluded , remain at the membrane for tens of seconds , and then leave tens of seconds before scission . During the period after departure of the ESCRT-IIIs , neutralizing the surface charge on the membrane accelerates the membrane scission . To determine the timing of ESCRT recruitment relative to membrane bending and scission , we quantified ESCRT recruitment during the assembly of HIV-1 virus like particles ( VLPs ) while assaying membrane bending and scission . Scission was assayed by monitoring the ability of protons to flow between the cytosol and the lumen of the virion . The pH in the lumen of the virion was monitored with a pH-sensitive GFP ( pHlourin ) ( Miesenböck et al . , 1998 ) fused to Gag ( Jouvenet et al . , 2008 ) while modulating the cytoplasmic pH by cycling the pCO2 every 10 s between 0 and 10% , thus an average of 5% pCO2 ( Figure 1—figure supplement 1 ) . CO2 rapidly diffuses across plasma membranes ( Hulikova and Swietach , 2014; Simon et al . , 1994 ) and is converted to carbonic acid by cytoplasmic carbonic anhydrase , altering the cytoplasmic pH . We have previously observed that Gag-pHlourin in a budded VLP is less sensitive to changes of pCO2 than in the cytosol , suggesting carbonic anhydrase is excluded from VLPs ( Jouvenet et al . , 2008 ) . At sites of VLP assembly , the average Gag-pHluorin fluorescence increase was similar to the increase of Gag-mEGFP ( Jouvenet et al . , 2008 ) , but the intensity oscillated in sync with switching the pCO2 ( Figure 1A , Figure 1—figure supplement 2 , Figure 1—video 1 and Figure 1—video 2 ) . At various times after Gag accumulation reached a steady-state maximum the magnitude of oscillations decreased , indicating a loss of the ability of protons to move between the VLP and cytoplasm due to scission . CO2 may still cross the VLP membrane after scission and enter via gaps in the lattice of the immature Gag lattice ( Briggs et al . , 2009; Carlson et al . , 2008; Woodward et al . , 2015; Wright et al . , 2007 ) . However , the greatly reduced sensitivity to pH indicates carbonic anhydrase is not present , consistent with analysis of cellular proteins in HIV particles by mass spectrometry ( Ott , 2008 ) . Not surprisingly , scission was never observed before Gag had finished accumulating at individual VLPs . Unexpectedly , the ESCRT-III CHMP4B both appeared ( Avg = 59 s , N = 30 out of 30 ) and disappeared ( Avg = 22 s , N = 27 out of 30 ) prior to scission ( Figure 1B ) . CHMP4B has been proposed to form a circular/spiral structure which supports assembly of a smaller CHMP2 ( A/B ) dome which generates fission by pulling the neck together ( Fabrikant et al . , 2009 ) . Thus , it is possible that CHMP4B may be removed prior to scission leaving CHMP2A or CHMP2B present to facilitate fission . To probe the timing of CHMP2 , endogenous CHMP2A and CHMP2B were lowered with siRNA ( Figure 1—figure supplement 3 ) to facilitate observation of mCherry-CHMP2A or mCherry-CHMP2B ( Figure 1C and E , Figure 1—figure supplement 2 ) . CHMP2A and CHMP2B both appeared ( Avg . = 77 s , N = 29 out of 29; and 63 s , N = 23 out of 23 , respectively ) and disappeared prior to scission ( Avg . = 23 s , N = 26 out of 29; and 27 s , N = 23 out of 23 , respectively Figure 1D , F ) . This observation indicates their assembly and disassembly is also not physically forcing scission . Next the dynamics of recruitment of VPS4 , the energy providing ATPase , was monitored ( Figure 1G and Figure 1—figure supplement 2 , Figure 1—video 3 and Figure 1—video 4 ) . Similar to the ESCRT-IIIs , VPS4A also appeared prior to scission ( Avg = 42 s , N = 28 out of 28 ) and disappeared prior to scission ( Avg = 17 s , N = 27 out of 28 , Figure 1H ) . On average VPS4A disappeared from the assembly site closer to the time of scission than CHMP4B ( 5 s ) , CHMP2A ( 6 s ) or CHMP2B ( 10 s ) . A simultaneous measurement of CHMP4B and VPS4A confirmed CHMP4B was recruited ~5 s prior to VPS4A ( N = 41 ) ( Figure 1—figure supplement 4 ) , which agrees with previous results in HeLa cells ( Bleck et al . , 2014 ) . Our results indicate that the ESCRT-IIIs and the ATPase VPS4 leave the membrane prior to scission . It is possible that the ESCRT-IIIs play an essential role in tightening the membrane neck , but then need to be cleared away to allow for the opposing membranes to come closer for the scission reaction . The specific lipid composition in the neck is not known . However , both HIV-1 Gag and the ESCRT complexes are recruited to regions rich in phosphatidylinositol-4 , 5-bisphosphate ( PIP2 ) , which has four negative charges at pH 7 ( Kooijman et al . , 2009 ) , with the net negative charge being the critical parameter for engaging ESCRTs ( Lee et al . , 2015 ) . Thus , ESCRTs at the neck may bias the composition toward more negatively charged lipids ( Chiaruttini et al . , 2015 ) . The pKa values for the two phosphate groups on the inositol of PIP2 are 6 . 5 and 7 . 7 ( van Paridon et al . , 1986 ) . The change in pHluorin intensity from 0 and 10% pCO2 , indicates the cytosolic pH ranges from ~7 . 5 to ~6 . 5 . Therefore , lowering the pH by raising the pCO2 to 10% should protonate and reduce the charge of these negatively charged lipids thus reducing the repulsive force between the membranes . To test if scission was affected by changes in cytosolic pH , we switched the pCO2 at a slower rate , every 120 s ( Figure 2 , Figure 2—video 1 and Figure 2—video 2 ) . Scission was ~3X more likely when the cytoplasm was in the low pH state ( 10% pCO2 ) than the high pH state ( 0% pCO2 ) ( Figure 2F ) consistent with the idea that scission is more likely when the net negative charge on phospholipids in the viral neck is reduced by protonation . Other common phospholipids like phosphatidic acid ( PA ) , phosphatidylethanolamine ( PE ) , phosphatidylcholine ( PC ) , phosphatidylserine ( PS ) , and phosphatidylglycerol ( PG ) , have pKa values outside the pH range used in these experiments so are not expected to be as sensitive to charge modulation during pH switching as PIP2 . However , other lipids or components with pH-dependent charge sensitivity might also or alternatively result in these observations . Multiple rounds of ESCRT-III/VPS4 recruitment were previously observed following completion of Gag accumulation ( Baumgärtel et al . , 2011; Jouvenet et al . , 2011 ) . It is possible that the first wave of ESCRT-III/VPS4 led to productive scission and the subsequent rounds are inconsequential . Alternatively , the initial waves could be non-productive , perhaps a consequence of a failure to recruit both ESCRT-IIIs and VPS4A concurrently , necessitating additional rounds . While simultaneously imaging Gag-pHluorin , CHMP4B and VPS4A , when there were multiple waves , both CHMP4B and VPS4A were recruited ( Figure 3 ) . Multiple waves of recruitment were observed in ~20% of traces , consistent with our previous observations of repeat rounds of recruitment of CHMP1B , CHMP4B , CHMP4C and VPS4A during assembly of HIV-1 and EIAV ( Jouvenet et al . , 2011 ) . Scission was only observed after the final wave of recruitment of ESCRT-III/VPS4 ( Figure 3B , C ) . Thus , not every cycle of arrival and then dispersal of ESCRT-III/VPS4 leads to subsequent scission . If scission does not occur then a subsequent cycle of recruitment and dispersal of the ESCRT-III/VPS4 is required to complete the process . If the function of VPS4A is to recycle ESCRT-IIIs after scission then only a single wave would be expected since ESCRT-IIIs would not be removed until after the single scission event has occurred . Next , we set out to determine when membrane bending occurs relative to the assembly of Gag and recruitment of ESCRT-IIIs . We expressed a fluorescent protein ( either EGFP or one of two circularly permutated superfolder variants , sf3 or sf11 ( Pédelacq et al . , 2006 ) as a fusion to Gag ( at the carboxyl terminus , p6-GFP , or in the matrix protein of Gag , MA-sf3 or MA-sf11 ) to be able to follow membrane bending in live-cell imaging via changes in anisotropy of the GFP tag . The orientation of the chromophore was characterized with a custom built polarized total internal reflection fluorescence ( TIRF ) illuminator ( Johnson et al . , 2014 ) . During accumulation of Gag at VLPs , the emission of the GFP was quantified while excitation alternated with polarization perpendicular ( p^ polarized ) followed by parallel ( s^ polarized ) to the glass surface . Orientation was characterized by the ratio of emission intensities ( P/S ) and total Gag was monitored by P+2S ( Figure 4A and Figure 4—figure supplement 1 , Figure 4—video 1 and Figure 4—video 2 ) ( Anantharam et al . , 2010 ) . As Gag accumulated , the ratio of P/S dropped from ~2 to ~1 . 4 , with little variation ( ±0 . 1 ) between the Gag-GFP versions ( Figure 4B ) . The drop in P/S correlated with the increase in Gag as would be expected if the plasma membrane was bending during Gag assembly . The halfway decrease of P/S occurred prior to the halfway increase of the total Gag fluorescence ( Figure 4C ) . Following Gag recruitment , as indicated by a plateau in Gag signal , there was no transition in P/S . This observation is inconsistent with the subsequent recruitment of the ESCRT-IIIs facilitating the transition from a flat lattice to a spherical particle . Bending was also investigated during the assembly of a Gag that is missing its carboxyl terminal p6 domain which functions to recruit early acting proteins like ESCRT-I/TSG101 or ALIX . A similar drop in P/S during assembly of Gag was observed with Gag-∆p6 ( Figure 4A , Figure 4—figure supplement 2 ) indicating that the ESCRTs recruited via p6 are also not necessary for the transition from a flat lattice to a spherical particle . In order to better understand the observed P/S ratio , we formulated an expected P/S ratio for a spherical cap growing out of a flat membrane ( Figure 4D ) . Briefly , we assumed the growing bud consisted of excitation dipoles uniformly distributed across the surface , with the dipoles oriented an angle β relative to the surface normal . A predicted P/S with respect to β and the normalized budded surface area ( area from 0 to 1 ) was then found by integrating over all defined dipole orientations and the extent budded surface area . More specifically , using coordinates described previously ( Anantharam et al . , 2010 ) , position on the surface of the sphere was given in terms of a polar angle θ and an azimuthal angle ϕ , and the current extent budded was defined by θ ( Figure 4E ) . Thus , when θ=0∘ there was no budding , when θ=90∘ the sphere was half budded with dipoles from θ=0∘→90∘ , and when θ=180∘ the sphere was fully budded with dipoles from θ=0∘→180∘ . A uniform distribution of excitation dipoles was assumed on the bud ( no dependence on ϕ or θ ) ; however , at any given position these dipoles had an angular distribution that depended on the polar angle β relative to the surface normal: ρ ( β ) . For instance , if all dipoles were oriented at β=45∘ then ρ ( β ) =δ ( β−45∘ ) where δ ( x ) is a delta function . The angular distribution was assumed to be uniform relative to the surface azimuthal angle ψ and the sphere was assumed to be smaller than the optical resolution of the microscope . A predicted P/S ( ρ ( β ) ) relative to extent budded θ was found by determining the average component of the dipole excitation in y^ ( parallel to glass surface ) and in z^ ( normal to glass surface ) . Note: Experimentally due to azimuthal scanning we excited in both x^ and y^ , each 50% the time , but for simplicity in this analysis 100% excitation in y^ was assumed since x^ and y^ are symmetric . The total collected fluorescence , S and P , in y^ and z^ were predicated by: ( 1 ) S=∫0θ∫02π∫0π∫02πQ∣∣∣Ey^μy^∣2 sin ( θ ) sin ( β ) dψdβdϕdθ ( 2 ) P=∫0θ∫02π∫0π∫02πQ⊥∣Ez^μz^∣2 sin ( θ ) sin ( β ) dψdβdϕdθwhere μy^ and μz^ are the components of the excitation dipole in y^ and z^ with respect to positon on the bud surface , Ey^ and Ez^ are the excitation electric field components in y^ and z^ , and Q∣∣ and Q⊥ are the light collection efficiencies of the microscope objective for dipoles parallel and perpendicular to the glass surface . The excitation field intensity in y^ and z^ were assumed to be the same , though this was an approximation since in reality p^ had a small component of s^ ( Sund et al . , 1999 ) . In addition , in TIR the excitation field decayed exponentially with distance from the glass surface , Ey^ or y^=Ee−z/2d=Ee− ( 1−cos ( θ ) ) /2d where the characteristic decay constant was defined in terms of a fraction of the radius of the VLP . The collection efficiency for emission parallel Q∣∣ versus perpendicular Q⊥were also assumed to be the same ( Q∣∣=Q⊥ ) , which was an approximation based on the use of a high numerical aperture objective ( Anantharam et al . , 2010 ) . From coordinate transforms described previously ( Anantharam et al . , 2010; Sund et al . , 1999 ) , the components of the dipoles in y^ and z^ are given by: ( 3 ) μy^=ρ ( β ) [cos ( θ ) sin ( ϕ ) sin ( β ) cos ( ψ ) +cos ( ϕ ) sin ( ψ ) sin ( β ) +sin ( θ ) sin ( ϕ ) cos ( β ) ] ( 4 ) μz^=ρ ( β ) [−sin ( θ ) sin ( β ) cos ( ψ ) +cos ( θ ) cos ( β ) ] P/SVLP ( θ , ρ ( β ) ) was solved computationally ( Mathematica , Wolfram ) with θ parameterized in terms of normalized surface area ( A:0→1 ) as θ=cos−1 ( 1−2A ) . In addition , contribution from fluorescence outside of the puncta was accounted for as follows: ( 5 ) P/S ( A , ρ ( A ) , Cback ) =A⋅P/SVLP ( A , ρ ( A ) ) +Cback⋅P/SBackgroundA+Cbackwhere Cback was the background intensity relative to final VLP intensity and P/SBackground was the ratio of fluorescence when puncta A=0 , that is P/SBackground=P/SVLP ( 0 , ρ ( 0 ) ) . We found an angle β=45∘ , Cback=0 . 45 , and d=1 approximately replicated the observed results ( Figure 4A and F ) , reproducing the t1/2 of P+2S−t1/2 of P/S of 0 . 16 ( normalized time ) ( Figure 4B ) . In Figure 4A an exponential fit to P+2S was assumed in order to directly equate the VLP area to the predicted P/S . Similar results were obtained by using a uniform distribution of dipoles over an angular range , such as β between 0° to 68 . 5° or 20° to 63 . 5° . Thus , our observation is consistent with formation of a spherical bud throughout the recruitment and multimerization of Gag ( Woodward et al . , 2015; Carlson et al . , 2008 ) , and is independent of the presence of early acting factors like ESCRT-I/TSG101 or ALIX ( Figure 4—figure supplement 2 ) . The bending also occurs many minutes before the recruitment of ESCRT-IIIs . In retroviral assembly , the late domains of Gag are believed to indirectly recruit ESCRT-IIIs which then polymerize into ring or spiral structures at the bud neck to drive bending and scission of the neck membrane ( Cashikar et al . , 2014; Fabrikant et al . , 2009; Hanson et al . , 2008 ) . However , our observations compel a new formulation of the role of ESCRTs in bending and scission . First , the initial bending of the membrane from a flat sheet to a spherical bud occurs during Gag assembly and does not require factors recruited by the late domains like ESCRT-1/TSG101 or ALIX . This bending also occurs prior to the arrival of ESCRT-IIIs and thus ESCRT-IIIs are also not required to initiate curvature . Instead , this process may be encouraged by Gag multimerization ( Briggs et al . , 2009; Wright et al . , 2007 ) . Second , the ESCRT-IIIs are only recruited after the accumulation of Gag is complete , indicating that Gag is not sufficient for recruitment . Third , the ESCRT-III/VPS4 are only recruited for tens of seconds and can no longer be detected at the moment of scission . While the ESCRTs and VPS4 cannot be detected in the tens of seconds prior to scission , it does not eliminate the possibility that a subpopulation of fewer than 20% of the ESCRT-III/VPS4 , which are not detectable in the background fluctuations , stay around for a longer period of time ( further discussed in Materials and methods ) . Factors beyond the late domains which might assist ESCRT-III recruitment include activation through ubiquitination or phosphorylation of Gag or one of the early ESCRTs , such as ESCRT-II ( Meng et al . , 2015 ) or ESCRT-I/TSG101 , which is co-recruited along with Gag ( Bleck et al . , 2014; Jouvenet et al . , 2011 ) . Additionally , ESCRT-III recruitment might be facilitated by high curvature in the neck or specific lipids recruited to these regions ( Lee et al . , 2015 ) . These conditions may also facilitate recruitment of ESCRT-IIIs even in the absence of ESCRT-I or ESCRT-II , although likely at lower rates , potentially accounting for the much slower viral particle release rates ( Bendjennat and Saffarian , 2016; Meng et al . , 2015 ) . The observation that the bulk of measured curvature of the nascent virion occurs prior to the recruitment of ESCRT-IIIs does not rule out any role for ESCRTs in membrane curvature . Indeed , the transient recruitment of ESCRT-IIIs may further constrict the neck linking the virion to the cell to prepare it for scission . The ESCRT-IIIs might facilitate scission by polymerizing into a spiral structure to constrict the neck ( ‘polymerization constriction’ ) ( Cashikar et al . , 2014; Wollert et al . , 2009 ) or into a ring which when removed constricts the neck ( ‘purse string constriction’ , Figure 5 ) ( Saksena et al . , 2009 ) . The VPS4 may remove the ESCRT-IIIs potentially acting as a unfoldase ( Yang et al . , 2015 ) . VPS4 may also remodel the ESCRT-IIIs throughout polymerization , potentially rearranging or tightening the structure ( Cashikar et al . , 2014 ) . Remodeling is consistent with our observation that recruitment of VPS4 was virtually contemporaneous with the ESCRT-III , with a slight 5 s lag in HEK293T cells and 10 s in HeLa cells ( Bleck et al . , 2014 ) . We suggest that fission can only occur when the neck is narrow and after the ESCRT-IIIs are removed ( Figure 5 ) . What is driving the scission event if ESCRT-IIIs are gone ? When the neck is sufficiently narrow ( a few nm ) fission may be a spontaneous event , possibly through a hemifission intermediate ( Fabrikant et al . , 2009; Kozlovsky and Kozlov , 2003; Liu et al . , 2006 ) . Narrowing of the neck by ESCRT-IIIs may allow additional Gag molecules adjacent to the neck to oligomerize , thereby keeping the constricted structure stable while the ESCRTs are displaced . However , cryo-EM images indicate a gap in the Gag lattice might be present at the site of the neck ( Carlson et al . , 2008 ) . Alternatively , constriction might encourage exchange or modification of phospholipids with shapes and charges that help to retain a narrow neck following ESCRT-III disappearance . The ESCRT-IIIs are dependent on phospholipids with negative charges . This could explain the greater than three fold increased scission rate at a lower pH ( 10% pCO2 ) , which would raise the proton concentration to that of the pKa ( 6 . 5 and 7 . 7 ) ( van Paridon et al . , 1986 ) , thereby reducing the surface charges ( Figure 2F ) . If scission does not occur sufficiently soon after removal of ESCRT-IIIs , a new round of ESCRT-III recruitment and assembly is required to achieve scission ( Figure 3 ) . Plasmids Gag-mEGFP , Gag-mTagBFP , pLNCX2-mEGFP-VPS4A and pLNCX2-mCherry-CHMP4B , Gag-∆p6 were described in Bleck et al . ( 2014 ) , and the plasmids pCR3 . 1/Syn-Gag and pCR3 . 1/Syn-Gag-pHluorin were described in Jouvenet et al . ( 2008 ) . The Gag in this study was based on the sequence from HIV-1 clone HXB2 ( Kotsopoulou et al . , 2000 ) . pLNCX2-mCherry-CHMP2A-siRNAres and pLNCX2-mCherry-CHMP2B-siRNAres were generated as follows . CHMP2A and CHMP2B were PCR amplified ( Platinum PCR SuperMix , Thermo Fisher ) from a cDNA library created from HEK293T cells ( made via Invitrogen SuperScript III CellDirect kit #46–6320 ) . PCR primers for CHMP2A ( NM_014453 . 3 ) were 5’-GCGCTCCGGACTCAGATCCCCGGAATTCATGGACCTATTGTTCGGGCG-3’ and 5’-GCGCCTCGAGTCAGTCCCTCCGCAGGTTCT-3’ and primers for CHMP2B ( NM_014043 . 3 ) were 5’-GCGCTCCGGACTCAGATCCCCGGAATTCATGGCGTCCCTCTTCAAGAA-3’ and 5’-GCGCCTCGAGCTAATCTACTCCTAAAGCCT-3’ . After PCR amplification , the fragments were digested with XhoI and BspEI ( New England BioLabs , Ipswich , MA ) and ligated into the plasmid pLNCX2-mCherry-CHMP4B ( which was first digested with the same restriction enzymes to remove CHMP4B ) using T4 DNA Ligase ( New England BioLabs ) yielding the plasmids pLNCX2-mCherry-CHMP2A and pLNCX2-mCherry-CHMP2B . Six silent coding mutations were then incorporated into these plasmids ( QuickChange Lightning Site-Directed Mutagenesis Kit , Agilent , Santa Clara , CA ) to make them insensitive to siRNA knockdown targeted at the cellular CHMP2A and CHMP2B RNA . The primers for the CHMP2A site-directed mutagenesis were 5'-ACCTGGGACACCACAGCATCGCTTTCTTCCTCGTCCTCCTCATCACCCATGGCATC-3’ and 5'-GATGCCATGGGTGATGAGGAGGACGAGGAAGAAAGCGATGCTGTGGTGTCCCAGGT-3’ and for CHMP2B were 5’-AACCGTGGATGATGTAATAAAGGAGCAAAACCGTGAATTACGAGGTACACAGAGGGCTAT-3’ and 5’-ATAGCCCTCTGTGTACCTCGTAATTCACGGTTTTGCTCCTTTATTACATCATCCACGGTT-3’ . The siRNA sequence for CHMP2A knockdown was 5’-rArArGrArUrGrArArGrArGrGrArGrArGrUrGrAdTdT-3’ ( begins at position 464 in NM_014453 . 2 ) and for CHMP2B was 5’-rGrGrArArCrArGrArArUrCrGrArGrArGrUrUrAdTdT-3’ ( begins at position 45 in NM_014043 . 3 ) ( Morita et al . , 2011 ) . The forward and reverse DsiRNA for CHMP2A knockdown was 5'-rArArGrArUrGrArArGrArGrGrArGrArGrUrGrArUrGrCrUdGdT-3' and 5'-rArCrArGrCrArUrCrArCrUrCrUrCrCrUrCrUrUrCrArUrCrUrUrCrC-3' . The forward and reverse DsiRNA for CHMP2B knockdown was 5'-rGrGrArArCrArGrArArUrCrGrArGrArGrUrUrArCrGrArGdGdT-3' and 5'-rArCrCrUrCrGrUrArArCrUrCrUrCrGrArUrUrCrUrGrUrUrCrCrUrU-3' . All ssDNA and ssRNA oligos were purchased from Integrated DNA Technologies ( Coralville , IA ) and manufacturers’ protocols were used for all preparations . pLNCX2-mEGFP-CHMP2A and pLNCX2-mEGFP-CHMP2B were generated by replacing mCherry in pLNCX-mCherry-CHMP2A and pLNCX-mCherry-CHMP2B with monomeric variant ( A206K ) of mEGFP from pEGFP-N1 ( Clontech/Takara Bio USA , Mountain View , CA ) . Restriction enzymes AgeI and BspEI ( New England BioLabs ) were used to digest the backbone and fragments , and these fragments were then ligated into the backbones with T4 DNA ligase . The clonal HeLa cell lines were generated from these plasmids using a previously described protocol ( Bleck et al . , 2014 ) . pLNCX2-mCherry-VPS4A was generated by replacing mEGFP in pLNCX2-mCherry-VPS4A ( Bleck et al . , 2014 ) with mCherry . mCherry was PCR amplified from pmCherry-N1 ( Clontech ) using In-Fusion recombination primers 5'-CTCTAGCGCTACCGGTCGCCACCATGGTGAGCAAGGGC-3’ and 5'-CTCTAGCGCTACCGGTCGCCACCATGGTGAGCAAGGGC-3’ . pLNCX2-mEGFP-VPS4A was digested with AgeI and BspEI and the fragment containing VPS4A was gel purified ( Thermo Fisher PureLink Quick Gel Extraction Kit ) . The PCR product was then inserted into purified backbone using In-Fusion HD Cloning Kit ( Clontech ) according to manufacturer’s instructions . Gag-∆p6-mEGFP was generated by deleting p6 and SP2 from Gag-mEGFP via site-directed mutagenesis ( QuickChange Lightning Site-Directed Mutagenesis Kit ) . The following primers were used for the deletion 5'- TACTGAGAGACAGGCTAATTCGGATCCACCGGT-3'; 5'- ACCGGTGGATCCGAATTAGCCTGTCTCTCAGTA −3' . Gag-MA-sf3 and Gag-MA-sf11 were generated by inserting the circularly permuted superfolder GFPs ( provided by Jeffrey Waldo lab ) ( Pédelacq et al . , 2006 ) into a variant of pCR3 . 1/Syn-Gag that has an EcoRV restriction enzyme site near the carboxy terminal of MA ( Asn-Gln-Val-Ser modified to Asn-Gln-Asp-Ile-Val-Ser ) . Circularly permuted version 3 was PCR amplified with recombination In-Fusion primers 5’-CACAGCAACCAGGATGGCAGCAGCCATCATCATC-3’ and 5’-GTTCTGGCTGACGATGGTACCTCCAGTAGTGCAAATAA-3’ . The primers for circularly permuted version 11 were 5’-CACAGCAACCAGGATGGCAGCAGCCATCATCATC-3’ and 5’-GTTCTGGCTGACGATGGTACCATCTTCAATGTTGTGG-3’ . After digesting pCR3 . 1/Syn-Gag-EcoRV with EcoRV ( New England BioLabs ) the In-Fusion HD kit was used to insert PCR fragments into the Syn-Gag backbone . With the exception of 120 s pCO2 switching data , all imagings were conducted in HEK293T cells grown in DMEM ( #11965 , Thermo Fisher Scientific , Waltham , MA ) with 10% FBS ( #F4135 , MilliporeSigma , St . Louis , MO ) . HEK293T cells were gift from P . Bieniasz Laboratory and were not further authenticated but tested negative for mycoplasma contamination . 120 s pCO2 switching experiments imaged in HeLa cells which were grown in the same growth medium . HeLa cells were from ATCC and were not authenticated or tested for mycoplasma contamination . For polarization excitation experiments , cells were grown on 35-mm glass bottom dishes ( #P35G-1 . 0–20 C , MatTek , Ashland , MA ) coated with fibronectin ( #33010 , Thermo Fisher ) by incubating the dish with 10 µg/ml fibronectin in PBS for 1 hr . For pCO2 switching experiments , perfusion slides ( #80186 µ-Slide , Ibidi , Martinsried , Germany ) were incubated with 60 µg/ml fibronectin in PBS for 1 hr before plating cells . For all experiments , cells at ~ 75% confluency were transfected with expression plasmids ~3 . 5 hr prior to beginning to image . For transfection 8 µl of Lipofectamine2000 ( #11668 , Thermo Fisher ) was incubated for 5 min in 250 µl of Opti-MEM I ( #31985 , Thermo Fisher ) and 2000 µg of DNA was incubated for 5 min in 250 µl of Opti-MEM I . Both solutions were then mixed and incubated for 20 min before adding to adhered cells on MatTek dish in 2000 µl of DMEM . The same procedure was used for cells in the flow slide , but the transfection mixture was added to 2000 µl of DMEM and then 1000 µl was perfused through the chamber . At least four cells were used for each experimental condition in order to account for potential variability between cells . Both CHMP2A and CHMP2B were knocked down with siRNA for mCherry-CHMP2A and mCherry-CHMP2B experiments . siRNA transfections were conducted 48 hr prior to DNA plasmid transfection using Lipofectamine RNAiMAX ( following manufacturer’s instructions; #13778 , Thermo Fisher ) . A second round of siRNA transfection was performed at the time of DNA transfection . The open-reading frame was verified via sequencing for all plasmids and the following DNA ratios were used for transfections: Gag:Gag-mEGFP ( 4:1 ) , Gag:Gag-MA-sf3 ( 4:1 ) , Gag:Gag-MA-sf3 ( 4:1 ) , Gag-∆p6:Gag-∆p6-mEGFP ( 4:1 ) , Gag:Gag-pHluorin:mCherry-VPS4A ( 12:3:5 ) , Gag:Gag-pHluorin:mCherry-CHMP4B ( 12:3:5 ) , Gag:Gag-pHluorin:mCherry-CHMP2A-siRNAres ( 4:1:5 ) , Gag:Gag-pHluorin:mCherry-CHMP2B-siRNAres ( 4:1:5 ) , Gag:Gag-TagBFP:mEGFP-VPS4A:mCherry-CHMP4B ( 24:6:5:5 ) . Note: The Gag to tagged Gag ratio was 4:1 in all experiments . Gas from compressed cylinders was bubbled into two imaging media reservoirs ( 140 ml open piston Monoject syringe , Medtronic , Minneapolis , MN ) partially filled with cell imaging media ( 10 mM HEPES , 9 . 7 g/L of Hanks BBS ( MilliporeSigma ) , and NaOH to adjust the pH to 7 . 4 ) with 1% FBS . One reservoir was equilibrated with compressed air ( labeled 0% in these experiments but actually contained 0 . 04% CO2 ) , and the other reservoir was equilibrated with 10% CO2 ( balanced with air ) ( Figure 1—figure supplement 1 ) . The reservoirs were mounted to a flow perfusion system ( ValveBank II , AutoMate Scientific , Berkeley , CA ) , which enabled automated selection of desired media via solenoid valves under the reservoirs . Tygon tubes ( R-3603 ) from each valve carried the media to a fluid combiner just prior to the perfusion slide chamber containing adhered cells . After the flow chamber , a single tube carried the discharge media to a collection container . This container was placed below the height of equilibration reservoirs so that fluid flow was driven by gravity . The flow rate ( ~3 ml/min ) was controlled with a clamp regulator attached to the discharge tube . At this flow rate , the response time of the fluorophores to a pCO2 change , characterized in terms of an exponential decay constant , was ~7 s . A peristaltic pump ( MS-Reglo , Ismatec/Cole-Parmer , Wertheim , Germany ) then passed the media from the collection reservoir back to the equilibration reservoirs so that the media could be recycled . The microscope and entire imaging media flow system were enclosed in a temperature control box held at 37°C . Valve regulation , camera trigger and laser excitation were all controlled via custom software ( https://github . com/SimonLab-RU/Microscope-Control; copy archived at https://github . com/elifesciences-publications/Microscope-Control ) written in LabView ( National Instruments , Austin , TX ) . For experiments with pCO2 switching every 10 s the media was continuously flowed through the imaging chamber , with the reservoir supplying the media being switched every 10 s . Two images were collected every 2 . 5 s ( 4 . 0 s for CHMP4B experiments ) with 488 nm excitation for pHluorin ( 100 ms exposure with power between 1 and 5 mW ) , followed by 594 nm excitation for mCherry ( 100 ms exposure with laser power between 5 and 20 mW; 100 mW DPSS laser , Cobolt , Solna , Sweden ) . A multipass emission filter ( zet405/488/594 m , Chroma Technology , Bellows Falls , VT ) enabled rapid sequential wavelength imaging . For experiments with pCO2 switching every 120 s the desired media was only applied to the chamber for 10 s , followed by no flow for 110 s . During this time sequential 488 nm ( 100 ms , 1 mW ) and 594 nm ( 100 ms , 1 mW ) excitation images were captured every 10 s . All experiments were conducted with 100 Hz azimuthal scanning TIRF microscopy illumination . The cytosolic pH was oscillated by switching the pCO2 from 0% to 10% ( for an average of 5% ) every 10 s or every 120 s . The cytosolic carbonic anhydrase ensures that the pH in the cytosol closely tracks the pCO2 ( Simon et al . , 1994 ) . At scission , the luminal pH of the VLP or virion is no longer continuous with the cytosol and no longer tracks the cytosolic pH . Individual VLPs were identified from the Gag-pHluorin images using Metamorph ( version 7 . 8 . 10 , Molecular Devices , Sunnyvale , CA ) and the peak pixel amplitude in a 15 pixel ( 975 nm ) diameter regions of interests centered on individual VLPs was found for all frames ( Figure 1 , Figure 1—figure supplement 2 ) . For switching every 10 s , we used a custom LabView software lock-in amplifier to find changes in VLP sensitivity to pCO2 modulation ( https://github . com/SimonLab-RU/CO2-switch-analysis; copy archived at https://github . com/elifesciences-publications/CO2-switch-analysis ) . The data were high-pass filtered ( 2-pole Butterworth with 0 . 01 Hz cutoff ) to bias the pCO2-dependent pHluorin intensity fluctuations ( with a period of 20 s ) around 0 . This signal was then multiplied by an in-phase sine wave with the same period and a moving average was calculated over a period of 60 s . A significant change in signal indicated a change in sensitivity to pCO2 modulation . At many of the identified VLPs , a clear transition from a high plateau to a low plateau in the lock-in signal was observed ( Figure 1—figure supplement 2 ) . The moment of scission was classified as the halfway amplitude between the two plateaus , with half of the moving window containing pre-scission data and the other half containing post-scission data ( Figure 1—figure supplement 2A ) . In the same regions of interest , the average mCherry signal ( tagged to VPS4A , CHMP4B , CHMP2A , or CHMP2B ) was analyzed and peaks were identified in which there was a clear increase followed by decrease in fluorescence intensity ( Figure 1A , C , E , G and Figure 1—figure supplement 2 ) . The appearance time was identified when a signal was first observed above cellular background , and the disappearance time when the signal dropped to cellular background . Appearance and disappearance relative to scission were then found by comparing these times to the scission time ( Figure 1B , D , F , H ) . Based on the observed distribution of appearance and disappearance times relative to scission , with an average of roughly a minute , we excluded data in which scission and ESCRT-III/VPS4A recruitment were more than 4 min apart . Excluded data was attribute to a failed or uncorrelated scission event . ~25 traces under each condition were collected to gain understanding of the distribution of events . The average signal-to-noise was approximately 7:1 ( peak signal:S . D . ) across all ESCRT-III/VPS4A measurements , and ESCRT-III/VPS4A recruitment was estimated to be detectable when the sustained signal deviated from the mean by ~ 1–2 standard deviations . Based on this deviation , it is estimated that ESCRT-III/VPS4A recruitment below ~20% of peak recruitment would be undetected . We estimate that there are between 10 and 100 ESCRTs in a complex at peak signal , with the possibility of < 10 ESCRTS undetectable in the background noise . Assuming a peak signal ( 100 ESCRTs ) decreases exponentially into the noise after 10 s ( 10 ESCRTS ) , we estimate only ~1 ESCRTs would remain after another 10 s . This time is comparable to the ~20 s measured between peak disappearance and scission . If the disappearance is faster than exponential , for example linear , the ESCRTs will be gone even sooner . This interpolation of ESCRT-IIIs or VPS4A disappearance profiles into the noise indicates the ESCRTs predominantly leave the budding site prior to scission ( Figure 1—figure supplement 2 ) . For a significant number ( >1 ) of either ESCRT-IIIs or VPS4A to be around following scission there would need to be a second , smaller population that follows much slower disappearance kinetics . For pCO2 switching every 120 s for each tracing we determined the first transition of pCO2 for which the fluorescence had decreased sensitivity , indicating protons were no long freely flowing between cytosol and lumen of the VLP . We assumed that scission must have occurred during the previous plateau of pCO2 ( Figure 2A–E ) . For example , if scission occurred during the 10% pCO2 cycle ( low fluorescence ) then after the next transition to 0% pCO2 , and all subsequent transitions , the fluorescence would be closer to that of the 10% ( low fluorescence state ) than the 0% pCO2 ( high fluorescence ) state . Conversely , if scission occurred during the 0% ( high fluorescence ) , future 10% pCO2 plateaus would be closer to the high fluorescence state . Occasionally traces appeared to be trapped in an intermediate pH state , which we attribute to scission occurring during the transition between pCO2 states . Individual VLP traces were categorized into being trapped in a high , middle , or low pH state by comparing the Gag-pHLuorin signal before and after the VLP became insensitive to pCO2 switching ( Figure 2F ) . Gag-mTagBFP , mEGFP-VPS4A and mCherry-CHMP4B were imaged with TIRF with sequential illumination of 594 nm ( 100 ms , 10–20 mW ) , 488 nm ( 100 ms , 4–10 mW ) and 405 nm ( 100 ms , 2–4 mW; 120 mW LuxX diode laser , Omicron ) . mEGFP-VPS4A and mCherry-CHMP4B images were acquired every 5 s and Gag-TagBFP images were acquired every 60 s . VPS4A and CHMP4B relative appearance time were conducted using the same method as described previously ( Figure 1—figure supplement 4 ) ( Bleck et al . , 2014 ) . HeLa cell lines stably expressing either mEGFP-CHMP2A or mEGFP-CHMP2B and transfected with the respective siRNA ( 48 hr in advance ) were imaged on an inverted microscope ( IX-70 , Olympus , Shinjuku , Japan ) with epi-illumination via a Xenon lamp and transmitted bright-field illumination ( Figure 1—figure supplement 3 ) . Images were collected on an inverted microscope ( IX-81 , Olympus ) with a custom built through-the-objective ( 100X UAPON 1 . 49 NA , Olympus ) polarized TIRF microscopy illuminator ( Johnson et al . , 2014 ) . Throughout imaging the excitation TIR light was azimuthally scanned at 200 Hz with mirror galvanometers ( Nutfield Technology , Hudson , NH ) in order to reduce spatial illumination nonuniformities . A multiband polychroic ( zt405/488/594/647rpc 2 mm substrate , Chroma ) was positioned between the galvanometers and objective in order to isolate the excitation light from the emitted light . Light from a 488 nm laser ( 100 mW LuxX diode laser , Omicron , Rodgau-Dudenhofen , Germany ) was modulated between being polarized perpendicular ( p^ ) or parallel ( s^ ) to the glass surface by passing the light through an electro-optic modulator ( EOM ) ( Conoptics , Danbury , CT ) and quarter-wave plate prior to the galvanometer scan-head . The polarization generated by the EOM was modulated in sync with the galvanometers such that during scanning a p^ or s^ state was maintained at all azimuthal positions . The scanning polar angle was selected such that the excitation light was just beyond the TIR critical angle , minimizing s^ polarized light contaminating p^ excitation ( Johnson et al . , 2014 ) . A combined p^/s^ ratio image was collected every 30 s . To generate this ratio image a sequential series of 10 p^ and s^ images were collected , divided ( after subtracting camera offset ) , and then averaged . Each p^ or s^ image had an exposure of 5 ms ( laser power between 25 and 50 mW ) , with a new image collected every 15 ms . Thus , in 30 ms , a single p^/s^ image was generated , and then 10 of these images ( over a 300 ms duration ) were averaged ( ImageJ ) ( Schneider et al . , 2012 ) to create the combined ratio image . The short period was utilized in order to minimize artifacts in the p^/s^ ratio image from VLP or cell movement . The galvonometers , EOM , camera shutter , and laser shutters were all driven/triggered by a multifunction data acquisition board ( PCIe-6323 , National Instruments ) and controlled from custom written software in LabView ( https://github . com/SimonLab-RU/Microscope-Control; copy archived at https://github . com/elifesciences-publications/Microscope-Control ) . Images were streamed from a CMOS camera ( Flash-4 . 0 , Hamamatsu , Hamamatsu City , Japan ) to a workstation ( T7500 , Dell , Round Rock , TX ) running image acquisition software ( Metamorph ) . A single band emission filter ( ET525/50 m , Chroma ) was used to isolate fluorophore emission . In order to characterize the amount of Gag in an assembling VLP , an average p^+2⋅s^ image was also generated at each time point . Puncta were found in the p^+2⋅s^ images which increased and then plateaued in amplitude . These puncta were then selected for orientation analysis with p^/s^ . A 2D Gaussian was fit to these assembling puncta to find a frame by frame subpixel peak location ( https://github . com/SimonLab-RU/Puncta-Fit; copy archived at https://github . com/elifesciences-publications/Puncta-Fit ) . Using bilinear interpolation , the p^/s^ and p^+2⋅s^ images were resampled 10X in the horizontal and vertical directions ( 65 to 6 . 5 nm wide pixels ) . On these resampled images P/S and P+2S values were found by averaging the resampled pixel intensities that are within 100 nm of the peak fit locations ( https://github . com/SimonLab-RU/Average-puncta-center; copy archived at https://github . com/elifesciences-publications/Average-puncta-center ) ( Figure 4A , Figure 4—figure supplement 1 ) . For frames prior to the appearance of a puncta P/S and P+2S values were determined using the fit location of the first frame in which there was a puncta fit . For each tagged version of Gag , the average P/S value for background and assembled VLPs was calculated by finding an average P/S for each trace before assembly and after a plateau was reached , and then averaging across all traces ( Figure 4B ) . A relative time to half growth for each VLP was determined by finding the halfway to assembly point ( i . e . the point where the intensity is halfway between the intensity at assembly beginning and plateau ) and then finding the normalized time at this point relative to the time assembly began and appeared to reach a plateau . This normalized time was between 0 and 1 . A normalized time to halfway drop in P/S was also found ( normalized to the same time scale ) and these values were subtracted to find: t1/2 of P+2S−t1/2 of P/S ( normalized ) . All fluorophore combinations were included in the P/S histogram ( Figure 4C ) since all combinations had similar P/S characteristics .	Viruses need to be able to enter and leave the cells of their hosts to multiply and spread infections . Once inside the cell , many viruses , including the HIV virus , hijack the cell’s genetic material to produce more HIV particles and release them back into the surroundings . As the new viruses leave the cell , they wrap themselves in the membrane of their host cell . The shell of HIV consists of thousands of copies of a protein called Gag , which helps to release the viruses . Gag aggregates inside of the host cell membrane , which then begins to bulge outwards forming a spherical bud that subsequently pinches off . These released virus particles are now able to infect other cells . Both assembling and budding of the virus particles requires the help of specific cell proteins called ESCRTs . Cells usually use ESCRTs to cut membrane-bound compartments during cell division . HIV can control ESCRTs to separate the bud from the cell , but where and how they are involved is still not fully understood . Some models propose that the ESCRTs bend the membrane , while others suggest that the ESCRTs are required during the split , providing the force for the cutting process . Now , Johnson et al . used fluorescent light to follow how individual viruses in human cells grown in the laboratory assemble , and polarized light to detect how the orientation of Gag changed as the membrane bends into a sphere . The experiments revealed that the virus particles started to form a sphere at the same time Gag began to gather at the cell membrane , before any ESCRTs were present . Once enough Gag proteins had accumulated , the ESCRTs were recruited to the spot and removed within tens of seconds , just before the virus was cut off from the host cell . This suggest that the ESCRTs may help prepare the membrane to split , but are not involved in the actual cutting process . The findings of Johnson et al . will help us to better understand the behavior of HIV viruses . Knowing how they assemble and leave the host cell may help to create new strategies for disrupting these processes in particular , so stopping the virus from spreading . Moreover , ESCRTs are involved in numerous activities of a cell , such as cell division or repair of the cell surface after rupture , but little is known about how they work . For the first time , it has been shown that ESCRTs are neither the driving force for bending the membrane nor are they present when the cutting of a virus happens . This may also apply to other processes in the cell that involve membrane-splitting by ESCRT proteins , and researchers may be compelled to reevaluate their proposed role in other systems .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "microbiology", "and", "infectious", "disease" ]	2018	Timing of ESCRT-III protein recruitment and membrane scission during HIV-1 assembly
Cytotoxic T lymphocytes ( CTLs ) are thought to arrive at target sites either via random search or following signals by other leukocytes . Here , we reveal independent emergent behaviour in CTL populations attacking tumour masses . Primary murine CTLs coordinate their migration in a process reminiscent of the swarming observed in neutrophils . CTLs engaging cognate targets accelerate the recruitment of distant T cells through long-range homotypic signalling , in part mediated via the diffusion of chemokines CCL3 and CCL4 . Newly arriving CTLs augment the chemotactic signal , further accelerating mass recruitment in a positive feedback loop . Activated effector human T cells and chimeric antigen receptor ( CAR ) T cells similarly employ intra-population signalling to drive rapid convergence . Thus , CTLs recognising a cognate target can induce a localised mass response by amplifying the direct recruitment of additional T cells independently of other leukocytes . Cytotoxic T lymphocytes ( CTLs ) constitutively migrate as single cells in search of infected or malignant cells ( Weninger et al . , 2014 ) . CTLs are key effectors of adoptive cell transfer immunotherapies ( Guedan et al . , 2019 ) , but their efficacy remains limited with solid tumours ( van der Woude et al . , 2017 ) , which they infiltrate in insufficient numbers ( Galon et al . , 2006 ) . Thus far , CTLs have been thought to arrive at target sites either via random search ( Krummel et al . , 2016 ) or following signals by other cell types ( Feig et al . , 2013; Harlin et al . , 2009 ) . Intravital imaging studies have been instrumental in uncovering some of the complex migration patterns of CTLs at various stages of an immune response . They have revealed , for instance , that CTLs employ highly evolved cell-intrinsic search strategies that are more efficient than Brownian motion ( Harris et al . , 2012 ) . During priming in lymph nodes , CD8+ T cells can follow local chemoattractant gradients to migrate directionally towards sites of antigen presentation ( Castellino et al . , 2006; Hickman et al . , 2011; Hugues et al . , 2007 ) . In the tumour microenvironment ( TME ) , the movements of CTLs and their interplay with various cells have been unveiled ( Boissonnas et al . , 2007; Deguine et al . , 2010 ) . Recent studies indicate that lymphocytes can recruit each other indirectly into tumours; natural killer ( NK ) lymphocytes produce chemokines that attract dendritic cells ( Böttcher et al . , 2018 ) , which in turn can recruit CTLs ( Spranger et al . , 2017 ) . Although intravital microscopy enables imaging of cellular interactions in the TME in situ , it is typically restricted to relatively short imaging periods and sub-millimetre fields of view ( Gabriel et al . , 2018 ) . Furthermore , the inherent complexities of the TME , its constituent cell populations and its biochemical landscape have limited our ability to uncover the contribution of an individual immune subset or signalling mechanism to progressive , large-scale phenomena . Here , we developed a 3D tumouroid model and in silico simulations to reveal independent collective behaviour in CTL populations attacking tumour masses . We show that CTLs coordinate their migration in a process reminiscent of the swarming observed in insects ( Avitabile et al . , 1975 ) and neutrophils ( Lämmermann et al . , 2013 ) . CTLs engaging tumour targets induce rapid chemotaxis in distant T cells through homotypic chemokine signalling . Newly arriving CTLs augment the chemotactic signal , further accelerating mass recruitment in a positive feedback loop . Furthermore , we show that local chemokine delivery triggers directed CTL movement through dense tumour tissue in vivo , and sustained secretion of CCL3 and CCL4 from tumours promotes CTL recruitment . Human effector and chimeric antigen receptor ( CAR ) T cells similarly employ intra-population signalling to drive rapid convergence . Our findings provide insights into how CTL populations amplify directed recruitment to an effector site independently of other leukocytes . We sought to investigate the population-wide movements and signals mediating interactions between CTLs during tumour clearance . To this end , we developed an ex vivo model enabling us to study the large-scale movements of primary CTLs around solid tumouroids embedded in three-dimensional ( 3D ) collagen matrices ( Figure 1A ) . We used primary murine CTLs isolated from OT1 ( Hogquist et al . , 1994 ) and gBT1 ( Coles et al . , 2003 ) T cell receptor transgenic mice that recognise ovalbumin ( SIINFEKL ) or herpes simplex virus glycoprotein B ( SSIEFARL ) residues , respectively , both in the context of the H-2Kb class I major histocompatibility complex . CTLs engaging a cognate tumouroid were rapidly recruited to its edge , where they accumulated over time ( Figure 1B and Video 1 ) . The marked accumulation of CTLs at the edge of the tumouroid is reminiscent of swarming , the coordinated collective convergence of large numbers of self-propelled individuals , which has been observed in as varied biological systems ( Okubo , 1986 ) as mammals , birds , fish , insects , and , at the cellular level , neutrophils ( Lämmermann et al . , 2013 ) . This behaviour was not observed with control tumouroids lacking cognate antigen ( Figure 1B and Video 2 ) . Cell movements were quantified using a novel swarming index ‘M’ ( see Materials and methods; Figure 1C ) . When an additional CTL population was embedded in the cognate tumouroid where it actively eliminated tumour target cells ( Video 3 ) , the surrounding CTLs amassed at the tumouroid ( Figure 1B ) and infiltrated it to a greater extent ( Figure 1D , E and Video 2 ) . This enhanced infiltration was also observed into masses constituted of cognate antigen-coated beads , which , in contrast to tumour cells , do not progressively give way to additional physical space due to lysis by CTLs ( Figure 1—figure supplement 1A–E ) . In order to assess whether the observed accumulation of CTLs around tumour masses involves emergent behaviour or is merely a cumulative result of randomly scanning CTLs arresting upon cognate antigen recognition , we modelled and simulated our experiments as agent-based processes in silico . Simulated cell motility characteristics were sampled from experimental data by bootstrapping ( Materials and methods; Figure 1—figure supplement 1F , G ) . We simulated the spatiotemporal scales of our experiments , with a diffusive chemoattractive signal progressively amplified by agents arriving at the tumouroid and inducing a directional bias in distant agent movements . The introduction of a desensitisation threshold in chemokine concentration , above which agents revert to unbiased motion , was necessary to recapitulate our experimental observations ( Figure 1F–G , Figure 1—figure supplement 1H and Video 4 ) . To further characterise how CTLs are recruited to the tumour mass , we next used an assay enabling tracking of individual CTL movements at higher spatiotemporal resolution ( Video 5 ) relative to a tumouroid exposing a straight interface in a constant direction from adjacent CTLs ( Figure 2A ) . Using this approach , we found that surrounding CTLs migrate rapidly and highly directionally towards the edge of a cognate tumour mass ( Figure 2B ) , containing pre-embedded CTLs at 1:1 or 1:5 ratios with tumour cells , or in the complete absence of pre-embedded CTLs ( Figure 2B ) . By contrast , CTLs adjacent to control tumouroids exhibited no chemotaxis , albeit , interestingly , some chemokinesis ( enhanced speed ) ( Figure 2B ) . The above results indicate that a factor diffusing from the site of engagement leads to distal recruitment . Indeed , cell-free supernatant from the co-incubation of CTLs with cognate tumour cells ( cognate supernatant ) attracts other CTLs in a transmigration assay ( Figure 2—figure supplement 1A ) and thus contains the soluble chemoattractants mediating the recruitment ( Figure 2—figure supplement 1B , C ) . Interestingly , 10-fold dilutions of cognate supernatants also induced CTL transmigration ( Figure 2—figure supplement 1B ) , indicating that the CTL response to the soluble chemotactic factors is highly sensitive . To distinguish whether the factors originate from CTLs or lysed tumour cells , we replaced tumour cells with polystyrene beads coated with cognate antigen ( Figure 2C ) and found that recruitment of CTLs was preserved ( Figure 2C , D ) . CTLs engaging cognate targets therefore produce factors that induce distal recruitment via homotypic signalling . Combined , our experimental and simulation data reveal emergence in CTL mass recruitment , which is progressively amplified by the arrival of further CTLs that contribute to a diffusive homotypic signal . We next tested whether the homotypic signal extends to T cells of different antigen specificities . Tumouroids containing monospecific CTLs and their cognate tumour targets ( Figure 2—figure supplement 1D ) are equally proficient at recruiting T cells of the same ( Figure 2E ) , different ( Figure 2F ) and polyclonal ( Figure 2G ) specificity . No recruitment is observed towards tumouroids containing CTLs engaging non-cognate target cells . Therefore , whilst production of the homotypic signal requires recognition of specific cognate antigen , the signal recruits CTLs irrespective of the T cell receptor ( TCR ) they express . CTLs of unrelated specificity would however not be able to contribute to the amplification of the signal upon arrival at the tumour mass . Next , we studied the movements of CTLs up to 1 . 6 mm from tumouroids ( Figure 3—figure supplement 1A ) . At the onset of the experiment ( 0 hr ) , CTLs were distributed uniformly adjacent to both cognate and non-cognate tumouroids ( Figure 3A , B and Figure 3—figure supplement 1B , C ) . After 2 hr , the distribution of CTLs was biased towards the cognate tumouroid ( Figure 3A , B ) , whereas the CTL distribution adjacent to a control tumouroid remained unchanged ( Figure 3—figure supplement 1B , C ) . Cell tracking analysis revealed that even the most distant CTLs show directional movement towards the cognate tumouroid ( Figure 3C , D and Video 6 ) and reveals a moving wave of chemoattraction over time ( Figure 3E ) . Cells close to the tumouroid start moving towards it at early timepoints but lose some of their directionality over time ( Figure 3E ) . It has previously been shown that chemotactic responses can be lost under excess local concentrations of ligand ( Lim et al . , 2018 ) , which could account for the behaviour observed here . For cells furthest from the tumouroid , directional migration commences at later timepoints ( Figure 3E ) . No such pattern is observed with a control tumouroid , where cells at all distances exhibit undirected migration ( Figure 3C–E and Video 6 ) . These results indicate that a gradient of diffusing factors is established following cognate interactions between CTLs and tumour cells to induce the long-range chemoattraction of other T cells . To identify the molecular mediators of the chemoattraction , we first tested CTL migration in the presence of pertussis toxin ( PTX ) , an inhibitor of Gαi-protein-coupled receptors ( GPCRs ) . PTX inhibits the enhanced average speed and directional migration of CTLs towards cognate tumouroids ( Figure 4A ) , indicating that GPCR ligands mediate the chemoattractive signal . This result is consistent with previous work where the formation of small CTL clusters around malaria-infected hepatocytes required GPCR signalling ( Cockburn et al . , 2013 ) . Unlike neutrophils that employ lipid signalling via leukotriene B4 to swarm ( Lämmermann et al . , 2013 ) , this result suggests that chemokines , which are sensed by GPCRs , underpin homotypic signalling in CTLs . We then compared the transcriptomes of CTLs engaging cognate and non-cognate targets ( dataset available in Supplementary file 1 ) . Differentially expressed genes were filtered for secreted factors , among which the following GPCR ligands were identified: Chemokine ( C-C motif ) ligand 1 ( CCL1 ) , CCL3 , CCL4 , CCL9 , and X-C motif chemokine ligand 1 ( XCL1 ) ( Figure 4B ) ; their upregulation was further validated by quantitative reverse transcription PCR ( Figure 4—figure supplement 1A , B ) . Mass-spectrometry-based proteomics analysis of cognate versus non-cognate secretomes reveals that CCL1 , CCL3 , CCL4 , and XCL1 are more abundant in cognate supernatant ( Figure 4C and Supplementary file 2 ) , confirmed by quantitative detection assays ( Figure 4—figure supplement 1C ) . The recombinant chemokines CCL3 , CCL4 , CCL5 , and CXCL12 could attract CTLs in transwell migration assays ( Figure 4—figure supplement 1D ) , in each case effectively inhibited by the corresponding neutralising antibody ( Figure 4—figure supplement 1E ) . However , only CCL3 inhibition consistently disrupted CTL transmigration in transwell chambers towards cognate supernatants ( Figure 4—figure supplement 1F ) . Directional migration towards a cognate tumouroid in more physiological 3D matrices was abolished only when both CCL3 and CCL4 were blocked using neutralising antibodies ( Figure 4—figure supplement 1G ) , indicating that these two chemokines act redundantly in the homotypic attraction . Therefore , CCL3 and CCL4 constitute the diffusive homotypic signal that attracts distant CTLs during target engagement . Secreted CCL3 forms a gradient from the site of antigen-specific target engagement , visualised by in situ capture and staining ( Figure 4—figure supplement 2A–C ) . Strikingly , CTLs sustain their ability to secrete CCL3 and CCL4 ( Figure 4—figure supplement 1H ) , and to recruit distant T cells ( Figure 4—figure supplement 1I ) , even after disengagement from cognate targets . To confirm that CCL3 and CCL4 secretion are sufficient to induce chemoattraction in distant CTLs , we engineered tumour cells that constitutively secrete both chemokines ( Figure 4—figure supplement 3A , B ) , or CCL3 or CCL4 alone . Secreting tumouroids induced enhanced rapid directional motility in CTLs ( Figure 4D ) , swarming and infiltration ( Figure 4—figure supplement 3C–H ) . CTLs infiltrate CCL3/CCL4-secreting cognate tumouroids as efficiently as tumouroids within which CTLs are actively engaging cognate targets ( Figure 1D ) . In the absence of cognate antigen , CTLs do not stop at the edge of secreting tumouroids and thus infiltrate them deeper ( Figure 4—figure supplement 3E and H ) . We next established an in vivo model to investigate if CCL3/CCL4-secretion influences endogenous leukocyte recruitment to tumours engrafted in mice ( Figure 4—figure supplement 4 ) , and showed that CCL3/CCL4-secreting tumours consistently recruit more endogenous NK cells than contralateral control tumours ( Figure 4—figure supplement 4B ) . We next sought to identify the receptor mediating the chemoattractive signal . To this end , we tested the chemokine receptors CCR1 , CCR2 , CCR3 , and CCR5 that have been associated with CCL3 or CCL4 sensing , and which are all expressed by CTLs ( Figure 5—figure supplement 1A ) . Pharmacological inhibition of CCR1 , CCR2 , or CCR3 ( Figure 5—figure supplement 1B , C ) did not have any effect , whereas targeting CCR5 with Maraviroc or the CCR2/CCR5 dual antagonist Cenicriviroc abrogated the homotypic recruitment of CTLs ( Figure 5—figure supplement 1B , C ) . Furthermore , polyclonal T cells isolated from Ccr5 knockout mice ( Ccr5-/-; Figure 5—figure supplement 1D ) were not efficiently recruited towards cognate tumouroids ( Figure 5A ) or cognate supernatant ( Figure 5—figure supplement 1E ) . When wild type ( WT ) and Ccr5-/- CTLs were co-embedded around a cognate tumouroid , the Ccr5-/- population exhibited severely impaired swarming ( Figure 5B , C ) , and infiltration depth ( Figure 5D ) . Moreover , Ccr5-/- CTLs exhibited a reduced capacity to eliminate cognate tumour masses ( Figure 5—figure supplement 1F , G ) , despite fully functional antigen-specific cytotoxicity ( Figure 5—figure supplement 1H ) and chemokine secretion ( Figure 5—figure supplement 1I ) . These findings demonstrate that CCR5 mediates the chemokine signal underpinning the homotypic recruitment of T cells . In vivo , Ccr5-/- CTLs that were co-transferred with WT CTLs showed impaired homing into CCL3/CCL4-secreting tumours ( Figure 5—figure supplement 2A , B ) and , in contrast to WT CTLs , did not bind fluorescent CCL3 ( Figure 5—figure supplement 2C , D ) . We also observed impaired CCR5-dependent recruitment through dense tumour tissue in response to the acute injection of a molecular hydrogel enabling shear-reversible containment of recombinant CCL3 ( Nisbet et al . , 2018; Figure 5—figure supplement 3 , Video 7 ) . To determine if tumour-reactive CTLs can enhance the recruitment of additional CTLs into tumours , we engrafted Rag-/- mice with SSIEFARL-expressing tumours and delivered a primary transfer of tumour-reactive gBT1 CTLs ( cognate ) . 48 hr later , a secondary cohort of WT and Ccr5-/- OT1 CTLs were co-transferred ( both non-cognate for the tumours ) ( Figure 5E ) . The presence of tumour-reactive CTLs ( Figure 5—figure supplement 2E , F ) markedly increased the overall recruitment of all non-cognate CTLs into the tumours ( Figure 5F ) . Furthermore , recruitment of Ccr5-/- OT1 CTLs into tumours containing activated gBT1 CTLs was compromised compared to WT CTLs . However , Ccr5-/- CTLs cleared tumours with comparable efficacy to WT CTLs ( Figure 5—figure supplement 2G ) . Together , these results indicate that sustained release of CCL3 and CCL4 is sufficient to promote CCR5-dependent homing into tumours in vivo , and that the presence of tumour-reactive CTLs in a tumour promotes the recruitment of distant CCR5+ CTLs . Homotypic CTL recruitment via CCR5 is , however , likely not the only or primary signalling circuit that results in CTL tumour infiltration or clearance . It remains to be explored whether adoptive transfers using lower CTL numbers or earlier following tumour engraftment ( Sharma et al . , 2013; Chheda et al . , 2016 ) would result in a more dominant role for CCL3 and CCL4-mediated homotypic signalling in recruitment into solid tumours . We next investigated whether swarming also occurs in primary human T cell populations . Human effector T cells activated by anti-CD3/CD28 antibody-coated beads were found to upregulate CCL3 and CCL4 expression ( Figure 6—figure supplement 1A ) as previously reported ( Cristillo et al . , 2003 ) and produced supernatant that attracts additional effector T cells ( Figure 6A ) . When co-embedded with activating beads in place of a tumouroid , human T cells recruit distant T cells , which exhibit increased directionality ( Figure 6B , C ) and swarming ( Figure 6D–F and Video 8 ) . This directional recruitment is abolished by Cenicriviroc ( Figure 6C ) . Finally , we evaluated whether the homotypic recruitment mechanism is inducible via chimeric antigen receptor ( CAR ) engagement of T cells in an immunotherapeutic context . To this end , we engineered CAR T cells targeting Ephrin type-A receptor 2 ( EphA2 ) that is abundantly overexpressed in glioblastoma ( Liu et al . , 2006 ) . A truncated CD19 domain co-transferred with the CAR construct allows enrichment of genetically-modified CAR-expressing human primary T cells by magnetic sorting ( Figure 6—figure supplement 1B ) . Human CD19+ EphA2-specific CAR T cells efficiently killed glioblastoma cells ( Figure 6—figure supplement 1C ) . When EphA2-specific CAR T cells were embedded in a tumouroid with glioblastoma cells , surrounding CAR T cells swarmed towards the tumouroid , whereas no recruitment was observed when CAR T cells were co-embedded with EphA2-negative fibroblasts in the tumouroid ( Figure 6G–I and Video 9 ) . Collectively , these results demonstrate that upon activation either via the TCR or an engineered CAR , human CTLs engage in homotypic chemokine signalling , thereby amplifying the recruitment of additional T cells to an effector site . Swarming is a common collective behaviour in the natural world ( Okubo , 1986 ) . At the immune cell level , swarming has been described for neutrophils , during parasite infection in the lymph node ( Chtanova et al . , 2008 ) and in response to tissue injury ( Lämmermann et al . , 2013 ) . Here , our data reveal that CTLs engaging targets employ homotypic signalling via secretion of the chemokines CCL3 and CCL4 to accelerate the direct recruitment of distant T cells . As additional antigen-specific CTLs are recruited and engage cognate targets , they contribute to the chemoattractive signal , thus further amplifying CTL recruitment in a positive feedback loop , which leads to swarming behaviour . Interestingly , it appears as though CTLs closest to the source of the chemotactic signal become desensitised over time and lose directional bias , which could be due to exposure to excess ligand concentrations ( Lim et al . , 2018 ) or rapid internalisation of the CCR5 receptor ( Escola et al . , 2010 ) . Pioneering work had uncovered that in lymph nodes , naïve CD8+ T cells follow CCR5 ligands , including CCL3 and CCL4 , to migrate towards dendritic cells interacting with CD4+ and CD8+ T cells ( Castellino et al . , 2006; Hickman et al . , 2011; Hugues et al . , 2007 ) , to optimise the efficiency of T cell priming . Maximal priming and generation of CTLs in infected or draining lymph nodes during antiviral and antitumour responses also depend on CCR5 expression on T cells ( González-Martín et al . , 2011; Hickman et al . , 2011 ) . Rather than the behaviour of naïve T cells within secondary lymphoid organs , our study focused on interactions between effector CTLs , T cells that are active in peripheral tissues following priming and emigration from lymph nodes . Our findings reveal that CTLs can independently induce mass recruitment around an antigen-specific target . Our data also reveal that lymphocytes can promote the recruitment of distant lymphocytes into peripheral tissues , thus far thought to only occur indirectly through the action of professional antigen-presenting cell ( APC ) intermediaries ( Spranger et al . , 2017; Böttcher et al . , 2018 ) . Although the concept of bacterial quorum sensing has been applied to coordinated population responses by T cells , the amplification of local T cell densities was shown to occur via differentiation or proliferation ( Antonioli et al . , 2019 ) , relatively long-term effects compared to the rapid directional recruitment of distal CTLs discussed here . CTLs engaging tumour targets that do not present cognate antigen induce chemokinesis in distant CTLs , which could be mediated via the induced secretion of netrins , autotaxin , or semaphorins ( Boneschansker et al . , 2016; Katakai et al . , 2014; Takegahara et al . , 2005 ) . Importantly , only CTLs that have directly engaged a cognate antigen-presenting target secrete the homotypic recruitment signal that induces chemotaxis in distant T cells . Unconjugated CTLs exposed to cognate supernatant ( containing abundant CCL3 , CCL4 , and pro-inflammatory cytokines such as interferon gamma and tumour necrosis factor ) did not themselves upregulate CCL3 or CCL4 expression ( Supplementary file 1 and 2 ) . This therefore excludes the possibility of a signal relay , whereby CTLs that sense the recruitment signal away from the target start secreting the homotypic signal themselves , which could presumably lead to aberrant local CTL clustering . We thus propose that the individuals of a CTL population exist in three distinct states: a ‘searcher’ state whilst scanning for cognate antigen unaffected by the homotypic signal; a ‘responder’ state when they sense the homotypic signal and move rapidly towards the effector locus; and an ‘engager/recruiter’ state upon target cell recognition when they concurrently deliver their effector function and secrete CCL3 and CCL4 ( Figure 7 ) . The first CTLs to recognise a cognate target directly transition from search into engagement and recruitment , whereas subsequent cells arrive in a responder state . It is interesting to note that CTLs can also remain in an exclusive recruiter state for nearly 24 hr following target elimination in the absence of active engagement ( Figure 4—figure supplement 1H , I ) . This multistate recruitment model explains prior observations of non-specific T cells only deeply infiltrating tumour tissues in the presence of tumour-reactive CTLs ( Boissonnas et al . , 2007 ) . Although antigen recognition is specific , the chemoattractive signal that is raised to amplify recruitment is generic and thus able to recruit a repertoire of CTLs regardless of their TCR specificity . Indeed , there have been prior observations of bystander recruitment of effector T cells to sites of inflammation or tumours initiated by unrelated antigen ( Boissonnas et al . , 2004; Topham et al . , 2001; Ariotti et al . , 2015; Hickman et al . , 2015 ) . During an immune response resulting in the clonal expansion of antigen-specific CTLs , which increases their frequency up to 104-fold ( Murali-Krishna et al . , 1998 ) , such a generic recruitment mechanism is likely efficient in their local amplification in sufficient numbers for an effective response . Interestingly , studies that compared tumours infiltrated by leukocytes ( ‘hot’ ) to those that excluded immune cell infiltration ( ‘cold’ ) identified more abundant CCL3 , CCL4 , and CCL5 in hot tumours ( Chakravarthy et al . , 2018; Spranger et al . , 2015 ) , including via single-cell transcriptomics that ascribed their expression to CD8+ T cells ( Jerby-Arnon et al . , 2018; Roider et al . , 2020 ) . Whilst we have identified CCL3 and CCL4 as the key mediators of homotypic signalling in CTLs isolated from mice , CCL5 may also have a central role in homotypic recruitment of human CTLs . We demonstrated that CTLs exhibit chemotaxis towards CCR5 ligands within dense tumour tissue by intravital imaging , and revealed that the presence of antigen-specific CTLs within a tumour promotes the recruitment of distant CTLs into the tumour in a manner partially dependent on CCR5 . The activation of intratumoural CTLs were found by others to promote the recruitment of not only CTLs , but also natural killer cells and myeloid cells ( Rosato et al . , 2019 ) . Alternate chemokine signalling circuits centred on CXCR3 and BLT1 have been implicated in CTL tumour trafficking and infiltration ( Chheda et al . , 2016; Chow et al . , 2019; Mikucki et al . , 2015; Sharma et al . , 2013 ) . Furthermore , IFNγ , produced by activated CTLs and abundant in inflamed tumours , can induce CXCL9 and CXCL10 secretion in various intermediary cells , including tumour-infiltrating myeloid cells ( Dobrzanski et al . , 2001; Gordon-Alonso et al . , 2017; Hickman et al . , 2015 ) , which may enhance the role of CXCR3 in homing . CCR5-expressing CTLs have been associated with severe autoimmune diseases ( Mackay , 2014 ) , and CCR5 ligands are abundant in the upper respiratory tract during acute hyperinflammation associated with viral infection ( Chua et al . , 2020 ) ; silencing homotypic recruitment may well prove beneficial in such contexts . Ultimately , the homotypic recruitment mechanism could be exploited to enhance control over positioning and tissue infiltration of T cells in adoptive cell transfer immunotherapies , not least given our findings that engineered CAR T cells , following antigen recognition via the CAR , are able to induce homotypic recruitment of distal T cell populations . All mice were maintained at Australian BioResources ( Moss Vale , NSW , Australia ) and in the animal facility of the Lowy Cancer Research Centre , University of New South Wales . All animal breeding and experimentation were conducted in accordance with New South Wales state and Australian federal laws and animal ethics protocols overseen and approved by the University of New South Wales Animal Care and Ethics Committee ( 16/83B and 19/133B ) . T cells were obtained from spleens of 7- to 12-week-old mice , either OT1 ( specific for the OVA257-264 peptide SIINFEKL in a H-2Kb major histocompatibility complex class I context ) , hereafter referred to as OT1 CTLs , or OT1 ×Lifeact EGFP Galeano Niño et al . , 2020 referred to as OT1GFP CTLs . Mice were euthanised by cervical dislocation . Splenocytes were stimulated ex vivo with OVA257-264 ( SIINFEKL ) peptide ( AusPep , Melbourne , VIC , Australia ) for 4 hr in T cell medium ( TCM ) consisting of Roswell Park Memorial Institute ( RPMI ) 1640 , 10% foetal calf serum , 1 mM sodium pyruvate , 10 mM HEPES , 100 U/ml penicillin , 100 μg/ml streptomycin , 2 mM L-glutamine and 50 μM β2-mercaptoethanol ( all from Gibco , ThermoFisher Scientific , Waltham , MA , USA ) . 100 ng/ml recombinant mouse IL-2 ( R & D Systems , Minneapolis , MN , USA ) were added on the following day and subsequently every 2 days . T cells were cryopreserved at day 3 after isolation according to established methods ( Galeano Niño et al . , 2016 ) . Cryopreserved T cells were recovered by quick thawing in a 37°C water bath , the thawed cryopreservation solution was diluted in 10 ml TCM and the cells were then pelleted by centrifugation and resuspended in TCM supplemented with 100 ng/ml recombinant mouse IL-2 , renewed every 2 days until use . All experiments were performed using T cells cultured to days 6 or 7 post-isolation . gBT1 TCR transgenic mice ( Coles et al . , 2003 ) that had been crossed to UBI-GFP mice ( expressing GFP under control of the human Ubiquitin C promoter ) ( gBT1-GFP ) were a kind gift from W . R . Heath . The gBT1 TCR is specific for HSV glycoprotein B498-505 peptide SSIEFARL in a H-2Kb context . SSIEFARL peptide was purchased from AusPep . gBT1 CTLs were generated following the same procedure as OT1 CTLs . To generate polyclonal CD8+ CD44+ effector T cells ( from wild-type ( WT ) or Ccr5-/- C57BL/6 mice , the latter kindly provided by G . Clarke ) , single splenocyte suspensions were stimulated ex vivo for 24 hr with 1 µg/ml anti-CD3 ( clone 145–2 C11; BioLegend , San Diego , CA , USA ) , 1 µg/ml anti-CD28 ( clone 37 . 51; BioLegend ) and 100 ng/ml IL-2 in 10 ml of TCM . Cells were then washed and sorted on the FACS-Aria III flow sorter ( BD Biosciences; Franklin Lakes , NJ , USA ) based on CD8 and CD44 surface expression detected by antibody staining with 1 µg/ml anti-CD8a-Pacific Blue ( clone 53–6 . 7; BD Biosciences ) and anti-CD44-APC ( clone IM7; BD Biosciences ) . After sorting , cells were expanded in TCM and IL-2 until day 6 . In some experiments , cells were frozen on day 3 and thawed for expansion until day 6 or 7 before experimental use . Ccr5-/- C57BL/6 mice were crossed with OT1 transgenic mice to obtain Ccr5-/- OT1 CTLs . The murine EL4 lymphoma cell line , originally obtained from the American Type Culture Collection ( ATCC; TIB-39 ) , was obtained at early passage ( P3 ) from the Alexander lab . EL4 were cultured in TCM , with routine passaging three times per week , maintained at cell densities under 1 × 106/ml . For preparation of cognate cells for OT1 or gBT1 CTLs , EL4 cells were pulsed for 16 hr with 1 μg/ml SIINFEKL ( SFKL ) or SSIEFARL ( SFARL ) peptide , respectively ( referred to as ‘cognate tumour cells’ ) . The cell line was further authenticated by engrafting into C57BL/6 immunocompetent mice , as well as verifying sensitivity to H-2Kb-specific cytotoxicity . Cells were free of mycoplasma contamination based on the MycoAlert Mycoplasma Detection Kit ( Lonza , Basel , Switzerland ) detected on a FLUOstar Omega microplate reader ( BMG Labtech GmbH , Ortenberg , Germany ) . The patient-derived glioblastoma cell line WK1 ( Day et al . , 2013 ) was cultured as previously described ( Stringer et al . , 2019 ) . NIH/3T3 mouse fibroblast cells were cultured in Dulbecco’s Modified Eagle’s Medium ( DMEM ) supplemented with 10% feotal calf serum . For in vivo experiments , we used either C57BL/6 mice , Rag-deficient B6 . 129S7-Rag < tm1mom > JAusb mice ( ‘RAG1’ , Australian BioResources , NSW , Australia ) , hereafter referred to as Rag-/- , or B6 . SJL-PtprcaPepcb/BoyJAusb mice , referred to as PTPRCA mice . Human peripheral blood mononuclear cells ( PBMCs ) were obtained from healthy donors after informed consent and were used in experiments under a Human Research Ethics Committee ( HREC ) approved protocol ( Sydney Children’s Hospitals Network , LNR/13/SCHN/241 ) . We isolated PBMCs by separation over Lymphoprep , and then enriched for T cell populations using a human T cell isolation kit according to the manufacturer’s instructions ( Stem Cell technologies , Tullamarine , Australia ) . T cells were activated with T Cell TransAct ( Miltenyi Biotec Australia , Sydney , Australia ) in CTS OpTmizer T cell expansion medium supplemented with CTS Immune cell SR ( 2 . 5% , ThermoFisher Scientific Australia ) and recombinant human IL-7 and IL-15 ( 10 ng/mL and 5 ng/mL respectively , Miltenyi Biotech Australia ) . Human T cells were subsequently expanded in the above medium supplemented with IL-7 and IL-15 every 48 hr for ~12 days prior to use in experiments . Human T cells were isolated and activated as indicated above and transduced with a pRRL lentivirus ( Dull et al . , 1998 ) encoding an EphA2-specific CAR consisting of the EphA2-specific 4H5 scFv ( Chow et al . , 2013 ) and a 41BB . CD3ζ endodomain , linked by a 2A sequence with a truncated form of CD19 ( Yi et al . , 2018 ) . The transduced T cells were assessed for CD19 expression at day 4 post-transduction and further expanded for another 6 days . CAR-expressing T cells were enriched using an EasySep release human CD19 positive selection kit ( StemCell Technologies , Vancouver , Canada ) . Purity of the enriched CAR T cell population was validated by flow cytometry ( see Figure 6—figure supplement 1B ) . CTLs at day 6 or 7 post-isolation were labelled with 5 μM CellTracker Orange 5- ( and-6 ) - ( ( ( 4-chloromethyl ) benzoyl ) amino ) tetramethylrhodamine ( CMTMR ) ( Invitrogen , California , USA ) and EL4 tumour cells were labelled with CellTracker Deep Red dye ( ThermoFisher Scientific ) either pulsed with cognate peptide ( cognate ) or unpulsed ( non-cognate ) . A total of 4–6 × 106 cells ( either tumour cells alone or mixed 1:1 with labelled CTLs ) suspended in phenol red-free TCM ( 20 μl ) were added to ice cold liquid-phase rat-tail collagen I ( 25 μl at ~3 mg/ml; Corning , New York , NY , USA ) containing 1 N NaOH ( 0 . 77 μl ) and 10 × PBS ( 5 μl ) for a total volume of 50 μl on ice . Where bead-containing side tumoroids were constructed , 20 µl of a collagen preparation containing 2–3 × 106 each of cognate particles and CTLs were deposited on one side of a 14 mm microwell . A volume of 20 μl of the solution was rapidly transferred to one side of a 35 mm Petri dish containing a 14-mm microwell with a precision glass coverslip ( MatTek , Ashland , MA , USA ) and incubated at 37°C and 5% CO2 for 10 min to allow the gel to polymerise into a 3D matrix with a high density of tumour cells ( with or without co-embedded CTLs ) that mimics a solid tumour mass . A second collagen gel was prepared containing 3 × 105 CTLs that was deposited on the total surface of the microwell and allowed to polymerise at 37°C and 5% CO2 for 10 min to create a 3D collagen matrix with dispersed cells ( see Figure 2A ) . Finally , 1 . 5 ml cold phenol red-free TCM were gently added to the dish and incubated at 37°C and 5% CO2 for 10 min , then four-dimensional confocal microscopy was performed as described below . For experiments with a central tumouroid , 2 μl of a collagen preparation containing 3 . 5 × 105 EL4 tumour cells ( prepared as above ) or 1 . 75 × 105 cells each of CTLs and EL4 tumour cells were deposited in the centre of a well in a Greiner Sensoplate glass bottom 96-well optical plate ( Sigma Aldrich ) . As above , a second collagen matrix ( 50 μl total volume ) containing dispersed CTLs was polymerised on top of the first gel ( see Figure 1—figure supplement 1A ) . The whole well was imaged in x , y over 14–16 hr by confocal microscopy as described below . For experiments where WT OT1 and Ccr5-/- OT1 CTLs were co-embedded around a cognate tumouroid ( containing WT OT1 and cognate tumour cells ) , the two CTL populations and the cognate tumour cells were labelled with CMTMR , CellTracker Green 5-chloromethylfluorescein diacetate ( CMFDA ) ( Invitrogen , California , USA ) or Deep Red dye ( different combinations were used for different experiments ) . For experiments with primary human T cells , cells that were embedded in the tumouroid were labelled with Deep Red dye and pre-stimulated with control or activating beads for 3 hr , whereas migrating cells were labelled with CMTMR . For experiments with CAR T cells , tumouroids were prepared with embedded CAR T cells and tumour cells ( WK1 glioblastoma cells or NIH/3T3 control fibroblasts , the former EphA2-expressing , the latter EphA2-negative ) labelled with Deep Red dye and migrating CAR T cells were labelled with CMFDA . Four-dimensional imaging data were collected through a 20 × water immersion objective , numerical aperture ( NA ) of 1 . 37 ( Leica Microsystems , Wetzlar , Germany ) , on a Leica TCS SP5 confocal microscope equipped with a resonant scanner and an incubator that maintains 37°C and 5% CO2 throughout imaging . LifeactGFP and CMFDA were excited at 488 nm , CMTMR at 561 nm and Deep Red at 633 nm wavelengths using a tunable white light laser ( Leica Microsystems ) . Images were obtained from the x , y and z planes , with a total z thickness of 65 μm ( with the lower 5 μm immediately above the glass rejected in order to avoid the inclusion of two-dimensional movements on the glass surface ) and step size of 1 . 6 μm every 20 s for 1 hr , with three fields of view imaged sequentially at every timepoint . Imaging of whole wells with a central tumouroid was performed through a 10×/0 . 30 NA dry objective on the same instrument with an image taken every 5–6 min over 14–16 hr and the Leica LAS AF software was used to automate x , y tiling and facilitate image stitching ( 9 × 9 tiles ) . For tumouroid samples that were imaged for 40 hr ( Video 1 ) , sample preparation was performed as described above , but in the centre of a 35 mm MatTek dish , using cognate tumour cells stained with CMTMR for the tumouroid and OT1GFP CTLs embedded in the surrounding collagen matrix . Time-lapse imaging data were collected through a 20×/0 . 8 NA air objective on a Nikon BioStation live-cell imaging platform ( Nikon , Tokyo , Japan ) . Images were taken every 20 min and the BioStation IM software was used to automate x , y tiling and facilitate image stitching ( 9 × 11 tiles ) . Image processing for visualisation was performed in Fiji/ImageJ ( US National Institutes of Health , Bethesda , MA , USA ) . Streptavidin-coated polystyrene particles ( 6 . 0–8 . 0 μm diameter; Spherotech Inc , Illinois , USA ) were incubated with monobiotinylated H-2Kb/SIINFEKL ( Tetramer Synthesis Service , John Curtin School of Medical Research , Australia National University , ACT , Australia ) in BSA solution ( phosphate-buffered saline with 20 mM HEPES , 150 mM NaCl and 2% bovine serum albumin ) at 37°C for 45 min . The beads were then washed three times in BSA solution before use in subsequent experiments . For experiments with human T cells , beads were coated with 1 μg/ml IgG isotype control antibody or with 1 μg/ml anti-human CD3ε and anti-CD28 antibodies ( ThermoFisher Scientific ) . For experiments where central tumoroids were constructed using Streptavidin-coated polystyrene particles , 2 . 5 µl of a collagen preparation containing 3 . 5 × 105 cognate ( prepared as above ) or non-cognate particles , or 1 . 75 × 105 each of cognate particles and CTLs were deposited in the centre of a well in a glass bottom 96-well optical plate . As above , a second collagen matrix ( 50 µl total volume ) containing GFP expressing CTLs was polymerised on top of the first gel . The whole well was imaged in x , y over 16 hr by confocal microscopy as described above . For experiments where CTLs were embedded around a cognate tumoroid containing H-2Kb/SIINFEKL coated polystyrene particles and Ccr5-/- OT1 CTLs , the embedded Ccr5-/- OT1 CTLs were labelled with Deep Red dye . 2 × 106 cells each of OT1 CTLs and EL4 tumour cells ( cognate or non-cognate ) or polystyrene beads ( coated with H-2Kb/SFKL or uncoated ) were resuspended in 1 . 5 ml RPMI 1640 medium containing 0 . 5% bovine serum albumin ( BSA ) in 24-well plates . The plates were centrifuged for 5 min at 300 × g followed by incubation at 37°C . After 3 hr , the samples were centrifuged and the supernatants collected and filtered through 0 . 22 μm pore polyethylsulfonate filters ( Millipore , Burlington , MA , USA ) . EL4 cells were pulsed with 1 µg/ml SIINFEKL peptide overnight for 16 hr . 4 × 106 OT1 Ccr5-/- CTLs were labelled with CMFDA ( Invitrogen , California , USA ) and resuspended with 4 × 106 cognate EL4 cells in 5 ml TCM in a 6-well plate . The plate was centrifuged for 5 min at 300x g followed by incubation at 37°C . After 4 hr , the cells were pelleted , and the supernatant collected and filtered through a 0 . 22 µM pore polyethylsulfonate filter ( Millipore , Burlington , MA , USA ) . The cells were collected , and 1 . 25 × 106 OT1 Ccr5-/- CTLs were sorted on the FACS-Melody flow sorter ( BD Biosciences , Franklin Lakes , NJ , USA ) based on CMFDA labelling . After sorting , the cells were split across seven wells in a 24-well plate , containing 620 µl of TCM each , and incubated at 37°C until supernatant collection . Supernatant from each well was collected at either 2 , 4 , 8 , 16 , 24 , 48 , or 72 hr post-sorting and filtered through a 0 . 22 µM pore polyethylsulfonate filter ( Millipore , Burlington , MA , USA ) for analysis by cytometric bead array ( CBA ) . Chemotaxis was assayed using a Transwell Chamber ( Corning ) with a 5 μm-pore size polycarbonate filter . Briefly , 600 μl of control medium ( serum-free RPMI 1640 medium containing 0 . 5% BSA ) or medium containing recombinant chemokines at 1 , 10 or 100 ng/ml ( Peprotech , Rocky Hill , NJ , USA ) or filtered supernatant were placed in the lower chambers . 1 × 106 CTLs were placed in the upper chamber in 100 μl RPMI + 0 . 5% BSA and incubated at 37°C and 5% CO2 for 3 hr ( Figure 2—figure supplement 1A ) . For experiments with collagen gels , cells were embedded in liquid-phase collagen on ice ( as per migration assay above ) . After 3 hr , cells were harvested from the lower compartment and analysed by flow cytometry using a BD LSR Fortessa X20 flow cytometer ( BD Biosciences ) . Cell numbers were enumerated using SPHERO AccuCount blank particles ( Spherotech Inc , Chicago , IL , USA ) . Results were analysed using FlowJo software ( FlowJo 10 . 2 , Tree Star Inc , Ashland , OR , USA ) . The transmigration index is represented as bars ± SEM and was calculated as follows:Transmigration index=number of cells transmigrated in samplenumber of cells transmigrated in control medium Cytotoxicity assays were performed using flow cytometry based on the ratio between live target and non-target cells . Effector CTLs ( GFP+ ) , cognate tumour cells and non-cognate tumour cells ( either CMTMR+ or CMTMR- in different combinations ) were mixed at a 1:1:1 ratio and the distribution of the three cell populations was measured by flow cytometry at 0 hr and 2 hr . Samples were run in duplicate in each experiment . The cytotoxicity index was calculated as:Cytotoxicityindex ( % ) =[1− ( cognate cells ( 2h ) non−cognate cells ( 2h ) ) ( cognate cells ( 0h ) non−cognate cells ( 0h ) ) ] × 100 In some experiments , the following inhibitors were used: 500 ng/ml Pertussis toxin from Bordetella pertussis ( PTX ) or inactive mutated version ( m-PTX ) ( Sigma Aldrich , St Louis , MO , USA ) ; 10 μg/ml Maraviroc ( CCR5 antagonist ) ( Sigma Aldrich ) ; 1 μM Cenicriviroc ( CCR2 + CCR5 dual inhibitor ) ( AdooQ Bioscience , Irvine , CA , USA ) ; 1 μM CCR2 antagonist ( CAS 445479-97-0 , Santa Cruz Biotechnology Inc , Dallas , TX , USA ) ; 50 μM UCB35625 ( CCR1 + CCR3 dual inhibitor ) ( R & D Systems ) ; or DMSO ( vehicle ) . In some experiments , the following neutralising antibodies were used: anti-CCL1 , anti-CCL3 , anti-CCL4 , anti-CCL5 , anti-CCL9 , and anti-XCL1 with corresponding isotype control antibodies ( all from R & D Systems ) . To prepare cells for RNA extraction , 2 × 106 CTLs were incubated in 1 . 5 ml TCM in 24-well plates with equal numbers of either cognate or non-cognate EL4 tumour cells for 3 hr . Cells were then collected and centrifuged ( 300 × g , 5 min ) and washed 2 × with PBS before being re-suspended in 2 ml cold TCM . A total of 1 × 106 GFP-expressing CTLs were sorted on the FACS Aria III flow sorter ( BD Biosciences ) . Cells were then centrifuged ( 300 × g , 5 min ) before RNA isolation using the RNeasy Mini Kit ( Qiagen ) , according to manufacturer’s instructions . 1 . 5 µg of RNA were reverse transcribed to cDNA using the High Capacity cDNA Reverse Transcription Kit ( Applied Biosystems , ThermoFisher Scientific ) according to manufacturer’s instructions . Real-time quantitative PCR was carried out using chemokine and chemokine receptor primers predesigned and synthetised by Sigma-Aldrich ( KiCqStart SYBR Green Primers ) . 20 ng cDNA were added to each well of a 96-well PCR plate with 1 µM forward and reverse primer for each gene . 40 cycles were performed , with denaturing temperature at 95°C for 15 s , annealing at 55°C for 30 s , and extension at 72°C for 30 s . The amount of amplicon was measured using SYBR Green and detected in a BIO-RAD CFX96 Real time system ( Bio-Rad Laboratories , Hercules , CA , USA ) . The expression of each gene was normalised to the expression of the housekeeping genes β2-microglobulin ( β2M ) and ribosomal protein L13A ( RPL13A ) . The primer sequences used for qRT-PCR analysis of chemokine and chemokine receptor expression are found in the table below . ChemokineForward primerReverse primerCCL1ATGCTTACGGTCTCCAATAGTCTTCAGGTGATTTTGAACCCCL3TTCTCTGTACCATGACACTCCTCTTAGTCAGGAAAATGACACCCL4GGTATTCCTGACCAAAAGAGTCCAAGTCACTCATGTACTCCCL5AGGAGTATTTCTACACCAGCCAGGGTCAGAATCAAGAAACCCL6CTTTCAAGACACTTCTTCAGACCTGCTGATAAAGATGATGCCCCL7CTCTCTCACTCTCTTTCTCCTCTGTAGCTCTTGAGATTCCCCL8CTTCAACATGAAGATCTACGCCTGGATATTGTTGATTCTCTCGCCL9/10AATGTTTCACATGGGCTTTCCAATGCATCTCTGAACTCTCCCL12TGTGATCTTCAGGACCATACCATGAAGGTTCAAGGATGAAGCCL17CATTCCTATCAGGAAGTTGGCAGTCAGAAACACGATGGCCL19TTCTTAATGAAGATGGCTGCCTTTGTTCTTGGCAGAAGACCCL22CACATAACATCATGGCTACCCAGAAGAACTCCTTCACTAAACCXCL4TAGCCACCCTGAAGAATGGACATTTAGGCAGCTGATACCXCL9GAGGAACCCTAGTGATAAGGGTTTGATCTCCGTTCTTCAGCXCL10AAAAAGGTCTAAAAGGGCTCAATTAGGACTAGCCATCCACCXCL11CGACAAAGTTGAAGTGATTGGCACAGAGTTCTTATTGGAGCXCL12GAAAGCTTTAAACAAGAGGCGTGAAAGTACAGCAAAACTGCXCL16CCATTCTTTATCAGGTTCCAGCTTGAGGCAAATGTTTTTGGChemokine receptorForward primerReverse primerCCR1ATACTCTGGAAACACAGACTCGTCAAATTCTGTAGTTGTGGGCCR2ACCACATGTGCTAAGAATTGCTGGTTTTATGACAAGGCTCCCR3TCACCAGAGACAAGTAGAATGACTCATATTCATAGGGTGTGGCCR5AGACCTAAATCCTACCACACTGGCTTCAAACTATGGAAACHousekeeping geneForward primerReverse primerβ2-microglobulinGTATGCTATCCAGAAAACCCCTGAAGGACATATCTGACATCRPL13ACCTATGACAAGAAAAAGCGGCAGGTAAGCAAACTTTCTGG T cell secretomes from interactions with cognate beads , cognate tumour cells , non-cognate-beads , non-cognate tumour cells , and secretomes from EL4 ( tumour cells only ) and OT1 ( T cells only ) ( n = 3 per group ) were prepared for mass spectrometry analysis as previously described ( 40 , 41 ) with the following modifications . For all our experiments with magnetic beads , we used a 1:1 combination mix of the two types of commercially available carboxylate beads ( Sera-Mag Speed beads , #65152105050250 , #45152105050250 , ThermoFisher Scientific ) . Beads were prepared freshly each time by rinsing with water three times prior to use and stored at 4°C at a stock concentration of 20 μg/μl . Samples were transferred to a 2 ml LoBind deep well plate ( Eppendorf , Hamburg , Germany ) and reduced with 2 M Dithiothreitol ( DTT , 50 mM final conc . ) for 1 hr at 37°C . Samples were then alkylated with 1M Iodoacetamide ( IAM ) ( 100 mM final conc . ) for 30 min in the dark at room temperature ( RT ) . Samples were quenched with 2M DTT ( 250 mM final conc . ) and 5 μl of the concentrated bead stock carboxylate beads ( 20 μg/μl ) were then added to each sample followed by the addition of acetonitrile ( ACN ) to a final concentration of 70% ( v/v ) . Mixtures were left to incubate upright at RT for 20 min to allow proteins to precipitate onto the beads . The beads were placed on a magnetic rack and washed twice with 70% ethanol and once with ACN ( 500 μl washes ) . ACN was completely evaporated from the plate using a CentriVap ( Labconco , Kansas City , MO , USA ) prior to the addition of 40 μl digestion buffer ( 10% 2-2-2-Trifluorethanol ( TFE ) /100 mM NH4HCO3 ) containing 1 μg Trypsin-gold ( Promega , V5280 ) and 1 μg Lys-C ( Wako ) . The plate was briefly sonicated in a water bath to disperse the beads , and the plate transferred to a ThermoMixer instrument ( Eppendorf ) for enzymatic digestion at 37°C for 1 hr ( 1200 rpm ) . The supernatant comprising of peptides was then collected from the beads using a magnetic rack ( Ambion , Thermo Fisher Scientific ) and an additional elution ( 40 μl 2% Dimethyl sulfoxide , DMSO , Sigma ) was performed on the beads . The eluates were pooled together and transferred to pre-equilibrated C18 stage tips for sample clean-up . Briefly , two plugs of C18 resin ( 3M Empore , 66883 U ) were prepared in 200 μl unfiltered tips , pre-wetted with 100 μl methanol followed by sequential washes with 100 μl 80% ACN/5% formic acid ( FA ) , 50% ACN/5% FA and 5% FA . The pooled peptides were then added to the spin tip and the eluate collected into a fresh lo-bind Eppendorf tube . Bound peptides were washed twice with 5% FA . Elutions ( 50 μl ) were performed sequentially with 50% ACN/5% FA followed by 80% ACN/5% FA and collected into fresh Eppendorf tubes . All spins were performed on a benchtop centrifuge at 500 × g ( 1000–2000 rpm ) speeds . The eluates were lyophilized to dryness in MS vials using a CentriVap ( Labconco ) prior to reconstituting in 20 μl Buffer A ( 0 . 1% FA/2% ACN ) ready for MS analysis . The absolute concentration of chemokines present in supernatants was assessed by ELISA for CCL1 and CCL9 ( Sandwich ELISA kits , OriGene Technologies , Rockville , MD , USA ) and Cytometric Bead Array ( CBA ) for CCL3 , CCL4 , CCL5 , and CXCL10 ( LEGENDplexTM Multi-Analyte Flow Assay Kit , BioLegend , San Diego , CA , USA ) as per manufacturers’ instructions . All samples were run in triplicate . Chemokine levels in the supernatants were interpolated from standard curves generated using recombinant proteins provided in the kits . IFN-γ and TNF-α were used as positive controls for CTL activation . For ELISA , 100 µl of diluted supernatant and 100 µl of assay diluent were incubated in a 96-well antibody-coated plate for 2 hr . After 3–5 washes , the samples were incubated with 100 µl of detection antibody for 1 hr followed by the addition of 100 µl of streptavidin-HRP secondary antibody for 30 min . Finally , the HRP substrate tetramethyl benzidine was added and plates were analysed on a FLUOstar Omega microplate reader ( BMG Labtech GmbH , Ortenberg , Germany ) by reading the absorbance at 450 nm . For CBA , 25 µl of each supernatant were mixed with 25 µl of captured beads against the desired chemokine for 2 hr in a V-bottom plate . After two washes , the samples were incubated with 25 µl of detection antibody for 1 hr followed by the addition of 25 µl of PE-conjugated secondary antibody . Data were acquired on a Fortessa X20 flow cytometer and analysed by FlowJo software . As described previously , a high density tumouroid of 1:1 cognate EL4s and OT1 CTLs in a 3D collagen matrix was prepared on one side of a 35-mm Petri dish containing a 14 mm microwell with a precision glass coverslip ( MatTek , Ashland , MA , USA ) . 1 ml of CBA beads from the LEGENDplex Multi-Analyte Flow Kit ( BioLegend , San Diego , CA , USA ) were pelleted at 250 x g for five mins and supernatant was removed . The beads were resuspended in a second collagen gel mixture that was deposited adjacent to the tumouroid , and allowed to polymerise at 37°C and 5% CO2 for 10 min to create a 3D collagen matrix with dispersed CBA beads . 2 ml of warm TCM was added to the dish , which was then returned to 37°C for either 2 or 3 hr to allow formation of a CCL3 gradient from the tumouroid . The TCM was then removed and 1 ml of 10% Neutral Buffered Formalin ( NBF , Sigma Aldrich ) was added to the dish to fix the gel , which was placed on a plate rocker to incubate for 1 hr at room temperature . The 10% NBF was then removed , and the dish washed 3x with 2 ml FACS wash for 30 min per wash , on the plate rocker at room temperature . The FACS wash was then removed , and the gel resuspended in 2 ml of 1 µg/ml anti-murine CCL3 biotinylated rabbit antibody in FACS wash ( PeproTech , Rocky Hill , NJ , USA ) and incubated for 3 hr at room temperature , rocking . The gel was then washed twice with FACS wash as described previously , and then a third time with an overnight incubation at room temperature . The wash buffer was then removed and replaced with Streptavidin-PE ( also from the LEGENDplexTM Multi-Analyte Flow Kit ) and incubated for 45 min at room temperature , rocking . The gel was then washed twice with FACS wash for 30 min per wash , rocking and then a third time overnight before imaging . The coding sequences for all fusion proteins were synthesized as gBlocks by Integrated DNA Technologies , Inc ( Skokie , IL , USA ) and cloned into the MSCV-based retroviral expression vector ‘LENC’ ( kind gift from Johannes Zuber ) that we modified to contain multiple cloning sites . The design of these sequences was based on published sequences of murine CCL3 ( GenBank accession # NM_011337 . 2 ) , murine CCL4 ( #NM_013652 . 2 ) , mScarletI ( # KY021424 . 1 ) and miRFP670 ( KX421097 . 1 ) . Sequences encoding the mCCL3-mScarletI and mCCL4-miRFP670 fusion proteins were separated by the 16-residue SGGGGSGGGGSGGGGS linker and cloned into the AgeI/HpaI sites . Sequence encoding bicistronic expression of mTagBFP2 fluorescent protein ( Subach et al . , 2011 ) with the SSIEFARL or SIINFEKL epitopes were linked by the viral F2A sequence VKQTLNFDLLKLAGDVESNPGP and cloned into the BglII/HpaI sites . The following coding sequences were synthesized as oligos for integration into retroviral expression vector ‘LENC’ . To transduce EL4 cells , retrovirus pseudotyped with the vesicular stomatitis virus ( VSV-G ) envelope was produced by polyethylenimine ( PEI , molecular weight 4000 , PolySciences Catalogue No 24885–2 , Warrington , PA , USA ) transfection of GP2-293 cells ( Clontech , Palo Alto , CA , USA ) . NaOH-neutralised PEI ( 1 mg/ml ) was complexed with 6 . 8 µg of rMSCV-mCCL3-mScarletI and rMSCDV-mCCL4-miRFP670 and 3 . 2 µg of pMD2 . G plasmid ( VSVG coding sequence expressed from the CMV promoter , kind gift of Didier Trono ) for 30 min at room temperature before addition to 7 × 106 GP2-293 cells . At 72 hr after transfection , viral supernatant was used to transduce EL4 cells and fluorescent EL4 cells were sorted ( BD FACS Aria III ) 72 hr after transduction . GP2-293 cells were maintained in DMEM ( Gibco ) containing 4 . 5 g/L glucose , 4 mM L-glutamine and 1 mM sodium pyruvate supplemented with 10% heat-inactivated FBS . Sorted CCL3/CCL4-secreting or mTagBFP2-expressing EL4 tumour cells were cultured in TCM with routine passaging three times per week , maintained at cell densities under 1 × 106/ml . A total of 1 × 106 EL4 cells transduced to express CCL3-mScarletI and CCL4-miRFP670 or WT EL4 cells were injected subcutaneously into contralateral flanks of 8-week-old Rag-/- or PTPRCA mice . The tumours were allowed to grow for 7–10 days before day 6 or 7 effector CTLs were adoptively transferred via tail vein injection in 200 μl PBS . 24 to 72 hr post T cell transfer , mice were euthanised by CO2 asphyxiation . The spleens and both tumours were collected and dissociated with 1 mg/ml collagenase IV ( Sigma-Aldrich ) for 30 min at 37°C ( shaking at 800 rpm ) . The samples were filtered through 70 µm cell strainers to obtain single-cell suspensions . Tumours or spleens were resuspended in final volumes of 2 ml or 5 ml FACS wash buffer ( 2% HI-FCS , 2 mM EDTA and 0 . 02% sodium azide in 1 × PBS ) , respectively . 100 µl of cells from these suspensions were mixed with 100 µl of buffer containing 2 × 104 AccuCount Blank Particles to determine the absolute number of infiltrating cells . Remaining cells were stained with fluorescent conjugated antibodies against CD11b ( Brilliant Violet 711; clone M1/70 ) , NK1 . 1 ( Alexa Fluor 488; PK136 ) , Ly-6C Antibody ( FITC; HK1 . 4 ) , CD90 . 2 ( Brilliant Violet 510; 30-H12 ) , CD64 ( Brilliant Violet 421; X54-5/7 . 1 ) and CD45 ( APC/Fire 750; 30-F11 ) , CD11c ( FITC; N418 ) , I-A/I-E or MHC II ( Pacific Blue , M5/114 . 15 . 2 ) , or CD45 . 1 ( APC/Fire 750; A20 ) for 30 min on ice in FACS wash buffer containing 10% normal mouse serum . Final cell suspensions were prepared in 200 µl cold FACS wash buffer containing 0 . 5 µg/ml 4’ , 6-diamidino-2-phenylindole ( DAPI ) and acquired on the BD Fortessa X20 flow cytometer . Flow cytometry data were analysed with FlowJo software . To quantify recruitment of WT and Ccr5-/- CTLs to non-cognate or CCL3/CCL4-secreting tumours , Rag-/- mice were inoculated subcutaneously with 1 × 106 WT EL4 or CCL3/CCL4-secreting EL4 tumour cells on contralateral flanks . On day 10 , equal numbers of WT and Ccr5-/- effector OT1 CTLs ( 12 . 5 × 106 each ) , distinguished by CellTracker Deep Red , CellTracker Green , or CellTracker Orange CMTMR labelling , were co-injected intravenously . OT1 CTLs in single-cell suspensions prepared from both tumours and the spleen were enumerated by flow cytometry 22 hr later . Ratios of Ccr5-/-:WT CTLs in tumours were normalised to the ratio of Ccr5-/-:WT in spleens for each mouse . Data in Figure 5F–H are from four independent experiments . 68-week-old Rag-/- mice were engrafted by subcutaneous injection of 1 × 106 EL4 . OVA tumour cells on both flanks . Tumour growth was monitored by daily caliper measurements in two orthogonal dimensions , where tumour volume in mm3 is calculated as = 0 . 5 x length ( mm ) x width ( mm ) 2 . On day 9 post-engraftment , 5 × 106 OT1 or 5 × 106 Ccr5-/- OT1 CTLs were adoptively transferred into the mice by tail-vein injection . Each tumour volume was normalised to volumes on the day of T cell transfer . A trifluoroacetic acid ( TFA ) -free , self-assembling peptide derivative , Fmoc-DDIKVAV , was synthesized using standard Fmoc chemistry procedure on a 1 . 5 mmol scale . Fmoc-Asp-Wang resin , Fmoc protected amino acids , 1-hydroxybenzotriazole ( HOBt ) , N , N-diisopropylethylamine ( DIPEA ) , and 2- ( 1H-Benzotriazole-1-yl ) −1 , 1 , 3 , 3-Tetramethyluronium hexafluorophosphate ( HBTU ) were purchased from GL Biochem ( Shanghai , China ) , with other reagents sourced from Sigma-Aldrich ( Australia ) . Solution-phase anion exchange with excess aqueous hydrochloric acid ( HCl ) was used to remove the TFA counterion , followed by lyophilization . Reverse phase high performance liquid chromatography ( RP-HPLC ) confirmed 95% purity . Gelation was initiated using a well-established pH switch ( Rodriguez et al . , 2013 ) . Briefly , one wt% hydrogels were prepared from amorphous Fmoc-DDIKVAV powder . This was suspended in deionized water , before the pH was subsequently raised with a minimal amount of 0 . 5 M NaOH to ensure solubilization . Gelation occurred spontaneously when the pH was lowered to 7 . 4 using dropwise 1 M HCl , and water used to ensure a concentration of 20 mg/ml . Fourier transform infrared ( FTIR ) , circular dichroism ( CD ) spectroscopy , and transmission electron microscopy were performed to verify synthesis and structure of the desired nanofibrillar structure ( data not shown ) . CCL3 was loaded within the molecular hydrogel via our recent shear-containment methodology ( Nisbet D . R . ; Nisbet et al . , 2018 ) . Briefly , the peptide hydrogels were disrupted via the application of shear force until a gel-solution transition was observed and then 10 µg/ml of recombinant CCL3 and 20 µg/ml of 10 kDa Dextran labelled with AlexaFluor647 ( Dextran-AF647 ) were added prior to syringe administration and spontaneous re-assembly . -week-old Rag-/- mice were engrafted by subcutaneous injection of 1 × 106 EL4 tumour cells into contralateral flanks . On day five post-engraftment ( 48 hr before imaging ) , a total of 40 × 106 OT1 and Ccr5-/- OT1 CTLs ( prepared 1:1 ) were adoptively transferred into the mice by tail-vein injection . The two CTL populations were labelled with CFSE and Cell Trace Violet ( Invitrogen ) at 100 mM final concentration ( dye selections were inverted for different experiments ) . Mice ( 17–18 g ) were anaesthetized by intraperitoneal injection of 100 mg/kg body weight ketamine and 15 mg/kg body weight xylazine . Tumours were surgically exposed and prepared for intravital microscopy by skin flap surgery . The tumour was stabilized on a coverslip on a microscope stage with intact vasculature and innervation . Intravital imaging was performed on a Nikon A1R inverted laser scanning confocal microscope fitted with a CFI APO LWD Lambda series 20×/0 . 95 NA water immersion objective , an Okolab humidified temperature-controlled microscope enclosure , objective heater and a custom-made stage insert . Heating was adjusted to maintain the temperature at 34°C within the chamber . Cell Trace Violet was excited with a 405 nm laser , CFSE with a 488 nm laser and Dextran-AF647 with a 640 nm laser . Images were taken using NIS Elements software ( Nikon ) . During imaging , tumours were injected with sub-µl volumes of hydrogel containing recombinant CCL3 chemokine and Dextran-AF647 by using a customized stereotaxic injection unit ( Kopf , Model 5000 microinjection unit , Tujunga , CA , USA ) equipped with a syringe and 29-gauge needle with 45° bevel angle ( Hamilton Company Inc , Reno , NV , USA ) . Intravital imaging data were analysed with Imaris 9 . 2 . 1 software ( Bitplane AG , Zurich , Switzerland ) by use of the Spots function to determine the number of cells in the Ccr5-/- and WT channels at each time point . Data were then graphed in Plot2 for Mac 2 . 6 . 1 ( apps . micw . org ) . In Figure 5—figure supplement 3B , densities were linearly interpolated at regular 10 min intervals in Matlab ( MathWorks ) to obtain the mean curve ( due to differences in sampling frequencies between the independent experiments ) . Image analysis for the tracking of CTL movement in 3D was performed with Imaris 9 . 2 . 1 software ( Bitplane AG , Zurich , Switzerland ) . Cells were segmented by creating surfaces with a filter below 100 µm3 to discard cell debris . Cells were tracked using autoregressive motion , applying a threshold of 10 min to filter out tracks of insufficient duration . Intensity , morphological and tracking data were then exported , yielding multiple motility parameters used to quantify population-wide migration behaviours . Track displacement is the net distance between first and last position:D=Dx ( tL , tF ) 2 + Dy ( tL , tF ) 2+ Dz ( tL , tF ) 2where D is net displacement , Di is the net displacement in the i-axis between tL , the position of the cell at the last timeframe of the track , and tF , the position of a cell at the first timeframe of the track . The mean speed was calculated by dividing the total length a cell travels by the duration of the track . The Forward Migration Index ( FMI ) is a measure of the directionality of the cell trajectory along the x axis ( i . e . towards the tumour ) . The FMI is defined as the displacement in x ( Dx ) divided by the net displacement ( D ) , therefore equivalent to the cosine of the angle between the displacement vector and the vector pointing towards the tumouroid . FMI=Dx D Image analysis for the population-wide behaviour of CTLs in whole wells was performed with Imaris 9 . 2 . 1 software . Tumouroids and CTLs were segmented using the Imaris ‘surface’ and ‘spots’ functions , respectively . The shortest distance between each CTL and the tumouroid surface was then calculated for each timepoint using the ‘Distance transform’ ImarisXT module employing MATLAB ( MathWorks Inc , Natick , MA , USA ) , for CTLs located either outside or inside the tumouroid surface . MATLAB was then used to calculate the fraction of total CTLs infiltrated into the tumouroid as well as the mean infiltration depth into the tumouroid ( measured as the shortest distance from the tumouroid edge for infiltrated CTLs ) . For visualisation and preparation of movies and figures , Fiji/ImageJ was employed in conjunction with Imaris . For the visualisation of experiments with human T cells , Fiji/ImageJ bleach correction ( histogram matching ) was applied to eliminate fluctuations in laser power that occurred over the 14 hr of imaging . Kymographs were generated to quantify the density of cells with respect to the tumouroid interface over time . Briefly , the cumulative sum of cells residing within radial distance r of the tumouroid interface was calculated for all r , for each imaging frame . These data were then normalised for cell counts over area of r , and smoothed over r using the ‘fda’ ( version 2 . 4 . 8 ) functional regression package in R . The smoothed functions are differentiated with respect to r to yield the density of cells with respect to distance from tumouroid . We quantify the spatial organisation of cells with respect to the tumouroid through a swarming index ( the 'M' metric ) ( code available ) . Briefly , this index ranges between −1 and 1 , with −1 indicating an even distribution of all cells along the well perimeter , and one indicating all cells as residing within the tumouroid . A value of 0 represents a uniform distribution of cells within the well , but outside of the tumouroid ( Figure 1C ) . The swarming index is independently quantified for each imaging frame , and the subsequent timeseries denotes the spatial evolution of cells with respect to the tumouroid over time . Custom code and notes for generating ‘M’ are available at https://github . com/marknormanread/TcellSwarming ( Niño et al . , 2020; copy archived at swh:1:rev:74c6678c55317a0aac98a70939e0c92fb29e58ad ) . Simulations of T cell motility and attraction were conducted through an adaptation of the agent-based simulation , ‘motilisim’ ( Read et al . , 2016 ) . Code is available at https://github . com/marknormanread/TcellSwarming . T cells are modelled as non-overlapping spheres of 12 µm diameter in a spatially explicit 3D environment . We modelled a whole well environment of 6 . 8 mm diameter containing a concentric tumouroid environment of 2 . 4 mm diameter , both with height 60 µm; these boundaries are impermeable . In all cases , a total of 35 , 000 cells are simulated within this environment . T cell motility patterns are sampled from the tracks of OT1 CTLs observed through 3D ex vivo imaging of space immediately adjacent to a tumouroid ( Figure 2A , B ) . The sampling methodology is performed via ‘bootstrapping’: each modelled T cell samples ( with replacement ) and then re-enacts 10 min blocks of observed ex vivo CTL motility ( Figure 1—figure supplement 1B , C ) . The pool of blocks to be sampled comprises all unique consecutive 10 min durations of track data , across all imaged CTLs in the source data . Each block describes a given CTL’s sequence of displacements , and their relative orientations , within the 10 min window . Modelled agents conducting ‘undirected’ motion sample their motilities from experiments involving random migration of CTLs in the absence of a tumouroid . Conversely , ‘chemotactic’ agents sample the motilities of CTLs imaged in the presence of a tumouroid containing cognate tumour cells and embedded CTLs ( Figure 1—figure supplement 1A ) . In this case , each block captures CTL motility with respect to the tumouroid , and the modelled agent reinterprets this with respect to the chemokine gradient ( Figure 1—figure supplement 1B ) ; ‘up’ ( 0x , 0y , 1z ) is maintained during these rotational translations . Agents entering the tumouroid environment cease bootstrapped motility reconstruction and instead move towards the tumouroid centre through the xy plane , holding a stable z-location , at 0 . 15 µm/min . Upon entering the tumouroid , agents in the ‘positive attraction’ simulation scenario ( Figure 1F ) secrete a soluble chemotactic factor . No such secretion takes place in the simulated ‘no attraction’ scenario . Chemokine-secreting agents do so at 1000 molecules/min . The chemokines have a diffusion coefficient of 250 µm2/s ( 2 . 5 × 10−6 cm2/sec ) , given CCL3 and CCL4 molecular weights of 10 kDa approximately . The chemokine concentration at a given point in space and time is resolved through applying the heat kernel to all prior secretion events , having recorded their location , time and quantity . The heat kernel is a numerical solution for the modelling of diffusion . Agents determine the chemokine concentration at six points around their spherical extremity ( where the sphere intersects the ±x , ±y and±z axes , relative to the agent ) , and from this determine the chemokine gradient direction . Agents are ‘chemotactic’ only whilst the maximum perceived chemokine concentration lies above a ‘chemotaxis actuation’ threshold , and below another ‘desensitisation’ threshold . Both thresholds are agent-specific , with each agent sampling a threshold from a log-normal distribution; agents hence differ in their sensitivity to chemokines . Whilst agents are permitted to move freely within this three-dimensional space during simulation , only the locations of agents intersecting an xy plane at depth 30 µm are recorded for downstream analysis; this is reflective of the restricted z-depth of the whole well imaging experiments , facilitating comparison of results . Agent location changes are updated in 20 s increments of simulated time , but recorded at 5 min intervals , in line with experimental whole well imaging . Agent diameters are enlarged to 48 µm to facilitate visualisation in simulations . Non-cognate or cognate EL4 tumour cells were labelled with 10 μM CMTMR for 20 min in serum-free RPMI 1640 medium and washed twice . Labelled cells were returned to TCM to recover for 2 hr at 37°C and 5% CO2 before use . Effector OT1 CTLs were stimulated with non-cognate or cognate tumour cells at a 1:1 ratio for a final cell concentration of 6 × 106/ml in 250 μl of TCM per well in 96-well U-bottomed plates . The plates were centrifuged at 300 × g for 3 min before incubation for 2 . 5 hr at 37°C and 5% CO2 . OT1 and CMTMR+ tumour cells were then sorted ( two-way sort ) on the FACS-Aria III flow sorter ( BD Biosciences ) . To prepare OT1 stimulated by cognate beads , OT1 CTLs were incubated with beads ( prepared as described above ) for 3 hr . Non-cognate ( OT1 CTLs conjugated with EL-4 ) or cognate ( OT1 CTLs conjugated with EL-4 pulsed with SIINFEKL as described above ) supernatants were prepared and filtered as described above and were then used to incubate 3 × 106/ml in 250 μl OT1 CTLs for 3 hr prior to RNA extraction . Total RNA was prepared from up to 106 cells using the RNeasy Mini RNA Isolation Kit ( Qiagen , Hilden , Germany ) as per manufacturer’s instructions . The library was prepared using the TruSeq Stranded mRNA and sequenced on the Illumina NextSeq500 ( Illumina , San Diego , CA , USA ) generating paired-end 75 bp read lengths ( Ramaciotti Centre for Genomics , University of New South Wales , Sydney , NSW , Australia ) . For data analysis of RNAseq data , quality check of the raw reads was performed with FastQC ( version 0 . 11 . 5 ) . Trimmomatic ( version 0 . 36 ) was used to trim low quality reads with low-quality scores using the following parameters: leading = 3; trailing = 3; window_len = 4; window_qual = 15; minlen = 50; avgqual = 20 . Low-quality reads that did not pass this step were removed from downstream analysis . The reads were aligned against the GRCm38 reference mouse genome with Tophat2 using default parameters . Read counts were quantified by FeatureCounts ( version 1 . 5 . 1 ) ( results provided in Supplementary file 1 ) . The distribution of read counts revealed a uniform distribution across samples; therefore , no further normalisation was necessary . Differential expression analysis was performed from the total read counts by first calculating the fold change between pairs of samples . For the pairwise differential expression analyses , only genes with at least 100 reads in at least one of the two analysed samples were considered . The detection of differentially expressed genes was based on the fold-change between two conditions . The heatmap in Figure 3B was obtained by filtering for genes that encode secreted proteins ( based on UniProtKB database; available at www . uniprot . org ) from the total RNAseq data ( Supplementary file 1 ) . Fold changes were used to identify genes differentially expressed between pairs of samples . Read counts corresponding to these genes were then utilised to cluster the samples using the package pheatmap in R with the option of ‘scale = row’ . Statistical tests and heatmaps were performed with the pheatmap R package . The heatmap shows the differential expression of genes via z-scores ( also called standard scores ) , obtained by pairwise comparisons between all conditions for each gene . For each gene , the z-scores are calculated by subtracting the mean of the pairwise comparisons from the read count of each condition and then dividing the difference by the standard deviation for the gene . Zi , jread count ( gene i , condition j ) −mean ( gene i ) standard deviation ( gene i ) where zi , j is the z-score for gene i in condition j . To identify genes that interact with GPCRs , an up-to-date list of GPCRs was obtained from UniProtKB ( all entries indicated as olfactory , odorant , taste or vomeronasal receptors were removed ) and fed into STRING DB ( https://string-db . org ) as multiple protein entries under Mus musculus along with gene entries from the heatmap in Figure 4B . Genes of secreted proteins with >0 . 9 combined STRING score for interactions with GPCRs are highlighted in red in Figure 4B . Peptides ( 5 μl ) were separated by reverse-phase chromatography on a C18 fused silica column ( I . D . 75 μm , O . D . 360 μm × 25 cm length ) packed into an emitter tip ( IonOpticks , Melbourne , VIC , Australia ) , using a nano-flow HPLC ( M-class , Waters , UK ) . The HPLC was coupled to an Impact II UHR-QqTOF mass spectrometer ( Bruker , Bremen , Germany ) using a CaptiveSpray source and nanoBooster at 0 . 20 Bar using acetonitrile . Peptides were loaded directly onto the column at a constant flow rate of 400 nl/min with buffer A ( 99 . 9% Milli-Q water , 0 . 1% FA ) and eluted with a 90 min linear gradient from 2% to 34% buffer B ( 99 . 9% acetonitrile , 0 . 1% FA ) . MS spectra were acquired in a data-dependent manner including an automatic switch between MS and MS/MS scans using a 1 . 5 s duty cycle and 4 Hz MS1 spectra rate followed by MS/MS scans at 8–20 Hz dependent on precursor intensity for the remainder of the cycle . MS spectra were acquired between a mass range of 200–2000 m/z . Peptide fragmentation was performed using collision-induced dissociation ( CID ) . For data analysis , raw files consisting of high-resolution MS/MS spectra were processed with MaxQuant ( version 1 . 5 . 8 . 3 ) for feature detection and protein identification using the Andromeda search engine . Extracted peak lists were searched against the Mus musculus database ( UniProt , October 2016 ) , as well as a separate reverse decoy database to empirically assess the false discovery rate ( FDR ) using strict trypsin specificity allowing up to two missed cleavages . The minimum required peptide length was set to seven amino acids . In the main search , precursor mass tolerance was 0 . 006 Da and fragment mass tolerance was 40 ppm . The search included variable modifications of oxidation ( methionine ) , amino-terminal acetylation , the addition of pyroglutamate ( at N-termini of glutamate and glutamine ) and a fixed modification of carbamidomethyl ( cysteine ) . The ‘match between runs’ option in MaxQuant was used to transfer identifications made between runs on the basis of matching precursors with high mass accuracy . Peptide-spectrum match ( PSM ) scores and protein identifications were filtered using a target-decoy approach at a false discovery rate ( FDR ) of 1% . Only unique and razor peptides were considered for quantification with intensity values present in at least two out of three replicates per group . Statistical analyses were performed using LFQAnalyst ( https://bioinformatics . erc . monash . edu/apps/LFQ-Analyst/ ) , whereby the LFQ intensity values were used for protein quantification . Missing values were replaced by values drawn from a normal distribution of 1 . 8 standard deviations and a width of 0 . 3 for each sample ( Perseus-type ) . Protein-wise linear models combined with empirical Bayes statistics were used for differential expression analysis using Bioconductor package Limma whereby the adjusted p-value cutoff was set at 0 . 05 and log2 fold change cutoff set at 1 . The Benjamini-Hochberg method of FDR correction was used . Statistical analyses were performed using Prism 8 . 0 ( GraphPad Software , La Jolla , CA , USA ) or R ( The R Project for Statistical Computing ) . D'Agostino and Pearson normality tests or Kolmogorov-Smirnov test were used to determine whether or not the data follow a Gaussian distribution . For data with a non-Gaussian distribution , Mann-Whitney U tests were used to compare medians between two groups and Kruskal-Wallis tests to compare medians between more than two groups followed by Dunn’s multiple comparison tests . For transwell experiments , statistical significance between means of two groups was determined by performing two-tailed , unpaired Student’s t tests , and multiple means were compared with one-way ANOVA and Tukey’s multiple comparisons test . For the FMI , we used a two-tailed Wilcoxon signed rank test to compare medians with a hypothetical median value of 0 . n = 3 independent experiments for each condition , 3 fields of view per condition per experiment . Box-plots ( as depicted in Figure 2B ) : the box represents the 25th to 75th percentiles of the data points . The interquartile range ( IQR ) is the difference between the 25th and 75th percentiles . The upper whisker indicates the 75th percentile plus 1 . 5 times the IQR and the lower whisker indicates the 25th percentile minus 1 . 5 times the IQR . The data points above or below the whiskers are outliers . Mean is indicated by thick red line; median by thin black line . Coloured circles represent the means of individual experiments . For each condition , cell numbers are indicated underneath the plot . Coloured ‘violin’ plots for infiltration depth into tumouroid depict the distribution probability density; black bar represents the median , grey dots are data points . Mann Whitney tests were performed to compared medians of violin plots . In statistical analysis , p>0 . 05 is indicated as not significant ( ns ) , whereas statistically significant values are reported in the figures . Instantaneous FMI are calculated from cellular FMI values at each timeframe . Surfaces fitted and plotted through generalised additive models ( such as in Figure 3E ) were generated through custom code . Instantaneous FMI values were extracted from spatiotemporal positional data exported from Imaris using the ‘motility_analysis’ python package , available at https://github . com/marknormanread/TcellSwarming . Generalised additive models were then fitted using the ‘mgcv’ package; code also available at the aforementioned repository . For statistical and biological robustness , each experiment was performed at least three times with cells from different mice , except when stated otherwise .	Immune cells known as cytotoxic T lymphocytes , or CTLs for short , move around the body searching for infected or damaged cells that may cause harm . Once these specialised killer cells identify a target , they launch an attack , removing the harmful cell from the body . CTLs can also recognise and eliminate cancer cells , and can be infused into cancer patients as a form of treatment called adoptive cell transfer immunotherapy . Unfortunately , this kind of treatment does not yet work well on solid tumours because the immune cells often do not infiltrate them sufficiently . It is thought that CTLs arrive at their targets either by randomly searching or by following chemicals secreted by other immune cells . However , the methods used to map the movement of these killer cells have made it difficult to determine how populations of CTLs coordinate their behaviour independently of other cells in the immune system . To overcome this barrier , Galeano Niño , Pageon , Tay et al . employed a three-dimensional model known as a tumouroid embedded in a matrix of proteins , which mimics the tissue environment of a real tumour in the laboratory . These models were used to track the movement of CTLs extracted from mice and humans , as well as human T cells engineered to recognise cancer cells . The experiments showed that when a CTL identifies a tumour cell , it releases chemical signals known as chemokines , which attract other CTLs and recruit them to the target site . Further experiments and computer simulations revealed that as the number of CTLs arriving at the target site increases , this amplifies the chemokine signal being secreted , resulting in more and more CTLs being attracted to the tumour . Other human T cells that had been engineered to recognize cancer cells were also found to employ this method of mass recruitment , and collectively ‘swarm’ towards targeted tumours . These findings shed new light on how CTLs work together to attack a target . It is possible that exploiting the mechanism used by CTLs could help improve the efficiency of tumour-targeting immunotherapies . However , further studies are needed to determine whether these findings can be applied to solid tumours in cancer patients .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "immunology", "and", "inflammation" ]	2020	Cytotoxic T cells swarm by homotypic chemokine signalling
Immunoglobulins ( Igs ) are a crown jewel of jawed vertebrate evolution . Through recombination and mutation of small numbers of genes , Igs can specifically recognize a vast variety of natural and man-made organic molecules . Jawless vertebrates evolved a parallel system of humoral immunity , which recognizes antigens not with Ig , but with a structurally unrelated receptor called the variable lymphocyte receptor B ( VLRB ) . We exploited the convergent evolution of Ig and VLRB antibodies ( Abs ) to investigate if intrinsic chemical features of foreign proteins determine their antigenicity and immunogenicity . Surprisingly , we find lamprey VLRB and mouse Ig responses to influenza A virus are extremely similar . Each focuses ∼80% of the response on hemagglutinin ( HA ) , mainly through recognition of the major antigenic sites in the HA globular head domain . Our findings predict basic conservation of Ab responses to protein antigens , strongly supporting the use of animal models for understanding human Ab responses to viruses and protein immunogens . A cornerstone of modern biology and medicine is that humans and other mammals generate a highly specific humoral immune response when confronted with microbes or toxins , providing the basis for vaccination against many pathogens . Originally defined as a functional principle ( e . g . , ability to protect animals against injection with a toxin ) , the responsible substance was termed an antibody ( Ab ) , which we now know to consist of immunoglobulins ( Igs ) in jawed vertebrates . After more than a century of steady progress , a basic molecular understanding of Ig function is at hand ( Ramaraj et al . , 2012; Georgiou et al . , 2014 ) ; surprisingly little is known , however , regarding the basic rules of immunogenicity . In responding to viruses , why are some proteins more immunogenic than others ? Why do Ig responses focus on certain regions of proteins ? To what extent is this due to the specific features of Ig structure and combining site chemistry or repertoire limitations imposed by self-tolerance ? Here we address these issues by comparing the Ab responses of lampreys and mice to influenza A virus ( IAV ) , a model antigen of high practical relevance . IAV is composed of four major structural proteins: two surface glycoproteins , hemagglutinin ( HA ) and neuraminidase ( NA ) , embedded in a lipid envelope , lined by the matrix protein ( M1 ) , which encases the nucleoprotein ( NP ) coated viral genome . Due to its importance as the target of protective Igs ( Couch and Kasel , 1983 ) , HA is probably the most intensively characterized immunogen/antigen . Most HA-specific Igs with virus neutralizing activity bind to or bridge 5 antigenic regions in the globular domain ( termed Sa , Sb , Ca1 , Ca2 , and Cb ) , which surround the sialic acid receptor site that attaches HA to host cells ( Gerhard et al . , 1981 ) . Variation in these sites as IAV evolves in the human population ( ‘antigenic drift’ ) prevents effective IAV vaccination , necessitating frequent changes in vaccine formulation . The recent discovery that humans can generate protective Igs to conserved structures in the membrane-proximal HA stem have raised hopes of more effective vaccination if stem responses can be augmented using appropriately designed vaccines ( Laursen and Wilson , 2013 ) . Better understanding the rules of immunogenicity could inform these critical efforts . We characterized the variable lymphocyte receptor B ( VLRB ) response of lampreys , which along with hagfish , are the only known living jawless vertebrates ( cyclostomes ) . Cyclostomes branched evolutionarily from jawed vertebrates approximately 550 million years ago ( Mya ) ( Figure 1 ) . Pioneering studies from the Good laboratory established that lampreys generate Ab responses to protein and carbohydrate immunogens ( Finstad and Good , 1964 ) . More than 40 years later , Cooper and colleagues discovered that lamprey Abs , VLRBs , bear no relationship to Igs , but rather are structurally similar to Toll-like receptors ( Pancer et al . , 2004 ) . In place of RAG-mediated V ( D ) J recombination , VLRB diversity is generated using a gene assembly mechanism reminiscent of the activation-induced cytidine deamine-catalyzed gene conversion mechanism used to diversify Ig genes in birds and some mammals ( Figure 1 ) . 10 . 7554/eLife . 07467 . 003Figure 1 . Origin of variable lymphocyte receptor B ( VLRB ) in jawless vertebrates . Jawless and jawed vertebrates last shared a common ancestor ∼550 Mya . VLR genes are only in jawless vertebrates , whereas Immunoglobulin ( Ig ) genes are only in jawed vertebrates . However , both jawed and jawless vertebrates have a lymphocyte-based adaptive immune system suggesting that the genetic programs necessary for lymphocyte development originated in a common ancestor before the antigen receptor genes . Cytidine deaminases are expressed by lymphocytes in both jawed and jawless vertebrates and may have originated in a common ancestor; activation-induced cytidine deaminase ( AID ) and cytosine deaminase ( CDA ) . Structures of prototypic VLRB ( Top , PDB: 3e6j ) and IgG ( Bottom , PDB: 1Igt ) are shown to the right , along with cartoons of their secreted forms . Regions of antigen recognition are shaded in blue or green . In red are the concave antigen binding residues of VLR and the complementarity determining regions ( CDRs ) of Ig . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 003 The germline VLRB gene is incomplete because the invariant 5′ and 3′ coding sequences are separated by non-coding intervening sequences ( Pancer et al . , 2004 ) . Several of the hundreds of leucine rich repeat ( LRR ) -encoding genes flanking the VRLB gene are copied into the gene to generate an in-frame , functional VLRB gene during lymphocyte development ( Nagawa et al . , 2007; Rogozin et al . , 2007; Alder et al . , 2008 ) . This generates a VLRB repertoire with diversity comparable to Igs ( Alder et al . , 2005 ) . VLRB encodes for single-chain , crescent-shaped proteins that bind to antigens with a concave surface composed of multiple LRR β-strands and a C-terminal variable loop ( LRRCT ) ( Kim et al . , 2007; Han et al . , 2008; Herrin et al . , 2008; Velikovsky et al . , 2009; Kirchdoerfer et al . , 2012; Deng et al . , 2013; Luo et al . , 2013 ) . In contrast , Igs consist of a heavy and a light chain , each of which contributes three complementarity determining region loops to form a structurally distinct antigen-binding site ( Figure 1 ) . To probe the VLRB response to IAV , we collected blood from lamprey larvae immunized three times with inactivated , purified prototypic H1N1 PR8 IAV . Polyclonal VLRB primarily migrates on an SDS-PAGE gel as disulfide-linked multimers under non-reducing conditions and as monomers in the presence of reducing agents ( Alder et al . , 2008; Herrin et al . , 2008 ) . As seen previously ( Alder et al . , 2005 ) , monitoring plasma VLRB by immunoblotting revealed that unlike mammalian Ig , where immunization induces only minor increases in substantial serum levels , VLRB levels increase ∼sevenfold ( Figure 2A ) . ELISAs revealed that each immunized lamprey generated VLRBs that bind PR8 , but not a similar amount of plate-bound parainfluenza-3 virus , which is a genetically and serologically completely distinct enveloped virus , but similar in architecture and complexity to IAV ( Figure 2B ) . 10 . 7554/eLife . 07467 . 004Figure 2 . Lamprey make VLRBs specific for influenza A virus ( IAV ) after immunization with non-adjuvented , UV-inactivated virus . ( A ) Left , whole lamprey plasma ( 5 µl of naïve or immunized three times with PR8 [L9] ) electrophoresed on a 4–12% SDS PAGE gel probed with anti-VLR monoclonal Ab ( mAb ) by immunoblotting . VLR monomers ( ∼35–45 kDa ) are naturally cross-linked by disulfide bonds to form VLR multimers >100 kDa . Right , lane intensity measured by ImageJ for immunoblots of 2 µl Naïve ( 3 animals ) or PR8 immunized ( 8 animals ) probed with anti-VLR 2° Ab . Each point represents one animal . Data were analyzed by two-tailed t-test using PRISM software ( p < 0 . 01 ) . The mean signal from immunized plasma was 7 . 4 ± 1 . 8 × greater than the naïve mean . ( B ) Equal protein quantities of purified virus were adsorbed to ELISA plates and probed with lamprey plasma from either immunized ( n = 9 ) or naïve ( n = 2 ) animals . Data were analyzed by two-way ANOVA followed by Bonferroni multiple comparison using PRISM software ( *p < 0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 004 To determine the immunogenicity of IAV structural proteins , we measured serum from PR8-immunized mice and lamprey via ELISA using either detergent soluble proteins from purified virus ( HA , NA , M1 ) , or the detergent insoluble core ( NP , M1 , small amounts of other non-glycoproteins [Hutchinson et al . , 2014] ) ( Figure 3A and Figure 3—figure supplement 1 ) . This revealed that in both mice and lamprey , more than 90% of the functional ELISA response is specific for HA and NA , as shown by the large difference in titers between detergent soluble proteins from PR8 ( H1N1 ) vs X31 ( H3N2 ) , a reassortant virus with the PR8 internal proteins but serologically distinct HK68 glycoproteins . Genetically isolating HA from NA using the J1 ( H3N1 , PR8 internal proteins ) and P50 ( H1N2 , HK internal proteins ) reassortants shows that upwards of 80% of ELISA-detected Abs are specific for HA in lamprey and mice ( Figure 3B ) . Low binding to X31 and HK soluble proteins , which contain significant amounts of M1 ( Figure 3—figure supplement 1 ) , indicate that M1 is negligibly immunogenic ( note that internal viral proteins from H3 and H1 viruses are antigenically highly conserved ) . Further , the low serum titers against PR8 cores confirm that only a small fraction of Igs are specific for NP or a low abundance internal virion component . 10 . 7554/eLife . 07467 . 005Figure 3 . Immunodominance hierarchy against IAV for lamprey and mice is the same . ( A ) Scheme depicting reassortant virus components used for experiments in this figure . ( B ) Equal protein quantities split ( HA/NA/M1 ) and core ( NP/M1 ) antigens bound to ELISA plates were tested for binding to anti-PR8 mouse sera or lamprey plasma . Mouse data are representative of two mice with n = 4 independent experiments . Lamprey data are from three pooled animals with n = 4 independent experiments . ( C ) Same as Figure 3B , but using anti-HK lamprey plasma . Data are from three pooled animals with n = 4 ELISA replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 00510 . 7554/eLife . 07467 . 006Figure 3—figure supplement 1 . Detergent-split reassorted viruses . Western blot of reassorted virus components probed with mouse mAbs specific for H1N1 hemagglutinin ( HA ) + neuraminidase ( NA ) or cross-reactive nucleoprotein ( NP ) , M1 , and M2 dimer . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 00610 . 7554/eLife . 07467 . 007Figure 3—figure supplement 2 . PR8 antibodies ( Abs ) bind HA and NA but not M influenza proteins . HeLa cells transfected with pDZ control plasmid or PR8 HA , NA , M1/2 , NP , or NS1 plasmids were stained with their respective mouse mAb , mouse sera , or lamprey plasma raised against PR8 virus . Both lamprey plasma and mouse sera bound HA and NA . The mouse sera bound NP , but the lamprey plasma did not . Neither bound M1/2 proteins . Neither immune sera stained NS1 protein . Note naïve lamprey plasma has less total VLRB and a lower background than immune lamprey plasma . Data is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 00710 . 7554/eLife . 07467 . 008Figure 3—figure supplement 3 . PR8 immunized lamprey plasma binds purified NP protein , but not purified M1 by ELISA . Purified NP or M1 protein were adsorbed to ELISA plates and probed with mouse sera ( n = 2 ) or lamprey plasma from either immunized ( n = 4 ) or naïve ( n = 2 ) animals . Data were analyzed by two-way ANOVA followed by Bonferroni multiple comparison using PRISM software ( p < 0 . 0001 ) . Both PR8-immune mouse sera and lamprey plasma bound NP better than unimmunized control . Neither bound M1 protein . mAbs M2-1C6 for M1 and HB65 for NP are shown to confirm proper folding of purified protein on plate . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 008 Reciprocal immunization of lampreys with HK virus ( Figure 3C ) confirmed the dominance of HA . This experiment also provides a direct control for the specificity of lamprey VLRB for H1N1 vs H3N2 glycoproteins . Flow cytometry of cells expressing either HA , NA , NP , M1/M2 or NS1 ( which is present in virions [Hutchinson et al . , 2014] ) from transfected cDNAs stained with lamprey plasma showed that PR8 induced detectable VLRB responses to HA and NA but not NP , M1 , M2 , or NS1 ( Figure 3—figure supplement 2 ) . Similarly , mouse serum Ig was positive against HA and NA and negative for M1 + M2 , although there was a response to NP . While lamprey plasma did not bind plasmid expressed NP by flow , in ELISA , both immune lamprey plasma and mouse sera bound plated NP , but neither bound M1 ( Figure 3—figure supplement 3 ) . The lack of NP binding in the flow assay is most likely spurious; due to limited VLRB access to NP within permeabilized cells , or low signal . Next we examined the functionality of the lamprey anti-HA response as revealed by hemagglutination inhibition ( HI ) or infectivity neutralization assays . HI measures the ability of Abs to block HA-mediated IAV attachment to erythrocyte surface terminal sialic acids . PR8-immunized lamprey plasma gave HI titers of 1:30 against PR8 , but <1:5 against an H3N2 IAV and B/Lee , an influenza B virus , which is serologically totally distinct from IAV ( Figure 4A ) . Immune lamprey plasma also significantly inhibited PR8 infectivity in MDCK cells relative to naïve plasma ( Figure 4B ) . 10 . 7554/eLife . 07467 . 009Figure 4 . Lamprey VLRBs bind to hemagglutinin and neutralize infection . ( A ) Plasma from PR8-immunized lamprey inhibits PR8 hemagglutination at a 1:30 plasma dilution , but did not inhibit hemagglutination by either HK or B/Lee at any dilution . Data are representative of two experiments . ( B ) MDCK cells were infected with an MOI 0 . 07 of PR8 in the presence of titrated mAb supernatants ( H17L2 against PR8 or control 1 . 2F4 against influenza B/Lee ) or lamprey plasma ( L9 vs Naïve ) . After 8 hr cells were fixed , double-stained with anti-HA and anti-NP Igs . Cells positive for either HA or NP by flow cytometry were considered infected . Data from four independent experiments were normalized to control for different percentages of infection between experiments and fit to a variable dose–response curve . The best-fit , calculated infectious dose 50 ( ID50 ) was significantly lower for both the immunized plasma and PR8 specific Ig ( **p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 009 The vast majority of Igs that inhibit IAV hemagglutination and viral infectivity bind the HA globular domain . To test if this is also the major target of lamprey VLRBs , we used a panel of PR8 viruses with 3 , 6 , 9 , or 12 amino acid substitutions located among the five defined antigenic sites ( Das et al . , 2013 ) . ELISAs using intact wild-type or mutant viruses as immunoadsorbents show that lamprey plasma similarly detect antigenic drift in the globular domain , with a significant loss of binding with six substitutions and a loss of ∼60% of binding with 12 substitutions ( Table 1 ) . Similar binding is seen with mouse , guinea pig , and chicken PR8 immune seras ( Table 1—source data 1 ) . Factoring in the contribution of anti-NA Abs to the signal , this indicates that the bulk of mouse , guinea pig , chicken and lamprey HA-specific Ig and VLRB recognize the globular head domain , and therefore that the globular domain is the immunodominant antigen for both mice and lamprey . As predicted , anti-PR8 VLRBs could easily distinguish antigenic drift in H1N1 isolates from the 1940s and 1950s ( Table 1 and Table 1—source data 2 ) . 10 . 7554/eLife . 07467 . 010Table 1 . Lamprey plasma binding is sensitive to drifted viruses by ELISA and HIDOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01010 . 7554/eLife . 07467 . 011Table 1—source data 1 . Other animals show similar binding to sequential virus series . ELISA binding curves for lamprey plasma from Table 1 are plotted alongside PR8 immunized guinea pig , chicken and mouse sera against the same plated Sequential virus series . Each graph shows representative data on a single animal's sera . Percent change in area under curve between wt PR8 and Sequential 12 is shown on each graph ( ΔAUC ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01110 . 7554/eLife . 07467 . 012Table 1—source data 2 . More H1N1 isolates . ELISA binding curves for lamprey plasma from Table 1 against plated H1N1 isolates are plotted along with two additional isolates omitted from the Table . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01210 . 7554/eLife . 07467 . 013Table 1—source data 3 . Anti-HA stem Ab binding curves used to normalize amounts of plated HA in Table 1 . ( A ) ELISA binding curves for serially diluted anti-HA stem Ab ( 3A01 ) added to the Sequential virus series are show from each of the four experiments . ( B ) Same as ( A ) for the H1N1 isolate panel . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 013VirusELISA % ΔAUCHI Titer†PR8–40# Substitutions in HA head 31510 6−2520 9−46<10 12−59<10H1N1 isolates A/Weiss/43−61– A/Cameron/46−50– A/Malaysia/54−77–Equal amounts of each virus were plated and probed with lamprey plasma ( L25 ) or mouse sera . Data from four independent experiments were normalized to the BMax of the stem binding mAb 3A01 to allow precise comparison between viruses and replicates ( Table 1—source data 3 ) . Data were fit to a hyperbola and the percent change in area under curve ( AUC ) between PR8 and the indicated virus is reported . All curve-means were significantly different from PR8 by Two Way ANOVA followed by Tukey Multiple comparisons test , p < 0 . 001 , except PR8 vs 3 . †Agglutination inhibition of four HAU of input virus by lamprey plasma ( L27 ) occurred at the dilution reported . HA , hemagglutinin . The loss of lamprey VLRB and mouse Ig binding in similar proportions to the PR8 HA-head domain mutants implies that each recognizes similar epitopes and is subject to similar physicochemical rules of binding . To compare Ig and VLRB footprints , we competed lamprey plasma against mouse monoclonal Abs ( mAbs ) specific for defined HA antigenic sites for binding to PR8-coated ELISA wells . Although IAV immunization elicited VLRB responses in all lampreys tested , we selected lampreys with the highest titers for the competition experiments because these experiments require larger amounts of VLRB . Comparing the relative titers of inhibitory activity provides an approximate measure of binding proximity . Such competition assays are complicated by steric hindrance between Abs binding physically adjacent epitopes as well as more subtle positive and negative conformational effects that occur upon Ig binding to HA ( Lubeck and Gerhard , 1982 ) . Anti-PR8 plasma from lampreys 7 and 9 ( L7 , L9 ) competed with the mAbs tested , whereas neither naïve lamprey plasma nor mAbs to an irrelevant antigen competed with the mAbs ( Table 2 and Table 2—source data 1 ) . Also , while both L7 and L9 have a similar ELISA titer to whole virus , L9 plasma competes with mAbs specific for each of the five sites , but L7 fails to compete with Sa and Sb mAbs . This demonstrates that VLRB binding to the HA head does not uniformly block binding of all head-specific Igs and , importantly , indicates that fine antigen specificity varies among individual lampreys . We also infer this from the different titers observed against the various mAbs , an effect that is unlikely to be based on mAb affinity , since VLRBs are allowed to bind to HA prior to mAb addition . 10 . 7554/eLife . 07467 . 014Table 2 . IC75 values for anti-PR8 lamprey plasmas and guinea pig serum ( positive control ) in competition with defined HA mAbs by ELISADOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01410 . 7554/eLife . 07467 . 015Table 2—source data 1 . Competition ELISA against α-Head HA panel Abs . Data from Table 2 shown in graph form . Serially diluted unlabeled lamprey plasma raised against PR8 ( L7 , L9 or Naïve ) was added to PR8 immobilized on 96 well ELISA plates . After 1 hr incubation , a fixed concentration of each indicated hybridoma supernatant ( PEG-1 , H28E23 , H18 S413 , H35 C12 , H2 4B1 , H18 S210 , and Y8 2D1 ) was added at a predetermined concentration—65% of maximum binding ( EC65 ) . Data from three independent experiments were analyzed by Two Way ANOVA followed by Bonferroni Multiple Comparisons against the Naïve plasma data using PRISM . ( p < 0 . 05; p < 0 . 01; p < 0 . 001; p < 0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01510 . 7554/eLife . 07467 . 016Table 2—source data 2 . Competition ELISA against α-Head HA Fabs . Same as Table 2—source data 1 but with Fabs instead of hybridoma supernatants . p-value measurements determined with One-Way Anova followed by Dunnett's Multiple Comparison Test against Naïve plasma values . Stars indicate differences among whole groups . Data collected from only one experiment due to shortage of lamprey plasma . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 01610 . 7554/eLife . 07467 . 017Table 2—source data 3 . Immune lamprey plasma does not compete against stem binding Abs by ELISA . Serially diluted naïve or immune lamprey plasma raised against PR8 ( L29 ) on 96 well ELISA plates immobilized with PR8 . After 1 hr incubation , a fixed concentration of purified monoclonal C179 or 2G02 was added at EC65 . As a positive control , the two stem Abs were competed against each other or against an anti-HA head Ab ( H28E23 ) . Data are from at least two separate experiments with four total replicates . There was no statistical difference between the lamprey plasma curves . ELISA signal from these Abs is low , thus the curves are noisy . In contrast , the ‘2G02 then C179 curve’ is statistically different from the ‘H28E23 then 2G02 curve’ by two-tailed t-test ( p< 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 017L7L9L29Guinea pigStemEpitope†IgGFabIgGFabIgGFabIgGC179SaNC‡NC250400—29036 , 000NCSbNCNC320410—53017 , 000—Cb280610600980—86026 , 000—Ca1250350360300—86040 , 000—Ca2370—600———25 , 000—Stem————NC——3 . 52G02–––––––nM*Data was fit to a Hill Slope . IC75 value was calculated from the curve using PRISM . IgG data are from three independent experiments , Fab data are from one experiment due to limited lamprey plasma . †mAbs used—Sa: PEG-1; Sb: H28E23; Cb: H36 C12 ( IgG ) , H9 D3 ( Fab ) ; Ca1: H2 4B1; Ca2: H18 S413; Stem: C179 and 2G02 . ‡NC , no competition; ‘—’ , not determined . mAbs , monoclonal antibodies . To minimize steric effects , we extended these findings using Fab mAb fragment ( 25 kDa vs 150 kDa for intact Ig ) . The patterns observed with L7 and L9 were highly similar to those obtained with intact Igs . Plasma from an additional lamprey ( L29 ) also effectively blocked each of the four Fabs tested ( Table 2 and Table 2—source data 2 ) . L29 plasma did not , however , block binding of either a representative mouse or human mAb specific for the stem region with broadly neutralizing activity , suggesting that , as in mammals , the stem region is poorly immunogenic in lampreys ( Table 2 and Table 2—source data 3 ) . Rather , as in man and mouse , the lamprey immune system focuses on the globular domain , and further , recognizes highly similar epitopes . Our findings support the conclusion that Ig and VLRB demonstrate similar specificity for HA head epitopes . Although steric issues limit interpretation of the competition experiments , the common effect of amino acid substitution on VLRB and Ig binding argues strongly for convergent recognition of highly overlapping epitopes . Equally surprising is the similarity of the overall immunodominance profile of IAV immunogens in mammals and lamprey responding to inactivated intact virus: the HA globular domain is favored over the stem; HA is favored over NA; and internal proteins , despite their relative abundance in the virion , are weakly ( NP ) or undetectably ( M1 ) immunogenic . Our findings extend understanding of the cyclostome VLRB response . We confirmed the Pancer group's finding that total plasma VLRB concentration increases after immunization ( Alder et al . , 2005 ) . Total Ig can also increase in jawed fish after immunization ( Castro et al . , 2013; Xu et al . , 2013 ) . This differs from mammals , where high levels are maintained constitutively . The ability of lampreys to respond to purified IAV without additional adjuvant indicates that lampreys recognize viral RNA or other viral innate immune activating molecules . We also show in competition assays that individual lampreys mount VLRB responses against different antigenic sites . Whether this is due to genetic , environmental , or stochastic factors is a question for further study . It is well worth noting that there is little information regarding how individuals among outbred groups vary in their Ig response to viruses or other complex antigens . It is surprising that such structurally disparate molecules as Ig and VLRB receptors recognize the same proteins with overlapping epitopes . These recognition systems have evolved independently for >500 Mya , presumably shaped by different selective pressures from the environment , self-tolerance , and microorganisms . These differences notwithstanding , the four available VLRB-antigen crystal structures reveal many similarities in Ig and VLRB antigen recognition . In the two VLRB structures with protein antigens ( hen egg lysozyme and anthrax coat protein , BclA ) , the contact area ( ∼1500 Å2 ) is in the same range as reported for Igs ( 1400–2300 Å2 ) , and utilizes the same non-covalent forces ( salt bridges , hydrogen bonds , and van der Waals contacts ) in similar proportions to mediate binding , which requires close shape complementarity to the antigen ( Velikovsky et al . , 2009; Kirchdoerfer et al . , 2012; Deng et al . , 2013 ) . In the VLRB-HEL structure , the relatively small , crescent-shaped VLRB binds to an epitope in the catalytic cleft , whereas larger , dimeric Ig VHVL Abs bind to flatter epitopes away from the catalytic site . Interestingly , structures of single-chain camelid VHH and shark IgNAR have revealed that they also favor the catalytic cleft of HEL that is presumably sterically inaccessible to dimeric VHVL Abs ( Velikovsky et al . , 2009 ) . Analysis of hundreds of VLRB sequences has previously revealed a bias towards aromatic amino acids at the variable positions on the concave surface ( Velikovsky et al . , 2009 ) . In these analyses , less than half of the variable position residues contact antigen . When only the antigen contacting residue frequency is quantified , the amino acids are biased towards Tyr , Trp , Asn , and Asp residues ( Figure 5 ) . A similar bias towards these residues in antigen-contacting positions of Ig has also been observed ( Mian et al . , 1991; Davies and Cohen , 1996; Ramaraj et al . , 2012; Robin et al . , 2014 ) . These similarities may account for the recognition of similar epitopes . If so , this may also represent the optimal general solution for producing a single family of receptors capable of recognizing what is essentially an infinite array of antigens with high specificity and affinity . 10 . 7554/eLife . 07467 . 018Figure 5 . Paratope signature of VLRBs . ( A ) Contact residues determined by the crystal structures of VLRBs in complex with their antigens are highlighted in orange . RBC36 against H trisaccharide ( 3E6J ) ; aGPA . 23 against TF disaccharide ( 4K79 ) ; VLR4 against BclA ( 3TWI ) ; and VLRB . 2D against HEL ( 3G39 ) . ( B ) Enrichment or shortfall of each amino acid in the contact residues relative to the total amino acids found in the full VLRB was determined from the ratio of frequency of each amino acid in contact residues vs the frequency in the total VLRB sequence . Leucine was excluded from the analysis as it is the major structural amino acid of VLRBs . No M , K , or P were found among the contact residues . Shortfall was determined by estimating , based on total VLRB frequency , how many amino acids would be there if the amino acid distribution was even throughout the VLRB . DOI: http://dx . doi . org/10 . 7554/eLife . 07467 . 018 The universality of Ig and VLRB antigenicity and immunogenicity illuminated by our findings provides strong support for the utility of animal models for understanding human Ab responses to vaccines and other medically relevant immunogens . Perhaps , when it comes to Ab responses , neither mice nor lamprey lie , after all . We used the following Abs for ELISA and immunoblot experiments: 1:3000 mouse α-HA1 mAb , CM1 ( Magadan et al . , 2013 ) ; 1:2000 α-M1 , M2-1C6 , anti-Mouse recognizes 9 kDa N-terminal fragment ( Yewdell et al . , 1981 ) ; 1:3000 α-NP C-Terminal rabbit polyclonal 2364 , 487–498; 1:10 , 000 α-NP HB-65 ( Yewdell et al . , 1981 ) ; 1:3000 α-NA C-Terminal anti-Rabbit polyclonal ( Dolan et al . , 2010 ) ; 1:2500 mouse α-lamprey VLRB 4C4; 1:100 C179 α-HA Stem anti-Mouse mAb ( Takara ) ; α-HA stem anti Human 3A01 and SFV005 2G02 g02; 1:5000 Donkey α-Mouse IRDye 800 nm ( Li-Cor ) ; 1:5000 Donkey anti-Rabbit IRDye 680 nm ( Li-Cor ) ; 1:5000 Mouse α-Flag M2 ( Sigma ) ; 1:2000 α-Mouse Kappa-HRP ( Southern Bioscience ) ; 1:2000 α-Rabbit Kappa-HRP ( Jackson ) ; 1:2000 α-Human-HRP ( Jackson ) ; and 1:2000 α-Guinea Pig-HRP ( Jackson ) . Viruses B/Lee , A/HK/68 ( HK ) , A/PR8/MCa ( PR8 ) , X31 ( HK HA and NA , PR8 background ) , J1 ( HK HA , PR8 background ) , and P50 ( PR8 HA , HK background ) were grown by pipetting 250 TCID50 viral units diluted in 50 µl 0 . 1% BSS/BSA into 10-day-old eggs . We collected allantoic fluid after 48 hr , clarified at 3000 RPM for 20 min , and pelleted virus through 20% sucrose by centrifuging for 2 hr at 26 , 000×g . We incubated pellets overnight in 2 ml PBS with calcium and magnesium ( PBS++ ) and purified by centrifuging virus on a discontinuous 15–60% sucrose gradient , collecting virus at the interface , and pelleting 34 , 000×g for 2 hr . After resuspending the pellet in 500 µl PBS++ overnight , we completely inactivated viral infectivity ( as determined by adding to MDCK cells ) by exposing for 20 min to 254 nm light at 2 . 4 mW/cm2 , and sterilized virus by passing through a 0 . 22-µm PDVF filter ( Millipore ) . We measured total viral protein with the DC Protein Assay ( Bio-Rad ) . We fractionated purified viruses by incubating 250 μl purified virus with 250 μl 15% octyl-β-glucoside . After pipetting until the opalescent solution became clear , we added 50 μl 10% NP-40and pipetted further before adding 950 μl PBS++ . We centrifuged virus at 50 , 000×g for 2 hr at 4°C , collected the top 1 ml and removed detergents with detergent removal spin column ( Pierce ) . We resuspended the pellet ( cores ) in 1 ml PBS++ , repelleted and sonicated cores into PBS ++ . Lamprey larvae ( Petromyzon marinus ) captured from the wild by commercial fishermen ( Lamprey Services , MI ) were housed in sand-lined aquariums maintained at 16–18°C using a water chiller and fed brewer's yeast . The lampreys were immunized three times by intracoelomic cavity injections spaced 2–3 weeks apart containing ∼10 µg virus diluted in 30 µl of 0 . 67× PBS ( to match lamprey tonicity ) . 2 weeks after final immunization , ∼200 µl lamprey plasma was collected in 300 µl of 30 mM EDTA , an anticoagulant and stored at 4°C in 20 mM MOPS pH 7 . 2 buffer and 0 . 025% sodium azide to prevent microbial growth . Plasma was also collected from non-immunized lampreys to serve as naïve controls . Leukocytes were harvested from the blood of each animal by collecting cells at the interface of a 55% Percoll gradient and banked in RNAlater ( Qiagen ) for future characterization . Guinea pigs and mice were immunized with an intramuscular injection of ∼10 μg virus in 25–50 μl PBS and boosted 2 weeks later . Serum was collected 2 weeks after the boost . Mouse and guinea pig studies were approved by and performed in accordance with the Animal Care and Use Committee of the National Institute of Allergy and Infectious Diseases . We incubated purified IAV ( ∼0 . 05 µg per well in 100 µl PBS++ ) , purified A/WSN/33 NP ( Ye et al . , 2013 ) generously provided by Dr Yizhi Jane Tao and Dr Yukimatsu Toh at Rice University ( 100 nmol/well ) , or purified matrix protein ( Oxford and Schild , 1976 ) ( 4 nmol/well ) in Immulon 4HBX 96 well plates for 12 hr to 7 days at 4°C on an orbital shaker ( all incubations were similarly shaken ) . Just before using , we washed plates 3× with PBS + 0 . 05% Tween-20 ( PBST ) . We then incubated 100 µl diluted primary sera for 1 hr at 4°C , washed with PBST , and for lamprey plasma , incubated with 100 µl 4 . 4 nM mouse anti-VLR 4C4 mAb for 1 hr at 4°C . After washing 3× with PBST , we incubated wells with 100 µl 1:2000 HRP anti-mouse κ chain ( Southern Biotech ) in PBS++ 1 hr at 4°C , washed 3× with PBST and added 100 µl/well SureBlue Peroxidase Substrate ( KPL ) . After 5 min at room temperature , we inactivated HRP with 50 µl/well 1 M HCl . And measured absorbance at 450 nm . The absorbance data was graphed and fit to Hill Equation using PRISM software . For competition ELISA , we first incubated wells with competing Ig or VLRB for 1 hr at 4°C , and then added mouse mAbs as supernatants or purified Fab fragments at a concentration equivalent to the Igs' EC65 binding . Data was fit to a Hill Equation from which the IC75 was calculated in PRISM . After an hour at 4°C , we washed the plates , and developed with anti-mouse HRP Ig and peroxidase substrate , as described above . We mixed purified IAV ( 0 . 5–4 µg protein ) with 4× NuPage loading buffer ( Invitrogen ) , with or without 4 mM DTT , and boiled for 15 min at 96°C . We electrophoresed samples with SeeBlue Plus2 ladder on 4–12% Bis-Tris Gels ( Invitrogen ) at 180 V for 90 min . To visualize proteins , we fixed gels for 10 min with 10 ml 10% acetic acid and 50% methanol , shaking at RT . After removing fixative we added 10 ml GelCode stain ( Pierce ) and shook for 30 min at room temperature , then destained the gels with water overnight . For immunoblotting , we transferred proteins from gels to PVDF membranes with the iBLOT at P3 setting for 7 min . We blocked membranes for 1 hr at room temperature with either 10% BSA in water for blots probed with lamprey plasma or with StartingBlock for mouse Abs ( Thermo ) . After incubating with primary Ig or VLRB 1 hr at room temperature , washing 5× for 5 min each in TBST ( 10 mM Tris , 150 mM NaCl , 0 . 1% Tween-20 ) , we added secondary and tertiary Ig , repeating the washing step after each incubation . We imaged blots on a Li-Cor Odyssey . We transfected HeLa cells using Lipofectamine LTX ( Life Technologies ) with PR8 proteins HA , NA , M , NP , or NS1 in a pDZ vector and cultured for 24 hr . To enable Ig or VLRB access to internal proteins , we fixed and permeabilized NP , NS1 , and M transfected cells with FoxP3 buffer ( eBiosciences ) . We stained all cells with appropriate dye-labeled Ig , mouse sera , or lamprey plasma and analyzed samples using a LSR II flow cytometer and fitted data to a one-site binding hyperbola model using PRISM software . For HI , we treated lamprey plasma for 56°C for 30 min and incubated with four HAU PR8 before adding human O+ erythrocytes . For neutralization , we cultured MDCK cells ( 350k ) in 24-well plates overnight . The next day we added PR8 at a MOI of 0 . 07 in the presence of either H17-L2 ( anti HA mAb ) or control 1 . 2F4 against influenza B , or lamprey plasma , animal L9 vs naïve plasma diluted in 0 . 1% BSS/BSA . After 1 hr incubation , we removed the supernatant and replaced with complete media . After 7 more hours , we trypsinized cells , fixed and permeabilized cells with FoxP3 Buffer and stained with anti-HA and anti-NP mAbs , then analyzed cells with a LSR II flow cytometer . Single cells positive for either HA or NP by flow cytometry were considered infected . Data were fitted to a variable dose-response curve and the best-fit infectious dose 50 ( ID50 ) calculated using PRISM software .	Influenza viruses infect ten of millions of people each year . To conquer a flu infection , the human immune system develops antibodies that hasten recovery and prevent future flu infections . Unfortunately , flu is constantly changing in response to the human immune response , and antibodies induced by previous infection or vaccination provide partial protection , at best , against new strains . An ideal flu vaccine would stimulate the immune system to produce antibodies that protect against all future strains of influenza . Most human antibodies that are induced by influenza target a part of the virus called the hemagglutinin , which attaches the virus to cells to start a flu infection . Some hemagglutinin-specific antibodies recognize many strains of influenza , but individuals do not produce enough of these antibodies to prevent infections with new strains . A basic understanding of what drives the production of different types of antibodies is important to devise vaccines that produce broadly effective antibodies for flu and for other viruses and pathogens that have proven to be difficult vaccine targets . To better understand the rules of antibody generation , Altman et al . compared antibodies produced in response to flu in mice and lampreys . Lampreys are a primitive fish that branched off from other vertebrates ( animals with a backbone , like you ) 550 million years ago and developed their own system of antibody recognition based on a completely different template . Despite this , Altman et al . found that the antibody response of mice and lampreys to flu is remarkably similar . Of several potential viral targets , antibodies from both mice and lampreys were predominantly directed against the hemagglutinin . Of numerous potential locations on the hemagglutinin , mouse and lamprey antibodies predominantly recognized the same region . These similarities suggest that the specificity of antibodies is based largely on the properties of the virus , and varies little with the properties of the responding organism . Most importantly , this supports the conclusion that studies in mice and other mammals are likely to accurately predict how humans will respond to vaccines for viruses and other pathogens .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "immunology", "and", "inflammation" ]	2015	Lamprey VLRB response to influenza virus supports universal rules of immunogenicity and antigenicity
Larval recruitment , the transition of pelagic larvae into reef-associated juveniles , is a critical step for the resilience of marine fish populations but its molecular control is unknown . Here , we investigate whether thyroid-hormones ( TH ) and their receptors ( TR ) coordinate the larval recruitment of the coral-reef-fish Acanthurus triostegus . We demonstrate an increase of TH-levels and TR-expressions in pelagic-larvae , followed by a decrease in recruiting juveniles . We generalize these observations in four other coral reef-fish species . Treatments with TH or TR-antagonist , as well as relocation to the open-ocean , disturb A . triostegus larvae transformation and grazing activity . Likewise , chlorpyrifos , a pesticide often encountered in coral-reefs , impairs A . triostegus TH-levels , transformation , and grazing activity , hence diminishing this herbivore’s ability to control the spread of reef-algae . Larval recruitment therefore corresponds to a TH-controlled metamorphosis , sensitive to endocrine disruption . This provides a framework to understand how larval recruitment , critical to reef-ecosystems maintenance , is altered by anthropogenic stressors . Life history transitions are critical for many animal species and often correspond to concomitant developmental and ecological shifts ( Bishop et al . , 2006 ) . Unfortunately , little is known about how internal and external cues act in concert during these events . The life cycle of most teleost reef fish include a major developmental and ecological transition . Adults reproduce in the vicinity of the reef , emitting eggs that disperse and hatch in the ocean , where the larvae grow ( Leis and McCormick , 2002 ) . Larvae migrate back and enter reefs where they become juveniles , a step called larval recruitment ( Leis and McCormick , 2002 ) . This step involves the perception of environmental cues for larvae to localize and settle in the reef , and is accompanied by major morphological changes ( McCormick et al . , 2002; Lecchini et al . , 2005b , 2005a; Lecchini et al . , 2013; Barth , 2015 ) . This transition of pelagic larvae into reef-associated juveniles is often referred to as metamorphosis and is critical for the maintenance of reef fish populations , but its molecular control remains largely unknown ( Dufour and Galzin , 1993; Doherty , 2002; Leis and McCormick , 2002; McCormick et al . , 2002; Dixson et al . , 2011; Barth , 2015 ) . Since larval recruitment is an ecological event coupled to a morphological transformation , reminiscent of the situation in amphibians , it may correspond to a thyroid hormones ( TH ) controlled metamorphosis ( Bishop et al . , 2006; Brown and Cai , 2007 ) . TH and their receptors ( TR ) trigger and coordinate metamorphosis of many species such as Xenopus ( Brown and Cai , 2007 ) or flatfish ( Solbakken et al . , 1999 ) . The thyroid gland produces mainly the thyroxine hormone ( T4 ) , which is peripherally transformed into triiodothyronine ( T3 ) , the active form ( Chopra , 1996 ) . The transformation of T4 into T3 , and the degradation of T4 and T3 are controlled by a family of enzymes called deiodinases ( Bianco and Kim , 2006 ) . TH levels peak at the climax of metamorphosis . Precocious treatment with TH triggers metamorphosis whereas goitrogen ( TH synthesis inhibitors ) treatment blocks it ( Tata , 2006 ) . Similar mechanisms have been described in several teleost fishes , but in contexts disconnected from the natural environment ( Brown , 1997; de Jesus et al . , 1998; Yamano and Miwa , 1998; Kawakami et al . , 2003; Marchand et al . , 2004; McMenamin and Parichy , 2013 ) . In the convict surgeonfish Acanthurus triostegus , larvae have a planktonic diet while juveniles and adults are herbivorous , and external morphological changes occurring at recruitment have been previously described ( Randall , 1961; McCormick , 1999; McCormick et al . , 2002; Frédérich et al . , 2012 ) . Interestingly , crest-captured larvae can delay their metamorphosis when relocated to the external slope of the reef ( open ocean ) , showing an influence of the environment on their morphological transformation ( McCormick , 1999 ) . Nevertheless , again , the molecular mechanisms remain unknown . Here we report that A . triostegus larval recruitment to the reefs corresponds to metamorphosis with major pigmentation changes , remodeling of the digestive tract and behavioral modifications . Our results show that TH levels and TR expressions control this remodeling process . Furthermore , recruiting larvae can freeze their TH signaling when relocated on the external slope , demonstrating an influence of the environment on metamorphosis . Chlorpyrifos ( CPF ) , a known reef pollutant and endocrine disruptor ( Roche et al . , 2011; Botté et al . , 2012; Juberg et al . , 2013; Slotkin et al . , 2013 ) , affects A . triostegus metamorphosis by impairing TH signaling , preventing intestine lengthening , and inhibiting fish grazing activity . CPF therefore impairs A . triostegus control on algae spreading , which is a major threat for reef conservation . We extend our observation upon TH levels to four other coral reef fish species . Our work provides a unifying framework that integrates the developmental , ecological and evolutionary perspectives of vertebrate life history transitions . The involvement of the TH signaling pathway in this key post-embryonic step , thus prone to endocrine disruption , provides an obvious entry point to study how anthropogenic stressors may affect reef fish populations . A . triostegus crest captured individuals ( n = 297 ) experience an important weight loss between day 1 ( 0 . 71 ± 0 . 01 g ) and day 8 ( 0 . 53 ± 0 . 01 g ) following their entry in the reef ( Figure 1A–B ) . Individuals also undergo extensive pigmentation changes ( Figure 1C–L ) , as exemplified by the very rapid appearance of black stripes ( less than four hours , MB pers . obs . ) after entering the reef ( Figure 1C–D , H-I ) . This is followed by the widening of these vertical black stripes and by the onset of the body’s white pigmentation ( Figure 1D–G , Figure 1I–L ) . Furthermore , using conventional X-Ray microtomography , we identified three different dental generations in crest larvae to day 8 juveniles ( red , blue and green , Figure 1M–Q ) . Crest larvae present tiny and poorly mineralized pointed teeth at the distal ends of the jaws ( red , Figure 1M , Figure 1—figure supplement 1R ) that are likely remnants of an oceanic larval dentition ( dentition A ) . Other medium-sized and more mineralized teeth are in function in medial areas of the jaws ( blue , Figure 1M ) . These teeth display pointed cusps and belong to a more advanced dental generation ( dentition B ) . Lastly , much larger teeth of the next generation ( green , dentition C ) are about to erupt from jaws in crest larvae ( Figure 1M ) and replace A and B previous dentitions in day 2 to day 8 juveniles ( Figure 1N–Q ) . These teeth are highly similar to the shovel-shaped adult teeth ( Figure 1—figure supplement 1S ) that serve for grazing algae on hard substrata . The rapid formation of dentition C is consistent with a diet shift from planktivorous to herbivorous at recruitment . This shift is precisely organized , since it begins in the sagittal area of all individuals , and secondarily extends to the distal parts of the jaw ( Figure 1M–Q ) . The fact that crest captured larvae already have teeth of the C dentition about to erupt ( Figure 1M ) indicates that fish are prepared for reef life conditions before reef entry . The intestine also goes through a drastic remodeling at recruitment . In particular , the gastro-intestinal tract lengthens from about 2 . 52 ± 0 . 07 cm in crest captured larvae to 7 . 38 ± 0 . 14 cm in day 8 juveniles ( Figure 1W ) . To characterize this change , we performed histological sections on the middle part of the intestine during the larval recruitment process ( Figure 1R–V ) . Intestines of crest larvae and day 2 juveniles have a thick wall with many muscle fibers and regular villi ( Figure 1R and S ) . At day 3 , intestines undergo a spectacular remodeling with the disappearance of the epithelium , leaving only the muscular layer ( Figure 1T ) , while epithelium reformation occurs later at day 5 ( Figure 1U ) . In day 8 individuals , the intestine is fully remodeled: the epithelium is thicker and exhibits villi again ( Figure 1V ) . Such spectacular remodeling events are also observed in the proximal and distal portion of the intestines ( Figure 1—figure supplement 1A–O ) . To assess whether and to what extent this intestine remodeling was correlated to a diet shift at larval recruitment , we sequenced A . triostegus digestive tract content ( 18S mass sequencing ) to identify its eukaryotic composition . We investigated crest captured individuals , day 8 juveniles and adults ( Figure 1W ) using a blocking primer to minimize amplification of host sequences . Crest larval intestines are largely empty although we identified sequences of fungi and algae . In day 8 juveniles , the intestinal content is different with more brown and red algae , as well as few protostome sequences . This indicates a shift in the eukaryotic gut content between crest individuals and day 8 juveniles . Adults gut content is less diversified and is mainly composed of brown algae , red algae and alveolates , different from those found in day 8 juveniles ( Figure 1W ) . Using 16S metagenomic sequencing , we found differences in the relative quantity of bacterial community between the juveniles at day 2 , 5 and 8 ( Figure 1—figure supplement 1P ) . The 16S profile of the adult is also different from the juvenile profile , with the notable presence of Lachnospiraceae , among which the giant bacteria Epulopiscium fishelsoni is a well-known symbiont of Acanthuridae ( Clements and Bullivant , 1991 ) . Given the role of TH in vertebrate metamorphosis ( Laudet , 2011 ) and the spectacular transformations that A . triostegus undergoes at larval recruitment , we analyzed TH levels in larvae captured in the far ocean ( from 5 to 10 km offshore , n = 3 ) , near ocean ( 2 km offshore , n = 5 ) , crest-captured larvae and juveniles ( raised in situ in lagoon cages , n = 297 ) up to day 8 after recruitment ( Figure 2A–B ) . T4 levels are low in far ocean larvae and rise in fish located closer to the reef ( near ocean larvae ) , before dropping 6-fold between the near ocean larvae and the day 8 juveniles ( p-value=0 . 001 , Figure 2A ) . This indicates a peak of T4 in near ocean and in crest larvae . Far ocean larvae exhibit a high level of T3 but with an important variability ( Figure 2B ) . Although these larvae live in the same ecological niche ( i . e . open ocean ) and are roughly of the same size ( 4 . 7 , 5 . 5 , and 6 . 3 mm ) , they were captured in different locations and are probably not at the same developmental stage . Near ocean and crest captured larvae have similar levels of T3 that are 6-fold higher than the level observed in day 8 juveniles ( p-value < 0 . 001 Figure 2B ) . We characterized the three TR of A . triostegus: TRα-A , TRα-B and TRβ . These receptors have between 86% ( TRα-A ) and 96% ( TRβ ) sequence identity with their zebrafish orthologs ( Figure 2—figure supplement 1A ) . The phylogenetic analysis indicates that each of the three TR clusters with its respective teleosts orthologs ( Figure 2—figure supplement 1B ) . Moreover , transactivation assays show that these receptors behave as genuine TR ( Figure 2—figure supplement 1C ) . We thus assessed the expression levels of TRα-A , TRα-B , and TRβ during A . triostegus development ( Figure 2C ) , as well as Klf9 , a conserved TR regulated gene ( Denver and Williamson , 2009 ) . Far ocean larvae have two profiles: one with a very low TR and Klf9 expressions ( ‘Far a’ ocean ) and one with very high levels of expression ( ‘Far b’ ocean ) ( Figure 2C ) . This suggests that there are , at least , two distinct situations in far ocean larvae: larvae with low TH/TR levels and larvae in which TH/TR levels surge . Far b ocean , near ocean , and crest larvae have high expression levels of the all the TR as well as Klf9 , indicating a strong activation of the TH signaling at these stages . We observed a drop of all TR expression between crest larvae and day 8 juveniles ( p-values=0 . 001 , 0 . 005 and 0 . 002 for TRα-A , TRα-B and TRβ respectively , Figure 2C ) . To link the TH pathway , the environment and the metamorphic changes occurring at larval recruitment , we disturbed the TH signaling and the location of crest-captured larvae during their metamorphosis ( from capture to up to day 8 after reef colonization ) . In order to study how the environment controls the metamorphosis processes , we first relocated crest-captured larvae back to the external slope immediately after their reef entry . Indeed , mimicking oceanic condition has been acknowledged to delay metamorphosis ( McCormick , 1999 ) ( Figure 3A , upper panel ) . Similar to this earlier study , we observed a striking delay in the white pigmentation appearance between the vertical black stripes in day 2 relocated fish compared to their lagoon raised counterparts ( Figure 3A , lower panel ) . The levels of T4 and T3 in fish relocated on the external slope at day 2 , day 5 and day 8 are overall higher than the levels of their control counterparts in the lagoon ( Figure 3B–C ) . The levels of T4 , in particular , remain similar in day 2 to day 8 juveniles to the levels in the crest condition ( Figure 3B ) . The expressions of TRs and Klf9 in the relocated juveniles were also higher than in control juveniles at the same age ( Figure 3E ) . This suggests that the environment effectively controls the metamorphosis by modulating the TH signaling . To assess the implication of the TH signaling pathway on metamorphic processes , we then raised crest-captured larvae in situ in the lagoon and injected them daily , in their ventral cavity , with 20 µl of different drug solutions that activate or disrupt their TH molecular pathway . Four drug treatments were applied: ( i ) solvent control ( DMSO diluted 10 . 000 times in Phosphate Buffer Saline 1X , as all drugs were made soluble in DMSO and diluted 10 . 000 times in PBS 1X ) ; ( ii ) T3 +iopanoic acid ( IOP ) both at 10−6 M , IOP being used as an inhibitor of the deiodinase enzymes that therefore prevents the degradation of injected T3 ( Galton , 1989; Little et al . , 2013; Lorgen et al . , 2015 ) ; ( iii ) NH3 at 10−6 M , NH3 being used as an antagonist of TR that prevents the binding of TH on TR ( Lim et al . , 2002; Renko et al . , 2012 ) therefore disrupting A . triostegus TH pathway ( Figure 2—figure supplement 1D ) ; and ( iv ) T3 + IOP + NH3 all at 10−6 M to ensure the non-toxicity of NH3 . We used intestine total length , internal structure of guts ( intestinal microvilli ) , dentition and pigmentation as markers of the advancement of metamorphosis since they strongly change after reef recruitment ( Figure 1 ) . T3 + IOP treatment increases the length of the intestine in day 2 and day 5 juveniles ( p-values<0 . 001 , Figure 3E , blue ) . On the contrary , NH3 treatment partially prevents the intestines lengthening occurring after recruitment ( p-values<0 . 001 , Figure 3E , green ) . Injections of T3 + IOP + NH3 result in an intestine length similar to the control individuals at day 2 and day 5 ( p-values=0 . 486 and 0 . 444 respectively , Figure 3E ) . This shows that T3 and NH3 effectively compete , and confirms the non-toxicity of NH3 . In parallel , intestines of fish relocated on the external slope are shorter than those of their lagoon counterparts at day 2 and day 5 ( p-value<0 . 001 and p-value=0 . 042 respectively , Figure 3E , orange ) . This shows that their development has been slowed down , similarly to what we observed with the NH3 treatment . A similar pattern was observed concerning the microvilli remodeling within guts . In the anterior part of the guts , while T3 + IOP treatment accelerates the loss of the intestinal microvilli at day 2 and accelerates the development and the thickening of a new epithelium at day 5 ( Figure 3F , second column ) , NH3 treatment and external slope relocation prevent microvilli renewal at both day 2 and day 5 ( Figure 3F , third and fourth column ) . These observations can be extended to the anterior , medium and posterior sections of the guts ( see Figure 3—figure supplement 1 ) . These changes are consistent with the notion that TH control remodeling of the guts at metamorphosis in A . triostegus . We did not observe any effect on teeth development neither with T3 + IOP or NH3 treatment , nor with external slope relocation ( Figure 3—figure supplement 1 ) . Given that teeth remodeling is a lengthy process that starts very early on in oceanic larvae and therefore before reef colonization , it is not surprising that the treatments performed on crest captured larvae do not affect ( or are too late to affect ) teeth development . Also , contrarily to individuals relocated on the outer slope , we did not observe any effect of T3 + IOP nor NH3 treatments on skin pigmentation at day 2 and day 5 ( Figure 3—figure supplement 2 ) . This suggests that pigmentation during the larva to juvenile transition is strongly coupled to the environment but is not directly under TH control that may not sustain such rapid pigmentation changes . Taken together: ( i ) intestine lengthening and remodeling are mediated by TH signaling; and ( ii ) disturbing the TH pathway disrupts the normal development of fish after larval recruitment , showing that TH effectively controls some major aspects of A . triostegus metamorphosis . Finally , we tested how TH signaling influences fish behavior . To achieve this , we used grazing activity ( number of bites on algal turf ) , as A . triostegus larvae start to graze a couple of days after colonizing the reef , to investigate how TH signaling and metamorphosis completion influence the behavior of recruited fish . T3 + IOP treatment increases the grazing activity by around 50% in day 5 juveniles when compared to the control individuals ( p-value=0 . 004 , Figure 3G , blue lanes ) . On the contrary , NH3 treatment significantly decreases the number of bites at day 2 by 33% ( p-value<0 . 001 , Figure 3G , green lane ) . The grazing activity of fish relocated on the external slope ( Figure 3A ) is also diminished by more than 4-fold at day 2 ( p-value<0 . 001 , Figure 3G , orange lanes ) . These data indicate that TH signaling promotes the biting activity of A . triostegus and subsequent grazing behavior adopted at larval recruitment . This shows that TH controls the transformations of the feeding process not only at the morphological and physiological levels but also at the behavioral level . To widen our understanding of coral reef fish metamorphosis , we investigated the TH signaling during larval recruitment in four other coral fish species ( Rhinecanthus aculeatus , Chromis viridis , Chaetodon lunula and Ostorhinchus angustatus ) from distant families ( Near et al . , 2012 ) . In these species , both T4 and T3 levels drop between day 1 and day 8 , up to 3-fold for T4 ( in O . angustatus , p-value=0 . 029 ) and up to 4-fold for T3 ( in R . aculeatus , p-value=0 . 05 ) ( Figure 4 ) . This drop of TH levels after reef entry strongly suggests that TH are also key determinants for the metamorphosis of these species . T4 levels in R . aculeatus and T3 levels in C . viridis also rise between near ocean and day 1 individuals ( Figure 4 ) , further confirming the model of a TH-mediated metamorphosis during coral reef fish larval recruitment . However , T4 levels are 3-fold higher in near ocean individuals than in day 1 individuals in C . viridis ( Figure 4 ) . This suggests some potential species-specific variation in coral reef fish TH profiles at recruitment that would be extremely interesting to decipher in future studies on additional species and with regards to other aspects of coral reef fish ecology ( e . g . , diets , size and pigmentation status at recruitment ) . Chlorpyrifos ( CPF ) is an agricultural insecticide that is widely used on tropical coastal crops , therefore one of the most common waterborne chemical pollutants encountered in coral reef surrounding waters ( Cavanagh et al . , 1999; Leong et al . , 2007; Roche et al . , 2011; Botté et al . , 2012 ) . Given the endocrine determinants of coral reef fish metamorphosis ( Figures 2–3 ) and the CPF endocrine disruption characteristics ( Juberg et al . , 2013; Slotkin et al . , 2013 ) , we assessed the impact of CPF on A . triostegus metamorphosis . To achieve this , we used exposure doses previously reported in coral reef fish ( 1 , 5 or 30 µg . l−1 of seawater ) ( Botté et al . , 2012 ) . CPF treatment at 30 µg . l−1 decreases the T3 levels by 50% in day 2 juveniles ( p-value=0 . 021 , Figure 5A ) . No significant effect was observed on T4 levels regardless of CPF concentration and duration ( p-value=0 . 205 for day 2 and p-value=0 . 496 for day 5 , Figure 5—figure supplement 1A ) . We did not observe any effect of CPF treatments on fish pigmentation at day 2 and day 5 ( Figure 5—figure supplement 1B ) . However , CPF exposure at 30 µg . l−1 prevents guts’ lengthening at day 2 ( p-value=0 . 024 , Figure 5B left panel ) , and both exposure at 5 and 30 µg . l−1 prevent intestines’ lengthening at day 5 ( p-values=0 . 047 and 0 . 029 respectively , Figure 5B right panel ) . Strikingly , CPF exposure at 30 µg . l−1 also decreases the grazing activity of day 2 juveniles with 2 . 63 ± 0 . 09 bites per fish per min contrasting with 3 . 62 ± 0 . 28 bites per fish per min in the solvent control condition ( p-value=0 . 014 , Figure 5C ) . In day 5 juveniles , exposure at 5 and 30 µg . l−1 decreases the grazing activity with respectively 2 . 68 ± 0 . 18 and 2 . 63 ± 0 . 05 bites per fish per min compared to 4 . 43 ± 0 . 33 bites per fish per min for the solvent control fish ( p-values=0 . 001 , Figure 5C ) . Such a behavior in fish treated with CPF results in up to a 14-fold reduction in the biomass of turf grazed after 5 days ( p-value=0 . 05 , Figure 5D ) . Our results highlight how coral reef fish larval recruitment , the transition between the open ocean and the reef , is a TH-regulated metamorphosis at the crossroads of ecological , developmental , physiological and behavioral transformations ( Holzer and Laudet , 2015 ) . These findings also demonstrate how coral reef fish TH signaling and larval recruitment processes can be altered by reef pollution . Since metamorphosis and larval recruitment are essential for the maintenance of fish populations and subsequent coral reef resilience , this study provides a general framework to better understand , at the molecular level , how global changes , water pollution and human activities can threaten reef ecosystems . Acanthurus triostegus , Ostorinchus angustatus , Chaetodon lunula , Chromis viridis and Rhinecanthus aculeatus larvae were captured at Moorea Island , French Polynesia . Larvae were captured during the night while colonizing the reef crest , using a crest net and hand nets ( Dufour and Galzin , 1993; Besson et al . , 2017 ) . Larvae were then kept in cages located in situ in the lagoon and provided with continuous food from the water column supplemented with coral rubble and algal turf . For the clarity of this study , the age in day of the juvenile corresponds to the number of day spend in the reef in cages since capture , not the absolute age . Near ocean A . triostegus larvae were sampled at 1–2 km away from the reef crest using drifting light trap and hand nets from 0 . 5 m to 5 m depth . Far ocean A . triostegus larvae were also captured offshore by trawl haul more than 10 km away from the reef between 1 and 60 m deep . Larvae were euthanized in MS222 at 0 . 4 mg . ml−1 in filtered seawater . For TH dosage , larvae were frozen dry and conserved at −20°C prior to extraction . For RNA extraction , larvae were lacerated and kept in RNAlater ( Sigma ) 1 hr on ice and stored at −20°C . For histology , larvae were rinsed in PBS 1X , dissected and the intestines were kept in Bouin’s fixative at room temperature . For µCT scan and metagenomics , Larvae were kept in 70% ethanol at room temperature . Acanthurus triostegus captured at the crest were injected with 20 µl of drugs in the ventral cavity . Drugs tested were as follows: ( i ) solvent control ( DMSO diluted 10 . 000 times in Phosphate Buffer Saline 1X , as all drugs were made soluble in DMSO and diluted 10 . 000 in PBS 1X ) ; ( ii ) T3 + iopanoic acid ( IOP ) both at 10−6 M , ; ( iii ) NH3 at 10−6 M; and ( iv ) T3 + IOP + NH3 all at 10−6 M . IOP was used as an inhibitor of deiodinase enzymes , as evidenced in mammals and amphibians ( Galton and Hiebert , 1987; Galton and Hiebert , 1988; Galton , 1989; Simonides et al . , 2008; Medina et al . , 2011; Renko et al . , 2012 ) , and as routinely used in fish to prevent the immediate degradation of injected T3 ( Little et al . , 2013; Lorgen et al . , 2015 ) . NH3 is a known TR antagonist in vertebrates ( Lim et al . , 2002 ) , a notion we confirmed in A . triostegus ( Figure 2—figure supplement 1D ) . NH3 prevents the binding of TH on TR and impairs the binding of transcriptional coactivators to TR , which therefore remain in inactive conformation ( Figueira et al . , 2011; Lim et al . , 2002 ) . Fish were then kept in the lagoon in in situ cages and injected daily following the same protocol to maintain pharmacological treatment . Fish were sampled on the second and fifth day after the beginning of this protocol . DMSO non toxicity was tested by comparing the pigmentation , the teeth development , the intestine length and internal structures between solvent-control individuals ( treated with DMSO diluted 10 . 000 times in PBS 1X ) , and control individuals ( simply raised in situ in lagoon cages as mentioned in the previous paragraph ) ( Figure 3—figure supplement 1 , Figure 3—figure supplement 2 ) . After crest capture , A . triostegus larvae were transferred in aquaria ( 30 L x 20 W x 20 cm H ) in groups of n = 10 fish . Each aquarium was filled with 9 liters of UV-sterilized and filtered - 10 µm filter , and was equipped with an air stone . Fish were then immediately exposed to: nothing ( control ) , acetone at a final concentration of 1:1 . 000 . 000 ( solvent control , as CPF was made soluble in acetone ) or CPF at a final concentration of 1 , 5 , and 30 µg . l−1 , as previously done in the literature in another coral reef fish ( Botté et al . , 2012 ) . All water was replaced every day , ensuring the maintenance of water quality as well as a continuous concentration of the pesticide or solvent . Fish were sampled the 2nd and 5th day after the beginning of this protocol . Acetone non toxicity was tested by comparing the TH levels , the pigmentation , and the intestine length between solvent-control and control individuals ( Figure 5A–B , Figure 5—figure supplement 1A–B ) . Muscle from adult fish and whole larvae were used . Samples were cut using sterile scalpel blades and crushed in a Precelyss with Qiagen extraction buffer . RNA extractions were performed using the Macherey-Nagel RNA extraction kit following the manufacturer’s instructions . Total pooled RNAs were retro-transcripted with the Invitrogen Super Script III enzyme following the manufacturer’s instructions . For TR cloning , actinopterygian universal primers ( Figure 2—figure supplement 2 ) were designed for TRα-A , TRα-B and TRβ to retrieve the full-length sequences of the genes . For polD2 and rpl7 , actinopterygian universal primers were designed to retrieve partial sequences including at least one exon-exon junction . PCR amplicon were cloned in the Invitrogen PCR II plasmid following the manufacturer instruction for sequencing . TRα-A , TRα-B and TRβ were subcloned in pSG5 between EcoRI sites for in cellulo expression . Far ocean to day 8 juveniles were assayed . For each larvae 1 µg of RNA were used for retro-transcription using the Invitrogen Super Sript III following the manufacturer’s instructions , including a DNAse I treatment . qPCR primers were designed to anneal on different exons , Rpl7 and Pold2 were used as normalization genes ( Figure 2—figure supplement 2 ) . qPCRs were performed in 96 well plate with the BioRad IQ Syber Green Super Mix in 10 µl of final reaction per well following the manufacturer’s ratios . qPCRs were assayed in a BioRad thermocyclers and analysed on BioRad CFX Manager software . Assays were performed on duplicates in at least two independent RNA extractions and retro-transcriptions . Klf9 , which is known to respond to TH signaling , was used as a TH signaling reporter gene . The amino acid sequences of A . triostegus cloned TRs as well as available TR sequences ( Figure 2—figure supplement 3 ) were aligned using MUSCLE software . Trees were generated using the Maximum Likelihood method with the Seaview four software under the JTT model with estimated gamma shape and eight rate categories ( RRID:SCR_015059 ) ( Gouy et al . , 2010 ) . Bootstrap analysis of 1000 replicates was carried out to support the tree . Human embryonic kidney 293 cells ( ATCC:CRL_11268 , RRID:CVCL_0045 ) ( Iwema et al . , 2007; Gutierrez-Mazariegos et al . , 2014; Sadier et al . , 2015 ) were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% of coal stripped foetal bovine serum and penicillin/streptomycin at 100 µg . ml−1 . Cell were maintained at 37°C , 5% CO2 and tests for mycoplasma contamination were negative . The transient transfection assays were carried out in 96-well plate with 30 000 cells per plate using Exgen500 according to the manufacturer’s instructions . For each well , cells were transfected with 50 ng of total DNA: 12 , 5 ng of full-length receptor encoding plasmid , 12 , 5 ng of reporter plasmid with four DR4 repeat in the luciferase promoter , 12 , 5 ng of β-galactosidase encoding plasmid and 12 , 5 ng of pSG5 empty plasmid . Drugs were incubated for 48 hr and cells were harvested using a passive lysis buffer and frozen at −20°C . On half of the lysate , luciferase activities were assayed with the luciferase reagent buffer from Promega on a Veritas Turner Biosystem luminometer . On the other half of the lysate , the β-galactosidase activity was measured using ONPG substrate and absorbance at 420 nm for normalization . Each assay was performed at least three times independently on well triplicates . Drugs from Sigma-Aldrich were diluted in DMSO à 10−2 M then in sterile PBS1X prior treatment . TH were extracted from dry-frozen fish following an extraction protocol adapted from previous publications relating TH level variations in teleost fish ( Tagawa and Hirano , 1989; Einarsdóttir et al . , 2006; Kawakami et al . , 2008 ) . Far ocean larvae to day 8 juveniles were assayed . At least three individuals per sampling point were used . Each fish was first crushed with a Precelyss in 500 µl methanol , centrifuged at 4°C and supernatant reserved , three times . Pooled supernatant were dried at 70°C . Hormones were re-extracted with 400 µl methanol , 100 µl chloroform and 100 µl barbital buffer twice from the first dried extract . Pooled supernatant was dried out and extract was reconstituted in 2 ml of PBS1X for quantification . The quantification was performed following the Roche ELICA kit on a Cobas analyser by a medical laboratory according to the manufacturer’s standardized method . The intestines in Bouin’s fixative were embedded in paraffin . Sections of 5 µm were performed in the proximal , medial and distal part using a microtome ( Leica ) . The histological sections were stained with Hematoxylin and Eosin . Photography were taken on a Leica microscope and the mosaic reconstructed using Image J software . For 16S mass sequencing , juveniles of 2 , 5 , 8 days after reef entry and adult A . triostegus gut were dissected in triplicates for a total of twelve samples . Total DNA was extracted with a Macherey-Nagel DNA extraction kit following the manufacturer’s recommendations . 16S library were constructed for each individual using the Ion 16S Metagenomic kit from Life technologies following the manufacturer’s recommendations . Sequencing of bacterial 16S was performed using a PGM Ion Torrent . Three controls for contaminations were performed . The environment control consisted of an open tube during the dissection . The extraction control monitored the DNA extraction process . The blank control monitored the library construction . The sequencing results were then analyzed using the Life technologies 16S pipeline . 18S mass sequencing was performed on crest larvae , 8 days juveniles and adults in triplicates . Intestine contents were extracted in the same condition as 16S . 18S library were constructed with the pre-amplified V7 region of the eukaryotic 18S ( Figure 2—figure supplement 2 ) . A blocking primer was designed to prevent the amplification of host sequences ( Figure 2—figure supplement 2 ) . The blocking primer was designed to partially overlap the 18S reverse primer and was modified with a C3 spacer . Sequencing was performed using a PGM Ion Torrent . The sequences were assembled and blasted against the PR2 database , completed with 18S sequenced from multi-cellular organisms for taxonomic affiliation . Fish samples were conserved in 70% ethanol , dehydrated in successive baths of 95% ethanol , twice 100% ethanol and in vacuum arena for at least 4 hr . X-ray microtomographies were performed on a Phoenix Nanotom ( General Electric ) at 70 kV of tension , 100 mA of intensity with a tungsten filament . 3000 images per sample were taken at 500 ms of exposure per image and at a resolution ranging from 2 . 5 to 2 . 8 µm . 3D volumes were reconstructed and analysed with VGI studiomax software . For each condition ( e . g . hormonal treatments , pesticide conditionings ) , 18 fish were placed in a 5 liter tank with coral rubbles covered with turf algae . After an acclimation period of 1 hr , we recorded the total number of bites made on turf algae during 10 min . This experiment was replicated with a new batch of fish 3 to 6 times depending on the condition . For each pesticide exposition conditioning , we weighed ( underwater ) 3 pieces of coral rubble prior the introduction of a group of 20 crest captured fish ( weight A ) , and after 5 days of grazing by this group of fish ( weight B ) . For each piece of coral rubble , the weight of grazed algae was estimated through the difference weight B minus weight A . All statistical analyses were conducted using the R-Cran project free software ( http://www . rproject . org/ , R-3 . 3 . 1 ) . Mean comparisons were performed using Wilcoxon-Mann-Whitney U-test when comparing two means , and using univariate analysis of variance ( ANOVA ) followed by Tukey post-hoc test ( should a significant difference be detected ) for multiple comparisons . Prior ANOVA , normality of values ( or residuals ) and variance homogeneity were assessed using Shapiro and Bartlett tests . In qPCR analyses , comparisons of gene expression were performed through Student's t cumulative distribution functions , automatically computed by qPCR software CFX Biorad Manager , ( CFX Manager SH , 2017 ) . All statistical information and data are available in each Figure source data ( see Figure captions ) .	Many animals go through a larval phase before developing into an adult . This transformation is called metamorphosis , and it is regulated by hormones of the thyroid gland in vertebrates . For example , most fish found on coral reefs actually spend the first part of their life as free-swimming larvae out in the ocean . The larvae usually look very different from the juveniles and adults . When these fish become juveniles , the larvae undergo a range of physical and behavioral changes to prepare for their life on the reef . Yet , until now it was not known what hormones control metamorphosis in these fish . To address this question , Holzer , Besson et al . studied the convict surgeonfish Acanthurus triostegus . This herbivorous coral-reef fish lives in the Indo-Pacific Ocean , and the results showed that thyroid hormones do indeed regulate the metamorphosis of its larvae . This includes changing how the larvae behave and how their adult features develop . Further , Holzer , Besson et al . found that this was also true for four other coral-reef fish , including the lagoon triggerfish and the raccoon butterflyfish . In A . triostegus , thyroid hormones controlled the changes that enabled the juveniles to efficiently graze on algae growing on the reef such as an elongated gut . When the fish larvae were then exposed to a pesticide called chlorpyrifos , a well-known reef pollutant , their hormone production was disturbed . This in turn affected their grazing behavior and also their metamorphosis . These fish had shortened , underdeveloped guts and could not graze on algae as effectively . Herbivorous fish such as A . triostegus play a major role in supporting coral reef ecosystems by reducing algal cover and therefore promoting coral recruitment . These new findings show that pollutants from human activities could disturb the metamorphosis of coral-reef fish and , as a consequence , their ability to maintain the reefs . A next step will be to test what other factors can disrupt the hormones in coral-reef fish and thus pose a threat for fish populations and the coral-reef ecosystem .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "ecology", "developmental", "biology" ]	2017	Fish larval recruitment to reefs is a thyroid hormone-mediated metamorphosis sensitive to the pesticide chlorpyrifos
Sox genes encode a set of highly conserved transcription factors that regulate many developmental processes . In insects , the SoxB gene Dichaete is the only Sox gene known to be involved in segmentation . To determine if similar mechanisms are used in other arthropods , we investigated the role of Sox genes during segmentation in the spider Parasteatoda tepidariorum . While Dichaete does not appear to be involved in spider segmentation , we found that the closely related Sox21b-1 gene acts as a gap gene during formation of anterior segments and is also part of the segmentation clock for development of the segment addition zone and sequential addition of opisthosomal segments . Thus , we have found that two different mechanisms of segmentation in a non-mandibulate arthropod are regulated by a SoxB gene . Our work provides new insights into the function of an important and conserved gene family , and the evolution of the regulation of segmentation in arthropods . Arthropods are the most speciose and widespread of the animal phyla , and it is thought that their diversification and success are at least in part explained by their segmented body plan ( Tautz , 2004 ) . In terms of development , insects utilise either derived long germ embryogenesis , where all body segments are made more or less simultaneously , or short/intermediate germ embryogenesis , where a few anterior segments are specified and posterior segments are added sequentially from a growth or segment addition zone ( SAZ ) ( Peel et al . , 2005; Davis and Patel , 2002 ) . It is thought that segmentation in the ancestral arthropod resembled the short germ mode seen in most insects ( Peel et al . , 2005; McGregor et al . , 2009 ) . Understanding the regulation of segmentation more widely across the arthropods is important for understanding both the development and evolution of these highly successful animals . We have a detailed and growing understanding of the regulation of segmentation in various insects , especially the long germ dipteran Drosophila melanogaster and the short germ beetle Tribolium castaneum . However , studies of other arthropods including the myriapods Strigamia maritima and Glomeris marginata , and chelicerates , such as the spiders Cupiennius salei and Parasteatoda tepidariorum , have provided important mechanistic and evolutionary insights into arthropod segmentation ( Peel et al . , 2005; Hilbrant et al . , 2012; McGregor et al . , 2008a; Leite and McGregor , 2016; Janssen et al . , 2004; Brena and Akam , 2012 ) . Previous studies have shown that different genetic mechanisms are used to generate segments along the anterior-posterior axis of spider embryos . In the anterior tagma , the prosoma or cephalothorax , the cheliceral and pedipalpal segments are generated by dynamic waves of hedgehog ( hh ) and orthodenticle ( otd ) expression ( Kanayama et al . , 2011; Pechmann et al . , 2009 ) . The leg-bearing segments are specified by gap gene like functions of hunchback ( hb ) and distal-less ( dll ) ( Pechmann et al . , 2011; Schwager et al . , 2009 ) . In contrast , the segments of the posterior tagma , the opisthosoma or abdomen , are generated sequentially from a SAZ . This process is regulated by dynamic interactions between Delta-Notch and Wnt8 signalling to regulate caudal ( cad ) , which in turn is required for oscillatory expression of pair-rule gene orthologues including even-skipped ( eve ) , and runt ( run ) ( McGregor et al . , 2009; McGregor et al . , 2008b; Schönauer et al . , 2016 ) . Interestingly , these pair-rule gene orthologues are not involved in the production of the prosomal segments ( Schönauer et al . , 2016 ) . Therefore , the genetic regulation of segmentation along the anterior-posterior axis in the spider exhibits similarities and differences to segmentation in both long germ and short germ insects . The Group B Sox ( Sry-Related High-Mobility Group box ) family gene Dichaete is required for correct embryonic segmentation in the long germ insect D . melanogaster , where it regulates pair-rule gene expression ( Nambu and Nambu , 1996; Russell et al . , 1996 ) . Interestingly , it was recently discovered that a Dichaete orthologue is also likely involved in segmentation in the short germ insect T . castaneum ( Clark and Peel , 2018 ) . This similarity is consistent with work inferring that these modes of segmentation are more similar than previously thought and provides insights into how the long germ mode evolved ( Clark and Peel , 2018; Clark , 2017; Verd et al . , 2018 ) . However , it appears that despite these similarities , Dichaete can play different roles in D . melanogaster and T . castaneum consistent with the generation of segments simultaneously via a gap gene mechanism in the former and sequentially from a posterior SAZ in the latter ( Clark and Peel , 2018 ) . We recently described the characterisation of 14 Sox genes in the genome of the spider P . tepidariorum ( Paese et al . , 2017 ) and that several of the spider Sox genes are represented by multiple copies likely produced during the whole genome duplication ( WGD ) in the lineage leading to this arachnid ( Paese et al . , 2017; Schwager et al . , 2017 ) . Interestingly , while Dichaete is not expressed in a pattern consistent with a role in segmentation ( Paese et al . , 2017 ) , we found that the closely related SoxB gene , Sox21b-1 , is expressed in both the prosoma and opisthosoma before and during segmentation ( Paese et al . , 2017 ) . Here , we report that in P . tepidariorum , Sox21b-1 regulates both prosomal and opisthosomal segmentation . In the prosoma , Sox21b-1 has a gap-like gene function and is required for the generation of the four leg-bearing segments . In addition , Sox21b-1 appears to act upstream of both Delta-Notch and Wnt8 signalling to regulate the formation of the SAZ , and knockdown of Sox21b-1 results in truncated embryos missing all opisthosomal segments . Therefore , while prosomal and opisthosomal segments are generated by different mechanisms in the spider , our analysis shows that Sox21b-1 is required for segmentation in both regions of the developing spider embryo . We previously identified and assayed the expression of the complement of Sox genes in the genome of the spider P . tepidariorum ( Paese et al . , 2017 ) ( and see Figure 1 ) . Our phylogenetic analysis indicates that P . tepidariorum Sox21b-1 and its paralog Sox21b-2 are members of the Sox group B , closely related to the Drosophila Dichaete and Sox21b genes ( Figure 1—figure supplement 1 ) . In insects ( McKimmie et al . , 2005; Wilson and Dearden , 2008 ) , Dichaete , Sox21a and Sox21b are clustered in the genome , however , both Sox21b paralogs are dispersed in the spider genome ( Paese et al . , 2017 ) . This suggests that Sox21b-1 and Sox21b-2 possibly arose from the WGD event in the ancestor of arachnopulmonates ( Schwager et al . , 2017 ) rather than by a more recent tandem duplication ( Figure 1—figure supplement 1 ) . In light of its interesting expression pattern , we elected to further analyse Sox21b-1 . Pre-vitellogenic P . tepidariorum oocytes contain a Balbiani’s body ( Jedrzejowska and Kubrakiewicz , 2007 ) , where maternally deposited factors are enclosed , and we found that Sox21b-1 is abundant in this region , indicating that it is maternally contributed ( Figure 1A ) . However , after fertilization we did not detect Sox21b-1 mRNA until early stage 5 , when weak expression is detected throughout the germ disc , with stronger expression in more central cells ( Figure 1B–C ) . At late stage 5 , expression becomes more restricted to the centre of the germ disc ( Figure 1D ) . During stages 5 and 6 , the cumulus migrates to the rim of the germ disc , opening the dorsal field and giving rise to an axially symmetric germ band ( Figure 1E ) ( see Mittmann and Wolff , 2012 ) . In early stage 6 embryos , Sox21b-1 is observed in the middle of the presumptive prosoma in a broad stripe ( Figure 1E ) , which develops further during stage 7 in the region where the leg-bearing segments will form ( Figure 1F ) . This expression pattern resembles the previously described expression of the gap gene hb ( Schwager et al . , 2009 ) . During these and subsequent stages , dynamic expression of Sox21b-1 is observed in the SAZ and the most anterior region of the germ band that will give rise to the head segments ( Figure 1H–I ) . Later in development , the expression of Sox21b-1 resembles that of SoxNeuro ( SoxN ) , another Group B Sox gene ( Paese et al . , 2017 ) . This expression is similar to that of both SoxN and Dichaete in the D . melanogaster neuroectoderm and segregating neuroblasts , however in spiders the neuroectoderm does not produce stem cell like neuroblasts , but instead clusters of delaminating cells that adopt the neural fate ( Paese et al . , 2017; Stollewerk and Chipman , 2006; McKimmie et al . , 2005 ) ( Figure 1H–I ) . Expression of the related group B Sox genes , Dichaete and Sox21b-2 are not detected in P . tepidariorum during embryonic development ( Paese et al . , 2017 ) . The expression of Sox21b-1 in the embryo suggests that it is involved in both anterior and posterior segmentation in this spider , and then later during nervous system development . To assay the function of Sox21b-1 during embryogenesis we knocked down the expression of the gene using a parental RNAi approach ( Akiyama-Oda and Oda , 2006 ) . We observed three phenotypic classes , which were consistent between both non-overlapping Sox21b-1 fragments we used for RNAi ( Figure 2 , Figure 2—figure supplement 1 , Supplementaryfile 1 ) . Class I embryos developed a presumptive head region ( Figure 2A–C ) , as well as normal cheliceral , pedipalpal and first leg-bearing ( L1 ) segments ( Figure 2C ) . The identity of these segments was confirmed by expression of labial ( lab ) in the pedipalps and L1 , and Deformed-A ( Dfd-A ) in L1 ( Figure 2—figure supplement 2A–D ) . However , the other three leg-bearing segments , L2 - L4 , as well as all of the opisthosomal segments were missing in Class I embryos . These embryos exhibited a truncated germ band , terminating in disorganised tissue in the region of the SAZ ( Figure 2C ) . In the case of Class II phenotypes , embryos only differentiated the head region and the cheliceral and pedipalpal segments ( Figure 2D , Figure 2—figure supplement 2A–B ) : all leg-bearing segments of the prosoma and opisthosomal segments produced from the SAZ were missing ( Figure 2D ) . In Class III embryos , the germ band did not form properly from the germ disc ( Figure 2E ) and we therefore looked earlier in development to understand how this severe phenotype arose . We observed that the formation of the primary thickening occurs normally at stage 4 ( Akiyama-Oda and Oda , 2003; Pechmann , 2016; Pechmann et al . , 2017 ) , but subsequently the cumulus , the group of mesenchymal cells that arise as the primary thickening at the centre of the germ disc , fails to migrate properly to the rim of the germ disc during stage 5 ( Figure 2—figure supplement 3 ) . Since migration of the cumulus is required for the transition from germ disc to germ band , this observation at least in part explains the subsequent Class III phenotype . Note that in this phenotypic class , the cells migrate towards the centre of the germ disc , creating a thick aggregation of blastomeres , whereas in embryos that we classified as ‘dead’ , the cells are scattered and stop migrating after the germ disc stage ( Figure 2—figure supplement 3B–C ) . We next examined the effect of Sox21b-1 depletion on cell death and proliferation at stages 5 and 9 between knockdown and control embryos using antibodies against Caspase-3 and phosphorylated Histone 3 ( PHH3 ) ( Figure 2—figure supplement 4 ) . At the germ disc stage , there is no detectable cell death in control embryos ( n = 10 ) , but we observed some small clusters of apoptotic cells in the Sox21b-1 knockdown embryos ( n = 10 ) ( Figure 2—figure supplement 4A–B ) . At stage 9 , a few cells expressed Caspase-3 in the posterior-most part of the SAZ ( Figure 2—figure supplement 4C ) , but we did not observe cell death in this region of Sox21b-1 knockdown embryos ( Figure 2—figure supplement 4D ) . However , we did detect pronounced cell death in the anterior extraembryonic layer of the same embryos ( n = 10 ) ( Figure 2—figure supplement 4D ) . Expression of PHH3 at stages 5 and 9 indicated that Sox21b-1 knockdown embryos show decreased cell proliferation compared to controls ( n = 10 for each ) ( Figure 2—figure supplement 4E–H ) . Interestingly , the cells were also clearly larger in Sox21b-1 knockdown embryos compared to controls , which may reflect perturbed cell proliferation ( Figure 2—figure supplement 4E–H ) . Thus , our functional analysis shows that Sox21b-1 is required for cell maintenance in several areas of the germ disc and is thus a key player in the transition from radial to axial symmetry . Moreover , Sox21b-1 is involved in two different segmentation mechanisms in spiders: it has a gap gene like function in the prosoma , as well as a requirement for the formation of the SAZ and subsequent production of opisthosomal segments . In P . tepidariorum , decapentaplegic ( dpp ) and Ets4 are required for cumulus formation ( Pechmann et al . , 2017; Akiyama-Oda and Oda , 2006 ) . To investigate if Sox21b-1 is involved in the formation of this cell cluster , we assayed the expression of dpp and Ets4 in Sox21b-1 RNAi knockdown embryos . However , both genes were expressed normally and cumulus formation was unaffected ( Figure 3E , F ) . The rim of the spider germ disc develops into the head structures and is regulated in part by hh , while the mesodermal and endodermal layers of the head are specified by the mesendodermal gene forkhead ( fkh ) ( Kanayama et al . , 2011; Akiyama-Oda and Oda , 2010 ) . To investigate if anterior expression of Sox21b-1 ( Figure 1 ) is involved in the formation of the head rudiment and differentiation of the mesodermal and endodermal layers in particular , we assayed the expression of hh and fkh in class I and II Sox21b-1 knockdown embryos . hh is expressed at the rim of the germ disc in the ectoderm ( Figure 3D ) ( Kanayama et al . , 2011 ) and remains unaffected by Sox21b-1 knockdown ( Figure 3H ) . fkh is also expressed in cells around the rim , as well as in the centre of the germ disc in mesendodermal cells ( Figure 3C ) . In Sox21b-1 knockdown embryos both of these fkh expression domains are lost ( Figure 3G ) , and it therefore appears that Sox21b-1 is required for specification of mesendodermal cells in the germ disc of spider embryos . Indeed , in the germ disc at stage 5 , when fkh expression commences , we observed invaginating cells forming a second layer ( Figure 2—figure supplement 2G ) . However , in Sox21b-1 knockdown embryos we observed a lower number of invaginating cells , which exhibit larger nuclei compared to controls ( Figure 2—figure supplement 2H ) . In both spiders and flies , the twist ( twi ) gene is involved in mesoderm specification ( Yamazaki et al . , 2005 ) and we therefore examined the expression of this gene after Sox21b-1 knockdown to further evaluate if the loss of fkh affects the formation of the internal layers . In the wild type , twi is expressed in the visceral mesoderm of the limb buds from L1 to L4 , in the opisthosomal segments O1 to O4 and in an anterior mesodermal patch in the central part of the developing head ( Yamazaki et al . , 2005 ) ( Figure 2—figure supplement 2E ) . While the head expression persists in Sox21b-1 class I embryos , expression in all the limb and opisthosomal segments is lower or absent ( Figure 2—figure supplement 2F ) . In orthogonal projections the anterior-most region of the embryo , three layers of cells can be identified in control embryos ( Figure 2—figure supplement 2I ) . However , in Sox21b-1 knockdown embryos the formation of these layers is perturbed ( Figure 2—figure supplement 2J ) . These data suggest that the ectodermal segmentation in the prosomal region occurs even when there is a reduction in the internal layers of the embryo . In P . tepidariorum , formation of the SAZ and production of posterior segments requires the Wnt8 and Delta-Notch signalling pathways ( McGregor et al . , 2008b; Schönauer et al . , 2016 ) . Interactions between these pathways regulate hairy ( h ) and , via cad , the expression of pair-rule gene orthologues including eve ( McGregor et al . , 2008b; Schönauer et al . , 2016 ) . To better understand the loss of segments we observe in Sox21b-1 knockdown embryos we analysed the expression of Dl , Wnt8 , h and cad in these embryos compared to controls . Dl is expressed at stage 7 in the forming SAZ , in the region of the L4 primordia and in the presumptive head ( Oda et al . , 2007 ) ( Figure 4A ) . Subsequently , at stage 9 , Dl expression is visible in clusters of differentiating neuronal cells and oscillates in the SAZ , an expression pattern associated with the sequential addition of new segments ( Figure 4B ) . In Sox21b-1 knockdown embryos , Dl expression is not detected at stage 5 ( Figure 4C ) and is absent in the posterior at stage 9 ( Figure 4D ) . However , expression in the anterior neuroectoderm appears normal up to the pedipalpal segment , although neurogenesis is apparently perturbed in the presumptive L1 segment ( Figure 4D ) . This suggests that the ectoderm up to the L1 segment differentiates normally , but the development of the SAZ and posterior segment addition controlled by Dl is lost upon Sox21b-1 knockdown . Wnt8 is initially expressed at stage 5 in the centre and at the rim of the germ disc ( Figure 4E ) . At stage 9 , striped expression of Wnt8 is seen from the head to the posterior segments and in the posterior cells of the SAZ ( Figure 4G ) . Knockdown of Sox21b-1 results in the loss of Wnt8 expression in late stage 5 embryos ( Figure 4F ) . At stage 9 , Wnt8 expression is observed in the cheliceral , pedipalpal and first walking limb segments of Sox21b-1 knockdown embryos , but no expression is detected in the remaining posterior cells ( Figure 4H ) . Consistent with the loss of Dl and Wnt8 , cad expression is also lost in stage 5 and stage 9 Sox21b-1 knockdown embryos ( Figure 4I–L ) . These observations indicate that Sox21b-1 acts upstream of Wnt8 and Delta-Notch signalling to regulate the formation of the SAZ and the subsequent production of posterior segments . In support of this regulatory relationship , we find that Sox21b-1 expression is still detected in the posterior regions of the truncated embryos produced by RNAi knockdown of either Dl or Wnt8 ( Figure 5 ) . The spider orthologue of h is expressed in the presumptive L2-L4 segments and dynamically in the SAZ ( McGregor et al . , 2008b ) ( Figure 4M , O ) . In late stage 5 Sox21b-1 knockdown embryos , the expression of h is lost throughout the entire germ disc ( Figure 4N ) . In addition , in Class I phenotype embryos at stage 9 , the expression of h is completely absent in the tissue posterior to the pedipalpal segment ( Figure 4P ) . Therefore , the loss of h expression is consistent with the loss of leg-bearing segments in the anterior gap-like phenotype that results from knockdown of Sox21b-1 as well as loss of segments produced by the SAZ . To look at the effect of Sox21b-1 knockdown on segmentation in more detail we examined the expression of engrailed ( en ) and hh . At stage 9 en is expressed segmentally from the cheliceral to the O3 segment in control embryos ( Figure 4Q ) . However , in Sox21b-1 knockdown embryos , expression of en was only observed in the cheliceral , pedipalpal and L1 segments , consistent with the loss of all the more posterior segments ( Figure 4S ) . hh has a similar expression pattern to en at stage 9 , except it exhibits an anterior splitting wave in the cheliceral segment and is also expressed earlier in opisthosomal segments and in the SAZ ( Figure 4R ) . Upon Sox21b-1 knockdown , hh is only detected in shortened stripes in the cheliceral and pedipalpal segments ( Figure 4T ) . Taken together , our analysis of P . tepidariorum embryos where Sox21b-1 is depleted by parental RNAi reveals an important role for this Group B Sox gene in both gap-like segmentation of the prosoma , as well as posterior segment formation from the SAZ . These experiments further emphasise the critical role this class of transcription factors play in arthropod segmentation . The Sox gene family encodes transcription factors that regulate many important processes underlying the embryonic development of metazoans ( Overton et al . , 2002; Wegner , 1999; Sinclair et al . , 1990; Lefebvre , 2010 ) . One such gene , Dichaete , is expressed in a gap gene pattern and is involved in regulating the canonical segmentation cascade in D . melanogaster ( Nambu and Nambu , 1996; Russell et al . , 1996 ) . Recently , the analysis of the expression of Dichaete in the flour beetle T . castaneum strongly suggests a role in short germ segmentation ( Clark and Peel , 2018 ) , further supported by knockdown of the Dichaete orthologue in Bombyx mori , which resulted in the loss of posterior segmentation ( Nakao , 2018 ) . Here we show that , while Dichaete is not involved in spider segmentation ( Paese et al . , 2017 ) , the closely related SoxB gene , Sox21b-1 , regulates formation of both prosomal and opisthosomal segments . In the prosoma Sox21b-1 has a gap gene like role and is required for the specification of L1-L4 segments ( Figure 6 ) , resembling the roles of hb and Dll in prosomal segmentation in this spider ( Pechmann et al . , 2011; Schwager et al . , 2009 ) and , at least superficially , gap gene function in D . melanogaster . In D . melanogaster the gap genes regulate pair rule gene expression , and while our results indicate that Sox21b-1 is required for the expression of h and the generation of leg-bearing prosomal segments ( Figure 4E , Figure 6 ) , in contrast to insects , in spiders this does not involve the orthologues of eve and runt because they are not expressed in the developing prosomal segments ( Schönauer et al . , 2016; Damen et al . , 2000 ) . In the posterior , Sox21b-1 knockdown perturbs SAZ formation and consequently results in truncated embryos missing all opisthosomal segments . Therefore , Sox21b-1 regulates development of the SAZ , and our observations indicate this is at least in part through roles in organising the germ layers and specification of mesendodermal cells during stages 5 and 6 . This is supported by the loss of fkh expression upon Sox21b-1 knockdown , which is required for mesoderm and endoderm formation in both spiders and insects ( Kanayama et al . , 2011; Feitosa et al . , 2017; Lan et al . , 2018 ) . Moreover , the subsequent dynamic expression of Sox21b-1 in the SAZ after stage 6 is suggestive of a role in segment addition . Our work on Sox21b-1 provides an important new insight into the gene regulatory network ( GRN ) underlying the formation of the SAZ and the sequential addition of segments from this tissue . We show that Sox21b-1 acts upstream of Wnt8 and Delta-Notch signalling in this GRN and is necessary for the activation of these important signalling pathways during posterior development ( Figure 6 ) . Note that while it is possible that Sox21b-1 could regulate Delta and Wnt8 and other segmentation genes directly or via intermediate factors , it remains possible that the loss of expression of these genes upon Sox21b-1 RNAi could be an indirect consequence of the loss of cells or incorrect cell specification when Sox21b-1 expression is knocked down . Further work is needed to determine if Group B Sox genes , such as Dichaete and Sox21b-1 , play a similar role in posterior segmentation in other arthropods . This could provide important new insights into the evolution of the regulation of segmentation in arthropods since a Wnt-Delta-Notch-Cad regulatory cassette was probably used ancestrally in arthropods to regulate posterior development ( McGregor et al . , 2009; Janssen et al . , 2004; Brena and Akam , 2012; McGregor et al . , 2008b; Schönauer et al . , 2016; Pueyo et al . , 2008 ) . Interestingly , SoxB genes also cooperate with Wnt and Delta-Notch signalling in various aspects of vertebrate development including the patterning of neural progenitors and maintenance of the stem state in the neuroepithelium ( Wegner , 1999; Holmberg et al . , 2008; Kormish et al . , 2010; Koch et al . , 2017 ) . Our study shows that Sox21b-1 is not only involved in segmentation but is also maternally supplied and regulates cell division in the early germ disc , as well as the transition from radial to axial symmetry during germ band formation . Further experiments with Sox21b-1 are required to fully elucidate the mechanisms by which it affects these early functions . Furthermore , while spider head development is less affected than trunk segmentation by knockdown of Sox21b-1 , it is clear from our experiments that Sox21b-1 regulates cell fate in this region . Interestingly , Sox2 is involved with the neuro-mesodermal fate choice in mice and Dichaete has a role in embryonic brain development in D . melanogaster ( Koch et al . , 2017; Soriano and Russell , 1998 ) : consequently , SoxB genes may play an ancestral role in the patterning of the head ectoderm and mesoderm in metazoans ( Koch et al . , 2017; Soriano and Russell , 1998 ) . The evolution and diversification of Group B Sox genes in insects is not fully resolved due to difficulties in clearly assigning orthologues based on the highly conserved HMG domain sequence ( Russell et al . , 1996; Wegner , 1999; Zhong et al . , 2011 ) . However , despite these ambiguities it is clear that the Dichaete and Sox21b class genes in all arthropods examined to date are closely related and likely arose from a duplication in the common ancestor of this phylum ( see Zhong et al . , 2011 for discussion ) . Note that in all insects characterised to date Dichaete , Sox21a and Sox21b are clustered in the genome ( McKimmie et al . , 2005 ) , however , while Dichaete and Sox21a are also clustered in P . tepidariorum , the Sox21b paralogs are dispersed in the genome of this spider ( Paese et al . , 2017 ) . We believe it is highly significant that two very closely related SoxB genes are involved in segmentation in both the spider P . tepidariorum and in insects , pointing to an ancient role for this subfamily of Sox genes in invertebrates . Given the close similarity between the HMG domains of Sox21b and Dichaete , it is possible that in some lineages the Dichaete orthologue assumed the segmentation role , whereas in others it was Sox21b . In spiders , Wnt8 is involved in posterior development while in other arthropods this role is played by Wnt1/wg ( McGregor et al . , 2008b ) , and therefore the evolution of Sox21b-1 may have led to the co-option of different genes and subsequent developmental systems drift of the regulation of posterior development . The spider contains an additional related SoxB gene , Sox21b-2 , that possibly arose as part of the whole genome duplication event in the ancestor of arachnopulmonates over 400 million years ago ( Schwager et al . , 2017 ) . It will be interesting to investigate if SoxB genes are involved in segmentation in other spiders and arachnids , including those that did not undergo a genome duplication . Finally , Blast searches of the Tardigrade Hypsibius dujardini genome reveal a single Dichaete/Sox21b class gene , and it will be of some interest to characterise the expression and/or function of this gene in this sister group to the arthropods . P . tepidariorum were cultured at 25°C at Oxford Brookes University . The spiders were fed with D . melanogaster with vestigial wings and subsequently small crickets ( Gryllus bimaculatus ) . Cocoons from mated females were removed and a small number of embryos were immersed in halocarbon oil for staging ( Mittmann and Wolff , 2012 ) . To identify the phylogenetic relationship of P . tepidariorum Sox genes the HMG domains of Anopheles gambiae , Mus musculus , D . melanogaster , P . tepidariorum and S . mimosarum Sox genes were aligned with ClustalW ( Figure 1—source data 1 ) ( Paese et al . , 2017; Larkin et al . , 2007 ) . Phylogenetic analysis was performed in RAxML , with support levels estimated implementing the rapid bootstrap algorithm ( 1000 replicates ) ( Stamatakis et al . , 2008 ) , under the PROTGAMMALG model of amino acid substitution , which was identified as best fitting model using the ProteinModelSelection . pl Perl script from the Exelixis Lab ( https://github . com/stamatak/standard-RAxML/blob/master/usefulScripts/ProteinModelSelection . pl ) . Embryos from pRNAi injected females were immersed under halocarbon oil to assess their stage . In the case of stage 9 and 10 embryos that showed class II and III phenotypes , it was difficult to assess the stage by light microscopy , consequently these embryos were also staged according to their age and the presence of walking limbs as well as the stage of sibling embryos from the same cocoons that did not show any phenotype . Embryos with phenotypes showing a failure to develop after the L1 limb were considered dead after that stage . For the difference in development seen in embryos from the same cocoon , we were careful to select embryos at the same stage in both control and RNAi treated embryos for subsequent gene expression analysis although the development of RNAi embryos was occasionally slower so they appeared slightly younger . At least 10 embryos from each different phenotypical class were used for every in situ hybridization experiment ( Supplementary file 2 ) and in situs were carried out on embryos from different cocoons and different mothers . Embryos ranging from the 1 cell stage to stage 13 were dechorionated and fixed as described previously ( Akiyama-Oda and Oda , 2016 ) , with a longer fixation time of 1 hr to facilitate yolk removal for flat-mounting . For immunohistochemistry , methanol steps were omitted . Ovaries from adult females were dissected in 1x PBS and fixed in 4% formaldehyde for 30 min . Probe synthesis and RNA in situ hybridisation were carried out as described previously with minor modifications ( Akiyama-Oda and Oda , 2003 ) , omitting the Proteinase K treatment and post-fixation steps . Poly-L-lysine ( Sigma-Aldrich ) coated coverslips were used for flat-mounting embryos . Nuclei were stained by incubating embryos in 1 μg/ml 4–6-diamidino-2-phenylindol ( DAPI ) in PBS with 0 . 1% Tween-20 for 15 min . For imaging of flat-mounted embryos after in situ hybridisation , an AxioZoom V16 stereomicroscope ( Zeiss ) equipped with an Axiocam 506-Mono and a colour digital camera were used . Immunostained embryos were imaged with Zeiss LSM 800 or 880 with Airyscan confocal microscopes . For live imaging , embryos were aligned on heptane glue coated coverslips and submersed in a thin layer of halocarbon oil . Bright-field live imaging was performed using an AxioZoom V16 stereomicroscope , while fluorescence live imaging was performed with confocal microscopes . Image stacks were processed in Fiji ( Schindelin et al . , 2012 ) and Helicon Focus ( HeliconSoft ) . Image brightness and intensity was adjusted in Corel PhotoPaint X5 ( CorelDraw ) and Fiji . Fragment of genes were amplified using PCR and cloned into pCR4-TOPO ( Invitrogen , Life Technologies ) . Oligonucleotide sequences are listed in Supplementary file 3 . Immunostaining was carried out as described previously ( Schwager et al . , 2015 ) with minor modifications: antibodies were not pre-absorbed prior to incubation and the concentration of Triton was increased to 0 . 1% . The following primary antibodies were used: mouse anti-α-Tubulin DM1a ( Sigma ) ( 1:50 ) , rabbit α cleaved caspase 3 ( Cell Signaling - 9661 ) ( 1:200 ) and rabbit Anti-phospho-Histone H3 ( Ser10 ) ( Merck Millipore - 06–570 ) . For detection the following secondary antibodies were used: donkey anti-mouse IgG Alexa Fluor 555 ( Invitrogen ) and goat anti-rabbit Alexa Fluor 647 ( Invitrogen ) . The counterstaining was carried out by incubation in 1 μg/ml 4–6-diamidino-2-phenylindol ( DAPI ) in PBS + Triton 0 , 1% for 20 min . Double stranded RNA ( dsRNA ) for parental RNA interference was synthesized according to ( Schönauer et al . , 2016 ) , dissolved in deionized water and injected following the standard protocol ( Akiyama-Oda and Oda , 2006 ) . Two non-overlapping fragments of P . tepidariorum Sox21b-1 were isolated from the 1134 bp coding sequence of the gene: fragment 1 spanning 549 bp and fragment 2 covering 550 bp . Double stranded RNA for P . tepidariorum Dl ( 853 bp ) , Wnt8 ( 714 bp ) and the coding sequence of GFP ( 720 bp ) as used previously ( Akiyama-Oda and Oda , 2006 ) , were transcribed using the same method . Synthesis of double stranded RNA was performed using the MegaScript T7 transcription kit ( Invitrogen ) . After purification the dsRNA transcripts were annealed in a water bath starting at 95°C and slowly cooled down to room temperature . dsRNA was injected at 2 . 0 μg/μl in to the opisthosoma of adult females every two days , with a total of five injections ( n = 7 for each dsRNA; n = 2 for GFP controls ) . The injected spiders were mated after the second injection and embryos from injected spiders were fixed for gene expression and phenotypic analyses at three different time points: stage 4 ( cumulus formation ) , stage 5 late ( germ disc with migrating cumulus ) and stage 9 ( head and limbs bud formation ) .	Insects , spiders , centipedes and lobsters all belong to a group of animals known as arthropods . A common feature of these animals is that their bodies are made up of repeated segments . However different arthropods build their segmented bodies in different ways . For example , the fruit fly makes all of its segments at the same time , while most other arthropods – including spiders – make a few segments at once and then add the rest , one or two at a time , to the rear end of their bodies . Recent research in different insects has shown that these two processes – adding segments simultaneously or sequentially – are more similar than previously thought . This research also showed that these processes involve a gene called Dichaete , which belongs to the Sox gene family . However it was not known if Sox genes also control the production of segments in other arthropods like spiders . Paese et al . have now found that , just like insects , the common house spider does indeed require a Sox gene to form its segments . Specifically , the experiments revealed that spiders need a Sox gene called Sox21b-1 to make both the segments that carry their legs ( which are made all at once ) , and the segments that make up the rear of their bodies ( which are added one at a time ) . Since spiders and insects both use a Sox gene to control the formation of their body segments , it is likely that the ancestor of arthropods used one too . However , because spiders and insects use a different Sox gene for these processes , Paese et al . suggest that one gene may have replaced the role of the other during the evolution of insects and spiders . Together these findings broaden the current understanding of how genes interact to organise cells to build organisms and how these processes evolve over time . Furthermore , since Sox genes direct many important events in all animals , including humans , the discovery of a new role for one of these genes may help scientists to better understand the development of other animals too .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "evolutionary", "biology" ]	2018	A SoxB gene acts as an anterior gap gene and regulates posterior segment addition in a spider
It remains challenging to relate EEG and MEG to underlying circuit processes and comparable experiments on both spatial scales are rare . To close this gap between invasive and non-invasive electrophysiology we developed and recorded human-comparable EEG in macaque monkeys during visual stimulation with colored dynamic random dot patterns . Furthermore , we performed simultaneous microelectrode recordings from 6 areas of macaque cortex and human MEG . Motion direction and color information were accessible in all signals . Tuning of the non-invasive signals was similar to V4 and IT , but not to dorsal and frontal areas . Thus , MEG and EEG were dominated by early visual and ventral stream sources . Source level analysis revealed corresponding information and latency gradients across cortex . We show how information-based methods and monkey EEG can identify analogous properties of visual processing in signals spanning spatial scales from single units to MEG – a valuable framework for relating human and animal studies . How do results from human magnetoencephalography ( MEG ) and electroencephalography ( EEG ) experiments relate to those obtained from animals in invasive electrophysiology ? It is generally well understood how potential changes in large populations of neurons can propagate through tissue types and lead to detectable electric potentials and associated magnetic fields outside the head ( Pesaran et al . , 2018 ) . Yet , in typical MEG and EEG experiments , we have little clue which specific cellular and circuit mechanisms contribute to the recorded signals ( Cohen , 2017 ) . This can be attributed to several factors . First , the reconstruction of cortical sources from non-invasive signals is generally limited and based on assumptions ( Darvas et al . , 2004 ) . Second , invasive and non-invasive electrophysiology are largely separate research fields . Comparable experiments performed on both levels and in the same species are rare , with few recent exceptions ( Bimbi et al . , 2018; Godlove et al . , 2011; Reinhart et al . , 2012; Shin et al . , 2017; Snyder et al . , 2015; Snyder et al . , 2018 ) . Third , studies employing invasive and non-invasive methods in parallel suffer from sparse sampling of recording sites . Massively parallel invasive recordings in multiple brain regions have only recently become viable ( Dotson et al . , 2017; Jun et al . , 2017; Siegel et al . , 2015 ) , and EEG recordings in awake behaving animals have so far been limited to relatively few electrodes . This sparsity limits specificity when drawing conclusions from one level to the other . In summary , the mapping between measurement scales is severely underconstrained , both theoretically when trying to infer cortical sources of non-invasively measured activity , and experimentally by the lack of sufficiently comparable data . Thus , key for linking different scales are comparable large-scale recordings on all levels to provide high specificity and eventually trace the origins of large-scale phenomena back to their underlying cellular mechanisms . Importantly , this includes non-invasive recordings in animals . These allow to bridge the gap between invasive animal electrophysiology and non-invasive human experiments by permitting to disentangle similarities and differences due to species membership from those due to measurement technique . An especially suitable candidate for this is monkey EEG , making use of evolutionary proximity and promising to better connect the rich literature in non-human primate neurophysiology with human studies . A powerful tool to link data from different measurement scales is the abstraction from measured activity itself to its information content , as enabled by multivariate decoding methods . Representational similarity analysis ( RSA ) compares the representational structure of signals ( Cichy et al . , 2014; Kriegeskorte et al . , 2008a ) . However , as decoding approaches have inherent difficulties to identify the sources of decodable information ( Carlson et al . , 2018; Liu et al . , 2018 ) , it is necessary to employ thoughtful control analyses or experiments ( Cichy et al . , 2015 ) to disambiguate different possible mechanisms underlying large-scale information structure . This crucially relies on empirical knowledge about processes on the circuit-scale . To bridge the gap between invasive and non-invasive electrophysiology , in the present study , we developed and employed fully human-comparable high-density monkey EEG . We presented identical color and motion stimuli to both human participants and macaque monkeys and combined large-scale recordings on multiple scales , including invasive electrophysiology from six areas across the macaque brain , monkey EEG and human MEG with multivariate decoding and representational similarity analysis . We found color and motion direction information not only in invasive signals , but also in EEG and MEG . We show how motion and color tuning in human MEG can be traced back to the properties of individual units . Our results establish a proof of principle for using large-scale electrophysiology across species and measurement scales to link non-invasive recordings to circuit-level activity . We found that information about both motion direction and color was present in all signal types ( Figure 2 ) . In LFP , multi-unit and single-unit data , motion and color information were strongest in areas MT and V4 , respectively , in line with their established functional roles . Nonetheless , both features were represented ubiquitously ( p<0 . 05 , cluster permutation , for most areas apart from motion in IT LFP ) . Importantly , monkey EEG ( Figure 2B and E ) and human MEG ( Figure 2C and F ) also contained information about motion direction and color ( p<0 . 05 for both features in both species , cluster-permutation ) . Our analysis of microelectrode recordings showed decreasing information strength along the cortical hierarchy . To test whether this phenomenon was also detectable non-invasively , we performed source-reconstruction of monkey EEG and human MEG data using detailed physical headmodels ( see methods , ‘Source reconstruction and searchlight analysis’ ) . We then repeated the MVPA in a searchlight fashion across the cortex . Indeed , for both monkey EEG and human MEG , this revealed gradients of information with strongest information in early visual areas ( Figure 2B , C , E , F; insets ) . To compare the dynamics of feature information , we estimated information latencies as the time point the decoding performance reached half its maximum ( Figure 3 ) . For the invasive recordings , latencies were in accordance with the visual processing hierarchy , with information rising earliest in MT for motion direction ( SUA: 78 ms , MUA: 81 ms , LFP: 98 ms ) , earliest in V4 for color ( SUA: 82 ms , MUA: 86 ms , LFP: 91 ms ) , and last in frontal areas . Generally , color information was available earlier than motion direction information in most areas where latencies could be estimated reliably for SUA ( V4: p=0 . 001; IT: p=0 . 13; MT: p=0 . 39 , random permutation ) , MUA ( V4: p<0 . 001; IT: p=0 . 12; MT: p=0 . 26 , random permutation ) and LFP ( V4: p=0 . 006; IT: p=0 . 75; MT: p=0 . 37 , random permutation ) , consistent with previous results from the same animals in a different task ( Siegel et al . , 2015 ) . These results translated to the noninvasive signals: Both for monkey EEG ( color: 91 ms , motion: 103 ms , p=0 . 03 , random permutation ) and human MEG ( color: 70 ms , motion: 97 ms , p<0 . 001 , random permutation ) , color information rose earlier , and the latencies were comparable with those found invasively . Using the searchlight decoding analysis , we again found gradients consistent with the cortical hierarchy , with lowest latencies in occipital and highest latencies in more frontal regions ( Figure 3B , C , E , F; insets ) , as confirmed by correlating source position and estimated latencies ( MEG color: p=10−15 , MEG motion direction: p=10−4 , monkey EEG color: p=0 . 017 , monkey EEG motion direction: p=10−20 ) . Is it plausible that the contents of sensory representations are accessible to noninvasive electrophysiology ? It has been shown that , in general , features represented at the level of cortical columns can propagate to decodable MEG and EEG signals ( Cichy et al . , 2015 ) . Recently , it was reported that information about the motion direction of random dot stimuli can be extracted from EEG signals ( Bae and Luck , 2019 ) . This study is , however , to our knowledge the first direct report of color decoding from MEG or EEG . It is conceivable that luminance confounds introduced by imperfections in the color calibration or individual variation in retinal color processing could explain color decoding . To exclude this possibility , we performed a control experiment in a single human subject , in which we manipulated luminance such that each stimulus was presented in a darker and a brighter version . We then used a cross-classification approach to test whether true color information dominated the artificially introduced luminance effect . To this end , we grouped trials such that , for each color , one luminance level was used for training and the other for evaluating the decoder , effectively creating a mismatch of information between test and training data . The color decoder could now , in principle , pick up three sources of information: true color differences , unknown , confounding luminance differences , and experimentally introduced luminance differences . In isolation , these luminance differences should lead to below-chance accuracy . Therefore , any remaining above-chance effect would either indicate that the luminance confound was even stronger than the control manipulation , or that true color information was present . Indeed , we found that classifier accuracy was still significantly above chance ( p<0 . 05 , cluster permutation ) , and undiminished by the luminance manipulation ( Figure 4A ) . Furthermore , we compared the confusion matrices of classifiers trained and tested on dark or bright stimuli , trained on dark and tested on bright stimuli , or vice versa ( Figure 4B ) . All confusion matrices were highly similar , indicating that the representational structure was comparable for low- and high luminance colors . Taken together , this suggests that in our main experiment , equiluminant and equisaturated color stimuli lead to discriminable MEG signatures , and luminance confounds had only a small , if any , effect . Our stimuli were generated in LCh-space , which is designed based on perceptual uniformity in humans . However , it has been shown that color sensitivities in macaque monkeys are highly similar , but not identical to humans ( Gagin et al . , 2014; Lindbloom-Brown et al . , 2014 ) . To ensure that color decoding in the monkey data was not driven by luminance differences , we performed a psychophysical control experiment in a third macaque monkey . Using a minimum-motion technique and eye movements as readout ( Logothetis and Charles , 1990 ) , we found that equiluminant colors generated in LCh-space were also close to perceptually equiluminant for this monkey ( Figure 4—figure supplement 1 ) . While in human MEG data , there was more information about motion direction than about color , monkey EEG data showed the opposite effect ( Figure 2B , C , E , F ) . In principle , this could be due to differences in species , measurement modality ( EEG or MEG ) , or differences in the visual stimulation that were beyond our control due to the separate recording environments . To exclude measurement modality as the relevant factor , we acquired simultaneous MEG and EEG data in one of the human participants and compared the amount of motion direction and color information across MEG and EEG data . All monkey EEG recording sessions contained more information about color , and all human MEG recordings contained more information about motion direction . Notably , the human EEG session was consistent with the MEG results . While information was generally lower for EEG than for simultaneous MEG , EEG showed the same dominance of motion information ( Figure 4C ) . This suggests that the differences of information dominance between human MEG and monkey EEG were not due to the recording modality . Having established the presence of information in all signal types , we next asked how the representational structure of motion direction and color varied across brain areas , species , and measurement scales . To address this , we performed representational similarity analysis ( Kriegeskorte et al . , 2008a ) ( RSA ) on the LDA confusion matrices averaged over a time window during which visual information was present ( 50–250 ms ) . In short , we used RSA to compare patterns of similarity between stimulus classes , as given by the confusion matrices , across areas and signal types . First , we sought to characterize the diversity of representations across the six areas measured invasively ( Figure 5A ) . For color information , we found that representations were highly similar between SUA , MUA and LFP , as well as between all six cortical areas ( p<0 . 05 for most pairs of areas and measures , uncorrected ) , indicating that a single representational structure was dominant across the brain . In the case of motion direction , areas were split into ventral stream visual areas ( IT and V4 ) and frontal and dorsal visual stream areas ( MT , LIP , FEF , PFC ) . Within each of these two groups , there were again high correlations between areas and measures , but we found no significant similarity between the groups . How does information contained in locally recorded neuronal activity relate to information in large-scale EEG signals ? We found that the color representation in macaque EEG was highly similar to those of SUA , MUA and LFP in all six areas , while the EEG motion direction representation reflected only the ventral stream areas V4 and IT ( Figure 5B , p<0 . 05 for IT SUA and MUA , V4 SUA , MUA and LFP , random permutation , corrected for number of areas ) . Notably , we found no motion direction similarity between area MT and EEG ( SUA: p=0 . 84; MUA: p=0 . 85; LFP: p=0 . 82 , uncorrected ) . This implies that , although MT contained a large amount of motion direction information , EEG signals were dominated by activity from areas with V4- or IT-like motion direction tuning . We found similar results when comparing invasive data to human MEG; again , there were strong similarities between color representations in all areas and human MEG , as well as between motion direction representations in V4 and IT and human MEG ( Figure 5C , p<0 . 05 ) . Furthermore , both color and motion direction representations were highly similar between monkey EEG and human MEG ( Figure 5D , color: r = 0 . 83 , p=0 . 0002; motion: r = 0 . 69 , p=0 . 0003 ) . Color representations were similar across the brain , while motion direction representations were divided into two categories , only one of which translated to non-invasive signals . To investigate what led to these effects , we examined the underlying representations more closely . Figure 6 shows the color and motion direction confusion matrices for MT and V4 multi-unit activity as well as for monkey EEG and human MEG . All color confusion matrices displayed a simple pattern decreasing with distance from the diagonal . This implies that neural activity distances in all areas , signals and both species approximately matched perceptual distances in color space . We found a similar representation of motion direction in area MT . However , motion direction representations in V4 , monkey EEG and human MEG displayed a distinct peak in similarity on the off-diagonal opposite to the true motion direction , indicating that these signals were , to some extent , invariant to motion in opposite directions . To assess the temporal dynamics of this effect , we collapsed the confusion matrices over stimuli , which results in prediction probabilities as a function of the angular difference between true and predicted stimuli ( Figure 6 ) . Here , the off-diagonal elements in the confusion matrices translated to an increased probability of a stimulus to be predicted as the one opposite in stimulus space . At all timepoints , color stimuli were least likely to be classified as the opposite color , whereas there was an increased probability for motion directions to be identified as the opposite . In terms of population tuning , this corresponds to bimodal tuning curves ( Figure 6 ) . We quantified the presence of such bimodal tuning across areas and measurement scales by calculating the slope in prediction probability between opposite ( 180-degree difference ) and next-to-opposite ( 135- and 225-degree difference ) stimuli , normalized by the range of prediction probabilities ( Figure 7 ) . This revealed that motion direction tuning was indeed significantly bimodal in V4 and IT as well as monkey EEG and human MEG , but not for any of the more dorsal or frontal areas . There was no significant bimodal color tuning for any area or measurement scale . We used linear regression to estimate the contribution of bimodality differences to the pattern of similarity between invasively measured areas and signal types ( Figure 7C and E ) . To this end , we computed differences in bimodality between each combination of SUA , MUA and LFP , and all areas . We then assessed to what extent these differences in bimodality accounted for the variance in representational similarity . Importantly , in the case of motion direction , bimodality could largely explain the pattern of representational similarity between areas and measures ( R2 = 0 . 28 , p=0 ) . This was not the case for the small bimodality differences in color tuning , which did not affect representational similarity ( R2 = 0 , p=0 . 99 ) . Thus , similar motion direction bimodality led to V4 and IT showing similar motion representations , which were also similar to those in monkey EEG and human MEG . Finally , we asked on which level the motion direction bimodality arose . The presence of a bimodality effect in MEG , EEG , LFP , multi-unit and sorted unit data suggests that it was not caused by anisotropies in the large-scale organization of motion direction tuning , but rather by properties of individual units: if individual single or multi-units , or LFP channels , were able to distinguish between opposite motion directions , a multivariate analysis of several channels would be expected to also reflect this separability . We therefore expected bimodal motion direction tuning curves to be prominent in those areas which exhibited a multivariate bimodality effect . To test this , we aligned all tuning curves in V4 and MT according to their preferred direction and calculated , for each area , their average . Indeed , direction tuning curves in areas V4 ( SUA: p=110−9 , MUA: p=810−11 , LFP: p=0 . 03 ) were bimodal , whereas direction tuning curves in area MT ( SUA: p=0 . 66 , MUA: p=0 . 31 , LFP: p=0 . 87 ) or color tuning curves in either area ( all p>0 . 42 ) were not ( Figure 7D and F ) . Consistent with earlier reports ( An et al . , 2012; Mendoza-Halliday et al . , 2014; Siegel et al . , 2015 ) , we found color and motion information in all areas we measured , rather than in a small amount of specialized areas . Nonetheless , the amount of information strongly depended on the area . Interestingly , the motion direction decoding accuracies we found were lower than previously reported in both area MT and prefrontal cortex ( Mendoza-Halliday et al . , 2014 ) . This can largely be attributed to differences in the paradigm and analysis strategy: First , rather than decoding from large pseudo-populations , we used small , simultaneously recorded populations . Second , we report averaged single trial probabilities , which tended to be smaller but more robust than the corresponding discrete classification results . Third , the rapid succession of very short stimuli likely limited cortical processing of each stimulus . Fourth , our paradigm only involved passive fixation . Especially in higher-order areas we would expect representations to be strengthened , and altered , according to task demands in a more engaging cognitive task . Stimulus features showing a spatial organization at the scale of cortical columns , such as orientation , can in principle be decoded from EEG and MEG ( Cichy et al . , 2015 ) . This implies that other , similarly topographical representations should be equally accessible . Indeed , a clustering of both color ( Conway and Tsao , 2009; Roe et al . , 2012; Tanigawa et al . , 2010 ) and motion direction ( Albright et al . , 1984; Li et al . , 2013; Tanigawa et al . , 2010 ) has been reported in several areas of the visual system . This suggests that our successful decoding of stimulus color and motion direction was not attributable to confounding factors , but likely stemmed from true feature-related signals . Crucially , even though we recorded invasively in many areas , our results do not unequivocally identify the sources of visual information in MEG and EEG . First , neither color nor motion direction representations are limited to the areas we recorded from . Secondly , partially due to the simple feature spaces used for our stimuli , many areas are expected to show highly similar tuning properties . Based on RSA , we can therefore only conclude that the non-invasively measured information stems from areas with tuning similar to V4 or IT . It is reasonable to assume that earlier visual areas strongly contributed to this , which is corroborated by our source level searchlight analysis revealing strong information peaks in occipital cortex . Furthermore , it has been shown that for example area V1 exhibits a more bimodal motion direction tuning ( i . e . orientation or axis tuning ) than area MT ( Albright , 1984 ) , matching the results found here in V4 . There is , however , previous evidence that the structure of color representations decodable from area V1 using fMRI is not in agreement with perceptual similarity ( Brouwer and Heeger , 2013 ) , contrary to area V4 , and contrary to the representations we found in MEG and EEG , suggesting that these color representations might not be explained by V1 activity alone . Notably , in area MT cortical columns with opposite preferred motion directions along the same axis lie spatially close to each other ( Albright et al . , 1984; Born and Bradley , 2005 ) . This could , in principle , lead to a diminished decodability of opposite motion directions from mass signals such as EEG , MEG or fMRI . In such a scenario , the source of bimodal motion direction tuning might still lie in area MT . However , this would require columns with opposite preferred motion directions to be close to uniformly distributed at the relevant spatial scale . While several recent fMRI studies have focused on motion axis decoding ( Schneider et al . , 2019; Zimmermann et al . , 2011 ) , motion direction has been successfully decoded from BOLD signals in area MT ( Kamitani and Tong , 2006 ) . Given that motion representations are prevalent across visual cortex ( An et al . , 2012 ) , we consider it unlikely that MT was a dominant source of the bimodally tuned motion signals we measured in EEG and MEG . In sum , this suggests that the information decoded from non-invasive signals originated in a mixture of several early visual areas . Recordings from additional visual areas using the same paradigm are required to further clarify this . Future studies may also expand the stimulus space - a limitation of the present proof-of-principle study . Manipulating other stimulus features in order to maximize differences between areas will allow to further dissociate representations in specific parts of the visual system . We utilized human-comparable monkey EEG as a bridge technology to link invasive animal electrophysiology to human MEG . High electrode density and methods identical to those used in human M/EEG enabled us to perform source reconstruction and directly relate measures across species . The few available previous studies measuring EEG in nonhuman primates were typically restricted to only a few electrodes ( Bimbi et al . , 2018; Snyder et al . , 2015; Snyder et al . , 2018 ) and used diverging methods such as skull-screw electrodes , or both ( Godlove et al . , 2011; Musall et al . , 2014; Reinhart et al . , 2012; Whittingstall and Logothetis , 2009; Woodman et al . , 2007 ) . We show how monkey EEG can serve as a missing link to enable the disentangling of species differences from differences in measurement modality . In isolation , our observation of bimodal motion direction tuning in human MEG could not directly inform conclusions about the relative contributions of dorsal and ventral stream areas . Finding the same result in monkey EEG allowed us to infer that it was not due to a decreased influence of MT-like areas in the human , but rather a sign of a general dominance of V4-like tuning in non-invasive signals . State-of-the-art animal electrophysiology requires large technical efforts and comes at a significant ethical cost . When applied in addition to ongoing invasive experiments , the marginal cost of monkey EEG is comparably small . It is non-invasive , depends mostly on standard human neuroscience tools , and does not necessitate further animal training . This is far outweighed by the potential benefits of establishing a database for linking invasive and non-invasive electrophysiology and for enhancing comparability between the fields . Notably , another possibility to achieve this goal is given by invasive electrophysiological recordings in human patients , that are however severely constrained by the requirement for medical necessity . In the current work , we used an information-based approach to compare brain areas , measurement scales , and species . Such analyses are powerful tools to relate very different signals based on their information contents . This may not only include data from different measurement techniques , such as MEG and fMRI ( Cichy et al . , 2014; Cichy et al . , 2016a ) , or species ( Cichy et al . , 2014; Kriegeskorte et al . , 2008b ) , but also cognitive or computational models ( Cichy et al . , 2016b; Wardle et al . , 2016 ) . Furthermore , instead of comparing representations of simple sensory stimuli , the same framework can be applied to complex task conditions ( Hebart et al . , 2018 ) . We would like to highlight that our framework of cross-species and cross-scale comparisons is not limited to information-based analyses . For example we anticipate that it will be highly interesting to compare and pinpoint specific spectral signatures of circuit activity in signals on all scales ( Donner and Siegel , 2011; Siegel et al . , 2012 ) . This has been successful in some cases ( Sherman et al . , 2016; Shin et al . , 2017 ) , but could significantly benefit from the present large scale approach to gain further specificity . In the long term , with sufficient knowledge about mechanistic signatures on all scales , this could facilitate the establishment of transfer functions between circuit activity and non-invasive human electrophysiology ( Cohen , 2017; Donner and Siegel , 2011; Siegel et al . , 2012 ) . It is important to note that such transfer can only be possible based on knowledge on all scales . As has been noted before ( Sprague et al . , 2018 ) , macro-scale signals alone always suffer from an ill-posed inverse problem when trying to infer micro-scale properties . The approach of dense recordings on all scales , as outlined here , allows to bridge this gap by constraining inferences . Such developments would allow quick and inexpensive access to circuit function in the human brain , both for basic research and in clinical practice ( Siegel et al . , 2012 ) . In sum , we show that color and motion direction can be decoded from non-invasive electrophysiology in humans and monkeys . Our results suggest that such simple stimulus representations are dominated by signals from early ventral stream areas . This inference serves as a proof-of-principle for , and was enabled by , using high-density monkey EEG as a bridge technology to link scales and species . To enable source reconstruction , we acquired anatomical MRI scans from both macaques and humans . T1-weighted images were obtained for each human participant and the two monkeys used for EEG recordings . We used linear discriminant analysis ( LDA ) to extract the content and structure of information about stimulus features from all signal types . Trials were stratified such that each combination of color and motion direction occurred equally often , grouped according to one of the two stimulus features and split into training and test sets . For each time-point , we trained multi-class LDA on the training set , and predicted stimulus probabilities in the test set , using the activity in single or multi-units , EEG electrodes or MEG sensors as classification features . From the predicted stimulus probabilities , we created confusion matrices indicating the probability of stimuli being labeled as any other stimulus by the classifier . We evaluated classifier performance as the hit rate , calculated as the mean of the diagonal of the confusion matrix . For EEG and MEG , we repeated this analysis in a 10-fold cross-validation scheme for each session , using all available sensors . For SUA , MUA and LFP , we used 2-fold instead of 10-fold cross-validation . Here , stimuli were presented in sequences of six or eight stimuli , and the occurrence of individual stimuli at each sequence position was not fully balanced . To prevent a potential confound of stimulus information with sequence position , we chose a stratification approach that kept the number of occurrences of each stimulus at each sequence position identical by oversampling the under-represented stimuli within each cross-validation fold . Due to the relatively low number of stimuli per recording session , 10-fold cross-validation would not have resulted in sufficient trials per fold for this approach . We therefore chose 2-fold cross-validation instead and performed classification independently for each of the six areas recorded . We restricted the analysis to five units per area at a time and repeated it for all or maximally 40 random combinations of the available units , to enable a comparison of information content in different areas . Results from these repetitions were averaged before statistical analysis . This analysis was performed for each time point from 250 ms before to 500 ms after stimulus onset , in steps of 10 ms , resulting in confusion matrices and classifier performances at 76 time points . In most of our recordings we presented eight different colors or motion directions . However , in the invasive recordings in stimulus configuration A there were 12 colors and directions . Therefore , we interpolated the confusion matrices of these recordings from a 12 × 12 to an 8 × 8 space . We assessed the presence of significant information using a cluster sign permutation procedure ( similar to Cichy et al . , 2014 ) . After subtracting chance performance ( 0 . 125 ) , we determined temporally contiguous clusters during which information was higher than 0 ( one-tailed t-test over recording sessions , p<0 . 01 ) . We then randomly multiplied the information time-course of each recording session 10 , 000 times with either 1 or −1 , resulting in an expected value of 0 . In each random permutation , we re-computed information clusters and determined the cluster-mass of the strongest cluster . Each original cluster was assigned a p-value by comparing its size to the distribution of sizes of the random permutation’s strongest clusters . Information latency was computed as the time point classifier performance reached half its peak . The peak was estimated as the first local maximum in performance that reached at least 75% of the global performance maximum . To avoid latencies being dominated by those recording sessions containing the most information , we normalized each session’s classifier performance and used only those sessions where the post-stimulus performance peak was at least 1 . 5 times higher than the largest deviation during pre-stimulus baseline . We estimated 95%-confidence intervals using bootstrapping . To statistically assess latency differences between color and motion direction , we used a random permutation procedure . True differences were compared to a distribution of latency differences generated from 10 , 000 random permutations of the group labels . To test whether latencies in the source-reconstructed monkey EEG and MEG systematically varied along the occipito-frontal gradient , we selected all sources containing significant information ( cluster permutation , p<0 . 05 ) . We then computed Pearson correlation coefficients between the physical location of those sources along the occipito-frontal gradient and the estimated latencies . To control for possible effects of luminance on color classification , we measured MEG as described above in one human participant during an additional control experiment . For this experiment , we used the same stimulus space as for the main experiment , but additionally included each color at a lower luminance level , such that the luminance contrast between colored dots and background was 20% lower . We then employed the same multivariate classification approach , but split training and test data according to their luminance levels . First , we used only either low-luminance or high-luminance trials for both training and testing . Second , we repeatedly split the color space into two halves , along each possible axis , trained on high-luminance stimuli from the one half and low-luminance stimuli from the other , and tested on the remaining stimuli . We then averaged confusion matrices over all axes , before extracting classification accuracies . To assess statistical significance , we repeated the analysis 100 times after shuffling the stimulus labels; the distribution of accuracies from shuffled data was used to compute p-values for the unshuffled data . As color vision in macaques and humans is slightly different , we performed a psychophysical control experiment in a third macaque monkey to assess if our stimuli were in fact perceptually equiluminant to macaque monkeys . To this end , we used an adapted minimum motion technique using eye-movements as a readout ( Logothetis and Charles , 1990 ) . We measured small eye movements while the monkey was required to hold fixation on sequentially presented grating stimuli . Each stimulus lasted 500 ms and consisted of a repeating sequence of 4 frames , where frames 1 and 4 contained luminance contrast gratings , whereas frames 2 and 3 contained a contrast between a reference gray of a defined luminance and the probe color we wanted to assess . The phase of each grating proceeded by a quarter cycle with respect to the previous one , such that a probe color of higher luminance than the reference gray would elicit a motion percept in one direction , whereas a probe color of lower luminance would elicit a motion percept in the opposite direction . A probe color of the same luminance as the reference gray should not elicit any consistent perceived motion . Each stimulus was presented in two conditions: In the first condition , a color of higher luminance would elicit upwards motion , in the second one it would elicit downwards motion . We showed stimuli in trials of four , where subsequent stimuli always belonged to the opposite condition . We computed the difference in eye trace curvature – the second derivative of the vertical eye position over time - between conditions as a measure of perceived luminance deviation from the reference gray . We used this procedure for colors of eight hues in LCh-space , as in the main experiment . Stimuli of each color were generated at 19 L values , centered around the L value of the reference gray . The reference gray was chosen as the center of the largest possible equiluminant circle in LCh-space , such that it was comparable in luminance to the stimuli used in the main experiment . Using linear regression , we assessed at which L value the luminance difference measure crossed 0 , which established the point of perceptual equiluminance . To assess whether the inverted relationship between color and motion information in monkeys and humans was due to differences between EEG and MEG , we simultaneously measured EEG and MEG in one of the eleven human participants . Identical analyses were performed on the human EEG data , and we compared maximal accuracies for color and motion decoding in all monkey EEG and human MEG sessions as well as the human EEG session . To assess the distribution of information in human and macaque brains , we performed source reconstruction on monkey EEG and human MEG data and repeated the multivariate classification in a searchlight fashion . We used structural MRI scans to create individual realistic head models . For MEG source reconstruction , we generated single-shell head models ( Nolte , 2003 ) . In the case of EEG source reconstruction , we manually extracted the skull and then segmented the brain into gray matter ( GM ) , white matter ( WM ) and cerebrospinal fluid ( CSF ) using SPM and fieldtrip toolboxes in combination with probabilistic tissue maps for the macaque brain ( Rohlfing et al . , 2012 ) . We determined the position of the titanium headposts with respect to the head surface using an optical tracking system , and incorporated 3D models of the headposts into our segmentation . These overall six tissue types ( WM , GM , CSF , skull , scalp , titanium ) were then used to generate detailed finite element head models ( FEM ) using the SimBio toolbox ( Fingberg et al . , 2003 ) as implemented in Fieldtrip . We estimated human MEG source activity at 457 and monkey EEG source activity at 517 equally spaced locations on the cortical surface , using linear spatial filtering ( Van Veen et al . , 1997 ) . We compared representational structure between brain areas , measurement methods and species using representational similarity analysis . To this end , we computed the temporal average of the confusion matrices over a time period in which stimulus information was available ( 50–250 ms ) . Each entry in the resulting matrix gave an approximation of the similarity between stimulus representations . We then performed RSA by correlating matrices , after removing the diagonal . To assess significant similarity , we used a permutation procedure in which we randomly reassigned stimulus labels to the rows and columns of the confusion matrices 10 , 000 times . P-values were computed as the probability that the similarity between shuffled matrices deviated from zero at least as strongly as the true similarity . From the time-averaged confusion matrices , we extracted several tuning parameters to identify the factors contributing to similarity across scales and species . First , we collapsed confusion matrices across stimuli to obtain tuning curves denoting classifier prediction probability as a function of distance between stimuli . In these tuning curves , a peak at zero indicates a high probability of a stimulus being correctly identified by the classifier , and a peak at 180 degrees indicates an elevated probability of a stimulus being identified as its opposite . We estimated population tuning bimodality by computing the difference between opposite ( 180 degrees ) and next-to-opposite ( 135 , 225 degrees ) stimuli normalized by the difference between maximal and minimal prediction probabilities . This bimodality-index is positive in case of a second peak at 180 degrees and zero or negative in case of a unimodal tuning curve . We used t-tests over sessions or subjects to test statistical significance of the bimodality ( bimodality-index>0 ) . To estimate the importance of bimodality for representational similarity , we computed the differences in bimodality between all invasively measured areas and signal types . We then used linear regression to determine the amount of variance in the representational similarities explained by these bimodality differences . To estimate average tuning curves of single units , multi-units and LFP channels in each cortical area , we performed one-way ANOVAs on each channel to select those containing information about color or motion direction , respectively , with a statistical threshold of p<0 . 05 . We then computed single-channel tuning curves and aligned them according to their preferred stimulus , determined as the stimulus for which firing rate , or LFP power , was highest . Finally , we computed the mean of all aligned tuning curves within one area , for each signal type . To assess single-unit bimodality , in a given area , we used one-sided t-tests to assess if the above described bimodality index was larger than 0 . All analyses were performed in MATLAB , using custom code as well as the Fieldtrip ( Oostenveld et al . , 2011 ) and SPM toolboxes .	Neurons carry information in the form of electrical signals , which we can listen to by applying sensors to the scalp: the resulting recordings are called an EEG . Electrical activity within the brain also generates a weak magnetic field above the scalp , which can be measured using a technique known as MEG . Both EEG and MEG only require a few dozen sensors , placed centimeters away from the brain itself , but they can reveal the precise timing and rough location of changes in neural activity . However , the brain consists of billions of neurons interconnected to form complex circuits , and EEG or MEG cannot reveal changes in activity of these networks in fine detail . In animals , and in patients undergoing brain surgery , scientists can use hair-thin microelectrodes to directly record the activity of individual neurons . Yet , it is difficult to know how activity measured inside the brain relates to that measured outside . To find out , Sandhaeger et al . had monkeys and healthy human volunteers perform the same task , where they had to watch a series of colored dots moving across a screen . The brain of the human participants was monitored using MEG; in the monkeys , EEG provided an indirect measure of brain activity , while microelectrodes directly revealed the activity of thousands of individual neurons . All three recordings contained information about movement and color . Moreover , the monkey EEG bridged the gap between direct and indirect recordings . Sandhaeger et al . identified signals in the monkey EEG that corresponded to the microelectrode recordings . They also spotted signals in the human MEG that matched the monkey EEG . Linking non-invasive measures of brain activity with underlying neural circuits could help to better understand the human brain . This approach may also allow clinicians to interpret EEG and MEG recordings in patients with brain disorders more easily .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	Monkey EEG links neuronal color and motion information across species and scales
Although it is a central question in biology , how cell shape controls intracellular dynamics largely remains an open question . Here , we show that the shape of Arabidopsis pavement cells creates a stress pattern that controls microtubule orientation , which then guides cell wall reinforcement . Live-imaging , combined with modeling of cell mechanics , shows that microtubules align along the maximal tensile stress direction within the cells , and atomic force microscopy demonstrates that this leads to reinforcement of the cell wall parallel to the microtubules . This feedback loop is regulated: cell-shape derived stresses could be overridden by imposed tissue level stresses , showing how competition between subcellular and supracellular cues control microtubule behavior . Furthermore , at the microtubule level , we identified an amplification mechanism in which mechanical stress promotes the microtubule response to stress by increasing severing activity . These multiscale feedbacks likely contribute to the robustness of microtubule behavior in plant epidermis . Epithelia have a crucial role during the development of most multicellular organisms . Consistently , several mechanisms ensure some level of coordination between epithelial cells . In addition to biochemical signals , such as morphogens that diffuse and provide regional coordination across several cell files ( Wolpert , 1969; Jonsson et al . , 2006; Jaeger et al . , 2008; Wartlick et al . , 2009 ) , mechanical stress also contributes to growth coordination , for instance by synchronizing cell proliferation rate ( Shraiman , 2005 ) and orientation ( Thery et al . , 2007 ) , or by prescribing cell polarity ( Asnacios and Hamant , 2012 ) and cell fate ( Engler et al . , 2006 ) . In theory , these coordinating mechanisms could lead to relatively homogeneous cell shapes . While this is observed in some classic cases , such as the cells of Drosophila ommatidia or Arabidopsis petals , most epithelia exhibit variable cell sizes and shapes , demonstrating that each cell retains the ability to regulate its own growth and shape ( Roeder et al . , 2010 , 2012 ) . This heterogeneity has been studied in several systems . In Drosophila embryos , stochastic actomyosin-dependent constrictions of cells occur during gastrulation ( Martin et al . , 2009 ) and dorsal closure ( Solon et al . , 2009 ) , and this stochasticity has been proposed to play a key role in invagination events ( Pouille et al . , 2009 ) . In Arabidopsis sepals , stochastic events including cell division and entry into endoreduplication also play a critical role in the distribution of cells of different shapes ( Roeder et al . , 2010 ) . Altogether this suggests that cell behavior results from both local and supracellular cues . The exact role of such heterogeneity remains poorly explored , and how cells can differentiate between local and global cues is completely unknown . In this study , we show that mechanical stress act as a common instructing signal for microtubule ( MT ) orientation at both subcellular and tissue scales . Mechanical forces have been proposed to provide directional information in control of MT orientation in plant cells and changes in mechanical forces are known to affect microtubule alignment ( Green , 1980; Williamson , 1990; Schopfer , 2006 ) . MT arrays have been proposed to align along maximal mechanical stress directions in the shoot apical meristem , as prescribed by tissue shape , assuming tension in the epidermis ( Hamant et al . , 2008 ) . Mechanical forces were recently found to modify MT organization in leaf epidermal cell layers ( Jacques et al . , 2013 ) . In Arabidopsis and most angiosperms , the cotyledon and leaf epidermal cells , also called pavement cells , exhibit typical jigsaw puzzle shapes , with indented regions and lobe-like outgrowths . The intracellular effectors of these morphologies are being described in many reports . In particular , indenting regions are enriched in cortical MTs , which are thought to restrain growth expansion via the presumptive localized deposition of stiff cellulose microfibrils ( CMF ) ( Fu et al . , 2005; Yang , 2008 ) . Although this model seems relatively parsimonious , these biophysical assumptions have not been tested . The MT severing enzyme katanin is required for local MT ordering in pavement cell indenting regions , downstream of the plant hormone auxin and Rho GTPases ( Lin et al . , 2013 ) . How robust shapes could derive from such regulation is however a subject of debate . The complex morphology of pavement cells is a system of choice to decipher the contribution of cell and tissue shape-derived mechanical stresses in MT behavior . In this study , we have combined computational models and experiments to determine the relation between physical forces , material elasticity , and the behavior of cortical MT . We first relate MT behavior to cell wall reinforcements . Second , we confirm ( in a different tissue than investigated in the past and at a different scale ) that MTs orient along the predicted maximal tensile stress direction—and in this case , that they can do so at a subcellular or a supracellular scale , depending on the stresses involved . Lastly , we take advantage of the large size of the pavement cells to show how the MT response to stress depends on MT severing-dependent self-organization events . Altogether , this provides a scenario , in which not only tissue shape , but also cell shape , depends on a mechanical feedback loop . Based on our results , we propose that cells sense mechanical stresses at the subcellular scale , and that they are hence able to integrate cell shape-derived stresses and tissue shape-derived stresses , with a single mechanism . The presence of parallel bundles of MTs in pavement cells is spatially correlated with indenting neck regions of the cell ( Figure 1A , Figure 1—figure supplement 1; Fu et al . , 2005 ) . However , this correlation is debated , as MT orientations can be very noisy and pavement cell growth has even been proposed to be rather isotropic ( Zhang et al . , 2011 ) . To quantify MT behavior in pavement cells , we used a nematic tensor-based tool to measure MT anisotropy ( Uyttewaal et al . , 2012; Boudaoud et al . , 2014 ) . This showed that MT arrays in indenting regions were more anisotropic than the MT arrays in lobes ( mean ± SE is 0 . 40 ± 0 . 02 for indenting region and 0 . 20 ± 0 . 02 for lobes , n = 18 cells; 3 seedlings; p<0 . 01 , t test; Figure 1A , B , Figure 1—figure supplement 1 ) . Time lapse imaging of pavement cells showed that the anisotropy of MTs was maintained in indenting regions after 3 hr ( mean ± SE is 0 . 47 ± 0 . 01 , n = 14 cells; 3 seedlings ) , consistent with previous studies ( Zhang et al . , 2011 ) ( Figure 1A , B , Figure 1—figure supplement 1 ) . Note that we focused our analysis on a stage where cells have already attained their jigsaw puzzle shape and are still growing; our conclusions do not necessarily apply to these cells at a younger , undifferentiated stage or at an older , fully grown stage . The correlation between pavement cell shape and the presence of stably aligned MTs in indenting regions suggests that pavement cell shape relies on the impact of MTs on the mechanical anisotropy of the cell wall in these regions . This however has never been demonstrated . To directly measure mechanical properties of the outer wall with sufficient resolution , we used atomic force microscopy ( AFM ) to probe different regions of pavement cells , using an established protocol ( Milani et al . , 2011 ) . First , we probed guard cells , in which the presence of transverse MT and CMF is unequivocally established ( Marcus et al . , 2001 ) . The stiffness map of the outer wall of guard cells revealed the presence of transverse wall reinforcements matching the CMF orientation ( Figure 1C ) . Therefore , our AFM approach has sufficient spatial resolution to reveal mechanical heterogeneities in cell walls . The AFM-based stiffness map of pavement cells also revealed the presence of mechanical heterogeneities , with apparent elastic moduli spanning the range 2–8 MPa . In particular , we found fibrous patterns of higher elastic modulus congregating at sites of indenting regions ( Figure 1D , Figure 1—figure supplement 2 ) . These lines are independent of small-scale topography ( Figure 1E ) . Note that with AFM , we are measuring the mechanical properties of the wall in the normal direction to the wall surface with an isotropic probe , and therefore , we are not assessing the very local mechanical anisotropy of the cell wall in-plane . Yet , our data reveal a spatial pattern of mechanical anisotropy at a larger scale , and thus can be used as a proxy for mechanical anisotropy in sub-regions of the cell’s outer wall . Staining for cellulose in mesophyll cells of wheat that exhibit alternating patterns of lobe like outgrowth and indenting regions has shown a strong presence of cellulose in the constricted regions of the cells ( Jung and Wernicke , 1990 ) . Altogether , this suggests that CMFs act as a brace around the indenting regions providing mechanical reinforcement of those regions . 10 . 7554/eLife . 01967 . 003Figure 1 . Mechanical heterogeneity of pavement cells correlates with microtubule patterns . ( A ) Microtubule bundling persits in indenting regions of pavement cells over time . Scale bars 20 μm . ( B ) Microtubule anisotropy over time , lines represent average orientation of microtubule arrays at different time points within the region of interest ( dashed box in A ) . Longer lines indicate higher degree of anisotropy . ( C ) Stiffness map of the outer walls in two guard cells obtained by atomic force microscopy revealing transverse wall reinforcements . Gray scale bar represents scale of the observed elastic modulus in MPa . ( D ) Siffness map of the outer walls in pavement cells obtained by atomic force microscopy ( same scale bar ) . Indenting regions exhibit striations with increased values of elastic modulus , reflecting regions with strong mechanical anisotropy . ( E ) 3D rendering of pavement cell topography as obtained by atomic force microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00310 . 7554/eLife . 01967 . 004Figure 1—figure supplement 1 . Microtubule organization and mechanical heterogeneity . Additional examples of microtubule bundling persisting along the indenting regions of pavement cells over time , far right panels shows lines representing the average orientation of microtubule arrays at different time points within the region of interest ( dashed box ) . Length of the line indicates strength of anisotropy ( longer lines indicate higher degree of anisotropy ) . Scale bars 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00410 . 7554/eLife . 01967 . 005Figure 1—figure supplement 2 . Microtubule organization and mechanical heterogeneity . Additional example of stiffness map of the outer walls in another cotyledon obtained with AFM . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00510 . 7554/eLife . 01967 . 006Figure 1—figure supplement 3 . Microtubule organization and mechanical heterogeneity . ( A and B ) Microtubule orientation on the outer ( A ) and inner ( B ) side of epidermal pavement cells Scale bar 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 006 Recent reports suggest that MT and cellulose synthase orientation could have different behaviors on the inner and outer face of hypocotyl epidermal cells ( Chan et al . , 2010; Crowell et al . , 2011 ) and stems ( Fujita et al . , 2011 ) . To test if this is the case in pavement cells , we visualized MT organization on the inner face of the pavement cells in confocal Z sections . The inner MT arrays appear to be identical to those observed on the outer face , thereby suggesting the cell walls of the indenting regions are reinforced on both faces of the cell ( Figure 1—figure supplement 3A , B ) . These data thus consolidate a model in which pavement cell shape is maintained by localized MT-dependent wall reinforcements in indenting regions ( Fu et al . , 2005 ) . Next , we investigated whether mechanical stress can act as an instructive signal in this MT organization . Physical models of the shoot apical meristem as a pressurized vessel have shown that the supracellular mechanical stress pattern can prescribe global MT organization and tissue morphology ( Hamant et al . , 2008 ) . Scaling down , this model implicitly predicts that mechanical cues also contribute to single cell morphology . However , this question still remains to be explored . To address this , we first computed the expected patterns of stress in pavement cells using a three-dimensional ( 3D ) finite element model representing the outer face of the wall as a curved surface . Tensile stress experienced by the walls is caused by turgor pressure within the cells , therefore the main question was whether cell shape affects the anisotropy of stress in the wall . The 3D geometries of the cells were extracted from confocal microscope images of pavement cells processed with MorphographX and meshed with quadrilateral shell elements ( Kierzkowski et al . , 2012 ) . The boundaries of the cells contain additional beam elements that account for increase in stiffness due to the presence of the anticlinal walls . For the material properties , we used a constitutive model of hyperelastic transversely isotropic material . To account for the mechanical anisotropy of the wall ( Baskin et al . , 1999 ) , elasticity of the tissue was represented by the behavior of an isotropic matrix combined with the resistance of CMF oriented in a single-preferred direction per element . In the simulations , we have assumed that stress pattern arises due to turgor pressure in individual cells accompanied by tension in the epidermal layer ( Dumais and Steele , 2000; Kutschera and Niklas , 2007; Hamant et al . , 2008 ) . The displacement of the anticlinal walls was restricted in z direction . The outcome of the model with mechanically isotropic cell walls showed a strong anisotropic arrangement of stresses focused in indenting regions of the pavement cells ( Figure 2A , C , Figure 2—figure supplement 1A ) . Despite the noisy behavior of MTs in pavement cells , visualization of MT organization with YFP:MBD ( Wightman and Turner , 2007 ) showed a good correlation between the largest stress direction and MT arrangement , aggregating in indenting regions of the pavement cells ( Figure 2B , D , Figure 2—figure supplement 1B ) . Furthermore , the clustering of MT arrays in the indenting regions also correlated with regions of predicted higher magnitude of stress ( Figure 1E , Figure 2—figure supplement 1C ) . Stress directions were also computed in a model representing a 3D pavement cell shape along with the anticlinal and bottom walls . The model generated principal stress tensor direction pattern similar to the curved surface model ( Figure 2—figure supplement 2A–C ) . Both cases showed a higher magnitude of tensile stresses in the indenting neck-like regions , matching the local MT pattern; compression forces at approximately 10% of the tension value were observed in some regions close to the cell boundary with no clear relation with MT behavior ( Figure 2—figure supplement 2D–F ) . This suggests that tensile stress influences MT organization at a subcellular level in pavement cells , thereby indicating that the perception of stress must involve a mechanism that acts locally within each cell , rather than on a cell-wide or tissue-wide basis . 10 . 7554/eLife . 01967 . 007Figure 2 . Microtubule patterns correlate with physical stress patterns . ( A ) Mesh showing stress directions , with the corresponding microtubule organization shown in panel B . Highlighted cells in green are represented in panels C–E and Figure 2—figure supplement 1A–C . ( C ) Largest stress direction ( red ) and second principal stress direction ( green ) in mechanical models of the pavement cell . White arrowheads indicate regions of convergence of directional tensile stresses in necks of pavement cells . ( D ) Microtubule ( YFP-MBD ) orientations correlate with the maximal stress direction predicted in the mechanical model . ( E ) Heat map showing the magnitude of stress distribution in the mechanical model . Arrowheads indicate regions of highest stress magnitude in neck regions . Scale bars 20 μm . ( F ) Circumferential distribution of microtubules surrounding elevated guard cells represented as a depth color-coded Z-stack . ( G ) Mechanical model of stress patterns around a stomata reproduce the observed arrangements of microtubules surrounding guard cells . ( H ) Microtubule organization around a non-elevated stoma and a mechanical model of stress patterns of the same ( I ) . Scale bars 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00710 . 7554/eLife . 01967 . 008Figure 2—figure supplement 1 . Microtubule organization and correlation with stress patterns . ( A–C ) Example cell from Figure 2A showing correlation between predicted physical stresses and microtubule organization . Scale bars 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00810 . 7554/eLife . 01967 . 009Figure 2—figure supplement 2 . Simulation of single pressurized pavement cell shape A and D . 3D cell with epidermal , bottom , and anticlinal wall . ( B and E ) The bottom wall removed and replaced with boundary conditions at the bottom of the anticlinal wall . ( C and F ) Surface model of epidermal wall only with anticlinal walls replaced by boundary condition . ( A–C ) The color map ( Blue to Red ) shows the value of the first principal stresses . ( D–F ) The color map shows the value of second principal stresses . Blue to red mark positive values ( tension ) . The black is negative values ( compression ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 00910 . 7554/eLife . 01967 . 010Figure 2—figure supplement 3 . Microtubule organization and correlation with stress patterns . Color map of the cosine of angle between first principal stress without ( white lines ) and with anisotropic material and feedback to stress direction ( black lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 010 To check whether the mechanical anisotropy of the wall could alter stress direction in pavement cell models , we introduced a CMF direction with five times higher stiffness and subjected the cells to elastic expansion again . A feedback model between the stress perceived by a cell and orientation of the CMFs , which assumes an alignment of the CMF in the direction of maximal stress results in a very minor deviation of the stress pattern from the stress orientation obtained in the case of an isotropic material ( Figure 2—figure supplement 3 ) . The simulations thus suggest that for small elastic deformations the overall stress pattern is affected more strongly by the geometry of the cells than by the mechanical anisotropy of the existing cell wall material . The model also predicts that the topography of the epidermis impact stress patterns and can have a dominant role on MT behavior as compared to cell shape . In particular , assuming that the epidermis is under tension , the largest stress direction should be circumferential around a local bump . Stomata being often in such an elevated position , we analyzed the MT arrays around guard cells and we found circumferential orientations , consistent with the predicted stress directions ( Figure 2F , G ) . Conversely , no circumferential MT pattern could be observed near stomata that were not elevated ( Figures 2H , I ) . The results suggest that , analogously to what was observed in the shoot meristem , patterns of cellular and supracellular mechanical stress and MT orientation are correlated . Furthermore , we find that this correlation holds down to the sub-cellular level . Recent observations of MT organization in epidermal cells of leaves shows a supracellular response of MTs after changes in physical forces ( Jacques et al . , 2013 ) . Application of compressive forces resulted in hyper-alignment of MTs . We tested whether mechanical stress can act as an instructive signal to organize MTs in cotyledon pavement cells by performing mechanical perturbations . Our finite element model predicts that when subjected to compression from above , there is an increase in overall stress in physical models ( Figure 3—figure supplement 1A , B ) . Note that this response would however largely depend on the ability of the epidermis epidermal wall to maintain a constant volume while under compression . We directly applied compressive forces to pavement cells by using a coverslip that was pressed on the surface of the cotyledons and kept in place using adhesive silicone applied on the margins ( Figure 3—figure supplement 1C ) . Depth color-coded Z stacks of MT organization and transects of the confocal Z-stack showed flattening of cells due to compression ( Figure 3—figure supplement 1C , D ) . Imaging of MTs in the cells immediately after and 7 hr after compression showed rearrangement of MTs into more aligned arrays by 7 hr ( Figure 3—figure supplement 1C , D , F; Video 1; mean ± SE is 0 . 37 ± 0 . 06 for 0 hr , n = 95 cells , 5 seedlings and 0 . 57 ± 0 . 06 for 7 hr , n = 95 cells; 5 seedlings; p<0 . 0001; Mann–Whitney U test; Figure 3H ) . This increased MT anisotropy persisted for longer periods when the tissue was maintained in the compressed condition ( Figure 3—figure supplement 2 ) . This response was also reversible: 24 hr following removal of compression , nematic tensor values for MT anisotropy were reduced to a value close to that obtained during the time point immediately after compression ( mean ± SE is 0 . 24 ± 0 . 02; Figure 3E , G , H ) . It should be noted that the nematic tensor values between the 0 hr time point and the recovery state are not identical which could be the result of changes in imaging conditions ( ‘Materials and methods’ ) . However , despite this the reported nematic tensor values clearly show a trend in the increase and decrease of the MT anisotropy in the compression experiments . These results on cotyledon epidermal cells are thus consistent with the recently published report on contribution of mechanical stress in controlling MT behavior in leaf epidermal cells ( Jacques et al . , 2013 ) . 10 . 7554/eLife . 01967 . 011Video 1 . Depth color-coded time series images showing changes in microtubule organization following compression . Scale bar 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01110 . 7554/eLife . 01967 . 012Figure 3 . Extrinsic perturbation of mechanical forces induce directional changes in microtubule arrays . ( A and B ) Mechanical models showing changes in stress directions upon ablation . ( C ) Large scale ablation of cotyledons result in circumferential distribution of microtubule arrays around the site of physical perturbations , reproducing the results of the physical model . ( D ) Microtubule arrays in pavement cells of mutant botero 1-7 shows random organization 7 hr after perturbation . Asterisk marks site of laceration . Scale bar 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01210 . 7554/eLife . 01967 . 013Figure 3—figure supplement 1 . Mechanical compression leads to increased microtubule anisotropy in pavement cells . ( A and B ) Mechanical models predict an increase in mechanical stress compression . ( C and D ) Depth color-coded Z-stack of microtubules , immediately after applying compressive forces ( C ) and after 7 hr of compression ( D ) . ( E ) Recovery of microtubule arrays 24 hr after release of compression . Lower panels show orthogonal projections along dashed lines showing flattening of cells due to compression , and recovery from compression . Arrowhead indicates flattened region . Far right panel shows color bar representing scale along the Z axis . ( F and G ) Scatter plots comparing nematic tensor values of individual cells showing increase of microtubule anisotropy 7 hr after compression ( F ) and decrease following release of compression ( G ) . Colored boxes represent identical cells in both plots . ( H ) Histogram of average microtubule anisotropy values under each condition . Error bars represent standard error . Asterisk shows significance ( p values derived from Mann–Whitney U test ) , N = 5 seedlings , 95 cells per treatment . Scale bar 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01310 . 7554/eLife . 01967 . 014Figure 3—figure supplement 2 . Compression of pavement cells results in stabilization of microtubule array orientation . Note the increased MT anisotropy during compression and reduced MT anisotropy upon release of compression . Scale bars 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01410 . 7554/eLife . 01967 . 015Figure 3—figure supplement 3 . Cotyledon epidermis is under tension . ( A ) Time series images of a cut cotyledon . Scale bar 500 μm . ( B ) Kymograph along the dashed red line in panel ( A ) showing gap opening immediately after physical laceration of the cotyledon , and consistent with release of tension . Dashed line indicates point of laceration and red arrows indicate displacement of the tissue in either direction . Scale bar 100 μm . ( C ) Removal of epidermis in the microtubule reporter line ( MBD:GFP ) by laser ablation shows an upward shift in the position of mesophyll cells of cotyledon tissue , consistent with mesophyll being under compression . Scale bar 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01510 . 7554/eLife . 01967 . 016Figure 3—figure supplement 4 . Microtubule response to changes in physical forces in katanin mutant . Depth color-coded Z stack of microtubule arrays before ( A ) and 8 hr after compression in botero 1-7 ( B ) . Scale bar 50 μm . ( C ) Transect along dashed line in ( B ) showing flattening of cell due to compression . Crosses represent strength of microtubule alignment in one direction or the other . Scale bar 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 016 A limit of the compression test is that water movements may be induced , and while it is extremely difficult to monitor , water flow may alter turgor pressure in the long term and thus the predicted stress pattern . So to further investigate the link between MT behavior and stress in pavement cells , we next induced large cuts to change stress patterns in cotyledons and observed the resulting MT pattern . The observation of tissue deformations after large cuts has long been used to deduce the stress pattern ( Dumais and Steele , 2000 ) and has recently been adapted to calculate solid stresses in animal tumors ( Stylianopoulos et al . , 2012 ) . Macroscopically , large cuts in cotyledons resulted in an immediate outward displacement of the cut edges , consistent with release of tension ( Figure 3—figure supplement 3A , B; Video 2 ) . We also observed upward movement of mesophyll cells from the layer below ( Figure 3—figure supplement 3C ) , consistent with the epidermis being under tension as observed in other plant tissues ( Kutschera and Niklas , 2007 ) . Simulation of such a large-scale nick in physical models was done with the assumption that laceration would lead to removal of turgor pressure from the cut cells followed by a reduction of wall elasticity , as previously published in more local ( Hamant et al . , 2008 ) and more global ( Dumais and Steele , 2000; Kutschera and Niklas , 2007 ) tissue contexts . Our physical model suggests that maximal stress directions become circumferential to the laceration after treatment and independent of cell geometry ( Figure 3A , B; Video 3 ) . To test the model predictions , we performed similar lacerations on cotyledons of seedlings expressing fluorescent reporters of MTs . Time series imaging of MT rearrangements showed a progressive change in the organization of MT arrays around the site . After 3 . 5 hr , we observed hyper-aligned MT arrays in cells adjacent to the damaged region and at 7 hr the alignment advanced to cell layers farther away from the site of laceration , independently of cell shape ( Figure 3C ) . Quantification of MT anisotropy showed a significant increase in MT anisotropy 7 hr after laceration ( mean ± SE is 0 . 080 ± 0 . 007 for 0 hr , n = 61 cells and 0 . 37 ± 0 . 02 for 7 hr , n = 65 cells; 3 seedlings each; p<0 . 0001; Mann–Whitney U test ) , indicating that changes in tissue-wide stresses could override cellular level control of MT anisotropy . 10 . 7554/eLife . 01967 . 017Video 2 . Laceration of cotyledon shows outward displacement of cut edges . Scale bar 500 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01710 . 7554/eLife . 01967 . 018Video 3 . Video of computational simulation showing circumferential distribution of stress and increase in MT anisotropy after ablation . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 018 To decipher if the response of MTs to mechanical perturbation depended on the intensity of stress imposed on the tissue , we performed simulation in which single cells were removed and the resulting patterns of stress computed . Similar to the large scale laceration experiments , a circumferential rearrangement of stresses was observed . However , the rearrangement was less pronounced than after cotyledon dissection , and observed in cells only adjacent to the site of perturbation ( Figure 4A , B ) . To test this experimentally , we ablated single cells using a pulsed dye-coupled laser to see if small-scale perturbations could decouple the cellular level control of microtubule organization . 7 hr after ablation , we observed that MT alignment had responded to treatment by tending toward a circumferential alignment around the ablated cell site . The MTs around the site of ablation were not completely circumferential , consistent with a scenario in which stress anisotropy is weaker in single cell ablation cases than in cases of large-scale dissection , therefore leading to a comparatively reduced alignment of MTs around the site of ablation ( Figure 4C , D , E ) . The alignment observed after single-cell ablation was significantly different from that resulting from large-scale perturbation , in which a strong circumferential response was obtained . 10 . 7554/eLife . 01967 . 019Figure 4 . Stress intensity regulates microtubule alignment . ( A and B ) Simulation showing less pronounced circumferential rearrangements of stresses after ablation of single cell . Images of microtubule reporter line before ( C ) and 7 hr after ( D ) ablation of single cell , showing aligned microtubule arrays not completely circumferential after ablation of single cell . Scale bars 50 μm . ( E ) Magnified images of cells in figure D ( Images not to scale ) . Asterisk shows the location of the ablated cell . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 01910 . 7554/eLife . 01967 . 020Figure 4—figure supplement 1 . Microtubule array organization in guard cells remains unaffected by changes in directional force field . Close up image of microtubule arrays in guard cell before ( A ) and 7 hr after ( B ) large-scale ablation . Majority of the guard cells retain the transverse pattern of microtubule arrays after laceration . Scale bars 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 02010 . 7554/eLife . 01967 . 021Figure 4—figure supplement 2 . Microtubule response to isoxaben treatment . Changes in microtubule arrays before ( A ) and after treatment with 40 μM isoxaben for 16 hr ( B ) . ( C ) Histogram showing increase in nematic tensor values after isoxaben treatment . Error bars represent standard error , asterisk shows significance ( Mann–Whitney U test; p<0 . 0001 ) . Scale bars 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 021 To further test if a directional force field could alter transverse MT organization observed in guard cells , we monitored MT arrays in guard cells adjacent to the region of large scale laceration . Single optical sections show that MT arrays still remain transverse in majority of the guard cells ( Figure 4—figure supplement 1 ) indicating that these cells retain a certain degree of control of their MTs while being under the influence of a directional force field . The observed rearrangement of MTs around the site of laceration supports our hypothesis but several other possibilities could exist . The changes observed could be due to a modified pattern of mechanical stress but also to the biochemical consequences of wounding of the sample . To investigate this we used the cellulose synthase inhibitor isoxaben , as a pharmacological means of increasing mechanical stress in pavement cells . Isoxaben was shown previously to induce hyperbundling and hyperalignment of cortical microtubules along the predicted directions of maximal stress in shoot meristem cells ( Heisler et al . , 2010 ) , consistent with increased stress levels in the cell walls compared to non-treated plants . Treatment of 3-day-old seedlings with 40 μM isoxaben for 16 hr led to a sizeable increase in the anisotropy of microtubules ( mean ± SE is 0 . 27 ± 0 . 04 for 0 hr , n = 53 cells and 0 . 54 ± 0 . 08 for 16 hr , n = 53 cells; 3 seedlings; p<0 . 0001 , Mann–Whitney U test; Figure 4—figure supplement 2A–C ) . These observations add further evidence that mechanical forces are responsible for MT rearrangement . However , it should be noted that a much more complex scenario , involving both mechanics and wound responses , could govern this aspect of MT rearrangements . Homozygotes for the katanin loss-of-function allele ktn1-3 exhibit severe pavement cell shape defects , consistent with the role of MTs in neck formation ( Lin et al . , 2013 ) . When quantifying MT anisotropy in the bot 1-7 katanin allele , we found a 60% reduction when compared to the wild type ( Mean ± SE is 0 . 09 ± 0 . 01 for bot 1-7 , n = 83 cells; 4 seedling and 0 . 23 ± 0 . 05 for WT , n = 53 cells; 3 seedlings ) , thus confirming the relation between MT ordering and pavement cell shape . It has been shown in the shoot meristem that the response of MTs to tissue stress relies on MT self-organization . In particular , MT rearrangement after changes in mechanical stress is promoted by katanin-dependent MT severing ( Uyttewaal et al . , 2012 ) . To test whether , as in the shoot meristem , katanin activity is required for the MT response to stress in pavement cells , we performed lacerations and compressions in the katanin mutant background bot1-7 . In both assays , MT arrays in pavement cells remained relatively isotropic after the micromechanical perturbations ( Figure 3D , Figure 3—figure supplement 4A–C; Video 4 ) . Measurement of MT anisotropy showed no significant difference before and 7 hr after ablation in bot 1-7 ( mean ± SE is 0 . 09 ± 0 . 02 for 0 hr , n = 83 cells and 0 . 10 ± 0 . 02 for 7 hr , n = 74 cells; 3 seedlings; p>0 . 05 , t test; Figure 3D ) . Only subtle alignments could be detected in the immediate vicinity of the ablated zone in large-scale lacerations , but no alignment over several cell files was observed ( Figure 3D ) . These results are consistent with observations made in the shoot apical meristem ( Uyttewaal et al . , 2012 ) and suggest that the MT response to stress is not only conserved in different tissues and at different scales , but also relies on a similar mechanism . 10 . 7554/eLife . 01967 . 022Video 4 . Depth color-coded time series images showing microtubule arrays in botero 1-7 does not induce hyper-alignment of microtubule arrays after compression . Scale bar 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 022 As quantification of severing events in meristems is impractical due to difficulties in accessing the cells for high-resolution imaging , we took advantage of the large size of pavement cells to further test this conclusion . To do so , we directly scored single MT-severing events in YFP-MBD cells adjacent to the lacerated region in pavement cells ( Figure 4A ) . Using time-series imaging of MT arrays , we could observe severing of MTs occurring at sites of crossover points of MTs ( Figure 5A; Wightman and Turner , 2007; Lindeboom et al . , 2013; Wightman et al . , 2013; Zhang et al . , 2013 ) . The severing rate in mock-treated control seedlings of 0 . 08 ± 0 . 02 × 10−3 events μm−2 min−1 ( mean ± SD , N = 16 cells , 4 seedlings , total area of 2 . 99 × 104 μm2 ) is in agreement with what has been previously published ( Wightman and Turner , 2007; Wightman et al . , 2013; Zhang et al . , 2013 ) . However , in seedlings subjected to changes in mechanical forces , cell files close to the site of ablation showed a significant increase in severing rate to 0 . 93 ± 0 . 45 × 10−3 events μm−2 min−1 ( mean ± SD , N = 16 cells , 4 seedlings , total area of 1 . 5 × 104μm2 , p<0 . 01 , t test ) 3 hr after ablation ( Figure 5B; Video 5 ) . Measurement of number of crossover points ( i . e . , two polymerizing MT crossing each other without experiencing severing or catastrophe ) immediately after and 3 hr after mock treatment showed a 9% increase ( mean ± SE is 11 . 0 ± 1 . 4 per 10 μm2 for 0 hr , n = 6 cells and 12 ± 1 . 5 per 10 µm2 for 3 hr ) whereas the cells experiencing changes in mechanical stress showed a 65% decrease in the number of cross over points ( mean ± SE is 15 ± 3 per 10 μm2 for 0 hr , n = 6 cells and 5 ± 2 per 10 μm2 for 3 hr ) . These measurements demonstrate an up-regulation in the severing of MTs at crossover sites after changes in mechanical stress . This mechanism could enrich the population of free MTs , while removing the MTs that do not align parallel to maximal tensile stress , thereby resulting in the generation of anisotropic MT arrays . 10 . 7554/eLife . 01967 . 023Figure 5 . Mechanical perturbations increase bundling by promoting severing . 3D surface plot of YFP microtubule time series images representing a typical microtubule severing event ( A ) , arrowheads indicate microtubule-severing at a crossover sites . Scale bar 5 μm . ( B ) Histogram representing microtubule severing rates of mock treated seedlings and in seedling of cells adjacent to site of ablation after 4 hours . Error bars represent standard deviation . Asterisk shows significance . ( Student’s t test ) . N = 16 cells , 4 seedlings , total area of 2 . 99 × 10 μm for mock and N = 16 cells , 4 seedlings , total area of 1 . 5 × 10 μm for ablation . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 02310 . 7554/eLife . 01967 . 024Video 5 . Video showing severing of microtubule immediately after and 4 hr post ablation of cells . Red dots mark sites of microtubule severing . Scale bar 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 024 The role of mechanical stress in animals has mainly been investigated in single cells . Some pioneering studies in Drosophila have investigated the role of mechanical stress in tissues , and most notably in germ band extension ( Lecuit and Lenne , 2007; Sherrard et al . , 2010 ) , dorsal closure ( Martin et al . , 2009; Solon et al . , 2009 ) , and gastrulation ( Pouille et al . , 2009 ) in relation to actomyosin reorganization . However the mechanical properties of these tissues have not been studied , notably because of the difficult physical access to these embedded tissues , and because the rapid growth dynamics of these tissues is not compatible with high resolution mechanical measurements , using for instance , atomic force microscopy . Because plant growth is much slower and because tissues are easily accessible , such measurements are possible in plants . Here , we could correlate the behavior of the microtubular cytoskeleton with the predicted stress pattern and with quantified mechanical properties . This validates a number of previous studies assuming that pavement cell shape depends on mechanical heterogeneities ( Fu et al . , 2005 ) , albeit relying only on MT behavior , and further supports plant tissues as facile systems to investigate the relation between the biophysics of growth and development . Current models on pavement cell morphology suggest an influence of a ROP-based signaling mechanism on the cytoskeletal network ( Fu et al . , 2005; Yang , 2008 ) . The findings suggest that growth is restricted in pavement cell indenting regions due to localized accumulation of anticlinal MT arrays by regulating CMF deposition , whereas local outgrowths are associated with the presence of an actin cytoskeleton that promotes growth . More recently it has been proposed that an auxin-dependent self-organizing mechanism controls the ROP-based signaling network ( Xu et al . , 2010 ) . This model of differential growth is challenged by time lapse-imaging studies of pavement cell development , which shows an initial phase of multiple lobe initiations , followed by a phase of isotropic expansion during which the cell shape is maintained ( Zhang et al . , 2011 ) . These studies do not take into account the mechanical aspects of the cell . In our study we show how subcellular mechanical stresses control MT organization , which in turn affects the mechanical anisotropy of the cell wall thereby contributing to cell shape irrespective of the previously proposed mechanisms . We believe that such a signaling module could add robustness to shape changes at the cellular scale . Our studies further shed light on two theories that are at the center of many debates in biology . The organismal theory of multicellular organisms states that comprehension of tissue properties is essential to understanding the development of the organism , as opposed the view of cell theory , that proposes that the function of individual cells is what dictates development of the entire organism ( Kaplan , 1992; Baluska et al . , 2004 ) . Previous findings on the shoot meristem suggest that MT organization depends on tissue shape-derived stresses ( Hamant et al . , 2008 ) . Here , we found that cells can also interpret mechanical signals that are generated by their own shape . As this response is lost in large-scale mechanical perturbations , we propose that MT behavior depends on stress intensity , which is cell autonomous as long as tissue stresses do not override it . Such a balance between cell autonomous and non-cell autonomous stresses could control subcellular events in animal systems too . Mechanical forces are known to cause changes to single MT dynamics in vivo by regulating the activity of MT-associated proteins ( Trushko et al . , 2013 ) . In several organisms tensile forces acting on the kinetochore complex-attached MT ends are known to influence MT elongation rates during mitosis ( Nicklas , 1988; Skibbens et al . , 1993; Gardner et al . , 2005; Franck et al . , 2007 ) . It remains to be tested if changes in mechanical forces could alter single MT dynamics in whole organism based studies . Previous studies on the shoot apical meristem show that katanin-mediated MT severing is required for MT response to changes in mechanical stresses ( Uyttewaal et al . , 2012 ) . Our data not only show that a similar mechanism is present in pavement cells , but also show that severing itself is promoted by mechanical stress . One possible mechanism for stress-dependent regulation of MT alignment , where each MT is stressed due to their coupling with CMFs potentially by means of the cellulose synthesizing complexes which ride along the MT synthesizing the CMFs parallel to the MTs ( Paredez et al . , 2006; Bringmann et al . , 2012 ) . In such a scenario , a MT under tension is strained producing a conformational change to the MT lattice , which then could bias the binding of katanin to the less strained MTs at crossover sites , leading to preferential severing of the MTs that are not aligned to the principal stress direction . Reports have shown that conformation changes to the MT lattice act as hotspots promoting katanin binding and activity ( Davis et al . , 2002; Díaz-Valencia et al . , 2011 ) . As a result , this strain in MTs would lead to a MT alignment parallel to the anisotropic stress , not to overall cellular strain . MT behavior would thus depend on two parameters of mechanical stress ( Figure 6 ) : stress direction from cell and tissue shape determine the dominant MT orientation , which dictates wall reinforcement , and thus in turn contributes to cell and tissue shape; and stress intensity modulates severing activity , which controls the response of MTs to stress . Based on these results , there is no need for several rules to explain MT behaviors at different scales or in different tissues in response to mechanical signals and perturbations , and this parsimony might be one of the defining features of mechanical signaling , when compared to molecular signaling . As shape rather relies on the actomyosin network in animals , the impact of mechanical forces on actin dynamics may in principle have a similar multiscale role in animals . 10 . 7554/eLife . 01967 . 025Figure 6 . Mechanical forces regulate pavement cell shape by controlling microtubule organization and cellulose deposition . DOI: http://dx . doi . org/10 . 7554/eLife . 01967 . 025 Arabidopsis thaliana lines expressing microtubule ( MT ) reporters YFP:MBD ( Landsberg erecta ) was previously used by Wightman and Turner ( 2007 ) and MBD:GFP ( Col-0 ) was used by Hamant et al . ( 2008 ) . botero 1-7 ( WS ) was isolated previously by Bichet et al . ( 2001 ) , YFP:MBD in a botero 1-7 background and YFP:MBD ( Wassilewskija ) ( Wightman et al . , 2013 ) was a kind gift from Simon Turner ( University of Manchester ) . All seeds were surface sterilized , stratified for 2 days , and grown vertically on plates containing half strength Murashige and Skoog ( MS ) media in light ( 16-hr photoperiod ) at 21°C for 5 days for confocal microscopy . 5 days after germination plants were transferred to sterile plastic boxes containing MS media . The plants were fixed by adding lukewarm 1% agarose to the hypocotyls and roots submerging them , thereby exposing the cotyledons . Plants were imaged using a Zeiss LSM-780 or Zeiss LSM-700 as described in Heisler et al . ( 2010 ) . Laser-induced ablation was performed using an Andor Micropoint ablation laser fitted to a Zeiss LSM-780 scanning confocal microscope as described ( Hamant et al . , 2008 ) using a 63X or 40X water dipping objective . Compression was achieved by placing cotyledons tightly mounted between a cover glass and slide with silicon grease . Imaging of the compressed tissue was performed using an oil immersion lens . To image MT recovery after compression , seedlings were transferred to MS media containing Petri dishes after carefully removing the coverslip and treated as described above before imaging with a water-dipping lens . Laceration of the cotyledon was performed using a sharp forceps or a scalpel . Isoxaben treatment was preformed as described in Heisler et al . ( 2010 ) . All images were processed and analyzed using the FIJI or ImageJ software . Background subtraction was performed using the ‘Subtract Background’ tool ( rolling ball radius 30–40 pixels ) , and the ‘StackReg’ plugin was used to correct focal drift of the sample . Depth color-coding was performed using the ‘Temporal-color code’ tool . Cell boundaries in Figure 2B are pseudo projection of the lower most stack in a different color . Walking average function was performed on time series data sets used for analysis of severing events . Quantification of MT alignments for anisotropy was done using the ImageJ macro described in Uyttewaal et al . ( 2012 ) and Boudaoud et al . ( 2014 ) for Figure 1 and for the laceration experiments in which a score of 0 indicates completely isotropic pattern and 1 , a case of completely aligned pattern . For all other experiments , the macro was modified in order to include in the output the visual and numerical measure of anisotropy of the nematic and texture tensors . The modified macro displays directions of both eigenvectors of the tensor of which relative length is an indication of the anisotropy of the signal . The score is calculated as a difference between the principal eigenvalues divided by the sum of diagonal elements of the tensor . The score gives 0 in case of completely isotropic pattern and 2 in case of completely aligned pattern . Analysis was carried out in each cell by drawing an outline of the region of interest , using the ‘Polygon selection’ tool in ImageJ . The outlines covered the entire area of the cell , without the anticlinal wall signal . These regions were then recorded using the ‘ROI manager’ tool in ImageJ and saved for automatic selection of the same cells in other time points . Severing quantification was preformed manually using ImageJ , where severing events are scored using the point picker tool in ImageJ based on the criteria described in Wightman et al . ( 2013 ) . Quantification was carried out in cell files extending up to four layers from the site of ablation . A total of 15 events were observed in the mock treated samples in a total of 27 . 2 min and 66 events 3 hr after ablation in a total of 20 . 9 min . Statistics is preformed in Excel or OriginPro . For data sets that do not have a normal distribution , we have preformed Mann–Whitney U test , a non-parametric test of the null hypothesis; for normally distributed data sets , we have used the t test to estimate statistical significance . AFM indentation experiments were carried out with a Catalyst Bioscope ( Bruker Nano Surface , Santa Barbara , CA ) , that was mounted on a optical macroscope ( MacroFluo , Leica ) using an objective ( 5x plano objective , Leica ) . To record surface topology and to create an elastic modulus map , PeakForce QNM AFM mode is used . A Nanoscope V controller and Nanoscope software versions 8 . 15 were utilized . All quantitative measurements were performed using standard pyramidal tips ( ScanAsyst Air , Bruker , Inc . ) . The tip radius is given by the manufacturer to be between 2 nm and 10 nm . The spring constant of cantilevers was measured using the thermal tuning method ( Hutter and Bechhoefer , 1993; Levy and Maaloum , 2002 ) and was ranged from 0 . 3–0 . 5 N/m . The deflection sensitivity of cantilevers was calibrated against a clean silicon wafer . All measurements were made under water at room temperature and the standard cantilever holder for operation in liquid was used . Cotyledons were detached from the stem with a fine blade and the abaxial ( lower ) side of the cotyledon was used for the measurement . The upper side of the sample was adhered to a Petri dish using a tissue section adhesive ( Biobond , Ted Pella , Inc . ) . Then the Petri dish was positioned on an XY motorized stage and held by a magnetic clamp . Then , the AFM head was mounted on the stage and an approximated positioning with respect to the cantilever was done using the optical macroscope . The foundation of material property mapping with PeakForce QNM is the ability of the system to acquire and analyze the individual force curves from each tap that occurs during the imaging process . In this mode , the probe is oscillated at a low frequency ( 0 . 5 kHz ) , capturing a force curve each time the AFM tip taps on the sample’s surface . The maximum force during imaging was 1 μN . For each sample , the topology and elastic modulus images were collected from different places of the sample over sizes of 70 × 70 to 150 × 150 µm2 and at a digital resolution of 128 pixels × 128 pixels . The 0 . 3 Hz scanning rate was used . In this QNM mode technique , the elastic modulus is derived from the force–indentation curves by using 2 different models: ( i ) the Hertz–Sneddon model ( Sneddon , 1965 ) or ( ii ) the Derjaguin-Muller-Toporov ( DMT ) model ( Derjaguin et al . , 1975 ) . Both assumed a rigid cone indenting a flat surface . Unlike the Hertz–Sneddon model ( Equation 1 ) that is applied on the approach curve , the DMT model is applied on the retract curve and accounts for adhesion ( Equation 2 ) . ( 1 ) F=2πE ( 1−ν2 ) tan ( α ) δ2 F is the force from force curve , E is the Young’s modulus , ν is the Poisson’s ratio , α is the half-angle of the indenter and δ is the indentation . ( 2 ) F−Fadh=43ER ( d−d0 ) 3 F − Fadh is the force on the cantilever relative to the adhesion force , R is the tip end radius , and d − d0 is the deformation of the sample . For this last model , the result of the fit is the reduced modulus E . Moreover , if the Poisson’s ratio is known , the software can use that information to calculate the Young’s Modulus of the sample by the equation ( Equation 3 ) : ( 3 ) E*=E ( 1−ν2 ) Here , we assumed our sample is perfectly incompressible so that the Poisson’s ratio used is 0 . 5 . However , since neither the Poisson’s ratio nor the tip shape is accurately known , we report in this work only an ‘apparent modulus’ ( Ea ) . For simulations of tissue mechanics , we have used a nonlinear Finite Element Method which provides an approximate solution of continuum elasticity problems on domains with complex geometry ( Zienkiewicz et al . , 2005 ) . Simulations involving pavement cell geometries were performed with in-house written software , which is specialized and optimized for cellular geometries . The software is based on a procedure of minimization of nonlinear strain energy for a given constitutive model of the material ( Bonet and Wood , 1997 ) . We have used a hyperelastic transversely isotropic model of the material , which is particularly suited for description of fibrillar tissues ( Weiss et al . , 1996 ) . For the matrix part of the material , we employed the Saint Venant–Kirchhoff model . The epidermal cell wall surface was modeled with shell elements , designed to handle thin curved surfaces . Additional beam elements accounting for the presence of the anticlinal cell walls were placed along the projections of the individual cell boundaries on the epidermal surface . These beam elements were assumed to have Young’s modulus of the matrix part of the material and thickness of 1/5 of the epidermal wall . For the model on feedback between wall anisotropy and mechanical stresses we have set the material anisotropy axis at each step of the Newton–Raphson iteration aligned with maximal principal stress measured in the previous step . The model parameter values were chosen from different experimental estimates of plant cell wall elasticity based on measurements of in vivo samples ( Suslov et al . , 2009; Hayot et al . , 2012; Nezhad et al . , 2013 ) and synthetic bio-composites ( Chanliaud et al . , 2002 ) . We set Young’s modulus equal to 40 MPa for the matrix part of the material . The Young’s modulus of the CMF was assumed to be five times larger than the matrix . Turgor pressure was set to 0 . 2 MPa and thickness of the epidermal wall was assumed to be 1 μm . We have also tested different values of these parameters varying them in the range 0 . 5–2 times the presented values and found the results of the simulations to be qualitatively consistent within these limits . To account for tension in the epidermal layer the outer boundary of the templates was expanded by 1% in the x-y plane . The inner boundaries of individual cells were free in the x-y plane and restricted in the z direction . Pressure was applied on the surface of each cell . The choice of low turgor pressure was mainly dictated by convergence requirements of our model . Large turgor pressures caused instability of the model with respect to x-y movement of the cell boundaries , as a result of our approximate treatment of anticlinal walls . We have found , however , that the stress pattern does not show major qualitative changes for larger pressures . Thus , we have used the turgor within low range of experimentally reported values assuring good convergence with reasonable computational cost . Most of the simulations involving the cell shapes extracted from experimental data were started from a flat geometry . The boundaries of the cells were obtained from the experimental data using MorphographX software ( Kierzkowski et al . , 2012 ) and the meshes constrained to those boundaries were constructed with the use of the CGAL algorithms ( CGAL , 2014 ) . For the stomata simulations ( Figure 2 ) , we have extracted information about the three-dimensional geometry of the cell surfaces using the three-dimensional confocal microscopy data and MorphographX software . In this case , the initial mesh was projected on the estimated real cell surface . Simulations involving simpler geometries were performed with Abaqus ( Simulia , Providence , RI ) finite element modeling software ( Figure 2—figure supplement 2 , Figure 3—figure supplement 1 ) , using the same material model as in other simulations and shell finite elements . These compression simulations were performed with a constant volume assumption and node to surface frictionless contact . Similar results were obtainable with a constant pressure assumption , but in that case the outcome of the simulation was dependent on material properties and the degree of deformation from pressing . To investigate the importance of the anticlinal walls on the stiffness of the epidermal wall model , we performed simulations with a simplified pavement cell shape including anticlinal walls in Abaqus ( Dassault Systemes; Figure 2—figure supplement 2 ) . The results were quantitatively consistent with our approximate model of the anticlinal walls . A detailed description of models is provided in supplemental information ( Supplementary file 1 ) .	The surfaces of plants are covered in epithelial cells that come in many different shapes , suggesting that individual cells must have some control over their own shape . An unusually shaped epithelial cell is the pavement cell , which looks like a jigsaw puzzle piece and is found in the leaves of many flowering plants . Relatively little was known about the exact contribution of mechanical properties of the wall to this shape . Furthermore , although it was known that parts of pavement cells are rich in microtubules—tubes of protein that act as a scaffold inside the cell— the possibility that shape impacts the behavior of microtubules was not fully addressed . Now , using a combination of computer modelling and experiments , Sampathkumar et al . reveal that the shape of the pavement cells relies in part on the response of the microtubules to stress . In an individual cell , microtubules align along the direction of the largest stress , with a protein severing those microtubules that are not aligned in this direction . As the stress inside a cell is determined in part by the cell’s shape , this sets up a feedback loop: the stress resulting from the cell shape aligns the microtubules that reinforce the cell wall , thus maintaining the shape of the cell . An external stress applied to the epithelium can override this internal stress . Because all of the plant cells are under turgor pressure from the inside , pressure from the outside , like squeezing a balloon , changes the stress pattern , causing the realignment of the microtubules so as to resist the new stress . This shows that the microtubules respond to local stresses within a cell , and are continually responsive to stress changes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "cell", "biology" ]	2014	Subcellular and supracellular mechanical stress prescribes cytoskeleton behavior in Arabidopsis cotyledon pavement cells
Perforin-2 ( MPEG1 ) is an effector of the innate immune system that limits the proliferation and spread of medically relevant Gram-negative , -positive , and acid fast bacteria . We show here that a cullin-RING E3 ubiquitin ligase ( CRL ) complex containing cullin-1 and βTrCP monoubiquitylates Perforin-2 in response to pathogen associated molecular patterns such as LPS . Ubiquitylation triggers a rapid redistribution of Perforin-2 and is essential for its bactericidal activity . Enteric pathogens such as Yersinia pseudotuberculosis and enteropathogenic Escherichia coli disarm host cells by injecting cell cycle inhibiting factors ( Cifs ) into mammalian cells to deamidate the ubiquitin-like protein NEDD8 . Because CRL activity is dependent upon NEDD8 , Cif blocks ubiquitin dependent trafficking of Perforin-2 and thus , its bactericidal activity . Collectively , these studies further underscore the biological significance of Perforin-2 and elucidate critical molecular events that culminate in Perforin-2-dependent killing of both intracellular and extracellular , cell-adherent bacteria . As the largest class of ubiquitin ligases , cullin-RING E3 ubiquitin ligases ( CRLs ) regulate numerous cellular processes including signal transduction , gene expression , development , and cell cycling ( Bosu and Kipreos , 2008; Metzger et al . , 2012 ) . CRLs are modular complexes that are assembled from a profusion of subunits . However , most share a similar architecture . At the core of each lies an elongated cullin upon which other CRL subunits assemble ( Bosu and Kipreos , 2008 ) . Adaptor molecules bind to the cullin's extended amino-terminal domain and are themselves bound by receptors that provide substrate specificity ( Wu et al . , 2003; Cardozo and Pagano , 2004; Petroski and Deshaies , 2005; Lydeard et al . , 2013 ) . The RING subunit binds to the cullin's globular carboxy-terminal domain and acts as an E3 ubiquitin ligase responsible for recruiting the complex's ubiquitin conjugating enzyme ( E2 ) . The placement of the substrate and E2 at opposite ends of the elongated cullin translates into a separation of ∼50 Å ( Wu et al . , 2003; Hao et al . , 2007; Merlet et al . , 2009 ) . This gap prohibits ubiquitylation of the substrate . This problem is solved by an additional E2 enzyme , such as UBC12 , that conjugates NEDD8 , an 8 . 6 kDa member of the ubiquitin family of proteins ( UniProt entry Q15843 , Pfam identifier PF00240 ) , to a conserved lysine within the carboxy-terminal domain of the cullin ( Petroski and Deshaies , 2005 ) . Cullin neddylation induces a conformational change that places the ubiquitin E2 and substrate in sufficient proximity for ubiquitylation to occur ( Duda et al . , 2008; Saha and Deshaies , 2008 ) . Thus , CRL-dependent ubiquitylation of a protein substrate is itself dependent upon cullin neddylation ( Morimoto et al . , 2000; Ohh et al . , 2002; Sakata et al . , 2007 ) . Cycle inhibiting factors ( Cifs ) are bacterial effector proteins that inactivate CRLs through deamidation of NEDD8 ( Cui et al . , 2010; Boh et al . , 2011; Crow et al . , 2012 ) . They are delivered to the cytosol of eukaryotic cells by type III secretion systems of some Gram-negative pathogens including Yersinia pseudotuberculosis and enteropathogenic Escherichia coli ( EPEC ) ( Marches et al . , 2003; Charpentier and Oswald , 2004; Jubelin et al . , 2009; Taieb et al . , 2011 ) . Upon entering the cytosol Cifs proceed to deamidate Gln40 of NEDD8 thus producing a Glu residue at that position ( Cui et al . , 2010 ) . Because CRL activity is dependent upon NEDD8 , this enzymatic modification prevents the ubiquitylation of CRL substrates ( Marches et al . , 2003; Saha and Deshaies , 2008; Toro et al . , 2013 ) . The discovery of Cif and elucidation of its enzymatic mechanism can be traced back to initial reports that certain pathogens cause cell cycle arrest ( De Rycke et al . , 1997; Nougayrede et al . , 2001 ) . It is now known that Cif causes the accumulation of cell cycle inhibitors by blocking their ubiquitylation and subsequent degradation by the 26S proteasome ( Marches et al . , 2003; Taieb et al . , 2006; Samba-Louaka et al . , 2008 ) . It has been proposed that Cif mediated cell cycle arrest provides enteric pathogens , such as EPEC and Y . pseudotuberculosis , with a stable platform by decreasing the turnover rate of intestinal epithelial cells ( Samba-Louaka et al . , 2009 ) . In theory this mechanism could promote colonization of the gastrointestinal tract . Although this is certainly a well reasoned hypothesis , to the best of our knowledge it remains untested . In addition , CRLs regulate diverse cellular processes . Although the majority of these processes probably have no role in limiting pathogen virulence , others may . Therefore , it is not yet known how the deamidation of NEDD8 by Cif contributes to pathogenicity . Recent studies suggest that Perforin-2 ( macrophage-expressed gene 1; MPEG1 ) is an effector of the innate immune system that limits the proliferation and spread of medically relevant Gram-negative , -positive , and acid fast bacteria . For example , expression of Perforin-2 in murine embryonic fibroblasts ( MEFs ) is associated with the killing of intracellular Salmonella enterica serovar Typhimurium ( hereafter Salmonella typhimurium ) , methicillin-resistant Staphylococcus aureus ( MRSA ) , Mycobacterium smegmatis , and Mycobacterium avium ( McCormack et al . , 2013b ) . Moreover , siRNA knockdown of Perforin-2 expression abolished the ability of MEFs to destroy intracellular bacteria unless the siRNA transfected cells were also complemented with a siRNA resistant Perforin-2-RFP expression plasmid ( McCormack et al . , 2013b ) . More recently , knockdown of Perforin-2 allowed Chlamydia trachomatis , an obligate intracellular pathogen that normally replicates within epithelial cells , to proliferate within mammalian macrophages ( Fields et al . , 2013 ) . In vivo , Perforin-2 is critical for protection against MRSA and S . typhimurium ( McCormack et al . , 2015 ) . Additional studies by McCormack et al . suggest that Perforin-2 has a primary role in the destruction of bacterial pathogens ( McCormack et al . , 2015 ) . Nevertheless , Perforin-2 remains a poorly characterized molecule . To address this deficiency , we sought to characterize the molecular events required for the activation and deployment of Perforin-2 through a variety of in vitro and in vivo infection models . In this study we show that Perforin-2 is ubiquitylated by a CRL complex containing CUL1 and βTrCP in response to infectious bacteria or pathogen associated molecular patterns ( PAMPs ) such as LPS . Ubiquitylation triggers a rapid reorganization of Perforin-2-RFP within the cytosol and is absolutely required for its bactericidal activity . Due to its ability to inactivate CRLs , the bacterial effector protein Cif blocks ubiquitylation of Perforin-2 and its subsequent redistribution within infected cells . This blockade prevents Perforin-2-dependent killing of bacteria in vitro . In vivo , wild-type Y . pseudotuberculosis is significantly more virulent than a cif mutant . This difference was not observed with Perforin-2 deficient mice which succumb to an otherwise non-lethal dose of either wild-type or Cif− bacteria . With regards to pathogenicity , these latter results suggest that the inhibition of Perforin-2-dependent killing is the primary function of Cif . Perforin-2 contains an amino-terminal membrane attack complex perforin ( MACPF ) domain ( Figure 1 ) . This domain is also present in the pore-forming components of complement and Perforin-1 ( Podack and Tschopp , 1982; Dennert and Podack , 1983; Podack and Dennert , 1983; DiScipio et al . , 1984; Lichtenheld et al . , 1988 ) . Moreover , crystallographic studies of MACPF domains have revealed structural similarities to bacterial pore-forming CDCs ( cholesterol dependent cytolysins ) ( Rosado et al . , 2007; Rosado et al . , 2008; Slade et al . , 2008; Law et al . , 2010 ) . Thus , the presence of a MACPF domain within Perforin-2 suggests that it may also have the ability to form lytic pores . This hypothesis is supported by the recovery of Perforin-2 from bacteria and images of pores with an average diameter of 100 Å in lipid membranes of mammalian cells expressing Perforin-2 and bacterial cell walls exposed to Perforin-2 ( McCormack et al . , 2015 ) . The MACPF domain of Perforin-2 is immediately followed by a domain of unknown function ( Figure 1 ) . We have dubbed this domain the Perforin-2 ( P2 ) domain because it has been conserved throughout evolution in Perforin-2 orthologs and is only associated with the MACPF domain of Perforin-2 . Although the function of the Perforin-2 domain remains to be elucidated , its conservation suggests that it is essential . The presence of a putative transmembrane alpha helix is an additional feature that distinguishes Perforin-2 from Perforin-1 and the pore-forming components of complement ( Figure 1 ) . Topological modeling and sequence analyses indicate that Perforin-2 is a type I transmembrane protein with a relatively short carboxy–terminal domain ( McCormack et al . , 2013b ) . The localization of Perforin-2 to vesicles staining with markers for endoplasmic reticulum , Golgi , early endosomes , and plasma membrane is consistent with this prediction ( McCormack et al . , 2015 ) . 10 . 7554/eLife . 06505 . 003Figure 1 . Domain organization of human C8a , C9 , Perforin-1 , and Perforin-2 . The pore forming components of complement ( C8α and C9 ) , Perforin-1 , and Perforin-2 all contain membrane attack complex perforin ( MACPF ) domains and amino-terminal signal peptides ( SPs ) . MACPF domains are also present in other components of complement ( C6 , C7 , and C8β ) . The presence of a MACPF domain within Perforin-2 suggest that it is also a mediator of innate immunity . Unlike C6-C9 and Perforin-1 , Perforin-2 is predicted to be an integral membrane protein because it alone contains a membrane spanning alpha helix ( TM ) followed by a short cytosolic tail . An additional distinguishing feature of Perforin-2 is the P2 domain which is of unknown function but conserved amongst Perforin-2 orthologs . Domain architecture was retrieved from UniProt entries P07357 , P02748 , P14222 , and Q2M385 . TSP1 , thrombospondin type-1 repeat; LDL , low-density lipoprotein receptor class A repeat; EGF , epidermal growth factor-like domain; C2 , calcium-dependent phospholipid binding domain . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 003 The topology of a type I transmembrane protein would orient the MACPF domain of Perforin-2 towards bacteria contained within the lumen of membrane vesicles or towards extracellular bacteria that adhere to the plasma membrane . This topology would also place Perforin-2's 40 amino acid long carboxy-terminal domain in the cytosol . Therefore , we hypothesized that regulatory proteins may interact with the carboxy-terminal domain of Perforin-2 to govern its bactericidal activity . To identify these proteins we used Perforin-2's carboxy-terminal domain as bait for macrophage-expressed gene products in a yeast two-hybrid system and identified UBC12 as a Perforin-2-interacting protein . We then confirmed this interaction by coimmunoprecipitation of UBC12 with Perforin-2-GFP ( Figure 2A ) . The addition of GFP , via a flexible linker , to the carboxy terminus of Perforin-2 was necessary because antibodies that immunoprecipitate native Perforin-2 are not available ( McCormack et al . , 2013a , 2013b ) . However , this also allowed us to use GFP as a negative specificity control and , as expected , UBC12 did not coimmunoprecipitate with GFP ( Figure 2A ) . 10 . 7554/eLife . 06505 . 004Figure 2 . Perforin-2 co-immunoprecipitates with CRL subunits and is ubiquitylated . RAW 264 . 7 cells were transfected with Perforin-2-GFP or GFP expression constructs and stimulated with IFN-γ prior to immunoprecipitation with anti-GFP . ( A ) Western blots probed with the indicated antibodies reveal that the NEDD8 E2 ligase UBC12 , CRL cullin scaffold CUL1 , and substrate receptor F-box protein βTrCP specifically coimmunoprecipitate with Perforin-2 . ( B ) Additional western blots reveal that IFN-γ and LPS , but not IFN-γ alone , stimulate ubiquitylation of Perforin-2 . P2 , Perforin-2; IP , immunoprecipitates; WCL , whole cell lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 004 UBC12 , also known as UBE2M , is an E2 enzyme that conjugates NEDD8 to cullins of multi-component CRLs ( Huang et al . , 2005 , 2009 ) . Therefore , we also probed coimmunoprecipitates for other CRL components that associate with Perforin-2 and found that CUL1 ( cullin-1 ) and βTrCP coimmunoprecipitate with Perforin-2-GFP but not with GFP ( Figure 2A ) . As with other cullins , CUL1 is a CRL scaffold . F-box proteins such as βTrCP1 and -2 provide substrate specificity ( Huang et al . , 2005 , 2009 ) . Because these molecular associations suggest that Perforin-2 is ubiquitylated by a CRL complex , we next sought to determine whether or not Perforin-2 is ubiquitylated . As reported previously , IFN-γ induces the expression of Perforin-2 ( Fields et al . , 2013; McCormack et al . , 2013b ) . However , ubiquitylation of Perforin-2 does not occur unless a PAMP , such as LPS , is also present ( Figure 2B ) . The ubiquitylation of Perforin-2 in response to a PAMP suggest that ubiquitylation may be essential for its bactericidal activity . To examine the biological relevance of CRL subunits that coimmunoprecipitate with Perforin-2 we used siRNA to ablate the expression of UBC12 and CUL1 ( Figure 3A ) . As expected from our coimmunoprecipitations , siRNA knockdown of either UBC12 or CUL1 blocked ubiquitylation of Perforin-2 ( Figure 3B ) . Although βTrCP also coimmunoprecipitated with Perforin-2 , there are two mammalian βTrCP paralogs ( βTrCP1 and -2 ) with overlapping activities . Therefore we did not target βTrCP due to the difficulty of simultaneously silencing both genes . 10 . 7554/eLife . 06505 . 005Figure 3 . Perforin-2 activity is dependent upon its ubiquitylation by a CRL complex . ( A ) siRNA efficiently and specifically knocks down expression of CUL1 , UBC12 , and Perforin-2 in IFN-γ stimulated CMT93 cells . ( B ) Western blots of immunoprecipitates from transfected CMT93 cells probed with the indicated antibodies demonstrate that Perforin-2 is ubiquitylated in cells stimulated with IFN-γ and LPS and that ubiquitylation is dependent upon UBC12 and CUL1 . ( C–E ) In contrast to scramble siRNA , siRNA knockdown of Perforin-2 , CUL1 , or UBC12 in CMT93 cells promotes survival of extracellular E2348/69 , an enteropathogenic Escherichia coli ( EPEC ) strain that does not express cycle inhibiting factor ( Cif ) . Solid and open symbols denote cotransfection with Perforin-2-GFP or GFP expression plasmids , respectively . Statistically significant ( p < 0 . 05 ) differences between ( C ) Perforin-2 siRNA + GFP and the other three conditions , ( D , E ) scramble siRNA + Perforin-2-GFP and the other three conditions by one-way ANOVA with Bonferroni multiple-comparisons post-hoc test; n ≥ 3 . P2 , Perforin-2 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 005 We next transfected CMT93 cells , a murine intestinal epithelial cell line , with a variety of siRNAs and expression plasmids to evaluate their effects in a bactericidal assay . Since the pathogen chosen for these assays , EPEC , is primarily extracellular we also chose to investigate extracellular killing rather than the intracellular killing that has been reported previously ( Fields et al . , 2013; McCormack et al . , 2013b ) . Killing of extracellular bacteria was evident within 2–4 hr after attachment to cells transfected with control scramble siRNA ( Figure 3C–E ) . Killing was reduced or eliminated by siRNA knockdown of Perforin-2 , CUL1 , or UBC12 . Moreover , siRNA knockdown of either CUL1 or UBC12 blocked Perforin-2-dependent killing of EPEC as efficiently as Perforin-2 siRNA ( Figure 3D , E ) . Killing was restored in cells transfected with Perforin-2 siRNA by cotransfection of a siRNA-insensitive Perforin-2-GFP expression plasmid ( Figure 3C ) . However , Perforin-2-GFP could not overcome the killing defect of CUL1 or UBC12 ablated cells ( Figure 3D , E ) . These results demonstrate Perforin-2 cannot kill bacteria in the absence of ubiquitylation . The carboxy terminal tail of Perforin-2 is the most likely site of ubiquitylation based on topological modeling ( Figure 1 ) . Human and mouse Perforin-2 both contain a cluster of four lysine residues; three of which are highly conserved across mammalian species ( Figure 4A ) . As predicted , site directed mutagenesis of the three most highly conserved lysines abolished ubiquitylation of Perforin-2-GFP ( Figure 4B ) . However , the triple mutation did not affect protein expression nor stability ( Figure 4B ) . Mpeg1 knockout MEFs transfected with a plasmid that expresses the triple mutant or GFP were unable to kill intracellular S . typhimurium ( Figure 4C ) . In contrast , knockout MEFs transfected with a Perforin-2-GFP expression plasmid cleared the pathogen as efficiently as wild-type MEFs . In aggregate , these results demonstrate that ubiquitylation of Perforin-2 is essential for Perforin-2-dependent killing of both extra- and intracellular bacteria . Moreover , the ubiquitylation of one or more conserved lysines within the carboxy-terminal tail of Perforin-2 is accomplished by a CRL complex in response to bacterial antigens . 10 . 7554/eLife . 06505 . 006Figure 4 . Conserved lysines within the carboxy-terminal domain of Perforin-2 are required for its ubiquitylation and bactericidal activity . ( A ) Alignment of the predicted cytosolic domains of Perforin-2 ( P2 ) molecules from select mammalian species . Dots denote identity to human Perforin-2 . The three most conserved lysine residues are highlighted in red . Numbering is relative to human Perforin-2 , GenBank accession number AAI12231 . ( B ) Site-directed mutagenesis was used to mutate the three conserved lysines to glutamine residues in murine Perforin-2 . Western blots of Perforin-2-GFP or Perforin-2-KQ-GFP expressed in CMT93 cells stimulated with IFN-γ and LPS demonstrate that both fusion proteins are expressed . However , only Perforin-2-GFP is ubiquitylated . ( C ) Murine embryonic fibroblasts ( MEFs ) isolated from wild-type and Perforin-2 −/− ( KO ) embryos were transfected with GFP , Perforin-2-GFP , or Perforin-2-KQ-GFP expression plasmids . Transfected cells were induced with IFN-γ ca . 24 hr before infection with Salmonella typhimurium . The results demonstrate that Perforin-2-GFP , but not the K-to-Q mutant nor GFP , restored killing in KO MEFs to wild-type levels . The differences between KO MEFs transfected with Perforin-2-GFP and Perforin-2-KQ-GFP or GFP are statistically significant , p < 0 . 05 , at hours 2 through 24 as determined by one-way ANOVA with Bonferroni post-hoc multiple comparisons; n ≥ 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 006 Since we have shown above that Perforin-2 is a CRL substrate , we evaluated the ability of Cif+ and Cif− pathogens to block ubiquitylation of Perforin-2 . As expected , wild-type Y . pseudotuberculosis blocked ubiquitylation of Perforin-2 whereas an isogenic cif::aadA mutant did not ( Figure 5 ) . Transformation of the cif mutant with a Cif expression plasmid restored the pathogen's ability to block ubiquitylation of Perforin-2 . Likewise , an EPEC strain ( E2348/69 ) carrying a naturally defective cif allele was unable to block ubiquitylation of Perforin-2 unless it was transformed with a Cif expression plasmid ( Marches et al . , 2003 ) ( Figure 5 ) . 10 . 7554/eLife . 06505 . 007Figure 5 . Cif blocks ubiquitylation of Perforin-2 . CMT93 cells were transfected with GFP or Perforin-2-GFP expression plasmids and stimulated with IFN-γ . As indicated , some transfected cells were also stimulated with LPS , infected with Yersinia pseudotuberculosis ( Yp ) , or EPEC strain E2348/69 . Unlike Y . pseudotuberculosis or other EPEC strains , E2348/69 harbors a defective cif locus . Species specific expression plasmids were used to complement cif mutants . Immunoprecipitates were then separated by SDS-PAGE and probed with the indicated antibodies . CifYp-C117A-FLAG denotes a point mutation within the enzyme's catalytic triad , conserved amongst Cif proteins , that is essential for deamidation of NEDD8 . The inability of CifYp-C117A-FLAG to block ubiquitylation of Perforin-2 is not due to a lack of expression ( Figure 5—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 00710 . 7554/eLife . 06505 . 008Figure 5—figure supplement 1 . CifYp expression . Western blot demonstrating that both CifYp-FLAG and CifYp-C117A-FLAG are expressed in Y . pseudotuberculosis cif::aadA . Cloning vector pFLAG-CTC was included as a negative control . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 008 Structural and mutagenic studies have shown that the enzymatic activity of Cifs is dependent upon a catalytic triad composed of Cys , His , and Gln residues ( Hsu et al . , 2008; Crow et al . , 2009; Jubelin et al . , 2009 , 2010; Yao et al . , 2009; Cui et al . , 2010 ) . Therefore , to exclude the possibility that the inhibition of Perforin-2 ubiquitylation is independent of NEDD8 deamidation we used site directed mutagenesis to change the codon of Cys117 , which along with His173 and Gln193 forms the catalytic triad to CifYp , to an Ala codon . In contrast to the CifYp-FLAG expression plasmid , the CifYp-C117A-FLAG expression plasmid was unable to complement a cif::aadA mutation in Y . pseudotuberculosis even though the two proteins were expressed equally well ( Figure 5 and Figure 5—figure supplement 1 ) . Thus , these results demonstrate that the deamidase activity of Cif is intrinsic to its ability to block ubiquitylation of Perforin-2 . In addition , these results provide further evidence that Perforin-2 is ubiquitylated by a CRL complex because deamidation of NEDD8 inhibits CRL activity . Having found that Cif blocks ubiquitylation of Perforin-2 and that ubiquitylation is essential to Perforin-2 activity , we next evaluated the biological relevance of these findings using in vitro and in vivo infection models . We found that transformation of a Cif− strain with a CifEc expression plasmid resulted in an EPEC strain that was able to block Perforin-2-dependent killing by stimulated Caco-2 cells ( Figure 6A ) . In addition , Cif blocked killing as effectively as Perforin-2 siRNA ( Figure 6A ) . In contrast , the cif mutant was rapidly killed unless expression of Perforin-2 was ablated with siRNA ( Figure 6B ) . Similar results were obtained with a murine mucosal epithelial cell line infected with wild-type Y . pseudotuberculosis and an isogenic cif::aadA mutant ( Figure 6C , D ) . Transformation of the cif::aadA mutant with a Cifyp-FLAG expression plasmid restored resistance to Perforin-2 while a vector control plasmid did not ( Figure 6E , F ) . In addition , the specificity of Perforin-2 knockdown was confirmed by transfection of siRNA-treated cells with a Perforin-2-RFP expression plasmid lacking the siRNA targeting region . As expected , Cif− bacteria were killed by siRNA ablated cells that express Perforin-2-RFP but not RFP ( Figure 6C–F ) . We also evaluated the impact of a point mutation within the catalytic triad of CifYp . Consistent with the inability of CifYp-C117A-FLAG to block ubiquitylation of Perforin-2 , we found that it was also unable to block Perforin-2-dependent killing ( Figure 7 ) . 10 . 7554/eLife . 06505 . 009Figure 6 . The bacterial effector protein Cif blocks the bactericidal activity of Perforin-2 . Caco-2 cells , a human intestinal epithelial cell line , were transfected with Perforin-2 specific or scramble siRNAs . The cells were subsequently infected with EPEC strain E2348/69 , which carries a naturally disrupted cif gene , transformed with a ( A ) CifEc expression plasmid or ( B ) vector control . Alternatively CMT93 cells , a murine epithelial cell line , were cotransfected with the indicated siRNA and RFP or Perforin-2-RFP expression plasmids . Transfected cells were subsequently infected with ( C ) wild-type Y . pseudotuberculosis , ( D ) a cif mutant , ( E ) the mutant transformed with a CifYp-FLAG expression plasmid , or ( F ) vector control . All mammalian cells were activated with IFN-γ for 24 hr prior to infection . p < 0 . 05 by Student's t-test , n ≥ 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 00910 . 7554/eLife . 06505 . 010Figure 7 . Enzymatically inactive Cif cannot block Perforin-2-dependent killing . CMT93 cells were transfected with the indicated siRNAs and RFP or Perforin-2-RFP expression plasmids . Following stimulation with IFN-γ , transfected cells were infected with a Y . pseudotuberculosis cif::aadA mutant transformed with a CifYp-C117A-FLAG expression plasmid . The C117A point mutation within the enzyme's conserved catalytic triad abolishes its ability to block ubiquitylation of Perforin-2 . Statistically significant ( p < 0 . 05 ) difference between Perforin-2 siRNA + RFP and scramble siRNA + RFP or Perforin-2 siRNA + Perforin-2-RFP by one-way ANOVA with Bonferroni multiple-comparisons post-hoc test; n ≥ 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 010 We next evaluated the role of Cif in vivo by using Y . pseudotuberculosis in a murine infection model ( Figure 8 ) . This human and rodent pathogen colonizes the distal ileum and proximal colon . Subsequent invasion of the underlying lymphatic tissue facilitates Y . pseudotuberculosis dissemination to the spleen and liver followed by death several days after inoculation ( Marra and Isberg , 1997; Mecsas et al . , 2001; Logsdon and Mecsas , 2003 ) . In our studies we found that 80% of C57Bl/6 mice perished 6–10 days after orogastric inoculation with 108 CFU of wild-type Y . pseudotuberculosis . These results are consistent with other studies that used the same inoculation method and similar infectious doses ( Mecsas et al . , 2001; Logsdon and Mecsas , 2003 ) . In sharp contrast to wild-type Y . pseudotuberculosis , all of the animals that were inoculated with the Cif− mutant survived even though they were inoculated at the same infectious dose ( Figure 8 ) . Thus , these results demonstrate that Cif is an important virulence factor of Y . pseudotuberculosis and by extension , EPEC and other Cif+ pathogens . 10 . 7554/eLife . 06505 . 011Figure 8 . Cif enhances pathogenicity in vivo . Survival curves of C57Bl/6 mice inoculated orogastrically with 108 CFU of wild-type Y . pseudotuberculosis or an isogenic cif::aadA mutant . Animals were weighed daily and euthanized if weight loss exceeded 20% . p < 0 . 05 by log-rank ( Mantel–Cox ) test , n = 9–10 mice per group . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 011 To determine if the effects of Cif in vivo are primarily through disruption of Perforin-2 bactericidal activity or another pathway ( s ) , we inoculated C57Bl/6 × 129X1/SvJ Mpeg1 +/+ , +/− and −/− mice with 106 CFU of Cif+ and Cif− Y . pseudotuberculosis . We note that these orogastric inocula are considerably lower than the ∼109 CFU that are typically used because we had previously determined that Perforin-2 deficient mice are hypersensitive to infectious agents ( Mecsas et al . , 2001; Logsdon and Mecsas , 2003 ) . As expected with such a low dose , wild-type mice survived equally well when infected with wild-type or mutant bacteria ( Figure 9A ) . Although no wild-type mice perished with this low infectious dose , some Cif+ bacteria were recovered from the spleens of infected animals ( Figure 10A ) . In contrast , Cif− bacteria did not disseminate to the spleens of wild-type mice . In Mpeg1 heterozygotes , Cif+ bacteria were clearly more virulent than Cif− bacteria and significantly more Cif+ than Cif− bacteria were recovered from the organs and blood of infected animals ( Figures 9B , 10B ) . Statistically insignificant differences were observed between Cif+ and Cif− bacteria in Mpeg1 knockout mice ( Figures 9C , 10C ) . However , a gene dosage effect is evident when organ loads are compared across murine genotypes ( Figure 10 ) . In this comparison there is a consistent trend of Mpeg1 −/− mice having the highest organ loads , wild-type mice the lowest , and intermediate loads in Mpeg1 heterozygotes . 10 . 7554/eLife . 06505 . 012Figure 9 . Bacterial inhibitors of neddylation abolish the bactericidal activity of Perforin-2 in vivo . Survival curves for two different lineages of Mpeg1 +/+ , +/− , and −/− mice following orogastric inoculation with 106 CFU of wild-type Y . pseudotuberculosis or an isogenic cif::aadA mutant . ( A–C ) n = 6–10 mice per group . ( D–F ) n = 22–28 mice per group . p < 0 . 05 by log-rank ( Mantel–Cox ) test . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 01210 . 7554/eLife . 06505 . 013Figure 10 . Cif diminishes the bactericidal activity of Perforin-2 in vivo . Representative organ loads of C57Bl/6 × 129X1/SvJ Mpeg1 ( A ) +/+ , ( B ) +/− and ( C ) −/− mice infected with Cif+ or Cif− Y . pseudotuberculosis . Animals were sacrificed 10 days after orogastric inoculation with 106 CFU . Samples were normalized by organ weight and each experiment was repeated twice . Horizontal lines represent the mean . Statistical analysis was performed by the nonparametric Kolmogorov–Smirnov test with Dunn's multiple comparison test; p < 0 . 05 , n ≥ 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 013 To confirm our observations with C57Bl/6 × 129X1/SvJ mice we conducted similar experiments with 129X1/SvJ Mpeg1 +/+ , +/− , and −/− mice . As with C57Bl/6 × 129X1/SvJ mice , there was not a significant difference between Cif+ and Cif− Y . pseudotuberculosis in wild-type 129X1/SvJ mice; although , a few animals died in both cases ( Figure 9D ) . In contrast , nearly all of the Mpeg1 knockout mice died ( Figure 9F ) . As with C57Bl/6 × 129X1/SvJ mice , there was not a significant difference between Cif+ and Cif− bacteria in 129X1/SvJ knockout mice . However a difference with Mpeg1 heterozygotes was again observed ( Figure 9E ) . In this latter group Cif+ bacteria killed 84% of the heterozygotes while Cif− bacteria killed none . Because our results with 129X1/SvJ mice replicate those obtained with C57Bl/6 × 129X1/SvJ mice , we conclude that Perforin-2 deficiency is the primary cause of increased sensitivity to bacterial pathogens . Our results also demonstrate that Cif enhances pathogenicity primarily through the disruption of Perforin-2 activity and not through other pathways , such as cell cycling , that are known to be disrupted by Cif . The molecular mechanism of this effect is undoubtedly the blockade of NEDD8-dependent ubiquitylation of Perforin-2 by Cif , as suggested by our in vitro studies . Having determined that ubiquitylation is necessary for Perforin-2 activity , we next sought to determine the number of and linkages between ubiquitin monomers as these properties determine the fate of ubiquitylated proteins . For example , polyubiquitylation through K48 linkages target proteins for degradation by the proteasome while K63 polyubiquitylation is often involved in cell signaling . Alternatively , ubiquitylation may terminate after the addition of a single ubiquitin molecule to the target protein . This is termed monoubiquitylation and typically serves as a sorting signal that directs the modified protein to one or more subcellular compartments . To determine the type of Perforin-2 ubiquitylation we probed Perforin-2-GFP immunoprecipitates with a linkage independent ubiquitin antibody ( P4D1 ) as well as antibodies specific for K48 and K63 polyubiquitylation ( Figure 11 ) . Consistent with previous results , the P4D1 antibody recognized Perforin-2-GFP immunoprecipitated from transfected knockout MEFs that were stimulated with IFN-γ and LPS . Ubiquitylation was not detected in the absence of LPS . In addition , MLN4924 , a small molecule that inhibits cullin neddylation , also blocked LPS stimulated ubiquitylation of Perforin-2-GFP ( Figure 11 ) ( Soucy et al . , 2009 ) . This latter result provides additional confirmation that Perforin-2 is ubiquitylated by a CRL complex . We estimate that ubiquitylation increases the mass of Perforin-2-GFP by ≤10 kDa ( Figure 11—figure supplement 1 ) . Given that ubiquitin has a mass of 8 . 5 kDa , this suggest that Perforin-2-GFP is monoubiquitylated . Since Perforin-2-GFP is a 105 kDa molecule , this relatively small increase in its mass also explains why mass shifts are not apparent on gels of less resolution; for example Figure 2B . Consistent with monoubiquitylation , antibodies specific for K48 or K63 polyubiquitylation did not recognize ubiquitylated Perforin-2-GFP ( data not shown ) . Although other forms of polyubiquitylation through K6 , K11 , K27 , K29 , K33 , or M1 linkages—in addition to chains of mixed linkages—are also possible , polyubiquitylation is inconsistent with the small increase in the mass of Perforin-2-GFP . This suggests monoubiquitylation serves as a sorting signal that directs Perforin-2 to a specific compartment within the cell or location on the plasma membrane . 10 . 7554/eLife . 06505 . 014Figure 11 . Perforin-2 is monoubiquitylated . Mpeg1 knockout MEFs were transfected with GFP or Perforin-2-GFP expression plasmids and stimulated with IFN-γ in the presence or absence of LPS . Some cultures were also treated with MLN4924 , a small molecule inhibitor of neddylation . The resulting GFP immunoprecipitates were probed with the indicated antibodies in Western blots . Consistent with previous results ubiquitylation of Perforin-2 was dependent upon LPS and neddylation when the immunoprecipitates were probed with a non-linkage specific ubiquitin antibody . When the same immunoprecipitates were probed with antibodies specific for K48 or K63 linkages , ubiquitylation was not detected . From additional Western blots we estimated that ubiquitylation increases the mass of Perforin-2-GFP by ∼10 kDa ( Figure 11—figure supplement 1 ) . These results indicate the Perforin-2 is a monoubiquitylated protein . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 01410 . 7554/eLife . 06505 . 015Figure 11—figure supplement 1 . Monoubiquitylation of Perforin-2 . CMT93 cells were transfected with expression plasmids and stimulated with IFN-γ and LPS as indicated . GFP immunoprecipates were then separated by SDS-PAGE and probed with a Perforin-2 antibody . The blot was then stripped and probed for ubiquitin . Ubiquitin increases the apparent mass of Perforin-2-GFP by ∼10 kDa which , given some uncertainty in mass estimation , is consistent the addition of a ubiquitin monomer ( 8 . 5 kDa ) ; that is , monoubiquitylation . Qualitative differences between the two images is the result of different imaging methods: The left image was collected by exposure to film while the right image was collected on a digital workstation . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 015 To determine if ubiquitylated Perforin-2 has a cellular distribution different than non-ubiquitylated Perforin-2 we transfected Mpeg1 knockout MEFs with Perforin-2-RFP and Perforin-2-KQ-RFP expression plasmids . Due to site directed mutagenesis of the three most highly conserved lysines within the carboxy-terminal tail of Perforin-2 the latter fusion protein cannot be ubiquitylated ( Figure 4 ) . Perforin-2-RFP was observed in distinct punctate bodies following stimulation of the transfected cells with IFN-γ and LPS ( Figure 12A ) . Three dimensional projections of the acquired confocal Z-stacks revealed the vesicular structure of the punctate bodies ( Video 1 ) . This is in sharp contrast to Perforin-2-KQ-RFP which had a diffuse , perinuclear distribution in LPS stimulated cells ( Figure 12B and Video 2 ) . Since we have already shown that LPS causes ubiquitylation of Perforin-2 and K-to-Q mutations in its cytosolic tail abolish ubiquitylation , these results suggest the subcellular distribution of Perforin-2 is determined by monoubiquitylation . 10 . 7554/eLife . 06505 . 016Figure 12 . Ubiquitylation determines the subcellular distribution of Perforin-2 . Mpeg1 knockout MEFs were transfected with ( A ) Perforin-2-RFP or ( B ) Perforin-2-KQ-RFP expression plasmids , the latter of which carries K-to-Q mutations of conserved lysines in the cytosolic tail of Perforin-2 that abolish ubiquitylation . Transfected cells were stimulated with IFN-γ for 24 hr prior to addition of LPS . Cells were fixed within 15 min of LPS addition and counter stained with DAPI . Images were acquired on a Leica confocal microscope with a 63× objective . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 01610 . 7554/eLife . 06505 . 017Video 1 . Perforin-2-RFP is located in vesicular bodies following stimulation with LPS . Mpeg1 knockout MEFs were transfected with a Perforin-2-RFP expression plasmid and stimulated overnight with IFN-γ . After stimulation with LPS for 15 min , cells were fixed and counter stained with DAPI . Three dimensional projections of the acquired confocal Z-stacks reveal that Perforin-2 is located in punctate bodies with vesicular structure in LPS stimulated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 01710 . 7554/eLife . 06505 . 018Video 2 . Redistribution of Perforin-2 is dependent upon ubiquitylation . Mpeg1 knockout MEFs were transfected with a Perforin-2-KQ-RFP expression plasmid . This fusion protein cannot be ubiquitylated because it carries three K-to-Q mutations in the carboxy-terminal tail of Perforin-2 . Transfected cells were stimulated overnight with IFN-γ then LPS . Cells were fixed 15 min after exposure to LPS and counter stained with DAPI . Three dimensional projections of the acquired confocal Z-stacks reveal that Perforin-2-KQ-RFP has a diffuse , perinuclear distribution in LPS stimulated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 018 To further test the hypothesis that ubiquitin serves as a sorting signal for Perforin-2 in response to PAMPs we infected the transfected MEFs with Cif+ and Cif− Y . pseudotuberculosis ( Figure 13 ) . When the MEFs were infected with wild-type bacteria that block ubiquitylation of Perforin-2 , the distribution of Perforin-2-RFP was diffuse and perinuclear ( Figure 13A ) . In contrast , Perforin-2-RFP was located in punctate bodies when the cells were infected with Cif− bacteria that are incapable of blocking ubiquitylation of Perforin-2 ( Figure 13B ) . To further test our hypothesis we also infected MEFs expressing Perforin-2-KQ-RFP . As expected , Perforin-2-KQ-RFP remained diffuse and perinuclear when the cells were infected with either Cif− or Cif+ bacteria ( Figure 13C , D ) . These results were not restricted to knockout MEFs because similar results were obtained with transfected CMT93 cells ( Figure 13—figure supplements 1 , 2 ) . Thus we conclude that CRL-dependent monoubiquitylation of Perforin-2 initiates a journey that ultimately terminates with the insertion and polymerization of Perforin-2 in a bacterial membrane . 10 . 7554/eLife . 06505 . 019Figure 13 . Cif blocks trafficking of Perforin-2 . Mpeg1 knockout MEFs were transfected with Perforin-2-RFP or Perforin-2-KQ-RFP expression plasmids and stimulated with IFN-γ for 24 hr prior to infection with ( A , B ) wild-type Y . pseudotuberculosis or ( C , D ) an isogenic cif::aadA mutant . Cells were fixed 15 min after infection and counter stained with DAPI . Similar results were obtained when transfected CMT93 cells were infected with Cif+ and Cif− bacteria expressing GFP ( Figure 13—figure supplements 1 , 2 ) . Images were acquired on a Leica confocal microscope with a 63× objective . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 01910 . 7554/eLife . 06505 . 020Figure 13—figure supplement 1 . Perforin-2 relocalizes from perinuclear to punctate bodies upon ubiquitylation . CMT93 cells , a murine rectal epithelial cell line , were transfected with murine Perforin-2-RFP or Perforin-2-KQ-RFP expression plasmids and induced with IFN-γ overnight . Cells were then infected with ( A , B ) wild-type Y . pseudotuberculosis or ( C , D ) an isogenic mutant that does not express Cif . Cells were fixed within 15 min of infection . Images were acquired on a Leica confocal microscope with a 63× objective . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 02010 . 7554/eLife . 06505 . 021Figure 13—figure supplement 2 . Perforin-2 perinuclear localization in noninfected CMT93s . CMT93 cells were transfected with ( A ) murine Perforin-2-RFP or ( B ) murine Perforin-2-KQ-RFP and induced overnight with IFN-γ . Images were taken on a Leica confocal microscope with a 63× objective . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 021 In this study we have shown that Perforin-2 is monoubiquitylated by a CRL complex in response to extracellular bacteria or PAMPs and that ubiquitylation is required for Perforin-2's bactericidal activity against both extracellular , cell adherent , and intracellular bacteria . Although we initially identified an interaction between UBC12 and Peforin-2 in a yeast two-hybrid screen , this result is rather curious as this NEDD8 conjugating enzyme is not thought to directly interact with CRL substrates . Thus , it is not clear if our initial two-hybrid hit reflects a bona fide or fortuitous interaction . Nevertheless , we subsequently demonstrated that Perforin-2 coimmunoprecipitates CUL1 , UBC12 , and βTrCP from lysates of mammalian cells . As an F-box substrate receptor , βTrCP may provide the primary and specific interaction between Perforin-2 and the CRL ( Figure 14 ) . It is also likely that phosphorylation of Perforin-2 precedes its ubiquitylation because βTrCP has been shown to recognize phosphorylated substrates ( Winston et al . , 1999; Strack et al . , 2000; Wu et al . , 2003 ) . However we have not yet determined whether or not this is the case for Perforin-2 . 10 . 7554/eLife . 06505 . 022Figure 14 . Monoubiquitylation of Perforin-2 and inhibition by Cif . ( A ) CRLs are modular complexes assembled upon cullin scaffolds such as CUL1 . F-box proteins such as βTrCP , which we have shown interacts with Perforin-2 , provide substrate specificity within the multisubunit complex . Because βTrCP is known to recognize phosphorylated substrates , it is likely that TLR signaling activates one or more kinases that phosphorylate Perforin-2 prior to its association with the CRL . SKP1 serves as an adaptor between βTrCP and the cullin scaffold . An ubiquitin E2 ligase and a RING protein—either RBX1 or 2—dock at the opposite end of the elongated cullin . The E1 enzyme NEDD8 activating enzyme ( NAE ) activates NEDD8 . The E2 ligase UBC12 then covalently binds NEDD8 to CUL1 . ( B ) Neddylation of CUL1 activates the CRL complex and is essential for subsequent monoubiquitylation of Perforin-2 . ( C ) The deamidation of NEDD8 by the bacterial effector protein Cif inhibits ubiquitylation of Perforin-2 and its subsequent trafficking to and destruction of extracellular bacteria at the plasma membrane or intracellular bacteria within an endosomal compartment . Ub , ubiquitin; N8 , NEDD8; N8-Q40E , deamidated NEDD8 . DOI: http://dx . doi . org/10 . 7554/eLife . 06505 . 022 We have demonstrated that siRNA knockdown of either CUL1 or UBC12 blocked ubiquitylation of Perforin-2 and Perforin-2-dependent killing of extracellular bacteria . These latter results confirm the biological relevance of the coimmunoprecipitates . We have also shown that MLN4924 , a small molecule inhibitor of the NEDD8 activating enzyme also blocks ubiquitylation of Perforin-2 ( Soucy et al . , 2009 ) ( Figure 11 ) . Thus , we conclude that PAMPs trigger CRL-dependent monoubiquitylation of Perforin-2's cytosolic domain . As described below , this conclusion is further supported by our studies with the bacterial effector protein Cif . Mutagenesis of conserved lysines within Perforin-2's carboxy-terminus also prevented its ubiquitylation and abolished its bactericidal activity . We have also shown that these mutations prevent the cellular redistribution of Perforin-2 in response to LPS . Likewise , we have shown that Cif blocks Perforin-2 trafficking . Because the destruction of bacteria involves a direct interaction between Perforin-2 and a bacterium , the inability of a cell to route Perforin-2 abolishes its bactericidal activity ( McCormack et al . , 2015 ) . Perforin-2 was originally identified as a potential marker of mature macrophages ( Spilsbury et al . , 1995 ) . However its expression is not limited to professional phagocytes . In human and murine cells , including non-immune cells , the expression of Perforin-2 can be induced by interferons or intracellular bacteria ( Fields et al . , 2013; McCormack et al . , 2013b ) . The gene encoding Perforin-2 is evolutionarily ancient and is present in all but one species of bony vertebrates ( D'Angelo et al . , 2012 ) . As with mammalian Perforin-2 , infectious bacteria also induce the expression of a Perforin-2 ortholog ( MPEG1 . 2 ) in zebrafish . Although ubiquitylation was not investigated , upregulation of MPEG1 . 2 has been shown to require NFκB and the TLR adaptor MyD88 ( Benard et al . , 2014 ) . In non-vertebrate species the expression of Perforin-2 orthologs is constitutive in some tissues and/or inducible by pathogenic bacteria or LPS ( He et al . , 2011; Kemp and Coyne , 2011; Bathige et al . , 2014 ) . In addition to similar patterns of expression , the function of Perforin-2 is also conserved across species . For example , Perforin-2 has been shown to limit the burden of Mycobacterium marinum and S . typhimurium in zebrafish embryos ( Benard et al . , 2014 ) . Additional studies indicate that Perforin-2 orthologs provide similar bactericidal activity in invertebrates ( He et al . , 2011; Kemp and Coyne , 2011; Bathige et al . , 2014 ) . Because Perforin-2 has broad bactericidal activity , is expressed in a variety of cell types and tissues , and is evolutionarily ancient , it is likely that evolution has endowed bacterial pathogens with novel strategies to defeat Perforin-2 . In this study we have identified the bacterial effector protein Cif as one such example ( Figure 14C ) . We have shown that Cif blocks ubiquitylation and trafficking of Perforin-2 . This blockade prevents Perforin-2 from reaching the bacterium and explains how Cif inhibits Perforin-2-dependent killing in vitro and enhances pathogenicity in vivo . These properties are abolished by a mutation in the catalytic site of CifYp . An equivalent mutation has been previously shown to abolish the deamidase activity of CifEc ( Marches et al . , 2003; Hsu et al . , 2008; Jubelin et al . , 2009 ) . Because we have also demonstrated that the bactericidal activity of Perforin-2 is dependent upon its ubiquitylation by a CRL complex , we can conclude that Cif abolishes the bactericidal activity of Perforin-2 through deamidation of NEDD8 ( Figure 14C ) . Moreover , the absence of Perforin-2 in Mpeg1 knockout mice negated the benefit of Cif in vivo . Similar results were observed by siRNA knockdown of Perforin-2 expression in vitro . These results indicate that the inactivation of Perforin-2 is not a secondary or ancillary function of Cif . Rather , our results demonstrate that Cif enhances virulence primarily through inactivation of Perforin-2 as opposed to other mechanisms such as reduced turn-over of epithelial cells . Although Cif is the first definitive example of a bacterial product that blocks the bactericidal activity of Perforin-2 , it is likely that others remain to be discovered . Our observation that Cif blocks monoubiquitylation of Perforin-2 and its trafficking within the cell is consistent with other transmembrane proteins for which monoubiquitylation is known to serve as a sorting signal directing their subcellular localization ( Hicke and Dunn , 2003; MacGurn et al . , 2012 ) . In our accompanying manuscript we provide evidence that suggest Perforin-2 polymerizes and forms pores on bacteria adherent to the plasma membrane or within the endosomal lumen ( McCormack et al . , 2015 ) . Thus , we propose a model in which monoubiquitylation of Perforin-2 initiates its trafficking towards bacteria at either location to facilitate Perforin-2-dependent destruction of bacterial pathogens . We also observed that the MACPF and P2 domains of Perforin-2 are cleaved from its transmembrane and cytosolic domains ( McCormack et al . , 2015 ) . Although it is not yet known when cleavage occurs within the Perforin-2-dependent killing pathway , cleavage may explain the lack of colocalization of Perforin-2-RFP and Cif− bacteria in confocal microscopy ( Figure 13 ) . Although bacterial pathogens such as Y . pseudotuberculosis and EPEC block the critical event—ubiquitylation—that initiates the trafficking of Perforin-2 to its final destination , other pathogens may target events further upstream or downstream of ubiquitylation in order to preserve their cellular integrity and flourish . RAW 264 . 7 ( TIB-71 ) , CMT93 ( CCL-223 ) , Caco-2 ( HTB-37 ) cell lines were obtained from American Type Culture Collection , Manassas , VA . MEFs and murine PMN were isolated as previously described ( Luo and Dorf , 2001; Scheuner et al . , 2001 ) . All cells were cultured at 37°C in a humidified atmosphere containing 5% CO2 following ATCC recommendations for culture conditions . S . typhimurium SL1344 ( gift from Dr J Galán , Yale University ) and EPEC strain E2348/69 were cultured in Luria–Bertani broth ( LB ) at 37°C . Y . pseudotuberculosis YPIII pIB102 ( Bolin and Wolf-Watz , 1984 ) was cultured in HIB at 27°C , or HIB plus 2 . 5 mM CaCl2 at 37°C when subculturing . Human and murine IFN-ɣ was purchased from Peprotech ( Rocky Hill , NJ , United States ) , LPS purchased from Invivogen ( San Diego , CA , United States ) , and MLN4924 was purchased from Active Biochem ( Maplewood , NJ , United States ) and EMD Millipore ( Billerica , MA , United States ) . Because EPEC is unable to translocate Cif of Y . pseudotuberculosis ( Jubelin et al . , 2009 ) , cif mutants were complemented with alleles cloned from their respective species . Accordingly , EPEC strain E2348/69 , a spontaneous Cif− mutant , was transformed with the CifEc expression plasmid pCifwt or vector control pBRSK ( gift from Drs E Oswald and F Taieb , University of Toulouse , France ) ( Marches et al . , 2003; Oswald et al . , 2005 ) . Plasmid pCifYP-FLAG expresses CifYp-FLAG . It was constructed by PCR amplification of cifYp from YPIII with primers 1118 ( ATGAAGCTTAGCCCTAATACCATTAGTCC ) and 1119 ( TCTGGTACCATTACAGTGAGTTTTAATG ) . Underlined sequences indicate primer-template mismatches . The PCR product was then digested with HindIII and KpnI then ligated into the same sites of pFLAG-CTC ( Sigma-Aldrich , St . Louis , MO , United States ) . Oligonucleotide directed mutagenesis of pCifYP-FLAG with primers 1136 ( GCGGGTGTCACGGCAAATACC ) and 1137 ( AACGGGTTCTATTATGCGC ) was used to construct pCifYP-CA-FLAG which expresses CifYp-C109A-FLAG . Strain GPM1769 ( cif::aadA ) was constructed by lambdaRED-mediated recombination in a multistep process . First , aadA was amplified from pHKLac1 with primers 238 ( CCTGGCAGTTTATGGCGGGCGT ) and 1131 ( GCGCATGCTGATCTTCAGATCCTC ) . The PCR product was then digested with ClaI and SphI then ligated into the BstBI and SphI sites within cifYp of pCifYP-FLAG . The cassette and cifYp flanking sequences were then amplified with primers 1133 ( GCCACCCTAAGTGCACG ) and 1134 ( GAGATGATCTGCACGCAG ) . Finally , the PCR product was recombined into YPIII/pSIM6 as previously described ( Datta et al . , 2006 ) . Recombinants were isolated by selection for simultaneous resistance to spectinomycin and streptomycin , and confirmed by PCR with primers 1131 ( GCGCATGCTGATCTTCAGATCCTC ) and 1132 ( GACGGATCCATGATAGCCTGAGCCAG ) . For the generation of Mpeg1 knockout mice the targeting vector was linearized and electroporated into RW-4 ES cells originating from the 129X1/SvJ strain , followed by selection in G418 . Targeted clones were screened by PCR . From 90 clones , 2 positive clones were selected that had undergone homologous recombination and were identified through Southern blot analysis . One ES clone was utilized for the generation of chimeric mice by injection using C57Bl/6J blastocysts as the host . The resulting female chimeras were further mated with C57Bl/6J male mice for germ line transmission . The heterozygous mice ( F1 mice ) were interbred to obtain wild-type , heterozygous , and homozygous littermates ( F2 ) . C57Bl/6 × 129X1/SvJ animals utilized in these experiments were backcrossed between 7–10 times for these experiments . Mouse genotype was determined by PCR using PCR probes MP10 and MP11 . To generate 129 pure animals without potential passenger mutations , chimeric mice were mated with 129X1/SvJ animals , and assessed for germ line transmission . The heterozygous mice were then inbred to obtain a genetically pure 129X1/SvJ strain . Mouse genotype was determined by PCR utilizing PCR probes MP10 and MP11 . Animals were bred at the University of Miami , Miller School of Medicine Transgenic Core Facility . Mice were allowed to freely access food and water and were housed at an ambient temperature of 23°C on a 12 hr light/dark cycle under specific pathogen-free condition . Animal care and handling were performed as per IACUC guidelines . All animal experiments were performed in accordance with the University of Miami Animal Care and Use Committee guidelines . The animal genotype was blinded prior to the experiment to limit bias . C57Bl/6 mice were infected orogastrically with 106 colony-forming units of Y . pseudotuberculosis as previously described ( Schweer et al . , 2013 ) . Mice were weighed daily throughout the experiment; the animals were euthanized after greater than 20% weight loss . Animal bacterial colonization was confirmed by collection of feces 12 hr after orogastric inoculation . Feces were homogenized in ddH2O , diluted , and plated on MacConkey agar plates ( Kanamycin 100 μg/ml ) . For CFU enumeration , mice were sacrificed 10 days after Y . pseudotuberculosis orogastric infection . At each time point , cardiac puncture was performed and intestines , liver , and spleen were harvested , weighed , and homogenized using a potter homogenizer in ddH2O with 0 . 05% Triton X-100 . The homogenates were diluted and plated on MacConkey agar plates ( Kanamycin 100 μg/ml ) . All organ samples were normalized based on weight . The following antibodies were used for Western blots: anti-murine Mpeg1 ( ab25146 ) , anti-human Mpeg1 ( ab176974 ) , anti-Cullin-1 ( Ab75817 ) , anti-GFP ( ab290 ) , and anti-β-TRCP ( Ab137674 ) ( Abcam , Cambridge , MA ) ; anti-GFP ( sc9996 ) ( Santa Cruz Biotechnology , Dallas , TX ) ; anti-β-actin ( 2F1-1 ) ; anti-Ubiquitin ( P4D1 ) ( BioLegend , San Diego , CA , United States ) ; anti-Ubc12 ( D13D7 ) ; anti-K63-linkage specific polyubiquitin ( D7A11 ) ; anti-K48-linkage specific polyubiquitin ( D9D5 ) ( Cell Signaling Technology , Danvers , MA , United States ) and anti-FLAG ( F-7425 ) ( Sigma-Aldrich ) . Densitometry analysis was performed where indicated utilizing ImageJ software . Co-IP and ubiquitylation assays were modified from ( Shembade et al . , 2010 ) utilizing Dynabeads Co-immunoprecipitation kit ( Thermo Fisher Scientific , Waltham , MA , United States ) . Whole-cell lysates were subjected to SDS-PAGE , transferred to nitrocellulose membranes , blocked in 5% milk , incubated with specific primary and secondary antibodies , then detected with SuperSignal West Pico Chemiluminescent Substrate ( Thermo Fisher Scientific ) by either X-ray film or Odyssey FC Imaging System ( LI-COR , Lincoln , NE , United States ) . The intracellular gentamicin protection assays were conducted as previously described for S . typhimurium experiments ( Lutwyche et al . , 1998; Laroux et al . , 2005; Law et al . , 2010 ) . Briefly , 100 ng/ml of species-specific human or murine IFN-γ was added 14 hr before infection where necessary to induce uniform Perforin-2 expression . S . typhimurium was added as indicated . After 30 min to allow for uptake/invasion , the culture media was removed and replaced with fresh medium supplemented with gentamicin . For gentamicin protection assays , the multiplicity of infection was between 20–50 bacteria per cell to allow for sufficient uptake of bacteria . For extracellular bacterial killing assays , mammalian cells were seeded so that >90% confluence was achieved at the time of infection . The expression of Perforin-2 was induced with species specific IFN-γ at 100 ng/ml for 14 hr prior to infection . Y . pseudotuberculosis or EPEC were added at an MOI of 5–10 . Bacteria were allowed to attach to mammalian cell membranes for 40 min . After the attachment phase nonadherent extracellular bacteria were removed by aspiration of culture medium followed by three to five washes with PBS . Fresh culture medium was added after the final wash and cultures were incubated in 5% CO2 at 37C . At selected time points the culture medium was aspirated and discarded . Adherent bacteria were recovered in ddH2O or PBS containing 0 . 1% ( vol/vol ) Triton X-100 . Bacteria were serially diluted in PBS and enumerated on MacConkey or LB agar plates . Although both extracellular and intracellular bacteria are recovered by this method , we determined that the load of intracellular bacteria is numerically insignificant compared to the far higher number of adherent extracellular bacteria . Perforin-2 −/− MEFs or CMT93 cells were nucleofected with either murine Perforin-2 RFP or Perforin-2-KQ-RFP , plated onto coverslips and stimulated overnight with IFN-γ . Cells were washed once , and stimulated with LPS or infected with wild-type or Cif− Y . pseudotuberculosis as described above . Infection or LPS stimulation was allowed to proceed for 15 min upon which cells were fixed with 3% paraformaldehyde for 15 min at room temperature and counter stained with DAPI . Images were taken on a Leica SP5 inverted confocal microscope with a motorized stage and 63× objective . Images were analyzed using Leica application suite advanced fluorescence software and ImageJ . Videos were constructed in ImageJ with 3D Projection of the confocal stacks with Y-axis rotation in the video . For murine cells , RNA interference and transfection were conducted as previously described ( Law et al . , 2010 ) . For human cells , the aforementioned murine system was modified through utilizing three human Perforin-2-specific silencer select siRNAs purchased from Ambion ( Waltham , MA , United States ) Silencer Select #s61053 , s47810 , s61054 . For Ubc12 three murine Ubc12-specific silencer select siRNAs were purchased from Ambion; Silencer Select #s75658 , s75659 , s75660 . For murine Cullin-1 knockdown , three MISSION predesigned siRNAs were selected SASI_Mm01_00112833 , SASI_Mm01_00112832 , and SASI_Mm02_00322009 was utilized . Silencer select negative control #1 and 2 from Ambion were utilized as negative controls . Student's t-test , multiple t-test with Holm-Sidak multiple comparisons correction , one-way ANOVA with Bonferroni multiple comparisons test , or Kruskal–Wallis non-parametric test with Dunn's multiple comparison test was used for comparisons ( GraphPad Prism Version 6 . 0b and SPSS 21 . 0 were utilized for statistical analysis ) .	A wide range of bacteria and other microbes can infect animals and cause disease . Throughout evolution , these microbes and their hosts have been fighting never ending arms races in which the microbes deploy ever more elaborate weapons , while the hosts adapt to defend themselves . An animal's first line of defense is provided by its ‘innate’ immune system . This system is activated by the general features of microbial cells; for example , the molecules that make up the walls surrounding most bacteria . Microbes must defeat the innate immune system in order to cause disease , and ultimately to spread from one host to the next . One component of innate immunity is a protein called Perforin-2 that is present in most , if not all , animal cells . This protein forms pores on bacterial cells , causing them to split open and die . However , it was not clear how Perforin-2 is switched on and what , if anything , bacteria do to counteract it . To address these questions , McCormack et al . infected human and mice cells with bacteria that cause serious diseases of the digestive tract . The experiments show that when animal cells detect bacteria , or merely a fragment of their cell wall , a specific group of proteins , called the CRL complex , attaches a molecule called ubiquitin to Perforin-2 . Ubiquitin works much like the shipping label of a package , enabling the efficient targeting of Perforin-2 to the invading bacteria . McCormack et al . also show that some bacteria use a protein called a cell cycle inhibiting factor ( or Cif for short ) to inhibit the CRL complex . This blocks the ubiquitin labeling of Perforin-2 , which renders it a useless weapon that can no longer be directed towards bacteria . Mice that are infected with a bacterium called Yersinia pseudotuberculosis become seriously unwell and often die . However , McCormack et al . found that mice infected with mutant Y . pseudotuberculosis that lacked Cif remained healthy . Also , mice that lacked Perforin-2 are highly susceptible to infectious diseases . McCormack et al . 's findings reveal how Perforin-2 is activated during the innate immune response and how some bacteria can defeat this pivotal defense . In the current age of antibiotic resistant bacteria , these studies may spur the development of new drugs that restore or increase the activity of Perforin-2 .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "immunology", "and", "inflammation" ]	2015	Enteric pathogens deploy cell cycle inhibiting factors to block the bactericidal activity of Perforin-2
Meiotic drivers are parasitic loci that force their own transmission into greater than half of the offspring of a heterozygote . Many drivers have been identified , but their molecular mechanisms are largely unknown . The wtf4 gene is a meiotic driver in Schizosaccharomyces pombe that uses a poison-antidote mechanism to selectively kill meiotic products ( spores ) that do not inherit wtf4 . Here , we show that the Wtf4 proteins can function outside of gametogenesis and in a distantly related species , Saccharomyces cerevisiae . The Wtf4poison protein forms dispersed , toxic aggregates . The Wtf4antidote can co-assemble with the Wtf4poison and promote its trafficking to vacuoles . We show that neutralization of the Wtf4poison requires both co-assembly with the Wtf4antidote and aggregate trafficking , as mutations that disrupt either of these processes result in cell death in the presence of the Wtf4 proteins . This work reveals that wtf parasites can exploit protein aggregate management pathways to selectively destroy spores . Meiotic drivers are selfish DNA sequences that break the traditional rules of sexual reproduction . Whereas most alleles have a 50% chance of being transmitted into a given offspring , meiotic drivers can manipulate gametogenesis to bias their own transmission into most or even all of an individual’s offspring ( Burt and Trivers , 2006; Lindholm et al . , 2016 ) . This makes meiotic drive a powerful evolutionary force ( Sandler and Novitski , 1957 ) . Meiotic drivers are widespread in eukaryotes and the evolutionary pressures they exert are thought to shape major facets of gametogenesis , including recombination landscapes and chromosome structure ( Bravo Núñez et al . , 2020b; Bravo Núñez et al . , 2020a; Crow , 1991; Dyer et al . , 2007; Larracuente and Presgraves , 2012; Schimenti , 2000; Pardo-Manuel de Villena and Sapienza , 2001; Hammer et al . , 1989; Zanders et al . , 2014; Grey et al . , 2018 ) . Harnessing and wielding the evolutionary power of meiotic drive has the potential to greatly benefit humanity . Engineered drive systems , known as ‘gene drives , ’ are being developed to spread genetic traits in populations ( Lindholm et al . , 2016; Burt , 2014; Gantz et al . , 2015; Esvelt et al . , 2014; Burt and Crisanti , 2018 ) . For example , gene drives could be used to spread disease-resistance alleles in crops . Alternatively , gene drives can be used to suppress human disease vectors , such as mosquitoes , or to limit their ability to transmit diseases ( Burt , 2014; Burt and Crisanti , 2018; Esvelt et al . , 2014; Gantz et al . , 2015; Lindholm et al . , 2016 ) . While there are many challenges involved in designing effective gene drives , natural meiotic drivers could serve as useful models or components for these systems ( Burt , 2014; Lindholm et al . , 2016 ) . However , the molecular mechanisms employed by most meiotic drivers are unknown . The recently characterized wtf gene family of Schizosaccharomyces pombe includes several meiotic drivers ( Bravo Núñez et al . , 2018a; Eickbush et al . , 2019; Hu et al . , 2017; López Hernández and Zanders , 2018; Nuckolls et al . , 2017 ) . The wtf coding sequences are small ( ~1 kb ) and encode autonomous drivers that specifically kill meiotic products ( spores ) that do not inherit the wtf + allele from wtf+/wtf - heterozygotes . These drivers carry out targeted spore destruction using two proteins: a poison ( Wtfpoison ) to which all spores are exposed , and an antidote ( Wtfantidote ) which rescues only the spores that inherit the wtf + allele ( Figure 1A and B ) . The two proteins of a given driver are encoded on largely overlapping coding sequences , but the antidote contains ~45 additional N-terminal amino acids ( Figure 1A ) . The small size and autonomy of the wtf drivers make them promising candidates for use in gene drive systems . It is important , however , to first understand more about the molecular mechanisms of the Wtf proteins and whether they are likely to be functional in other species . Here , we investigate the mechanisms of wtf drive using the wtf4 allele as a model . We demonstrate that the Wtf4 proteins are functional outside of gametogenesis and in the budding yeast Saccharomyces cerevisiae , despite over 350 million years since the two yeasts shared a common ancestor ( Hoffman et al . , 2015 ) . We also show that the two Wtf4 proteins assemble into distinct aggregated forms . Wtf4poison forms toxic aggregates that are dispersed throughout the cytoplasm . The Wtf4antidote forms aggregates that are recruited to the vacuole and vacuole-associated inclusions and are largely non-toxic . When the two Wtf4 proteins are expressed together , the Wtf4antidote and Wtf4poison co-assemble and are trafficked to the vacuole . This work adds to our understanding of how wtf meiotic drivers work . In addition , the conserved function of Wtf4poison’s toxicity and the fact that the Wtf4antidote exploits conserved aggregate management processes suggests that wtf genes represent good candidates for gene drive systems . The wtf4 meiotic driver used in this work is from S . kambucha , an isolate that is almost identical ( 99 . 5% DNA sequence identity ) to the commonly studied lab isolate of S . pombe ( Rhind et al . , 2011; Singh and Klar , 2002 ) . Our previous work demonstrated that the Wtf4antidote localizes to a region within the spores that inherit the wtf4 gene . The Wtf4poison protein , however , is found in all four spores and throughout the sac ( ascus ) that holds them ( Nuckolls et al . , 2017 ) . Here , we explored the localization of these proteins in greater depth to gain insight into their mechanisms . We used fluorescently tagged alleles of wtf4 to visualize the proteins . The two Wtf4 proteins have different translational start sites and thus different N-termini ( Figure 1A , Figure 1—figure supplement 1A ) . We took advantage of this feature to visualize the proteins separately . For the Wtf4antidote , we used an allele with an mCherry tag immediately upstream of the first start codon . This mCherry-wtf4 allele tags only the Wtf4antidote ( mCherry-Wtf4antidote ) but still encodes an untagged Wtf4poison . We previously demonstrated that this allele is fully functional ( Nuckolls et al . , 2017 ) . To visualize Wtf4poison , we used the wtf4poison-GFP allele . This separation-of-function allele encodes only a C-terminally tagged poison but no Wtf4antidote protein . We previously demonstrated that this tagged allele is functional but has a slightly weaker phenotype than an untagged wtf4poison separation-of-function allele ( Nuckolls et al . , 2017 ) . We integrated the tagged alleles at the ade6 locus in separate haploid S . pombe strains . We then crossed those two haploid strains to create heterozygous mCherry-wtf4/wtf4poison-GFP diploids and induced these diploids to undergo meiosis . We imaged the asci using both standard and time-lapse fluorescence microscopy ( Figure 1C , Figure 1—figure supplement 1B ) . We confirmed our previous observations that the mCherry-Wtf4antidote was enriched in two spores , whereas Wtf4poison-GFP was found throughout the ascus and often formed puncta of various sizes . In the spores that did not inherit the antidote , Wtf4poison-GFP also appeared dispersed throughout the spores . In the spores that inherited and thus expressed mCherry-wtf4 , however , the localization of Wtf4poison-GFP was more restricted . Specifically , we observed that the Wtf4poison-GFP largely colocalized with mCherry-Wtf4antidote in a limited region of the spore ( Figure 1C , Figure 1—figure supplement 1B ) . In time-lapse microscopy , it was evident that the two Wtf4 proteins consistently colocalized in a defined region of the spore , even as this region changed shape over time . This co-diffusion suggests the two proteins are either physically interacting or are present in the same compartment ( Figure 1—figure supplement 1B ) . It also appeared that the level of Wtf4poison-GFP protein is reduced in spores containing the antidote . We did not distinguish if this was due to technical reasons ( i . e . quenching of the GFP molecules ) or biological reasons such as degradation of Wtf4poison-GFP in spores with mCherry-Wtf4antidote and/or due to a higher expression of Wtf4poison-GFP in the spores that inherit it ( non-antidote spores ) ( Figure 1C ) . We completed Pearson correlation analysis ( Adler and Parmryd , 2010 ) of mCherry-Wtf4antidote and Wtf4poison-GFP in the spores ( where a result of >0 is positive correlation; 0 , no correlation; <0 anti correlation ) and obtained a coefficient of 0 . 61 , indicating strong colocalization between the two Wtf4 proteins ( Figure 1—figure supplement 1C ) . The limited distribution of the Wtf4 poison and antidote proteins within wtf4+ spores suggested they may be confined to a specific cellular compartment . To test this idea , we looked for colocalization of Wtf4 proteins with the vacuole , endoplasmic reticulum ( ER ) and nucleus ( see below ) . For these experiments , we used the fully functional wtf4-GFP allele , which tags both the poison and antidote proteins ( Nuckolls et al . , 2017 ) . To assay the localization of the Wtf4 proteins relative to the vacuole , we imaged asci produced by diploids that were heterozygous for both wtf4-GFP and cpy1-mCherry . Cpy1-mCherry localizes to the lumen of the vacuole in vegetative cells ( Sun et al . , 2013 ) but has not , to our knowledge , been imaged in spores . We could observe mCherry in two of the four spores – presumably the two that inherited the cpy1-mCherry allele ( Figure 1D ) . This 2:2 spore localization pattern has been previously observed in budding yeast for vacuolar proteins and several other organelles ( Neiman , 2011; Roeder and Shaw , 1996; Suda et al . , 2007 ) . We found that the Wtf4-GFP and Cpy1-mCherry proteins colocalized within the spores that inherited both tagged alleles , suggesting the Wtf4 proteins are found within the vacuole ( Pearson coefficient of 0 . 89 , Figure 1D , Figure 1—figure supplement 2A–B ) . Interestingly , we also saw colocalization of Wtf4-GFP proteins with an ER marker , pbip1-mCherry-AHDL ( Zhang et al . , 2012; Figure 1—figure supplement 2C–D ) . We speculate this colocalization with the ER is due to nitrogen starvation which is required to induce meiosis and promotes organelle autophagy in S . pombe ( Kohda et al . , 2007; Zhao et al . , 2016 ) . Because we could not distinguish the vacuole and ER within spores , we assayed the localization of the Wtf4 proteins in the absence of nitrogen starvation . To do this , we fluorescently tagged the coding sequence of wtf4poison ( wtf4poison-GFP ) and wtf4antidote ( wtf4antidote-mCherry ) separation-of-function alleles under the control of β-estradiol-inducible promoters ( Ohira et al . , 2017 ) . We then integrated the wtf4poison-GFP allele at the ura4 locus and the wtf4antidote-mCherry allele at the lys4 locus of the same haploid strain . Next , we observed the localization of the Wtf proteins relative to the vacuole ( visualized using the CellTracker Blue CMAC lumen stain ) or the ER ( using Sec63-YFP ) following β-estradiol induction . Similar to our observations in spores , we saw that the Wtf4poison-GFP and Wtf4antidote-mCherry proteins largely colocalized , with a Pearson coefficient of 0 . 68 ( Figure 1—figure supplement 3D and E ) . We also found that the Wtf4 proteins colocalized with the CMAC stain ( Figure 1E ) , which suggests that the Wtf4 poison and antidote proteins are largely within the vacuole . However , there were Wtf4poison-GFP puncta that lined the periphery of the cell and a circle in the middle of the cell , reminiscent of ER localization . These puncta were devoid of Wtf4antidote-mCherry ( Figure 1—figure supplement 3D ) . We also attempted to assay the localization of the Wtf4 antidote and poison proteins individually to test if the localization of the Wtf4poison was altered in the presence of the Wtf4antidote , as we observed in spores ( Figure 1C ) . We found that the localization of the Wtf4antidote-mCherry to the vacuole was similar in the absence of the Wtf4poison ( Figure 1F , Figure 1—figure supplement 3B ) , with a Pearson coefficient of 0 . 69 ( Figure 1—figure supplement 3C ) . This is analogous to previous observations of the localization of the slightly different Wtf4antidote protein ( 82 . 2% amino acid identity ) found in the S . pombe lab strain ( Matsuyama et al . , 2006 ) . We failed , however , to generate cells carrying the wtf4poison-GFP allele without the wtf4antidote-mCherry allele by transformation , or by crossing the strain carrying both wtf4poison-GFP and wtf4antidote-mCherry to a wild-type strain ( Figure 1—figure supplement 3A ) . This is likely due to leaky expression of the wtf4poison-GFP from the inducible promoter even without addition of β-estradiol . Overall , our results suggest that the Wtf4poison protein is toxic in vegetative cells , but the antidote is still capable of neutralizing the poison , as we could obtain cells carrying both the Wtf4 poison and antidote proteins . In the process of trying to understand the localization patterns of Wtf4 proteins , we assayed the localization of the Wtf4 proteins relative to the nucleus . For this experiment , we imaged asci produced by wtf4-GFP/ade6+ heterozygotes also carrying a tagged histone allele , hht1-RFP ( Tomita and Cooper , 2007 ) . Although we did not observe colocalization of Wtf4 proteins and the nucleus , we frequently ( 24/38 asci ) observed that the nuclei in the wtf4- spores appeared more condensed ( Figure 1G , younger ascus ) . Additionally , in 11 out of 38 asci , one or both of the nuclei in the wtf4- spores were disrupted and the nuclear contents were dispersed throughout the spores ( Figure 1G , older ascus ) . To address the timing of these nuclear phenotypes , we imaged diploids undergoing gametogenesis using time-lapse microscopy . We saw that all four nuclei tended to look similar shortly after the second meiotic division . As spores matured , however , we observed nuclear condensation sometimes followed by fragmentation in the spores that did not inherit wtf4 ( i . e . in spores lacking the enriched GFP expression and antidote function ) ( Figure 1—figure supplement 4A and B , see Materials and methods ) . This nuclear condensation and fragmentation are reminiscent of apoptotic cell death ( Carmona-Gutierrez et al . , 2010; Kerr et al . , 1972 ) . Our experiments in S . pombe suggest that the Wtf4 proteins can act when expressed outside of gametogenesis . However , our inability to induce expression of the Wtf4poison in the absence of the Wtf4antidote limited our ability to explore their mechanisms of action in this system . We , therefore , tested if the Wtf4 proteins functioned in the budding yeast Saccharomyces cerevisiae . To do this , we cloned the coding sequences of wtf4poison-GFP and wtf4antidote-mCherry under the control of β-estradiol inducible promoters on separate plasmids ( Ottoz et al . , 2014 ) . We then introduced these plasmids into S . cerevisiae individually and together . We found that cells carrying the wtf4poison-GFP plasmid were largely inviable when wtf4poison-GFP expression was induced , indicating the poison is also toxic to S . cerevisiae ( Figure 2A ) . However , cells expressing Wtf4antidote-mCherry had only a slight growth defect relative to control cells carrying empty plasmids ( Figure 2A ) . Importantly , expression of the Wtf4antidote-mCherry plasmid largely ameliorated the toxicity of Wtf4poison-GFP ( Figure 2A ) . Given that S . pombe and S . cerevisiae diverged >350 million years ago ( Hoffman et al . , 2015 ) , our results suggest that the target ( s ) of Wtf4poison toxicity are conserved and the Wtf4antidote does not require cofactors that are specific to S . pombe or gametogenesis to neutralize Wtf4poison’s toxicity . We assayed the localization of the Wtf4 proteins in S . cerevisiae using the inducible wtf4poison-GFP and wtf4antidote-mCherry alleles described above . Similar to our observations in S . pombe gametogenesis , we saw that Wtf4poison-GFP localized as puncta of varying sizes throughout the cytoplasm ( Figure 2B ) . We also observed some of the Wtf4poison-GFP protein localized to the ER ( Figure 2—figure supplement 1A; Friederichs et al . , 2011 ) . Analogous to our observation in S . pombe spores , we saw nuclear condensation in cells expressing Wtf4poison-GFP relative to wild-type cells ( Figure 2—figure supplement 1B–1D ) . Wtf4antidote-mCherry , on the other hand , generally localized to one or two large amorphous regions adjacent to the vacuole ( Figure 2C ) . When co-expressed , Wtf4poison-GFP and Wtf4antidote-mCherry co-localized to this region next to the vacuole ( Figure 2D ) . In some cells , a faint circle of Wtf4poison-GFP could also be observed ( likely ER localization ) ; however , the majority colocalized with the antidote in the vacuole-associated region ( Figure 2—figure supplement 1E ) . This localization was similar but not identical to our observations in S . pombe cells , where the Wtf4 proteins localize within , rather than adjacent to , the vacuole . To ensure the difference in localization ( sequestration to a single puncta ) and cell viability of the Wtf4poison-GFP protein observed in the cells co-expressing Wtf4antidote-mCherry was not due to the mCherry tag , we also confirmed these results with an untagged Wtf4antidote ( Figure 2—figure supplement 2 ) . Given that we failed to detect the Wtf proteins enter the vacuole , as we saw in S . pombe , we were interested in testing if the Wtf proteins were entering the vacuole but being rapidly degraded in S . cerevisiae . To test this , we prepared protein samples from cells expressing Wtf4antidote-mCherry and Wtf4poison-GFP and performed western blots using α-mCherry and α-GFP antibodies . We analyzed protein from both the pellet and supernatant of our protein preparations because the predicted proteins are likely to be hydrophobic with six predicted transmembrane domains ( Figure 1—figure supplement 1A ) and their solubility is unknown ( TMHMM model , Krogh et al . , 2001 ) . Using α-mCherry antibodies , we observed a prominent band at ~65 kDa that showed some smearing consistent with protein degradation . This band was not observed in samples prepared from cells not expressing Wtf proteins , suggesting it is likely full-length Wtf4antidote-mCherry protein ( expected size 65 kDa ) . The band was more prominent in the pellet than the supernatant consistent with low solubility ( Figure 2—figure supplement 3 ) . Using α-GFP antibodies , we observed multiple bands ranging from ~15–85 kDa , although the expected size of Wtf4poison-GFP is 60 kDa . The bands were not observed in control samples prepared from cells not expressing Wtf proteins , suggesting the signal is specific to Wtf4poison-GFP . The large apparent size of some of the Wtf4poison-GFP bands suggest the protein may have post-translational modifications . In addition , the small size of some of the bands and the overall smeary appearance of the blot is consistent with degradation of the protein . Like the Wtf4antidote-mCherry , a considerable amount of the Wtf4poison-GFP was found in the pellet , suggesting both proteins have low solubility ( Figure 2—figure supplement 3 ) . Given that the Wtf4poison and Wtf4antidote proteins colocalize and are both found in the pellet fraction of protein preparations , we tested if the proteins physically interact by using acceptor photobleaching Fluorescence Resonance Energy Transfer ( FRET , Sekar and Periasamy , 2003 ) in cells expressing both Wtf4poison-GFP and Wtf4antidote-mCherry proteins . This process involves bleaching the fluorescence of a tagged protein ( the acceptor ) and looking for a corresponding increase in fluorescence of another tagged protein ( the donor ) . If an increase in fluorescence of the donor is observed , the proteins are said to be physically interacting , as they are in close enough proximity ( less than 10 nanometers ) to transfer energy to each other ( Sekar and Periasamy , 2003 ) . When we bleached Wtf4antidote-mCherry , we saw a corresponding increase in Wtf4poison-GFP emission , supporting the idea that the two proteins physically interact ( Figure 2E and F , Figure 2—figure supplement 1F ) . The Wtf4 proteins localize as puncta of varying sizes , so we hypothesized that the proteins assemble into aggregates . To explore the nature of the Wtf4 protein assemblies , we utilized the recently developed Distributed Amphifluoric FRET ( DAmFRET ) assay ( Khan et al . , 2018 ) . This approach looks for FRET between red and green versions of the same fluorophore in a partially photoconverted population of mEos3 . 1-tagged proteins as a measure of the protein’s tendency to self-assemble ( Figure 2—figure supplement 4A ) . We generated wtf4antidote-mEos3 . 1 and wtf4poison-mEos3 . 1 alleles , both under β-estradiol inducible promoters on ARS CEN plasmids . Both tagged constructs encoded functional proteins in S . cerevisiae , but the mEos3 . 1-tagged wtf4poison allele was not as toxic as the GFP-tagged allele ( Figure 2—figure supplement 4B ) . We then carried out DAmFRET analyses on cells producing either Wtf4antidote-mEos3 . 1 , Wtf4poison-mEos3 . 1 , or on cells producing both proteins simultaneously . We observed high ratiometric FRET signal ( AmFRET ) between Wtf4antidote-mEos3 . 1 proteins and between Wtf4poison-mEos3 . 1 proteins . In fact , all cells expressing Wtf4poison-mEos3 . 1 and/or Wtf4antidote-mEos3 . 1 proteins exhibited FRET as compared to mEos3 . 1 negative control , regardless of the expression level of the proteins ( Figure 2—figure supplement 4C ) . The level of AmFRET did not change when both proteins were expressed simultaneously . Collectively , these experiments show that the Wtf4 proteins self-assemble , and that the proteins do not inhibit each other’s assembly . Interestingly , we did not detect any cells producing purely monomeric Wtf4poison in our DAmFRET analyses , indicating that self-assembly does not require a rate-limiting nucleation step that is characteristic of prion-forming proteins ( Khan et al . , 2018 ) . This , combined with the irregular shape of the GFP puncta in our images , suggests that the toxic species of Wtf4poison is a poorly-ordered assembly of the protein . To explore this idea , we tested if overexpression of chaperones that promote protein homeostasis could neutralize Wtf4poison aggregates . First , we independently transformed multicopy 2-µm plasmids carrying various galactose-driven chaperones ( SIS1 , YDJ1 , HSP42 , JJJ2 , HSC82 , HSP82 , HSP104 ) into a strain carrying the wtf4poison-mEos3 . 1 allele ( Figure 2—figure supplement 5; Li et al . , 2016b ) . Overexpression of any of these chaperones failed to ameliorate the toxicity of Wtf4poison-mEos3 . 1 , even when we reduced the levels of induction of the Wtf4poison ( Figure 2—figure supplement 5 ) . This suggests that neutralization of Wtf4poison may require the induction of multiple chaperones or stress response pathways at one time . The Wtf4poison and Wtf4antidote proteins share the same 293 C-terminal amino acids ( Figure 1A , Figure 1—figure supplement 1A ) . All the known active Wtfantidote proteins are highly similar to the Wtfpoison they neutralize ( Bravo Núñez et al . , 2020a; Bravo Núñez et al . , 2020b; Hu et al . , 2017 ) . In addition , mutations that disrupt the similarity between a given Wtfantidote and Wtfpoison can eliminate the ability of the Wtfantidote to neutralize the Wtfpoison ( Bravo Núñez et al . , 2018a ) . Here , we tested the mechanism underlying that requirement using Wtf4 proteins . Given that each Wtf4 protein self-assembles ( Figure 2—figure supplement 4 ) , we hypothesized that homotypic interactions between Wtf4poison and Wtf4antidote mediate their co-assembly and neutralization of the poison . To test this idea , we mutated sequences at the C-termini of the inducible wtf4poison-GFP and wtf4antidote-mCherry alleles in the S . cerevisiae plasmids described above . Specifically , we targeted our mutagenesis to a seven amino acid sequence ( IGNAFRG ) that is found in varying copy numbers in many members of the wtf gene family ( Eickbush et al . , 2019 ) . We previously showed that a mismatched number of these repeats between Wtf poison and antidote proteins is enough to disrupt their specificity ( Bravo Núñez et al . , 2018a ) . The wild-type S . kambucha wtf4 allele contains ~1 . 5 repeat units ( Figure 3A ) . To make the mutants , we inserted 18 additional codons into the repeat region of wtf4 to make a total of four repeats . We denote these repeat insertion mutants with an * ( Figure 3A ) . As expected , the Wtf4poison-GFP protein is functional ( i . e . toxic ) in S . cerevisiae and localizes similarly to the tagged wild-type Wtf4poison-GFP ( Figure 3B , Figure 3—figure supplement 1A ) . Wtf4poison-GFP is neutralized by the matching Wtf4antidote-mCherry protein , and the two mutant proteins colocalized as vacuole-associated assemblies , just like the tagged wild-type proteins in S . cerevisiae ( Figure 3B and C ) . The Wtf4antidote-mCherry protein on its own also resembled the wild-type Wtf4antidote-mCherry protein localization ( Figure 3—figure supplement 1B ) . The Wtf4antidote-mCherry could not , however , suppress the toxicity of the wild-type Wtf4poison-GFP ( Figure 3B ) . Similarly , the wild-type Wtf4antidote-mCherry could not neutralize the Wtf4poison-GFP’s toxicity ( Figure 3B ) . The poison and antidote proteins did not colocalize in cells with incompatible poison and antidote proteins , and instead the poison proteins formed distributed aggregates , similar to cells expressing no antidote ( Figure 3D and E ) . We next used transmission electron microscopy ( TEM ) to analyze the environment of Wtf proteins within the vacuole-associated aggregates . Similar to our observations made using fluorescence microscopy , we found using immuno-gold labeling that Wtf4poison-GFP largely clustered near the vacuole in cells also producing untagged Wtf4antidote ( Figure 4A ) . These images also revealed that the Wtf4 protein aggregates appeared within a cluster of lightly staining organelles resembling lipid droplets ( Figure 4A , Figure 4—figure supplement 1A ) . Very few immunogold particles were found in the cells carrying only empty vectors , suggesting minimal background and high specificity of the GFP antibody used for the immunolabeling ( Figure 4—figure supplement 1B ) . To look at these Wtf aggregate-associated organelles at higher resolution , we used TEM with a sample preparation method that better maintains cellular morphology ( see Materials and methods ) . We found that the organelles were in fact a mix of lipid droplets and large vesicles with bilayer membranes ( Figure 4B arrows , Figure 4—figure supplement 2A–C ) . We quantified the number of lipid droplets and large vesicles in cells carrying empty vectors and in cells carrying both β-estradiol inducible Wtf4antidote and β-estradiol inducible Wtf4poison-GFP . We found that cells expressing Wtf4 proteins had significantly more lipid droplets and large vesicles ( Figure 4D , Figure 4—figure supplement 2F ) . These results indicate that the large aggregates that form in cells expressing Wtf4antidote are embedded in a cluster of large vesicles and lipid droplets . This phenotype is reminiscent of another aggregation prone protein , α-synuclein , a protein associated with Parkinson’s disease in humans , that when expressed in yeast forms cytoplasmic accumulations in association with clusters of vesicles ( Soper et al . , 2008 ) . The α-synuclein vesicles , however , appear smaller and more numerous than the Wtf4-associated vesicles . To test if the increase in vesicles and lipid droplets was a common feature of aggregation prone proteins , we expressed ( using the β-estradiol system ) a different vacuole-associate prion aggregate , Rnq1-mCardinal ( Figure 4—figure supplement 2E ) . We did not observe elevated vesicles or lipid droplets , suggesting that the increase in vesicles is due to the Wtf4 aggregates , not a consequence of the over-expression system or a general feature of aggregation prone proteins . We also imaged cells expressing only Wtf4antidote or Wtf4poison . The morphology of cells expressing only the Wtf4antidote was indistinguishable from cells expressing both Wtf proteins ( Figure 4—figure supplement 2D ) . It is difficult to interpret the observations from the ( dying ) cells expressing only Wtf4poison because the majority of the cells did not maintain cellular integrity during sample preparation ( Figure 4—figure supplement 3A ) . In the few cells we could image , we observed diverse morphologies . We generally did not observe clustering of large vesicles and lipid droplets , as we saw in cells expressing Wtf4antidote . Instead , organelle integrity often looked disrupted and many cells expressing Wtf4poison appeared to have undergone extensive autophagy ( Figure 4—figure supplement 3B–D ) . Our results were reminiscent of other studies in which toxic aggregated proteins were neutralized via sequestration at cellular inclusions ( Kaganovich et al . , 2008; Liu et al . , 2010; Taylor et al . , 2003; Chen et al . , 2011; Hill et al . , 2017; Tyedmers et al . , 2010; Kryndushkin et al . , 2012; Bagola and Sommer , 2008; Arrasate et al . , 2004 ) . In S . cerevisiae , stable , misfolded proteins are generally sequestered to the Insoluble PrOtein Deposit ( IPOD ) , a compartment located near the vacuole and pre-autophagosomal site ( PAS ) ( Kaganovich et al . , 2008; Tyedmers et al . , 2010; Suzuki and Ohsumi , 2010; Rothe et al . , 2018 ) . This compartmentalization of damaged/misfolded proteins mitigates their toxic effects and facilitates their disposal , some of which occurs via autophagy ( Marshall et al . , 2016 ) . Given that the Wtf4antidote and Wtfpoison+Wtf4antidote aggregates localize adjacent to the vacuole , we hypothesized that they could be at the IPOD in S . cerevisiae . To test this idea , we looked for the localization of the Wtf4 proteins relative to Rnq1-mCardinal and GFP-Atg8 . Rnq1 localizes to the IPOD and Atg8 is a component of the pre-autophagosomal structure that is adjacent to the IPOD ( Kaganovich et al . , 2008; Tyedmers et al . , 2010; Rothe et al . , 2018 ) . Consistent with our hypothesis , we found that Wtf4antidote-mCherry either colocalized or was adjacent to Rnq1-mCardinal ( Figure 5A , Figure 5—figure supplement 1A–1C ) . Wtf4poison-GFP did not colocalize with Rnq1-mCardinal on its own , supporting the idea that Wtf4antidote recruits the poison to the IPOD ( Figure 5—figure supplement 1D ) . To visualize the localization of Wtf4antidote relative Atg8 , we expressed Wtf4antidote-mCherry from a Gal-driven promoter on a plasmid . This is a different expression system than our other experiments described so far , but the protein is functional ( Figure 5—figure supplement 2A–B ) . We found that Wtf4antidote-mCherry colocalized or was adjacent to GFP-Atg8 ( expressed from a plasmid under its endogenous promoter ) when both proteins localized outside the vacuole ( Figure 5—figure supplement 3D I ) Guan et al . , 2001 . This supports the idea that Wtf4antidote localizes at the IPOD . However , in the majority of cells expressing both Wtf4antidote-mCherry and GFP-Atg8 , the proteins both localized inside the vacuole ( Figure 5—figure supplement 3D , F H ) . This was different than cells expressing either protein alone , where Atg8-GFP localizes as a focus outside the vacuole ( Figure 5—figure supplement 3E ) and Wtf4antidote-mCherry localizes as a large aggregate outside the vacuole ( Figure 5—figure supplement 3G ) . This suggests that the overexpression of GFP-ATG8 enhances the degradation of Wtf4antidote-mCherry . Additionally , these results could suggest that the increased vacuolar localization of GFP-Atg8 in cells producing Wtf4antidote-mCherry is due to enhanced recruitment and degradation of autophagosomes . Indeed , we saw increased vesicles in cells expressing Wtf4antidote ( Figure 4 ) . Proteins in the IPOD tend to be insoluble ( Bagola and Sommer , 2008; Kaganovich et al . , 2008 ) . To test if the Wtf4antidote shared this property in S . cerevisiae , we used half punctum-Fluorescence Recovery After Photobleaching ( half-FRAP ) ( Khan et al . , 2018; Zhang et al . , 2015 ) . This analysis revealed that the Wtf4antidote-mCherry aggregate has very low internal mobility and is thus more solid-like than liquid-like ( Figure 5B ) . We were curious if the Wtf4antidote behaved similarly in its native context . To test this , we performed the half-FRAP assay on the Wtf4antidote-mCherry in S . pombe spores and , consistent with our result in S . cerevisiae , found very low protein mobility ( Figure 5—figure supplement 4A B ) . To better understand how toxic Wtf4 protein aggregates are neutralized , we screened for genes necessary for survival after induction of wtf4antidote and wtf4poison . Briefly , we screened the S . cerevisiae MATa , haploid deletion collection for mutants that failed to survive on galactose media when they carried plasmids containing galactose-inducible wtf4antidote-mCherry and wtf4poison-GFP genes ( Figure 5—figure supplement 2A B ) . We found 106 mutants that could grow on galactose when carrying empty vector plasmids , but not when carrying both wtf4 plasmids ( Figure 5—figure supplement 2—source data 1 ) . We also provide results of similar screens for comparison ( Figure 5—figure supplement 2—source data 1; Willingham et al . , 2003 Enyenihi and Saunders , 2003 ) . Amongst our hits , the only significantly enriched ( FDR p<0 . 05 ) gene ontology groups were mitochondrial translation and organization ( Figure 5—figure supplement 2—source data 1 ) . We speculate this enrichment is due to two known roles of mitochondria in managing protein aggregates . The first is the Mitochondria As Guardian In Cytosol ( MAGIC ) mechanism in which mitochondria help degrade protein aggregates ( Ruan et al . , 2017 ) . The second is that mitochondria mitigate the impact of toxic aggregates by promoting asymmetric aggregate segregation in mitosis ( Zhou et al . , 2014 ) . We also identified genes involved in cell wall integrity pathways ( POP2 , MPT5 , SLT2 and BCK1 ) as necessary for survival after induction of Wtf4antidote and Wtf4poison ( Jin et al . , 2015; Li et al . , 2016a; Stewart et al . , 2007 ) . The cell wall integrity pathway is triggered by diverse stress stimuli ( Fuchs and Mylonakis , 2009 ) and can promote stress-response gene expression and nuclear release of cyclin-C ( ( Ssn8 ) , also a hit in our screen ) ( García et al . , 2009 ) . Release of cyclin-C into the cytoplasm promotes mitochondrial hyper-fission , stress response gene activation , and either apoptosis or repair of the stress-induced damage ( Jin et al . , 2015 ) . Consistent with this , we observed separated , significantly smaller mitochondria in cells expressing the Wtf4 proteins ( Figure 4—figure supplement 2C and G ) . Altogether , our screen hits suggest links between the cell wall integrity stress response pathways , mitochondrial fission , and Wtf4antidote function . Several other screen hits were genes with known roles in maintaining protein homeostasis and/or aggregate management . For example , we found that several genes involved in vesicle transport , endocytosis , and trafficking to the vacuole ( e . g . ATG11 , SNF7 , and multiple VPS genes ) are also required for survival when the Wtf4 proteins are expressed ( Figure 5—figure supplement 2—source data 1 ) . These hits suggest vacuolar trafficking pathways contribute to the neutralization of Wtf4 protein aggregates . This is consistent with our EM analyses showing that the Wtf4antidote inclusion site is enriched with vesicles . Previous work demonstrated these pathways are also important for trafficking other proteins to the IPOD and for neutralizing the toxicity of the aggregation prone TDP-43 ( Rothe et al . , 2018; He et al . , 2006; Liu et al . , 2017 ) . Given our results , which suggest that Wtf4 protein localization is an important factor in mitigating toxicity , we next imaged the localization of Wtf4poison-GFP and Wtf4antidote-mCherry in all of the screen hits . We found that the localization of the Wtf4poison-GFP and Wtf4antidote-mCherry proteins was disrupted in all 106 hits relative to wild type ( where the proteins coalesce to the IPOD ) . In 81 mutants , the Wtf4poison-GFP and Wtf4antidote-mCherry proteins localized as dispersed aggregates throughout the cell . These mutants included deletions of YNL170W , a reported dubious open reading frame , and PHD1 , a transcriptional activator ( Figure 5—figure supplement 2C D ) . We noted there were often cells with dispersed Wtf4antidote-mCherry aggregates or cells with dispersed Wtf4poison-GFP aggregates , but rarely cells with both . We speculate this is due to toxicity of distributed aggregates and cells expressing both aggregates at the same time being destroyed quickly . Another common feature we observed throughout the screen hits was Wtf4antidote-mCherry signal in the vacuole . We also observed this vacuolar localization in the C-terminal mutants depicted in Figure 3D and E , so this appears to be a common feature of the Wtf4antidote-mCherry protein in cells being destroyed by Wtf4poison . Five mutants appeared to have wild-type looking Wtf4antidote ( single inclusion outside the vacuole ) but dispersed Wtf4poison , suggesting that the mutations may disrupt the interaction of the poison and antidote ( Figure 5—figure supplement 2E F ) . Twenty hits showed very little Wtf4 signal and soluble cytoplasmic localization of both proteins . ( Figure 5—figure supplement 2G ) . Because the Wtf4antidote protein is quite similar to the Wtf4poison and also assembles into aggregates , we were curious if the Wtf4antidote alone was toxic in the absence of any of our screen hits . We therefore assayed the viability of the 106 deletion mutants when only Wtf4antidote was expressed . We saw that in approximately half ( 44/106 ) of the deletion stains Wtf4antidote expression reduced viability ( Figure 5—figure supplement 2—source data 1 ) . These results are consistent with the idea that active aggregate management pathways are often required for cells to mitigate the toxicity of even the Wtf4antidote protein in S . cerevisiae . We also investigated one hit from our screen , VPS1 , more thoroughly using our β-estradiol-inducible wtf4 plasmids ( described above ) . Vps1 is a dynamin-like GTPase that is necessary for trafficking of aggregates to the IPOD and/or other inclusion sites ( Kumar et al . , 2016; Kumar et al . , 2017; Hill et al . , 2016; Marshall et al . , 2016 ) . In the absence of VPS1 , we found that the Wtf4antidote-mCherry and Wtf4poison-GFP proteins still physically interact ( Figure 5C , Figure 5—figure supplement 5A ) . The Wtf4 protein aggregates did not , however , coalesce to form large inclusions ( Figure 5D , Figure 5—figure supplement 5A and B ) and Wtf4antidote-mCherry failed to neutralize the toxicity of Wtf4poison-GFP in vps1Δ cells ( Figure 5E ) . Together , these experiments indicate that the physical interaction between the Wtf4poison and Wtf4antidote proteins is insufficient to neutralize the toxicity of Wtf4poison protein aggregates . Trafficking the aggregates to a vacuole-associated inclusion is also required . Interestingly , we also observed enhanced toxicity of the Wtf4antidote-mCherry protein in the absence of Vps1 and many of our other screen hits ( Figure 5E , Figure 5—figure supplement 2—source data 1 ) . These results suggest that the antidote aggregates are more detrimental to cells when they are distributed in the cytoplasm . Importantly , however , even in the vps1Δ mutant , expression of Wtf4antidote-mCherry is less toxic to cells than Wtf4poison-GFP . This , and the fact that not all of the 106 hits caused Wtf4antidote-mCherry to become toxic , suggests there are fundamental differences in the poison and antidote aggregates beyond their propensity to be trafficked to a vacuole-associated inclusion . Here , we explored how the Wtf4poison protein kills cells and how the Wtf4antidote protein neutralizes the toxicity of the Wtf4poison . We used a combination of genetics and cell biology to study these proteins in three contexts: ( 1 ) their endogenous context of S . pombe gametogenesis , ( 2 ) vegetatively growing S . pombe cells , and ( 3 ) vegetatively growing S . cerevisiae cells . In all three contexts , expression of Wtf4poison alone kills cells and expression of the Wtf4antidote rescues the toxicity . The simplest interpretation of these observations is that Wtf4poison exploits or disrupts a conserved aspect of cellular physiology that is important during both vegetative growth and gametogenesis . Similarly , the Wtf4antidote neutralizes the Wtf4poison using conserved cofactors that can act in both vegetative growth and gametogenesis . This conservation suggests that wtf-derived gene drives could be a useful tool for genetically altering populations . In S . pombe gametogenesis and in vegetative S . cerevisiae cells , we observed the Wtf4poison-GFP proteins assembled into small foci ( aggregates ) in the absence of Wtf4antidote . The aggregates were largely dispersed throughout the cytoplasm , with some ER localization . The assembly of Wtf4 proteins is reminiscent of another meiotic drive element , Het-s , which employs prion-like amyloid polymerization to convert Het-S proteins to a lethal form ( Dalstra et al . , 2003; Riek and Saupe , 2016 ) . We therefore evaluated whether Wtf4poison proteins exhibit prion activity in S . cerevisiae using DAmFRET ( Khan et al . , 2018 ) . We found that Wtf4poison-mEos proteins assembled with themselves even at very low expression levels ( Figure 2—figure supplement 4C ) . In fact , we were unable to detect cells that lacked self-assemblies , revealing that the toxic form of the protein is not appreciably supersaturated , as would be required for Wtf4antidote to detoxify it through a simple prion-like mechanism where the antidote templates the deposition of poison monomers . Nevertheless , the sequence-dependent self-assembly of Wtf4 remains consistent with amyloid polymerization . However , given its intimate association with vesicles , extensive testing would be required to further evaluate the structural basis of Wtf4 activity . The significance of the Wtf4poison aggregation is not clear . We speculate that the aggregation propensity is intimately tied to the toxicity of Wtf4poison . We propose that distributed Wtf4 aggregates interact broadly with other proteins and disrupt their folding or localization . Compounding effects of these hypothesized interactions could disrupt protein homeostasis or cellular integrity , leading to cell death . This death may occur via a programmed cell death pathway , as in both S . pombe gametogenesis and in vegetative S . cerevisiae , cells succumbing to the Wtf4poison exhibit nuclear condensation ( followed by nuclear fragmentation in S . pombe ) . The death may also be related to loss of cell wall integrity , as cell wall integrity pathways are necessary for cell survival upon expression of the Wtf4 proteins . Testing these ideas may be challenging , especially if understanding Wtf4poison toxicity proves to be as elusive as understanding the intensely studied neurotoxic aggregating proteins TDP-43 and α-Synuclein ( Johnson et al . , 2009; Cookson and van der Brug , 2008 ) . Like Wtf4poison , the Wtf4antidote also assembles into aggregates in both S . pombe and S . cerevisiae cells . Unlike the Wtf4poison , however , the Wtf4antidote aggregates have little effect on the viability of wild-type vegetative cells . This is surprising given the similarity of the two proteins ( the Wtf4poison shares 292 of the Wtf4antidote’s 337 amino acids ) . Our data suggest that the localization of the aggregates and/or the exposed aggregate surface area could underlie their differences in toxicity . The Wtf4poison aggregates ( without Wtf4antidote ) remain largely dispersed in the cytoplasm , whereas the Wtf4antidote proteins are trafficked to a confined region near or within the vacuole . In S . pombe cells , the Wtf4antidote aggregates enter the vacuole . In S . cerevisiae cells , the Wtf4antidote accumulates outside the vacuole in the IPOD , but could also be trafficked into the vacuole at some rate . We observed signs of protein degradation , but this could be due to vacuolar degradation or other degradation mechanisms ( Figure 2—figure supplement 3 ) . The different modes of cell division in S . pombe and S . cerevisiae may contribute to the differences between how the species handle Wtf4 aggregates . S . cerevisiae divides asymmetrically by budding and tends to retain aggregates , including Wtf4 proteins , in the mother cells ( Spokoini et al . , 2012 ) . S . pombe , however , generally divides symmetrically making it difficult for new cells to be born free of aggregates found in the progenitor cell . In some cases , S . pombe can asymmetrically divide damaged proteins to inclusions ( Coelho et al . , 2014 ) , but the mechanism is not as efficient as S . cerevisiae’s exclusion of aggregates from buds . Due to this , it may be more important for S . pombe to destroy aggregates in the vacuole . The structure of the vacuoles is also fundamentally different in the two species , as S . cerevisiae tends to have one large vacuole per cell , while S . pombe has many small , distributed vacuoles . Additionally , the mechanisms determining the site of inclusions in S . pombe remain to be elucidated . Interestingly , S . pombe wtf4 , the homolog of the S . kambucha wtf4 allele used in this study , was previously shown to localize both inside the vacuole and as large cytoplasmic inclusions when overexpressed in S . pombe vegetative growth ( Matsuyama et al . , 2006 ) . This suggests that some Wtf proteins can be recruited to both vacuoles and inclusions in S . pombe . When we disrupted the ability of S . cerevisiae cells to transport the Wtf4antidote aggregates with the vps1∆ mutation , we found that the Wtf4antidote aggregates were distributed and more toxic than in wild-type cells . This is consistent with the idea that a key feature of Wtf4 protein toxicity relies on the aggregates being widely dispersed in the cytoplasm . When Wtf4poison and Wtf4antidote are found together in wild-type cells , the proteins co-assemble into aggregates . The co-assembled aggregates then behave similarly to the Wtf4antidote aggregates and are trafficked into the vacuole ( in S . pombe cells ) or to or near the IPOD adjacent to the vacuole ( in S . cerevisiae cells ) where they cause limited toxicity . Also , like the Wtf4antidote aggregates , the toxicity of the Wtf4poison+Wtf4antidote co-assembled aggregates is greatly enhanced if aggregate transport to the vacuole is disrupted by mutations ( e . g . vps1∆ ) . Together , our observations suggest a mechanistic model for wtf4 function . In this model , wtf4 exploits protein aggregation control pathways to induce selective cell death . The Wtf4poison forms distributed toxic aggregates and the Wtf4antidote co-assembles with the Wtf4poison and neutralizes the aggregate’s toxicity via trafficking to the vacuole ( Figure 6 ) . This mechanism is unlike the mechanism of any other meiotic driver described to date ( Grognet et al . , 2014; Didion et al . , 2015; Long et al . , 2008; Dawe et al . , 2018; Rhoades et al . , 2019; Dalstra et al . , 2005; Hammond et al . , 2012; Vogan et al . , 2019; Chen et al . , 2008; Akera et al . , 2017; Bauer et al . , 2012; Pieper et al . , 2018; Herrmann et al . , 1999; Shen et al . , 2017; Yu et al . , 2018; Bauer et al . , 2007; Wu et al . , 1988; Xie et al . , 2019; Kruger et al . , 2019; Lin et al . , 2018; Svedberg et al . , 2020 ) , but there are very few mechanistically characterized killer meiotic drive systems ( reviewed in Bravo Núñez et al . , 2018b ) . This study focused on the wtf4 meiotic driver . There is , however , an incredibly diverse array of wtf genes that cause meiotic drive . For example , the poison protein encoded by wtf35 ( from the FY29033 isolate ) shares less than 23% amino acid identity with Wtf4poison ( Bravo Núñez et al . , 2020a ) . Despite that extreme divergence , both genes cause essentially the same phenotype in S . pombe meiosis: drive of the gene into >90% of the progeny of a heterozygote ( Bravo Núñez et al . , 2020a ) . The conserved protein aggregation model offers an explanation for how such a diverse array of proteins can cause the same phenotype . Under our model , the mechanism of the Wtfpoison proteins is dependent upon their aggregation propensity . Presumably , the evolution of a protein that must self-aggregate could be less constrained than the evolution of a protein that must maintain a specific enzymatic activity or interaction partner . Exon 1 of the wtf4 driver encodes the antidote-specific residues ( 45 amino acids ) that facilitate the recruitment of Wtf4 aggregates to the vacuole in S . pombe ( or the IPOD in S . cerevisiae ) . This antidote-specific function likely relies on unidentified interacting partners ( perhaps amongst our screen hits ) . This specific functional requirement could explain the greater conservation of exon 1-encoded residues amongst bona fide wtf drivers ( 68–100% amino acid identity ) compared to the conservation amongst the remaining exons ( 30–90% amino acid identity ) ( Bravo Núñez et al . , 2020a ) . Importantly , our model also suggests that aggregate management may be a major feature of gametogenesis in S . pombe . The number of wtf genes varies between different isolates , but most have 30 or more wtf genes ( Eickbush et al . , 2019 ) . Four of these genes are widely diverged from wtf4 and are either not expressed in gametogenesis or their proteins exhibit distinct cellular localization from known Wtfpoison and Wtfantidote proteins ( Bravo Núñez et al . , 2020a ) . The rest of the genes that have been tested are similar to wtf4 in expression and localization ( Bravo Núñez et al . , 2018a; Bravo Núñez et al . , 2020a ) . It is not clear how many wtf drivers are expressed in a given cell , but they all appear to be transcribed at some level ( Eickbush et al . , 2019; Kuang et al . , 2017 ) . It will be interesting to explore the direct and indirect impacts of these Wtf proteins on S . pombe gametogenesis . The genetic screen presented in this work identified a number of factors required for cell viability in S . cerevisiae cells expressing Wtf4poison and Wtf4antidote . Many of these genes informed our model for Wtf4 protein function and therefore fit nicely within our proposed model . For example , our screen implicated genes involved in the Cytoplasm-to-Vacuole Targeting ( CVT ) pathway as necessary for survival of the Wtf4 proteins . This pathway has been previously implicated in aggregate management ( Kumar et al . , 2016; Kumar et al . , 2017 ) . Not all of our screen hits , however , are in genes or pathways with annotated roles that clearly fit our model . Some of the genes have no annotated functions . It is possible that at least some of these genes are not directly involved in aggregate management , but the mutants are especially sensitive to the stresses imposed by Wtf4 aggregates . It is also possible that some of the genes do have roles in mitigating the effects of toxic aggregates . Indeed , in deletions of some genes with unknown functions , we saw distributed Wtf4 aggregates , suggesting these unknown proteins could play a role in sequestration of aggregates . Interestingly , other hits are in well-studied genes , such as multiple acetyltransferases and various kinetochore proteins . Future analysis of these hits will be essential to refine or to potentially reject our current model . Studying how parasites manipulate their hosts can uncover unexpected insights on the host’s biology . For example , studies of the mouse t-haplotype meiotic driver revealed that gene expression in spermatids can create sperm-autonomous phenotypes , even though spermatids are connected by intercellular bridges ( Herrmann et al . , 1999 ) . Under our model , a fine line exists between protein aggregates that cells can manage ( i . e . Wtf4antidote ) and lethal aggregates that are not effectively managed ( i . e . Wtf4poison ) . We propose that the wtf4 meiotic driver has exploited this feature for its own selfish advantage . Future studies can now exploit the Wtf4 proteins to learn about protein aggregate toxicity and cellular aggregate management strategies . Z3EV promoter system is a titratable inducible promoter system ( Ohira et al . , 2017 ) . The system requires the Z3EV transcription factor and a Z3EV-responsive promoter ( Z3EVpr ) . β-estradiol induces nuclear import of the Z3EV protein; therefore , genes placed immediately downstream of Z3EVpr in a strain expressing Z3EV become expressed upon β-estradiol addition to the media . The LexA-ER-AD system ( Ottoz et al . , 2014 ) utilizes a heterologous transcription factor containing a LexA DNA-binding protein , the human estrogen receptor ( ER ) and an activation domain ( AD ) . β-estradiol binds the ER and tightly regulates the activity of the LexA-ER-AD transcription factor . The LexA DNA-binding domain recognizes lexA boxes in the target promoter . To create the Sec63-YFP strain , we PCR amplified the C-terminus of sec63 ( using oligos 939+941 ) and the sequence downstream of sec63 ( using oligos 945+946 ) using SZY643 as a template . We also amplified a YFP-HIS3 cassette from pYM41 ( Janke et al . , 2004 ) using oligos 944+943 . We then used overlap PCR ( using oligos 939+943 ) to unite those three PCR products . We then transformed this PCR product into GP1163 with standard lithium acetate protocol ( Gietz et al . , 1995 ) ( selecting for His+ ) to integrate the tagged sec63-YFP at its endogenous locus to generate SZY1277 . We confirmed the strain via PCR using oligos 2037+2038 . We PCR amplified the vps1Δ::kanMX locus out of strain YKR001C from the haploid yeast knockout MATa collection ( Open Biosystems ) ( using oligos 1850+1851 ) and transformed the PCR product into SLJ769 using high efficiency lithium acetate protocol ( selecting for G418 resistance ) to create strain SZY2539 . We used PCR ( using oligos 1712+1713 ) and sequenced the locus to confirm the deletion . To add the PACT1-LexA-ER-haB42 transcription factor , we digested FRP718 ( Addgene #58431 , Ottoz et al . , 2014 ) with NheI and integrated it into SZY2539 at the his3-11 , 15 locus ( using standard lithium acetate protocol [Gietz et al . , 1995] and selecting for His+ ) to create SZY2552 . We induced samples of SZY2072 , SZY2070 , SZY2159 , and SZY2059 , with β-estradiol as described above . We then aliquoted these induced samples into a 96-well plate . We then partially photoconverted the mEos3 . 1 protein by exposing the plate , while shaking at 800 RCF , to 405 nm illumination for 25 min using an OmniCure S1000 fitted with a 320–500 nm ( violet ) filter and a beam collimator ( Exfo ) , positioned 45 cm above the plate . This exposure yielded a total photo dose of 16 . 875 J/cm2 . This photo dose reproducibly achieves the maximum fluorescence of the acceptor ( red ) form of mEos3 . 1 while minimizing photobleaching of the green form ( Khan et al . , 2018 ) . For Figure 2—figure supplement 4C , we assayed the photoconverted samples on a Bio-Rad ZE5 cell analyzer with high-throughput automation . We analyzed 20 µL of each sample to collect approximately 100 , 000 events per well . We excited the mEos3 . 1 donor ( green form ) with a 488 nm laser at 100 mW and collected with 525/35 nm and 593/52 nm bandpass filters , respectively . We excited the acceptor fluorochrome with a 561 nm laser at 50 mW and collected with a 589/15 nm bandpass filter . We performed manual compensation on-instrument at acquisition . We used DeNovo FCS Express for data analysis and visualization and calculated ratiometric FRET as FRET/acceptor signals . For imaging during S . pombe gametogenesis ( Figure 1C , D and G , Figure 1—figure supplement 1B , Figure 1—figure supplement 2A and C ) , we crossed the two haploid yeast strains to generate heterozygous diploids as previously described ( Nuckolls et al . , 2017 ) . We placed the diploids on sporulation agar ( SPA , 1% glucose , 7 . 3 mM KH2PO4 , vitamins , agar ) for 2–3 days . We then scraped the cells off of the SPA plates and onto slides coated with 0 . 2 mg/mL lectin ( Sigma ) for imaging ( Tomita and Cooper , 2007 ) . For vegetatively growing samples ( Figure 1E and F , Figure 1—figure supplement 3B and D ) , we induced gene expression with β-estradiol as described above . If we used vacuole staining , we took 1 mL of the induced culture , spun to pellet , and resuspended in 1 mL of 10 mM HEPES buffer , pH 7 . 4 , containing 5% glucose with 100 μM CellTracker Blue CMAC ( Component B; Invitrogen C2110 ) . We incubated these cells at room temperature for 30 min . We then washed with YEL media and imaged . For imaging , we used the LSM-780 ( Zeiss ) microscope with a 40x C-Apochromat water-immersion objective ( NA 1 . 2 ) in photon-counting channel mode . For GFP , we used 488 nm excitation and collected through a 491–552 bandpass filter . For mCherry , we used 561 nm excitation and collected through a 572 longpass filter . For YFP , we used 514 nm excitation and collected through a 500–589 nm bandpass filter . For CMAC , we used 405 nm excitation and collected through a 411–509 nm bandpass filter . Brightness and contrast are not the same for all images . We analyzed at least 20 cells for each strain and chose a representative image . For experiments assaying meiosis/gametogenesis , we used at least two independent progenitor diploids . For cells that were imaged during vegetative growth , we used at least three different starting cultures . We carried out Pearson correlation analysis ( Adler and Parmryd , 2010 ) as previously described ( Slaughter et al . , 2013 ) . Briefly , we drew a segmented line ( width of two pixels ) throughout the spore , randomly covering as much of the spore as we could . We then used an in-house custom written plugin for ImageJ ( https://imagej . nih . gov/ij/ ) to generate a two-color line profile . We calculated the Pearson correlation of the line profile with varying degrees of shifts in at least eight spores or six vegetatively growing cells per sample . We then combined and averaged the trajectories with standard error . To quantify nuclear size , we calculated the full width at half maximum of the fluorescence intensity of RFP . We quantified 42 spores that inherited wtf4-GFP and 19 that did not , all from a wtf4-GFP/ade6+ heterozygote after 2 days on SPA media . We excluded any nuclei that appeared to have already fragmented . For the nuclear timelapse ( Figure 1—figure supplement 4 ) , we grew diploid cultures to saturation at 32°C overnight in YEL media . We then plated 100 µL of the cultures on a SPA plate , cut a circle punch of agar from the plate , and placed this punch upside down ( cells facing down ) in a 35 mm glass bottom poly-D-lysine coated dish ( MatTek corporation ) . We placed grease around the edge of the MaTeK dish and a moist kim wipe inside to control for humidity . We then imaged the cells using the Nikon Ti Eclipse coupled to a Yokogawa CSU W1 Spinning Disk , using the 60x oil objective , acquiring images every ten minutes . Here , we excited RFP at 561 nm and collected its emission through a 605–70 nm bandpass filter . For the gametogenesis timelapse ( Figure 1—figure supplement 1B ) , we grew diploid cultures to saturation at 32°C overnight in YEL media . The next day , we diluted 100 µL of the saturated diploid culture into 5 mLs of PM media ( 20 mLs of 50x EMM salts , 20 ml 0 . 4 M Na2HPO4 , 25 mL 20% NH4Cl , 1 mL 1000x Vitamins , 100 µL 10 , 000x mineral stock solution , 3 g potassium hydrogen phthalate , 950 mL ddH2O , 25 mL of sterile 40% glucose after autoclaving , supplemented with 250 mg/L uracil ) . We grew the PM culture overnight at 32°C . The next day , we spun to pellet and resuspended the pellet in PM-N media ( 20 mLs of 50x EMM Salts , 20 mL 0 . 4 M Na2HPO4 , 1 mL 1000x Vitamins , 100 µL 10 , 000x mineral stock solution , 25 mL of sterile 40% glucose after autoclaving , supplemented with 250 mg/L uracil , volume up to 1 L with ddH2O ) . We shook the PM-N cultures for 4 hr at 28°C . Then , we took 100 µL of the PM-N culture and mixed it with 100 µL of lectin ( Sigma ) . We took 150 µl of this mixture and added it to a 35 mm glass bottom poly-D-lysine coated dish ( MatTek corporation ) . We waited 5 min to allow the cells to adhere . We then added 3 mL of fresh PM-N to the dish ( protocol modified from Klutstein et al . , 2015 ) . We imaged using a Zeiss Observer . Z1 wide-field microscope with a 63x ( 1 . 4 NA ) oil-immersion objective and collected the emission onto a Hamamatsu ORCA Flash 4 . 0 using μManager software . We acquired the mCherry with BP 530–585 nm excitation and LP 615 emission , using an FT 600 dichroic filter , acquiring images every 10 min . For all budding yeast images except for the three experiments described below , we induced samples as described above and imaged on an LSM-780 ( Zeiss ) microscope , with a 40x C-Apochromat water-immersion objective ( NA 1 . 2 ) in photon-counting channel mode . For GFP and mCherry , we used the same conditions as we did in S . pombe . For mCardinal , we used 633 nm excitation and collected through a 632–696 nm bandpass filter . Brightness and contrast are not the same for all images . We imaged at least 20 cells from at least three starting cultures and chose a representative image for each figure . For imaging vps1Δ cells ( Figure 5 ) , IPOD cytology ( Figure 5—figure supplement 1A ) , and the nuclear timelapse ( Figure 2—figure supplement 1B and C ) , we placed samples in a Millipore Onix 2 Cellasic system to allow for a constant flow of media . We initiated flow of inducing media ( SC with 500 nM b-estradiol for vps1Δ and SC galactose for the nuclear timelapse ) and took images every 10 min . We used a Perkin Elmer Ultraview Vox spinning disc microscope with a Hamamatsu EMCCD ( C9100-23B ) with a 40x C-Apochromat water-immersion objective ( NA 1 . 2 ) . We collected GFP and mCherry with 488 and 561 nm excitation as above but collected GFP through a 525–550 nm bandpass filter and mCherry through a 615–670 nm bandpass filter . We had two independent starting cultures for the sample . We chose representative cells and timepoints . Brightness and contrast are not the same for all images . To quantify nuclear size ( Figure 2—figure supplement 1D ) , we fit each nucleus to a two-dimensional Gaussian function and found the full width at half maximum of the fluorescence intensity of RFP per cell . We quantified at the beginning of the timelapse ( early ) and 14 hr into the timelapse ( late ) . We quantified 72 Wtf4poison-GFP expressing cells and 65 wild-type cells at the early timepoint . We quantified 62 Wtf4poison-GFP expressing cells and 79 wild-type cells at the later timepoint . We grew overnight cultures of SZY1954 ( cells carrying inducible Wtf4poison-GFP , Wtf4antidote-mCherry ) and SZY1821 ( empty vector ( EV ) , untagged control ) in SC -His -Ura -Trp ( without agar ) . The next day , we diluted 10 mL of the saturated culture into 90 mL of media of the same type . We then added β-estradiol to a final concentration of 500 nM . We shook the cultures at 30°C to induce . After 2 and 5 hr of induction , we prepared whole-cell lysates as previously described ( Gerace and Moazed , 2014 ) , with a few exceptions . First , we used 1% tritonX-100 instead of NP40 in the lysis buffer . For bead beating , we used the FastPrep-24 5G bead beater with Lysing Matrix Y bead tubes ( MP Biomedicals ) and the S . cerevisiae fast prep program ( 40 s ) . We took both the supernatant and pellet protein samples for analysis . Samples were boiled for 1 min at 95°C before loading . Next , we ran the proteins on NuPAGE 4–12% Bis-Tris Protein Gels ( Invitrogen , NP0321 ) and then transferred to Trans-Blot Turbo Mini PVDF membranes ( Bio-RAD #1704156 ) . We stained the membranes with a monoclonal , α-GFP antibody ( from cell signaling technology #2956 ) at 1:1000 and monoclonal , α-mCherry antibody ( from EnCor Biotechnology Inc #MCA-1C151 ) at 1:1000 , overnight at 4°C with agitation in Odyssey blocking buffer ( TBS , from LI-COR biosciences ) . A secondary , α-rabbit antibody ( 800 cW ) and secondary , α-mouse antibody ( 600 cW ) was used for fluorescent visualization of the proteins . We imaged the blot on the Odyssey-CLx ( LI-COR biosciences ) . Three independent starting cultures were analyzed to confirm results . We carried out acceptor photobleaching FRET with β-estradiol induced ( described above ) SZY1954 ( wild type ) using a LSM-780 ( Zeiss ) microscope , with a 40x C-Apochromat water-immersion objective ( NA 1 . 2 ) in photon-counting channel mode . For vps1Δ cells ( SZY2570 ) , we used a Perkin Elmer Ultraview Vox spinning-disc microscope with a Hamamatsu EMCCD ( C9100-23B ) with 488 and 561 nm excitation . For both samples , we photobleached the acceptor ( mCherry ) with 561 nm excitation ( for bleaching images , see Figure 2—figure supplement 1F for wild type and Figure 5—figure supplement 5A for vps1Δ ) . We analyzed 22 wild-type and 76 vps1Δ cells . We made 50 mL saturated overnight cultures of SZY1821 , SZY1952 , SZY1954 , and SZY2731 in SC media lacking histidine , tryptophan , and uracil ( to select for retention of the plasmids ) . The next day , we diluted 10 mL of the saturated cultures into 90 mL of the same media with 500 nM β-estradiol . We shook these cultures for four hours at 30°C , reaching log phase . We then pelleted the yeast cells by filtering and carried out high pressure freezing with the Leica ICE system ( Leica Biosystems ) . We further processed the frozen cell pellets by freeze substitution ( FS ) using acetone containing 0 . 2% uranyl acetate ( UA ) and 2% H2O was used as FS medium . The FS program was −90° to −80° over 70 hr , −80° to −60° over 6 hr , −60° for 5 hr , −60° to −50° over 6 hr , and −50° to −20°C over 4 hr . After washing extensively with acetone , we then infiltrated , embedded and polymerized the samples into resin . For Immuno-EM , we used HM-20 resin . We cut 60 nm sections with a Leica Ultra microtome ( Leica UC-6 ) and picked up onto a carbon-coated 150 mesh nickel grid . The grids were labeled with anti-GFP primary antibody ( a gift from M . Rout , Rockefeller University , New York , NY ) and 12 nm colloidal gold goat anti-rabbit secondary antibody ( Jackson Immuno Research Laboratories , Inc ) . After immuno labeling , we post-stained the samples with 1% UA for 3 min . We acquired images using a FEI Tecnai Biotwin electron microscope . For non-immuno-EM , we used Epon resin to better maintain morphology , but the rest of the procedure was the same . We analyzed the tomographs of at least 10 cells per condition . For quantification purposes , we also completed array tomography . For array tomography , we cut 60 nm serial sections with a Leica Ultra microtome ( UC-6 ) using an Ultra 35 Jumbo diamond knife ( Diatome ) and picked up on ITO coated coverslips using the ASH-100 Advanced Substrate Holder ( RMC Boeckeler ) . We post-stained serial sections with Sato’s triple lead stain for two minutes , 4% UA in 70% methanol for two minutes , and Sato’s lead stain again for two minutes . The coverslips were coated with 5 nm of carbon and imaged in a Zeiss Merlin Gemini 2 SEM with 4QBSD detector at 10 kV and 700 pA using Atlas 5 Array Tomography software ( Fibics ) . The obtained dataset was aligned with Midas of the IMOD software package ( Kremer et al . , 1996 ) and manually quantified . For better visualization , an image series of an individual yeast cell was cropped and further aligned with registration tools in ImageJ . For model building , the segmentation was done based on intensity and known organelle structure with Microscopy Image Browser ( Belevich et al . , 2016 ) and with IMOD . We used Amira ( Thermo Fisher Scientific ) software for model rendering and visualization . Further quantification of mitochondrial volumes was performed on selected cells after training a Unet ( Ronneberger et al . , 2015 ) . Hand annotation of training data was performed in Fiji . A suite of internally developed Fiji plugins , macros and CherryPy scripts called DeepFiji ( see below ) sent training data to a pair of in-house NVIDIA Tesla-equipped deep learning machines running Tensorflow . Representative cells were selected , and segmented images inferred using the same macros and deep learning machines before being aligned using a StackReg variant . Mitochondrial volumes were quantified in Fiji using the 3D Segmentation tools . DeepFiji is a suite of macros and plugins in Fiji , Python , and CherryPy ( a Python web framework ) that enable end users on any machine with a reasonable amount of RAM to request deep learning training and inference on a remote deep learning box as long as both machines have access to a shared file system . First , a user selects example sub-images that span the realm of potential objects , background levels and signal levels . Manual annotations are made using Fiji's Region of Interest ( ROI ) tools and manager , and individual ROI files are saved for each image ( in our case the ROIs were each individual cell and the total of the mitochondria inside ) . A user chooses a small subset of the annotated image/ROI pairs to be used as a validation set , while the remainder becomes the training set . For each image , two binary channels are added from the associated ROIs: a mask channel and an outline channel . The mask channel has all pixels contained within an ROI painted true , while the outline channel only paints true the pixels that were in the ROI's outline . As the training network expects standard image sizes and runs more efficiently with smaller images , the macro next makes image stacks for both annotated training and validation images that break the original images up into 512 × 512 sub-images with 50% overlap between sub-images . For the training set , it also applies a series of random rotations and translations to help the neural network generalize . Both the validation and new training images are saved to a shared file system and the deep learning boxes are notified to begin processing through a call to a webserver running on the box . The sub-image size , file location , and other parameters are configurable by the user at the time of running . For our in-house system , the deep learning boxes are running ubuntu with NVIDIA Tesla boards and configured with Tensorflow 1 . 13 . 1 and CUDA 10 . 0 . CherryPy is configured to listen for web calls on each and , once initiated from a user , begins processing files from the selected directory by calling trainer . py . The trainer first finds the standard deviation ( STDEV ) and mean ( MEAN ) of the non-zero pixel intensities and stores those values . The training , validation , and all future inference sets will be processed by ( Intensity-MEAN ) /STDEV+0 . 5 first to keep numbers roughly between 0 and 1 . The model used is a modified Unet ( ref ) : convolutional layers are alternated with max pool layers , doubling the channel depth at each layer while halving the resolution . The final layer is 32 × 32×512 . At each convolutional layer , the image is passed through a leaky rectified linear unit . Convolutional up-sampling brings the image back to its original resolution where it is passed through a tanh ( ) function and the mean square error is calculated with respect to the ground truth image for back-propagation . Every 100th iteration of the training is applied to the validation set and the images are visualized using TensorBoard ( which opens automatically on the user's host computer ) . Training proceeds with the learning rate adjusting over time , until 2200 iterations have passed at which point most networks have either converged or never will . Once training is completed , and a reasonable iteration point is found in TensorBoard , the user can run Inferer . ijm in Fiji on their host machine to apply their model to a new dataset . Inferer will similarly parse images into sub-images and contact a deep learning box to initiate processing . Once processing is complete the user can run a second script to blank out border regions and de-window their images . The output outline and mask channels are ranged from 0 to 1 and represent probabilities . Typically , thresholding pixel values above 0 . 5 in the mask channel will suffice for finding objects of interest . However , in cases with frequent object touching , one can subtract the outline probability from the mask . If consistent mistakes are found in the inferred data , the user can annotate them properly using Fiji and use Retrain . ijm to retrain the network using the new data together with the old to generate a new training set . Retraining starts from the original model so that it does not have to relearn from scratch . Plugins necessary to run in Fiji are available from the Stowers update site within Fiji . Macros , python code , and CherryPy configurations are available at https://github . com/cwood1967/DeepFiji .	Meiotic drivers are genes that break the normal rules of inheritance . Usually , a gene has a 50% chance of passing on to the next generation . Meiotic drivers force their way into the next generation by poisoning the gametes ( the sex cells that combine to form a zygote ) that do not carry them . Harnessing the power of genetic drivers could allow scientists to spread beneficial genes across populations . One group of meiotic drivers found in fission yeast is called the 'with transposon fission yeast' , or 'wtf' gene family . The wtf drivers act during the production of spores , which are the fission yeast equivalent of sperm , and they encode both a poison that can destroy the spores and its antidote . The poison spreads through the sac holding the spores , and can affect all of them , while the antidote only protects the spores that make it . This means that the spores carrying the wtf genes survive , while the rest of the spores are killed . To understand whether it is possible to use the wtf meiotic drivers to spread other genes , perhaps outside of fission yeast , scientists must first establish exactly how the proteins coded for by genes behave . To do this , Nuckolls et al . examined a member of the wtf family called wtf4 . Attaching a fluorescent label to the poison and antidote proteins produced by wtf4 made it possible to see what they do . This revealed that the poison clumps , forming toxic aggregates that damage yeast spores . The antidote works by mopping up these aggregates and moving them to the cell's main storage compartment , called the vacuole . Mutations that disrupted the ability of the antidote to interact with the poison or its ability to move the poison into storage stopped the antidote from working . Nuckolls et al . also showed that if genetic engineering was used to introduce wtf4 into a distantly related species of budding yeast the effects of this meiotic driver were the same . This suggests that the wtf genes may be good candidates for future genetic engineering experiments . Engineered systems known as 'gene drives' could spread beneficial genetic traits through populations . This could include disease-resistance genes in crops , or disease-preventing genes in mosquitoes . The wtf genes are small and work independently of other genes , making them promising candidates for this type of system . These experiments also suggest that the wtf genes could be useful for understanding why clumps of proteins are toxic to cells . Future work could explore why clumps of wtf poison kill spores , while clumps of poison plus antidote do not . This could aid research into human ailments caused by protein clumps , such as Huntington’s or Alzheimer’s disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "cell", "biology" ]	2020	The wtf4 meiotic driver utilizes controlled protein aggregation to generate selective cell death
Autophagy is an important intracellular catabolic mechanism involved in the removal of misfolded proteins . Atg14L , the mammalian ortholog of Atg14 in yeast and a critical regulator of autophagy , mediates the production PtdIns3P to initiate the formation of autophagosomes . However , it is not clear how Atg14L is regulated . In this study , we demonstrate that ubiquitination and degradation of Atg14L is controlled by ZBTB16-Cullin3-Roc1 E3 ubiquitin ligase complex . Furthermore , we show that a wide range of G-protein-coupled receptor ( GPCR ) ligands and agonists regulate the levels of Atg14L through ZBTB16 . In addition , we show that the activation of autophagy by pharmacological inhibition of GPCR reduces the accumulation of misfolded proteins and protects against behavior dysfunction in a mouse model of Huntington's disease . Our study demonstrates a common molecular mechanism by which the activation of GPCRs leads to the suppression of autophagy and a pharmacological strategy to activate autophagy in the CNS for the treatment of neurodegenerative diseases . Autophagy is an important intracellular catabolic mechanism that mediates the turnover of cytoplasmic constituents via lysosomal degradation . In multi-cellular organisms , autophagy serves important functions in mediating intracellular protein degradation under normal nutritional conditions . Defects in autophagy lead to the accumulation of misfolded proteins in the central nervous system , an organ that is protected from nutritional deprivation under physiological conditions ( Hara et al . , 2006 ) . How cells regulate autophagy under normal nutritional condition is an important unsolved question in the field . In mammalian cells , adaptor protein Atg14L/Barkor in complex with Vps34 , the catalytic subunit of the class III PI3K , and the regulatory proteins Beclin 1 and p150 , function as a key driver in orchestrating the formation of autophagosomes by regulating the formation of Vps34 complexes and for targeting to the isolation membrane involved in initiating the formation of autophagosomes ( Obara and Ohsumi , 2011 ) . However , it remains to be determined how Atg14L is regulated in response to extracellular signaling . G-protein ( heterotrimeric guanine nucleotide–binding protein ) -coupled receptors ( GPCRs ) are important regulators of cellular responses to diverse stimuli with major clinical implications ( Foord et al . , 2005 ) . While the activation of GPCRs is known to lead to numerous downstream events , the role and mechanism of autophagy regulated by GPCRs is not yet clear . Furthermore , it is also not clear how the signaling of GPCRs controls the levels of PtdIns3P . ZBTB16 , also known as promyelocytic leukemia zinc finger or Zfp145 , is a member of ‘BTB-POZ’ protein family and mediates the binding of CUL3 , a core component in multiple cullin-RING-based BCR ( BTB-CUL3-RBX1 ) E3 ubiquitin-protein ligase complexes and its substrates ( Furukawa et al . , 2003; Geyer et al . , 2003; Xu et al . , 2003 ) . In this study , we investigated the mechanism by which ZBTB16 regulates autophagy . We show that CUL3-ZBTB16 regulates autophagy by mediating the proteasomal degradation of Atg14L , which is controlled by GPCR ligands through GSK3β phosphorylation . Furthermore , we show that inhibiting GPCRs by pharmacological means leads to the activation of autophagy in the central nervous system ( CNS ) and ameliorates neural dysfunction in a mouse model of Huntington's disease . Our study identified a common mechanism by which multiple GPCR ligands mediate autophagy through regulating the levels of Vps34 complexes and a pharmacological strategy to activate autophagy in the CNS to inhibit neural dysfunction induced by the accumulation of misfolded proteins . A genome-wide siRNA screen identified ZBTB16 as one of the hits that when its expression is knocked down can lead to the activation of autophagy and increased production of PtdIns3P ( Lipinski et al . , 2010 ) . Since PtdIns3P is produced by Vps34 complexes , the class III PtdIns3 kinase , we hypothesize that ZBTB16 might affect the levels/activity of Vps34 complexes . Although ZBTB16 is known to be involved in regulation of transcription in nucleus ( Mathew et al . , 2012 ) , ZBTB16 is predominantly localized in the cytoplasm ( Costoya et al . , 2008 ) ( Figure 1A ) , suggesting that ZBTB16 might have a non-nuclear function . To test this hypothesis , we screened for the effects of ZBTB16 knockdown on the levels of Atg14L , Vps34 , Beclin1 , and UVRAG . Interestingly , we found that knockdown of ZBTB16 led to specific increases in the levels of Atg14L , a key component of Vps34 complexes specifically involved in regulating autophagy , and corresponding increases in the ratio of LC3II/tubulin and a reduction in the levels of p62 , indicating the activation of autophagy ( Figure 1B and Figure 1—figure supplement 1A ) ( Matsunaga et al . , 2009; Zhong et al . , 2009 ) . Overexpression of Atg14L has been shown to induce autophagy under normal nutritional conditions ( Matsunaga et al . , 2010; Fan et al . , 2011 ) . On the other hand , the levels of Beclin1 , Vps34 , UVRAG were not affected by the knockdown of ZBTB16 ( Figure 1B ) . We compared the changes of LC3II/tubulin in the presence or absence of chloroquine ( CQ ) , an inhibitor of lysosomal degradation . We found that compared to that of ZBTB16 knockdown or CQ treatment , the presence of CQ with ZBTB16 knockdown led to a further increase in the ratio of LC3II/tubulin , suggesting that ZBTB16 deficiency led to an increase in autophagic flux ( Figure 1C ) . 10 . 7554/eLife . 06734 . 003Figure 1 . ZBTB16 mediates the proteasomal degradation of Atg14L . ( A ) The cytoplasmic and nuclear fractions of HeLa cells cultured in normal media were separated by using Paris Kit ( Ambion ) and analyzed by western blotting using indicated antibodies . ( B ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting ZBTB16 and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( C ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting ZBTB16 and cultured for 72 hr . Before harvesting , the cells were treated with or without 10 μM CQ ( chloroquine ) for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( D ) The levels of ZBTB16 , Atg14L , LC3 , and actin ( control ) in the lysates isolated from wt and zbtb16−/− littermates were analyzed by western blotting using indicated antibodies . ( E ) HeLa cells were transfected with the Xpress-tagged ZBTB16 expression vector and cultured for 36 hr . The cell lysate was analyzed by western blotting using anti-Xpress and anti-ATG14L antibodies . Anti-tubulin was used as a loading control . ( F ) 293T cells were transfected with Myc-Atg14 , Myc-Cul3 , HA-ROC1 , and FLAG-ZBTB16 expression vectors and cultured for 24 hr . MG132 ( 25 µM ) was added in the last 8 hr as indicated . The cell lysates were then harvested and analyzed by western blotting using indicated antibodies . ( G ) 293T cells were transfected with Myc-Cul3 , HA-ROC1 , and FLAG-ZBTB16 expression vectors and cultured for 24 hr . MG132 ( 25 µM ) or E64D ( 10 µM ) was added in the last 8 hr as indicated . The cell lysates were then harvested and analyzed by western blotting using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 00310 . 7554/eLife . 06734 . 004Figure 1—figure supplement 1 . ZBTB16 mediates the proteasomal degradation of Atg14L . ( A ) Quantification for Figure 1B . HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting ZBTB16 and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) ; p < 0 . 001 ( * ) . ( B ) Quantification for Figure 1D . The tissues of wt and zbtb16−/− littermates were isolated and homogenized in lysis buffer . The levels of ZBTB16 , Atg14L , LC3 , and actin ( as a loading control ) were analyzed by western blotting using indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) ; p < 0 . 001 ( * ) . ( C ) HeLa cells were transfected with control FLAG vector and FLAG-ZBTB16 expression vector and cultured for 48 hr . Transfection efficiency was monitored by co-transfection of GFP vector . The mRNA was extracted and analyzed by RT-PCR . ( D ) 293T cells were transfected with expression vectors of Myc-Cullin3 , HA-ROC1 , and FLAG-ZBTB16 and cultured for 24 hr . The cell lysates were analyzed by western blotting using indicated antibodies . Anti-tubulin was used as a loading control . ( E ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting Cul3 and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( F ) Left: representative images of GFP-LC3 puncta ( autophagosomes ) in H4-GFP-LC3 expressing control non-targeting ( N . T . ) or ZBTB16 siRNA cultured in normal conditions and the quantitation . Bars are mean ± SEM of triplicate samples ( 500 cells analyzed per sample ) . Similar results were observed in three independent experiments . p < 0 . 01 ( ) . Middle: representative images of GFP-LC3 puncta ( autophagosomes ) in H4-GFP-LC3 expressing control non-targeting ( N . T . ) or Cullin3 siRNA cultured in normal conditions and the quantitation . Bars are mean ± SEM of triplicate samples ( 500 cells analyzed per sample ) . Similar results were observed in three independent experiments . p < 0 . 01 ( ) . Right: representative images of GFP-LC3 puncta ( autophagosomes ) in H4-GFP-LC3 expressing control non-targeting ( N . T . ) or ROC1 siRNA cultured in normal conditions and the quantitation . Bars are mean ± SEM of triplicate samples ( 500 cells analyzed per sample ) . Similar results were observed in three independent experiments . p < 0 . 05 ( * ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 004 To confirm the role of ZBTB16 in regulation of Atg14L and autophagy in vivo , we examined the levels of Atg14L and LC3II in ZBTB16−/− mice ( Barna et al . , 2000 ) . Interestingly , we found that the levels of Atg14L were dramatically increased in the kidney , heart , liver , and brain of ZBTB16−/− mice compared to that of wt ( Figure 1D and Figure 1—figure supplement 1B ) . The levels of autophagy as indicated by the ratio of LC3II/tubulin were significantly higher in ZBTB16−/− mice than that of wt mice . From these results , we conclude that ZBTB16 is an important regulator of Atg14L and autophagy . To further characterize the impact of ZBTB16 on the levels of Atg14L , we transfected an expression vector of ZBTB16 into HeLa cells . Overexpression of ZBTB16 led to down-regulation of Atg14L protein in a dose-dependent manner but not that of its mRNA ( Figure 1E and Figure 1—figure supplement 1C ) . Thus , ZBTB16 regulates the protein levels of Atg14L but not its mRNA . Since ZBTB16 is known to function as an adaptor for CUL3-ROC1 complex of ubiquitin ligase ( Xu et al . , 2003; Mathew et al . , 2012 ) , we next tested the effect of this complex on the levels of Vps34 complexes . Overexpression of CUL3-ROC1-ZBTB16 led to the reduction in the levels of transfected as well as endogenous Atg14L ( Figure 1F–G ) . Overexpression of ZBTB16 also had a minor reducing effect on the levels of Vps34 and Beclin1 , but not that of UVRAG ( Figure 1—figure supplement 1D ) , consistent with a coordinated regulation of Vps34 complexes ( Itakura et al . , 2008; Matsunaga et al . , 2009; Zhong et al . , 2009 ) . The down-regulation of Atg14L by CUL3-ROC1-ZBTB16 can be rescued by MG132 , which inhibits the proteasomal degradation , but not by E64d , an inhibitor of lysosomal degradation ( Figure 1F–G ) . Finally , we found that knockdown of CUL3 led to increases in the levels of Atg14L and autophagy ( Figure 1—figure supplement 1E ) . The effect of knockdown ZBTB16 , Cullin3 , and ROC1 was also confirmed using image-based LC3-GFP assay ( Figure 1—figure supplement 1F ) . Taken together , these results suggest that CUL3-ROC1-ZBTB16 complex may promote the degradation of autophagic-specific Vps34 complexes through proteasomal pathway . Since ZBTB16 is known to function as an adaptor for CUL3-ROC1 ligase complex ( Furukawa et al . , 2003; Geyer et al . , 2003; Xu et al . , 2003 ) , we next investigated the possibility that Atg14L might be able to bind to ZBTB16 . We found that while the interaction of Atg14L and ZBTB16 was detectable under normal condition , such interaction was dramatically enhanced in the presence of proteasomal inhibitor MG132 ( Figure 2A; Figure 2—figure supplement 1A ) . In contrast , while the interactions of Vps34 and Beclin1 with ZBTB16 were detectable , the presence of MG132 had only minimum effect ( Figure 2—figure supplement 1B–C ) , suggesting that the interaction of ZBTB16 with Vps34 and Beclin1 might be indirect . In addition , we found that the endogenous interaction of ZBTB16 with Atg14L can be detected in both 293T and HeLa cells and can also be enhanced in the presence of proteasomal inhibitor MG132 ( Figure 2B–C ) . Consistent with the regulation of ZBTB16-Cullin3-ROC1 complex , the interaction of endogenous Atg14L with this complex was also detected ( Figure 2D ) . 10 . 7554/eLife . 06734 . 005Figure 2 . Ubiquitination of ATG14L by ZBTB16 . ( A ) 293T cells were transfected with expression vectors of FLAG-ZBTB16 and Myc-Atg14 and cultured for 24 hr . The cells were treated with MG132 ( 10 µM ) for the last 4 hr before harvesting . The cell lysates were immunoprecipitated with anti-ZBTB16 antibody , and the immunocomplexes were analyzed by western blotting using anti-Myc antibody . ( B ) 293T cells were cultured for 24 hr and then harvested and lysed in Buffer II . The lysates were immunoprecipitated with anti-ATG14L antibody , and the immunocomplexes were analyzed by western blotting using anti-ZBTB16 antibody . ( C ) HeLa cells were treated with MG132 ( 10 µM ) for 4 hr and then harvested and lysed in Buffer II . The lysates were immunoprecipitated with anti-ATG14L antibody , and the immunocomplexes were analyzed by western blotting using anti-ZBTB16 antibody . ( D ) HeLa cell lysates were immunoprecipitated with anti-ATG14L antibody , or a control IgG , and the immunocomplexes were analyzed by western blotting using indicated antibodies . ( E ) 293T cells were transfected with the expression vectors of FLAG-ZBTB16 , Myc-ATG14L , truncated Myc-Atg14L ( ΔBATS ) , Myc-Atg14L ( ΔCCD ) as indicated and cultured for 24 hr . The cells were then harvested and lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-Flag antibody , and the immunocomplexes were analyzed by western blotting using anti-Myc antibody . ( F ) 293T cells were transfected with the expression vectors of Myc-ATG14L , FLAG-ZBTB16 , truncated FLAG-ZBTB16-dZF , ZBTB16-BTB as indicated and cultured for 24 hr . The cells were then harvested and lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-Myc antibody , and the immunocomplexes were analyzed by western blotting using anti-Flag antibody . ( G ) 293T cells were transfected with expression vectors of Myc-Atg14 , HA-ROC1 , FLAG-ZBTB16 , and Myc-Cul3 expression vectors as indicated and cultured for 24 hr . MG132 ( 25 µM ) was added in the last 6 hr as indicated . The cell lysates were harvested and immunoprecipitated with anti-ATG14L . The immunocomplexes were analyzed by western blotting using anti-Ub antibody for ubiquitin . ( H ) Myc-Cullin3 , FLAG-ZBTB16 , ROC1 , and ATG14L proteins were individually purified from 293T cells transfected with indicated expression vectors by immunoprecipitation . The eluted proteins were incubated with recombinant E1 , E2 , and Ub . The reactions were terminated by boiling for 5 min in SDS sample buffer . The sample was analyzed by western blotting using anti-Myc . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 00510 . 7554/eLife . 06734 . 006Figure 2—figure supplement 1 . Ubiquitination of ATG14L by ZBTB16 and Cullin3 . ( A ) 293T cells were transfected with expression vectors of FLAG-ZBTB16 and Myc-Atg14L and cultured for 24 hr . The cells were treated with MG132 ( 10 µM ) for the last 4 hr before harvesting and then lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-ATG14L antibody , and the immunocomplexes were analyzed by western blotting using anti-Flag antibody . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ** ) . ( B ) 293T cells were transfected with expression vectors of FLAG-ZBTB16 and HA-Vps34 and cultured for 24 hr and then treated with MG132 ( 10 µM ) for an additional 4 hr . The cells were lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-Vps34 antibody , and the immunocomplexes were analyzed by western blotting using anti-Flag . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . ns , no significance . ( C ) 293T cells were transfected with expression vectors of FLAG-ZBTB16 and GFP-Beclin1 for 24 hr and then treated with MG132 ( 10 µM ) for 4 hr and then lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-Beclin1 antibody , and the immunocomplexes were analyzed by western blotting using anti-Flag antibody . Statistical analysis was performed on biological repeats of three-independent sets of experiments using ImageJ . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . ns , no significance . ( D–E ) HeLa cells were first transfected with control or Cul3 siRNA for 48 hr and then transfected with the indicated constructs and cultured for another 24 hr . The cells were then treated with MG132 for 4 hr before harvesting . Fully denatured lysates were diluted with 0 . 5% NP-40 lysis buffer and IP with anti-Flag antibody or anti-Myc antibody as indicated . The lysates were WB with indicated antibodies . ( F ) Expression vectors for FLAG-tagged Cul3 or Cul3∆C deleted C-terminal E2 binding domain were cotransfected with that of Myc-tagged ATG14 , ROC1 , Xpress-tagged ZBTB16 , and HA-tagged ubiquitin into 293T cells as indicated for 24 hr . The cells were then treated with MG132 for 2 hr before harvesting . Fully denatured lysates were diluted with 0 . 5% NP-40 lysis buffer and IP with anti-Myc antibody . The lysates were WB with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 006 We next characterized the domains of Atg14L and ZBTB16 involved in their interaction . We found that deletion of the C-terminal Barkor/Atg14 ( L ) autophagosome targeting sequence ( BATS ) domain , known to be involved in binding of Atg14L to autophagosome membrane ( Fan et al . , 2011 ) , completely eliminated the binding to ZBTB16 ( Figure 2E ) . On the other hand , deletion of the center domain of ZBTB16 eliminated the binding to Atg14L ( Figure 2F ) . Thus , the C-terminal BATS domain of Atg14L likely interacts with the center domain of ZBTB16 . Consistent with regulation of Atg14L by ZBTB16-mediated ubiquitination , the expression of ZBTB16 , Cullin3 , and ROC1 led to the ubiquitination of Atg14L ( Figure 2G ) . Furthermore , ubiquitination of ATG14L can be observed in the presence of purified CUL3 , ROC1 , and ZBTB16 proteins in vitro ( Figure 2H ) . The role of Cullin3 is important as knockdown of CUL3 or expressing CUL3ΔC , a dominant negative mutant ( Jin et al . , 2005; Mathew et al . , 2012 ) , blocked the ubiquitination of Atg14L , as well as auto-ubiquitination of ZBTB16 ( Figure 2—figure supplement 1D–F ) . Taken together , we conclude that CUL3-ROC1-ZBTB16 complex controls the ubiquitination and proteasomal degradation of ATG14L to regulate autophagy . Since regulation of Atg14L and autophagy in cells and mutant mice that are either deficient or overexpressing ZBTB16 occurred under normal nutritional conditions ( Figure 1B–E ) , we hypothesized that ZBTB16 might respond to growth factor stimulation , rather than cellular nutritional status . Indeed , we found that serum starvation of HeLa cells for as short as half an hour induced the degradation of ZBTB16 and a concomitant increase in the levels of ATG14L with no effect on its mRNA , while the levels of Beclin1 and Vps34 showed no significant change ( Figure 3A and Figure 3—figure supplement 1A ) . Similar changes in the levels of Atg14L and ZBTB16 were found in multiple cell lines , including 7721 , SK-OV-3 , H4 , and HCT116 cells , upon serum removal ( Figure 3—figure supplement 1B–E ) . On the other hand , no such change in Atg14L or ZBTB16 was observed even after removal of glucose or amino acids ( Figure 3—figure supplement 1F–G ) . These results suggest that ZBTB16 and Atg14L are regulated by serum factors rather than the availability of nutrients , such as glucose and amino acids , or the status of mTOR , a major regulator of autophagy in response to nutritional starvation . 10 . 7554/eLife . 06734 . 007Figure 3 . Serum starvation regulates ZBTB16 activity through GSK3β . ( A ) HeLa cells were cultured in serum-free condition for indicated periods of time , and then the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( B ) HeLa cells were serum starved for indicated periods of time with or without MG132 ( 10 µM ) , and then the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( C ) HeLa cells were transfected with the expression vectors of HA-Ub and Myc-ATG14L and cultured for 36 hr and then cultured either in the presence or absence of serum ( SD ) for an additional 4 hr . The cell lysates were harvested and immunoprecipitated with anti-ATG14L . The immunocomplexes were analyzed by western blotting using anti-HA antibody for ubiquitin or anti-Myc as indicated . ( D ) HeLa cells were transfected with control non-targeting or ZBTB16 targeting siRNA . The cells were cultured in the presence of serum for 46 hr after transfection , subject to serum deprivation ( SD ) for 24 hr , and then re-stimulated by 10% FBS as indicated for another 2 hr . The total cell lysates were analyzed by western blotting with indicated antibodies . ( E ) HeLa cells were treated with PI3 kinase inhibitor LY294002 at indicated concentrations ( μM ) and time periods , and the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( F ) HeLa cells were transfected with control siRNA or siRNA targeting GSK3β and cultured for 72 hr . The total cell lysates were analyzed by western blotting with indicated antibodies . ( G ) HeLa cells were treated with or without SD for 4 hr and then treated with or without MG132 ( 10 µM ) , LiCl ( 10 mM ) , or SB216763 ( 20 µM ) for an additional 4 hr . The cells were lysed in NP-40 buffer with phosphatase inhibitors and immunoprecipitated with pan-phospho-Thr antibody or pan phospho-Ser antibody . The immunocomplexes were analyzed by western blotting with rabbit anti-ZBTB16 . ( H ) The expression vectors of Flag-tagged wild type or ZBTB16 point mutants were transfected into HeLa cells and cultured for 24 hr . Then , the cells were subject to normal or SD conditions as indicated with MG132 ( 10 µM ) for 8 hr . The cells were lysed in NP-40 lysis buffer with phosphatase inhibitors and immunoprecipitated with anti-Flag antibody . The immunocomplexes were analyzed by western blotting with pan phospho-Ser antibody . ( I ) The expression vectors of Flag-tagged wild type or mutants of ZBTB16 were transfected into HeLa cells and cultured for 24 hr . The cells were subject to serum deprived or normal condition with MG132 for 8 hr . Then , the cells were lysed in NP-40 lysis buffer with phosphatase inhibitors and immunoprecipitated with anti-Flag antibody . The immunocomplexes were analyzed by western blotting with pan-phospho-Thr antibody . ( J–K ) The expression vectors of Flag-tagged wild type or mutant ZBTB16 were transfected into 293T cells and cultured for 24 hr . The cell lysates were collected and analyzed by immunoprecipitation with anti-Flag antibody . The immunoprecipitated ZBTB16 was incubated with or without recombinant GSK3β kinase in vitro at 30°C for 1 hr and then analyzed by western blotting analysis with anti-phosphor-Ser ( J ) or anti-phosphor-Thr ( K ) . Quantification data are expressed as mean of 3 biological replicates ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 00710 . 7554/eLife . 06734 . 008Figure 3—figure supplement 1 . Serum starvation regulates ZBTB16 activity through GSK3β . ( A ) HeLa cells were serum starvation for the indicated periods of time , and then the mRNA was extracted and subjected to RT-PCR . ( B–E ) 7721 cells ( B ) , SK-OV-3 cells ( C ) , H4 cells ( D ) and HCT116 cells ( E ) were cultured in serum-free condition for indicated periods of time , and then the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( F ) HeLa cells were treated without glucose for the indicated periods of time , and the cell lysates were analyzed by western blotting with indicated antibodies . ( G ) HeLa cells were cultured in the medium with or without amino acids for the indicated periods of time . The cell lysates were analyzed by western blotting with indicated antibodies . ( H ) HeLa cells were transfected with control non-targeting or ZBTB16 targeting siRNA . The cells were cultured in the presence of serum for 46 hr after transfection , subject to serum deprivation ( SD ) for 24 hr , and then re-stimulated by 10% FBS as indicated for another 2 hr . The total cell lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) , ns , no significance . ( I ) Top: representative images of GFP-LC3 puncta ( autophagosomes ) in H4-GFP-LC3 cells expressing control or ATG14L siRNA and cultured under serum starvation condition for 8 hr . Bottom: quantitation of GFP-LC3 puncta in serum starvation conditions is shown as above . Bars are mean ± SEM of triplicate samples ( 500 cells analyzed per sample ) . Similar results were observed in three independent experiments . p < 0 . 001 ( * ) . ( J ) HeLa cells were treated with 10 μM LY294002 or 10 μM CQ for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 00810 . 7554/eLife . 06734 . 009Figure 3—figure supplement 2 . Serum starvation regulates ZBTB16 activity through GSK3β . ( A ) HeLa cells were cultured without serum for the indicated periods of time , and then the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( B ) HeLa cells were treated with LY294002 ( 10 µM ) or GSK3 inhibitor SB216763 ( 10 µM ) for indicated periods of time in the presence of FBS . The cell lysates were harvested and analyzed by western blotting using indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ** ) . ( C ) HeLa cells were transfected with the expression vectors of Flag-ZBTB16 , constitutively active HA-GSK3β ( CA ) for 36 hr and then treated with or without SB216763 ( 10 µM ) , MG132 ( 10 µM ) for an additional 4 hr . The cells were lysed in NP-40 buffer with phosphatase inhibitors and immunoprecipitated with anti-Flag antibody . The immunocomplexes were analyzed by western blotting with pan phospho-Ser antibody or phospho-Thr antibody . ( D ) 293T cells were transfected with expression vectors for HA-GSK3β and XP-ZBTB16 for 24 hr . The cell lysates were harvested and immunoprecipitated with anti-HA . The immunocomplexes were analyzed by western blotting using indicated antibodies . ( E ) HeLa cells were transfected with the expression vectors of Flag-ZBTB16 , mutant Flag-ZBTB16 , constitutively active HA-GSK3β ( CA ) for 36 hr and then treated with MG132 ( 10 µM ) for an additional 4 hr . The cells were lysed in NP-40 buffer with phosphatase inhibitors and immunoprecipitated with anti-Flag antibody . The immunocomplexes were analyzed by western blotting with pan phospho-Ser antibody . The lysates were WB with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) . ( F ) HeLa cells were transfected with expression vectors of Flag-ZBTB16S184A/T282A and cultured for 24 hr . Before harvesting the sample , the cells were treated with mTOR inhibitor Torin1 ( 500 nM ) in nutrient-rich conditions for the indicated periods of time . The cell lysates were analyzed by western blotting with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 009 The degradation of ZBTB16 in serum-free condition was blocked by the addition of MG132 , suggesting serum starvation may lead to increased proteasomal degradation of ZBTB16 ( Figure 3B ) . Consistent with this hypothesis , serum starvation also decreased the ubiquitination of ATG14L ( Figure 3C ) . However , in HeLa cells deficient for ZBTB16 , the levels of Atg14L remained elevated even after the addition of serum ( Figure 3D; Figure 3—figure supplement 1H ) . Conversely , elevated levels of autophagy as indicated by increased numbers of GFP-LC3 upon serum removal were suppressed after Atg14L knockdown ( Figure 3—figure supplement 1I ) . Thus , we conclude that the expression of ZBTB16 is important for the down-regulation of Atg14L levels in the presence of serum . To examine the potential involvement of class I PI3 kinase signaling in the degradation of ATG14L , we treated HeLa cells with LY294002 , an inhibitor of PI3 kinase . As expected , the addition of LY294002 led to the loss of S473 phosphorylation in Akt , a marker for its activation , and the loss of an inhibitory phosphorylation on Ser9 of glycogen synthase kinase-3β ( GSK3β ) ( Sutherland et al . , 1993 ) , indicating the activation of GSK3β . Interestingly , the treatment with LY294002 led to significant reductions in the levels of ZBTB16 and concomitant increases in the levels of ATG14L and LC3II/tubulin ratio and autophagic flux , suggesting the induction of autophagy ( Figure 3E and Figure 3—figure supplement 1J ) . Thus , the inhibition of class I PI3 kinase and AKT and activation of GSK3β may contribute to the reduction in the levels of ZBTB16 and the increases that of Atg14L . Since phosphorylation of substrates provides an important targeting mechanism for Cul3-Roc1 E3 ubiquitin ligase complexes ( Petroski and Deshaies , 2005 ) , we hypothesized that serum starvation induced degradation of ZBTB16 is regulated by phosphorylation . As inhibition of class I PI3 kinase by LY294002 led to the activation of GSK3β , which is known to be activated by serum starvation to regulate cell survival ( Frame and Cohen , 2001 ) , we examined the possibility that GSK3β be involved in mediating the degradation of ZBTB16 upon serum starvation . The phosphorylation of Ser9 , an inhibitory phosphorylation event , in GSK3β is strongly inhibited in serum- starved HeLa cells as reported ( Figure 3—figure supplement 2A ) ( Sutherland et al . , 1993 ) . Consistent with a role of GSK3β in regulating ZBTB16 , knockdown of GSK3β , or the addition of SB216763 , an inhibitor of GSK3β , restored the levels of ZBTB16 ( Figure 3F; Figure 3—figure supplement 2B ) . Thus , GSK3β might be involved in regulating ZBTB16 in response to serum removal . We tested the possibility that GSK3β might phosphorylate ZBTB16 . Indeed , we found that the expression of constitutively active GSK3β ( Stambolic and Woodgett , 1994; Zhao et al . , 2012 ) could lead to phosphorylation of ZBTB16 , which can be inhibited by SB216763 ( Figure 3—figure supplement 2C ) . In addition , the interaction of ZBTB16 and GSK3β could be detected , suggesting that ZBTB16 might be a substrate of GSK3β ( Figure 3—figure supplement 2D ) . A search of Scansite ( http://scansite . mit . edu ) identified three possible sites in ZBTB16 for GSK3β: S184 , T282 , and S347 . Interestingly , we could detect increased phospho-ZBTB16 under serum starvation condition , and the phosphorylation of ZBTB16 could be inhibited by inhibitors of GSK3β ( Figure 3G ) . We found that S184A mutation , but not S347A mutation , eliminated the phosphorylation of ZBTB16 induced by serum starvation condition detected by pan-phospho-Ser antibody ( Figure 3H ) . On the other hand , T282A mutation-eliminated phosphorylation induced by serum starvation condition detected by pan-phospho-Thr antibody ( Figure 3I ) . Furthermore , S184A and T282A mutations also eliminated phosphorylation by GSK3β in vitro ( Figure 3J–K ) or in cells expressing constitutively active GSK3 ( Figure 3—figure supplement 2E ) ( Stambolic and Woodgett , 1994; Zhao et al . , 2012 ) . Blocking of ZBTB16 phosphorylation , however , had no effect on the mTOR pathway as Torin1 , an inhibitor of mTOR , could still induce autophagy in HeLa cells expressing ZBTB16 S184A/T282A ( Figure 3—figure supplement 2F ) . Taken together , we conclude that S184 and T282 of ZBTB16 may be phosphorylated by GSK3β in cells under serum starvation condition . To determine if ubiquitination of ZBTB16 might be affected by serum starvation , we expressed wt and S184A/T282A mutant ZBTB16 in HeLa cells . We found that serum starvation significantly enhanced auto-ubiquitination of wt ZBTB16 , but not S184A/T282A mutant ( Figure 4A ) . Consistently , auto-ubiquitination of S184D/T282E ZBTB16 mutant was significantly enhanced compared to that of wt ( Figure 4B ) . On the other hand , the abilities of S184A/T282A and S184D/T282E ZBTB16 mutants to mediate the ubiquitination of Atg14L were higher and lower than that of wt , respectively ( Figure 4C ) . Since both S184 and T282 are localized in the region of ZBTB16 that is important for binding to Atg14L , we tested the effect of phosphor-mimetic mutation in ZBTB16 on the interaction with Atg14L . Consistently , the interaction of S184D/T282E ZBTB16 mutant with Atg14L was significantly weaker than that of wt ( Figure 4D ) . Finally , to examine if S184/T282 phosphorylation of ZBTB16 might be critical functionally to regulate autophagy under serum starvation condition , we analyzed the response of HeLa cells expressing wt or S184A/T282A mutant to serum starvation . Since endogenous and overexpressed wt ZBTB16 proteins are degraded upon serum starvation while S184A/T282A mutant is not , this experiment provides an opportunity to test if the persistent levels of S184A/T282A mutant under serum starvation condition might be able to suppress the increases in Atg14L and autophagy . Interestingly , we found that the increases in the levels of Atg14L and reduction in that of p62 were blocked by the expression of S184A/T282A ( Figure 4E ) . Thus , we conclude that phosphorylation of S184/T282 of ZBTB16 by GSK3β is functionally important for regulating Atg14L under serum starvation condition to promote autophagy . 10 . 7554/eLife . 06734 . 010Figure 4 . Phosphorylation of S184/T282 of ZBTB16 by GSK3b is functionally important for regulating Atg14L under serum starvation condition to promote autophagy . ( A ) The expression vectors of Flag-tagged wild type or mutants ZBTB16 and HA-Ub were transfected into HeLa cells and cultured for 24 hr . HeLa cells were treated with or without serum-deprivation ( SD ) for 4 hr . The cell lysates were collected and subjected to immunoprecipitation with anti-Flag antibody . The immunocomplexes were analyzed by western blotting using anti-HA for ubiquitin or anti-Flag as indicated . ( B ) 293T cells were transfected with the expression vectors of wide-type FLAG-ZBTB16 , mutant FLAG-ZBTB16 , HA-Ub as indicated and cultured for 20 hr . The cells were then treated with or without MG132 for 4 hr . Fully denatured lysates were diluted with 0 . 5% NP-40 lysis buffer and IP with anti-Flag antibody . The lysates were WB with indicated antibodies . ( C ) 293T cells were transfected with the expression vectors of wide-type FLAG-ZBTB16 , mutant FLAG-ZBTB16 , HA-Ub , Myc-ATG14 , HA-ROC1 , Myc-Cul3 as indicated and cultured for 36 hr . The cells were then treated with MG132 for 2 hr . Fully denatured lysates were diluted with 0 . 5% NP-40 lysis buffer and IP with anti-ATG14L antibody . The lysates were WB with indicated antibodies . ( D ) HeLa cells were transfected with expression vectors of wide-type FLAG-ZBTB16 , mutant FLAG-ZBTB16 , Myc-ATG14 as indicated and then lysed in NP-40 buffer . The lysates were immunoprecipitated with anti-Flag antibody , and the immunocomplexes were analyzed by western blotting using anti-Myc antibody . ( E ) The expression vectors of Flag-tagged wild type or ZBTB16 point mutants were transfected into HeLa cells and cultured for 12 hr . The cells were cultured in serum-free condition for indicated periods of time . The cell lysates were then harvested and analyzed by western blotting using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 010 Next , we proceeded to identify the factor ( s ) in the serum that can reduce the levels of ATG14L . Surprisingly , we found that heat-inactivated fetal bovine serum ( FBS ) was equally effective in reducing the levels of ATG14L induced by serum starvation ( Figure 5A and Figure 5—figure supplement 1A–B ) . The presence or absence of serum , however , had no effect on the predominant cytoplasmic localization of ZBTB16 ( Figure 5—figure supplement 1C ) . The effect of serum starvation on the increased ATG14L and decreased ZBTB16 levels can also be suppressed by the addition of bovine serum albumin ( BSA ) with different degree of purity , SDF1 , or lysophosphatidic acid ( LPA ) , which are known GPCR ligands ( Figure 5B–D and Figure 5—figure supplement 1D–G ) . In addition , the treatment of other GPCR ligands such as endothelin 1 , high density lipoprotein ( HDL ) could all block the increases in the levels of Atg14L induced by serum starvation ( Figure 5E and Figure 5—figure supplement 1H ) . Since the signaling of SDF1 ( Bleul et al . , 1996 ) , LPA ( Hecht et al . , 1996 ) , endothelin ( Horinouchi et al . , 2013 ) , HDL ( Whorton et al . , 2007 ) is all known to be mediated through GPCRs , these results suggest that the levels of Atg14L are under the control of GPCR signaling . 10 . 7554/eLife . 06734 . 011Figure 5 . Regulation of Atg14L by GPCR mediated signaling pathways . ( A ) HeLa cells were cultured in the presence ( no treatment ) or absence of serum ( serum starvation ) for 24 hr and then re-stimulated by 10% FBS or boiled FBS ( boiled for 30 min at 95°C ) for 1 hr . The cell lysates were analyzed by western blotting with the indicated antibodies . ( B ) Serum-starved HeLa cells were stimulated with 10% FBS or 10 mg/ml BSA ( A2058 , Sigma ) for indicated periods of time and then harvested and analyzed by western blotting using indicated antibodies . ( C ) H4 cells were cultured in the presence or absence of serum for 24 hr . SDF1 was added to serum-starved cells at indicated concentrations for 1 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . ( D ) HeLa cells were serum starved for 24 hr and then stimulated with 20 µM LPA or 10% FBS for 1 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . ( E ) HeLa cells were serum deprived for 24 hr and then stimulated with endothelin 1 or HDL at indicated concentrations for 4 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . ( F ) HeLa cells were transfected with siRNAs for control , Gαq/11 , or GαS and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( G ) HeLa cells were transfected with siRNAs for control or Gα12/13 and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( H ) HeLa cells were serum-deprived ( SD ) for 24 hr and then stimulated with forskolin for 2 hr . The cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( I ) HeLa cells were treated with Gαi/o inhibitor Pertussis toxin ( PTX ) at 1 or 2 μg/ml for 6 hr , and the cell lysates were harvested and analyzed by western blotting using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 01110 . 7554/eLife . 06734 . 012Figure 5—figure supplement 1 . Regulation of Atg14L by ligands and agonists of GPCR . ( A ) Quantification for Figure 5A . HeLa cells were cultured in the presence ( no treatment ) or absence of serum ( serum starvation ) for 24 hr , and then re-stimulated by 10% FBS or boiled FBS ( boiled for 30 min at 95°C ) for 1 hr . The cell lysates were analyzed by western blotting with the indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) . p < 0 . 001 ( * ) . ( B ) HeLa cells were serum starved for 30 hr and then stimulated with serum for the indicated periods of time . The cell lysates were analyzed by using western blotting with the indicated antibodies . ( C ) HeLa cells were cultured in the presence or absence of serum ( SD ) for 6 hr and then the cell lysate was harvested and analyzed by western blotting using indicated antibodies . Antibodies for lamin A/C and tubulin were used to determine purity of nuclear and cytoplasmic samples , respectively . ( D ) HeLa cells were serum starved for 24 hr , and then stimulated with for different bovine serum albumin ( BSA ) batches for 1 hr , and the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( E ) Quantification for Figure 5B . Serum-starved HeLa cells were stimulated with 10% FBS or 10 mg/ml BSA ( A2058 , Sigma ) for indicated periods of time and then harvested and analyzed by western blotting using indicated antibodies . The data are expressed the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) ; p < 0 . 001 ( * ) . ( F ) Quantification for Figure 5C . H4 cells were cultured in the presence or absence of serum for 24 hr . SDF1 was added to serum-starved cells at indicated concentrations for 1 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) ; p < 0 . 001 ( * ) . ( G ) Quantification for Figure 5D . HeLa cells were serum-starved for 24 hr , and then stimulated with 20 µM LPA or 10% FBS for 1 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) , p < 0 . 001 ( * ) . ( H ) Quantification for Figure 5E . HeLa cells were serum-deprived for 24 hr , and then stimulated with endothelin 1 or HDL at indicated concentrations for 4 hr . The cells were lysed , and the lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) , p < 0 . 01 ( ** ) . ns , no significance . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 01210 . 7554/eLife . 06734 . 013Figure 5—figure supplement 2 . Regulation of Atg14L by GPCR mediated signaling pathways . ( A ) Quantification for Figure 5F . HeLa cells were transfected with control siRNA , Gαq/11 , GαS and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ** ) . ( B ) Quantification for Figure 5G . HeLa cells were transfected with control siRNA , Gα12/13 and cultured for 72 hr . The cell lysates were analyzed by western blotting with indicated antibodies . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ** ) . ( C ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting GαS and cultured for 72 hr . Before harvesting the sample , the cells were treated with or without 10 μM CQ for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( D ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting Gαq/11 and cultured for 72 hr . Before harvesting the sample , the cells were treated with or without 10 μM CQ for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( E ) HeLa cells were transfected with control siRNA ( N . T . ) or siRNA targeting Gα12/13 and cultured for 72 hr . Before harvesting the sample , the cells were treated with or without 10 μM CQ for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( F ) Quantification for Figure 5H . HeLa cells were serum-deprived ( SD ) for 24 hr , and then stimulated with forskolin for 2 hr . The cell lysates were harvested and analyzed by western blotting using indicated antibodies . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) , p < 0 . 001 ( * ) . ( G ) Quantification for Figure 5I . HeLa cells were treated with Gαi/o inhibitor PTX at 1 or 2 μg/ml for 6 hr , and the cell lysates were harvested and analyzed by western blotting using indicated antibodies . The data are expressed as the mean of 3 biological replicates ( mean ± SD ) . Statistical significance was determined by a two-tailed , unpaired Student's t-test . p < 0 . 05 ( * ) ; p < 0 . 01 ( ** ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 013 Heterotrimeric G-proteins , composed of non-identical α , β , and γ subunits , are the critical mediators of GPCR signaling ( Malbon , 2005 ) . To directly test the role of GPCR signaling on Atg14L , we examined the effect of knocking down multiple Gα proteins . Interestingly , knocking down the expression of multiple Gα proteins led to increases in the levels of ATG14L and autophagic flux ( Figure 5F–G; Figure 5—figure supplement 2A–E ) . Thus , we conclude that at least a subset of GPCR signaling regulates autophagy through controlling the levels of ATG14L . The activation of GPCR triggers multiple signal transduction pathways including an increase in the activities of downstream mediators such as Rho-associated coiled-coil containing protein kinase , phospholipases C , which leads to the synthesis of lipid-derived second messengers , activation of adenylyl cyclase , and subsequent activation of protein phosphorylation cascades ( Lappano and Maggiolini , 2011 ) . Activation of these pathways is known to negatively regulate the activity of GSK3β ( Forde and Dale , 2007 ) . We next investigated the impacts of these intracellular signaling pathways on Atg14L . Activation of Gαs by a large group of GPCRs stimulates adenylyl cyclase and production of cAMP ( Malbon , 2005 ) . To directly test the role of cAMP in the induction of Atg14L in serum-starved cells , we treated HeLa cells with forskolin , an activator of adenylyl cyclase , and thereby stimulating the production of cAMP . Interestingly , the treatment of forskolin suppressed the increases of Atg14L under serum starvation condition ( Figure 5H; Figure 5—figure supplement 2F ) . Conversely , we treated HeLa cells with Pertussis toxin ( PTX ) , which inactivates all members of the Gαi family of G proteins , and found that the treatment of PTX led to reduction of ZBTB16 and increases in Atg14L in the presence of serum ( Figure 5I; Figure 5—figure supplement 2G ) . Taken together , these results suggest that ZBTB16 regulated Atg14L degradation is a common downstream mechanism by which multiple signaling pathways that are activated by at least a subset of GPCRs in regulation of autophagy . AMD3100 is a FDA approved GPCR antagonist used in clinics as an immunostimulant to mobilize hematopoietic stem cells in cancer patients ( Donzella et al . , 1998 ) . The treatment of HeLa cells with AMD3100 led to reduction in the levels of ZBTB16 and increases in the levels of Atg14L ( Figure 6—figure supplement 1A ) . Consistent with increases in the levels of Atg14L , AMD3100 treatment also increased the ratio of LC3II and tubulin , which is further stimulated in the presence of CQ , suggesting that AMD3100 promotes autophagic flux ( Figure 6—figure supplement 1B ) . Furthermore , the treatment with AMD3100 reduced the levels of mutant Huntington ( mHTT ) , which can be blocked by CQ , suggesting that AMD3100 can promote the degradation of mutant Htt through lysosomal-dependent autophagic degradation ( Figure 6—figure supplement 1C ) . We next tested the effect of AMD3100 in mice . We found that in mice dosed with AMD3100 , the phosphorylation of AKT ( S473 ) and GSK3 ( S9 ) was reduced , and the levels of Atg14L were increased ( Figure 6A ) as that cells under serum starvation in culture . Corresponding to the activation of GSK3β , the levels of ZBTB16 were decreased , while its phosphorylation was increased ( Figure 6A–B ) . Autophagy was activated in the brains of these mice treated with AMD3100 as the levels of autophagy marker p62 in the brain were reduced , and ratio of LC3II to tubulin was increased ( Figure 6A ) . In addition , we used immunostaining of p62 to measure the levels of autophagy . We found that the levels of p62 in the cortex and striatum of N171-82Q mice treated with vehicle or AMD3100 once daily for one month were significantly lower than that of vehicle-treated mice ( Figure 6—figure supplement 1D ) . These data suggest that pharmacological inhibition of GPCR signaling can lead to down-regulation of ZBTB16 and increases in the levels of Atg14L and up-regulation of autophagy in vivo . 10 . 7554/eLife . 06734 . 014Figure 6 . Up-regulation of Atg14L and autophagy by GPCR antagonist AMD3100 in vivo ameliorated the neural dysfunction of HD transgenic mice . ( A ) WT mice were dosed once intraperitoneally with AMD3100 at 10 mg/kg body weight and 24 hr later , the brain tissues were collected and analyzed by western blotting using indicated antibodies . Anti-Tubulin was used as a loading control . The levels of ZBTB16 , p62 , and LC3-II were quantified as graphs on the right side . ( B ) WT mice were dosed once intraperitoneally with AMD3100 at 10 mg/kg body weight and 24 hr later , the brain tissues were lysed in NP-40 buffer with phosphatase inhibitors and immunoprecipitated ZBTB16 antibody . The immunocomplexes were analyzed by western blotting with phospho-Ser antibody and anti-ZBTB16 antibody . The levels of ZBTB16 phosphorylation were quantified shown as a graph on the right . ( C ) Dosing of AMD3100 , but not saline alone , for two weeks reduced the clasping of the hindlimbs of N171-82Q mice at 13 weeks of age . ( D ) AMD3100 improved motor function of N171-82Q mice as determined by rotarod testing . The mice received training on rotarod for 10 min each day for 3 days before being tested on the first day of 11 week of age before dosing of AMD3100 started ( Day 1 ) . The mice received rotarod training again on day 15–17 during AMD3100 or saline dosing course and were tested again on day 18–21 as shown . For testing , the speed of the rod was set to 5 rpm and increased by 0 . 5 rpm/s . The data were collected as an average of three trials for each mouse everyday . Mice were allowed to rest for 20 min between trials . N = 13 for each group . The data are presented as mean values ± s . e . m . * . p < 0 . 05 by t-test for significance . AMD3100 ( 10 mg/kg body weight ) or saline treatment was delivered by intraperitoneal injection from 11 weeks of age daily ( every 24 hr ) . The experiments were carried out in double-blind manner ( the person who conducted rotarod testing was unaware if the mouse had received saline or AMD3100 , which was provided by another person ) . ( E ) Survival of wild-type ( WT ) , N171-82Q injected with saline ( HD + saline ) , N171-82Q mice injected with AMD3100 ( HD + AMD3100 ) . N = 13 for each group . p < 0 . 05 by log-rank test for significance . The experiments were carried out in double-blind manner ( the person who monitored mouse survival was unaware if the mouse had received saline or AMD3100 , which was provided by another person ) . ( F ) The brain tissues of WT mice or N171-82Q mice dosed for 4 weeks by saline alone or AMD3110 at the age of 15 weeks were isolated and analyzed by western blotting using indicated antibodies . Anti-tubulin was used as a loading control . AMD3100 ( 10 mg/kg body weight/day ) or saline treatment was delivered by intraperitoneal injection from 11 weeks of age every 24 hr . The levels of soluble mHtt were quantified as graphs on the right side . ( G ) Immunostaining of tissue sections from the cortex and striatum of N171-82Q mice treated with saline or AMD3100 daily for 4 weeks using rabbit EM48 . DAPI staining is for nucleus . Quantitative assessment of neuropil aggregate density is as graphs below . ( H ) A model for regulation of autophagy under normal nutritional conditions: inhibiting GPCR signaling leads to the activation of GSK3β , which mediates the phosphorylation of ZBTB16 to promote its auto-ubiquitination and inhibit the ubiquitination and proteasomal degradation of Atg14L that in turn leads to increased activity of class III PI3 kinase and production of PI3P , and activation of autophagy . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 01410 . 7554/eLife . 06734 . 015Figure 6—figure supplement 1 . Up-regulation of Atg14L and autophagy by GPCR antagonist AMD3100 in vivo ameliorated the neural dysfunction of HD transgenic mice . ( A ) HeLa cells were treated with AMD3100 at 10 µM for indicated periods of time , and then the cell lysates were harvested and analyzed by western blotting using indicated antibodies . ( B ) HeLa cells were treated with 10 µM AMD3100 for 4 hr . Before harvesting , the cells were treated with or without 10 µM CQ for 4 hr . The cell lysates were analyzed by western blotting with indicated antibodies . ( C ) HeLa cells were transfected with the expression vectors of GFP-140Q and cultured for 12 hr , and then treated with different doses AMD3100 for another 12 hr . The cells were treated with or without 20 μM CQ for 4 hr before harvesting . The lysates were WB with anti-GFP antibodies to detect soluble mHTT . ( D ) Anti-p62 immunostaining of tissue sections from the cortex and striatum of N171-82Q mice treated with vehicle or AMD3100 daily for 4 weeks . DAPI staining is for nucleus . ( E ) EM48 immunostaining of hippocampal sections from Q171-N82 mice at the age of 15 weeks after 1 month of treatment with vehicle or AMD3100 at 10 mg/kg body weight daily by intraperitoneal injection . DOI: http://dx . doi . org/10 . 7554/eLife . 06734 . 015 To directly examine the impact of activating autophagy by inhibiting GPCR signaling on the accumulation of expanded polyglutamine repeats and disease progression in animal models of expanded polyglutamine disorders , we used N171-82Q transgenic mice , a mouse model of Huntington's disease that expresses the N-terminal 171 amino acids of human htt with 82Q repeat under the control of the prion promoter ( Schilling et al . , 1999 ) . Untreated N171-Q82 mice develop intranuclear expanded polyglutamine inclusions and neuritic aggregates and behavioral abnormalities , including loss of coordination , tremors , hypokinesis , and abnormal gait , before dying prematurely . We found that the treatment of AMD3100 significantly reduced the frequency of hindlimb clasping when suspended by tail ( Figure 6C ) . We monitored the effect of AMD3100 on motor function using rotarod test . The motor deficit in N171-Q82 mice is detectable at 10–11 weeks of age ( Schilling et al . , 1999 ) . Beginning at 11 weeks of age , AMD3100 or vehicle control was administered daily to HD mice until the end of their life . When tested at 13 . 5 weeks of age , the N171-Q82 mice received AMD3100 showed significantly improved motor function as indicated by the extended time on rotarod ( Figure 6D ) . The N171-Q82 mice received AMD3100 daily survived significantly longer than that of control group ( 119 ± 3 . 82 days vs 104 ± 4 . 63 days . p < 0 . 05 ) ( Figure 6E ) . We characterized the accumulation of expanded polyglutamine in vehicle and AMD3100 treated N171-Q82 mice by western blotting and immunostaining . We found that the levels of expanded polyglutamine in brain tissues of N171-82Q mice at the age of 15 weeks were dramatically reduced compared to that of vehicle-treated mice as shown by western blotting ( Figure 6F ) . Furthermore , the accumulation of expanded polyglutamine in the cortex , striatum , and hippocampus was significantly reduced after the treatment of AMD3100 as shown by immunostaining ( Figure 6G and Figure 6—figure supplement 1E ) . Taken together , we conclude that the activation of autophagy by inhibiting GPCR signaling can promote the degradation of expanded polyQ and preserve neuronal functions by inhibiting Atg14L degradation through ZBTB16-mediated ubiquitination and proteasomal degradation ( Figure 6H ) . Our study reveals a novel mechanism by which GPCRs and the associated intracellular signaling mediators regulate autophagy by controlling the ubiquitination and proteasomal degradation of Atg14L mediated by ZBTB16-CUL3-ROC1 . Our study provides a mechanism of crosstalk between autophagy and proteasome , two major intracellular degradative machineries . Since GPCRs represent the largest family of membrane-bound receptors that can interact with a broad range of ligands , including small organic compounds , eicosanioids , peptides , and proteins ( Lagerström and Schiöth , 2008 ) , our results suggest that autophagy can be regulated by a much wider range of extracellular signals and conditions beyond starvation condition than currently appreciated . Furthermore , we demonstrate as a proof-of-principle that suppression of GPCR signaling by AMD3100 can activate autophagy to reduce the accumulation of expanded polyglutamine repeats , extend survival , and ameliorate neurodysfunction in a mouse model of Huntington's disease . Since our study suggests that ZBTB16-regulated Atg14L ubiquitination and degradation is a common downstream mechanism modulated by GPCR signaling in control of autophagy , our study suggests the possibility to explore regulators of additional GPCRs to manipulate autophagy in the CNS . Since GPCRs represent the most common ‘druggable’ targets , our results demonstrate a potentially wide range of possible targets for safely modulating the levels of autophagy through pharmacological manipulations of GPCR signaling . Our data , however , do not rule out that GPCRs might regulate additional pathways in control of autophagy; nor do we exclude the possibility that ZBTB16 has substrates , in addition to Atg14L , in regulating autophagy . The phosphorylation of targeting substrates often serves as a signal and prelude for binding , ubiquitination , and proteasomal degradation mediated by Cullin-RING ligase complexes ( Petroski and Deshaies , 2005 ) . In this case , we found that phosphorylation of S184/T282 in the center domain of ZBTB16 by GSK3β may provide a signal to disrupt the inhibitory folding and thereby activate its auto-ubiquitination activity . On the other hand , the phosphorylation of S184/T282 , which is in the domains of ZBTB16 involved in binding to Atg14L , inhibits its interaction with Atg14L and therefore , reduces the ubiquitination and degradation of Atg14L . Since the interaction of ZBTB16 and Atg14L is mediated by the binding to the BATS domain , a lipid-binding domain that binds to highly curved autophagosome structures enriched with PI3P ( Matsunaga et al . , 2010; Fan et al . , 2011 ) . Consequently , the interaction of ZBTB16 with Atg14L is potentially competitive with the role of Atg14L in promoting the formation of autophagosomes , consistent with its negative role in regulating autophagy . Thus , the degradation of ZBTB16 upon inhibition of GPCR signaling may not only release Atg14L from proteasomal degradation but also provide a permissive signal to allow the binding of the BATS domain with PI3P , an important event during the formation of autophagosomes . From Sigma Aldrich: LY294002 , MG132 , E64d , SB216763 , Forskolin , LPA , endothelin 1 , AMD3100 , U0126 , BSA , Rapamycin , cycloheximide . From Gibco: Pertussis Toxin . Recombinant proteins from Boston Biochem: Ubiquitin Activating Enzyme ( UBE1 , cat . E-305 ) , UbcH5c ( cat . E2-627 ) , Ubiquitin ( cat . U-100H ) . From ProSpec: human SDF1 . HEK293T , HeLa , human neuroglioma , H4 cells were cultured in Dulbecco's modified eagle medium ( DMEM ) supplemented with 10% FBS and 1% penicillin-streptomycin ( Gibco-BRL ) . For serum starvation , cells were incubated in DMEM without FBS . Cells were transfected with plasmid DNA using PolyJet DNA In Vitro Transfection Reagent ( Signagen Laboratories , Rockville , MD ) . siRNA transient transfections were performed using Lipofectamine 2000 ( Invitrogen ) . Anti-ATG14L ( MBL , PD026 , Lot . 003 ) , anti-ZBTB16 ( ab39354 , abcam ) , anti-Beclin1 ( sc-H-300 , Santa Cruz ) , anti-Vps34 ( 12452-1-AP , PTG ) , anti LC3B ( L7543 , Sigma ) , anti-P62 ( PM045 , MBL ) , anti-P-GSK3β ( ser9 ) ( #9323 , CST ) , anti-GSK3β ( #9832S , CST ) , anti-p44/42 MAPK kinase ( #9102 , CST ) , anti-p-p44/42 MAPK ( T202/Y204 ) ( #9101S , CST ) , anti-pAKT ( S473 ) ( #4060S , CST ) , anti-AKT ( sc-H136 , Santa Cruz ) , anti-Tubulin ( PM054 , MBL ) , anti-Myc ( M4439 , Sigma ) , anti-Flag M2 ( F1804 , Sigma ) , anti-Cullin3 ( ab75851 , abcam ) , anti-Xpress ( R910-25 , Invitrogen ) , anti-phosphoserine ( 44911M , Invitrogen ) , anti-p-Threonine-proline ( #9391S , CST ) , Gαq/11 antibody ( C-19 ) ( sc-392 , Santa Cruz ) , Gαs antibody ( A-16 ) ( sc-26766 , Santa Cruz ) , anti-Gα13 ( sc-410 , Santa Cruz ) , anti-Ubiquitin ( Z0458 , Dako ) , mouse EM48 ( MAB5374 , Millipore ) . Cells were lysed with NP-40 buffer ( 10 mM Tris-HCl , pH 7 . 5 , 150 mM NaCl , 0 . 5% NP-40 , protease inhibitors cocktail [Sigma] , 5% glycerol , 10 mM NaF , 1 mM PMSF ) , or Buffer II ( 1 mM EDTA , 0 . 1% NP-40 , 10 mM Tris-HCl pH 7 . 5 ) . Whole cell lysates obtained by centrifugation were incubated with 5 mg of antibody and protein G agarose beads ( Invitrogen ) overnight at 4°C . The immunocomplexes were then washed with NP-40 buffer for 3 times and separated by SDS-PAGE for further western blotting assay . Ubiquitination detection was conducted on fully denatured proteins . The cells were lysed with a 1% SDS lysis buffer and boiled for 15 min . Denatured lysates were then diluted with 0 . 5% NP-40 lysis buffer and immunoprecipitated with appropriate antibody . Myc-Cullin3 , FLAG-ZBTB16 , HA-ROC1 , Myc-ATG14 proteins were immunopurified from 293T cells transfected with the expression vectors encoding above proteins individually . The eluted proteins were incubated with recombinant E1 ( 60 ng ) , E2-Ubc5c ( 300 ng ) , and ubiquitin ( 1 μg ) in an in vitro ubiquitin ligation reaction containing 50 mM Tris-HCl pH 7 . 4 , 5 mM MgCl2 , 10 nM okadaic acid , 2 mM ATP , 0 . 6 mM Dithiothreitol ( DTT ) , 1 μg ubiquitin ( final volume 30 μl ) . The reactions were incubated at 37°C for 60 min , terminated by boiling for 5 min in SDS sample buffer . In vitro kinase reactions were performed using Flag-ZBTB16 proteins immunoprecipitated from HEK293T cells as substrates in kinase reaction buffer ( NEB ) , 200 mM ATP , and 500 units of GSK3 enzyme ( NEB p6040S ) at 30°C for 1 hr . Reactions were terminated by addition of SDS sample buffer . HeLa cells were washed with Phosphate Buffered Saline ( PBS ) in 4°C and incubated with Buffer A ( HEPES [pH 7 . 9] , 20 mM; 10 mM KCl , 1 mM EDTA , 0 . 5 mM PMSF , 1× protease inhibitor cocktail; 1 mM DTT , 5 mM NaF , 0 . 1 mM Na3VO4 , 10% Glycerol ) at 4°C for 10 min . The cell lysates were collected into cold 1 . 5-ml EP tube , and NP-40 was added to the final concentration of 0 . 5% . After a brief vortexing and incubation in ice ( 1 min ) , the lysates were centrifuged at 12 , 500 rpm for 3 min . The supernatant was collected as the cytoplasmic fraction . 50 ml Buffer B ( HEPES [pH 7 . 9] , 20 mM; 10 mM KCl , 1 mM EDTA , 0 . 5 mM PMSF , 1× protease inhibitor cocktail , 1 mM DTT , 5 mM NaF , 0 . 1 mM Na3 VO4 , 420 mM NaCl , 10% Glycerol ) was added into to the sediment . The mixture was mixed by vortex and incubated on ice for 40 min before centrifuged at 12 , 500 rpm for 3 min . The supernatant was collected as the nuclear fraction . Total RNA was prepared using RNeasy mini kit ( QIAGEN ) . RNA ( 1 . 25 mg ) was used for cDNA synthesis using SuperScript First-Strand Synthesis System for Reverse transcription polymerase chain reaction ( RT-PCR ) ( Invitrogen ) with oligo dT primers . Mutagenesis was performed using Quik-Change mutagenesis kit ( Stratagene ) . cDNAs for ZBTB16 , ATG14L , and their deletion mutants were cloned into pcDNA3 . 1 using ClonExpressTM II cloning kit ( Vazyme biotech ) . The specific primers were: Vps34 , 5′-GAACTTATCCCGTTGCCTTTA-3′ ( forward ) , and 5′-CATGACCTCAGCACTAATCCC-3′ ( reverse ) . Beclin1 , 5′-CCGCAAGATAGTGGCAGAAA-3′ ( forward ) , and GCGACCCAGCCTGAAGTTAT-3′ ( reverse ) . Atg14L , 5′-GCTGGTCAACATTCTGTCTCA-3′ ( forward ) , and 5′-CTCCTCAAGGTCTGCTCGTAC-3′ ( reverse ) . Actin , 5′-AGCGAGCATCCCCCAAAGTT-3′ ( forward ) , and 5′-GGGCACGAAGGCTCATCATT-3′ ( reverse ) . The sequences of siRNA Oligos used: 5′ to 3′:Non-target:UUCUCCGAACGUGUCACGUZBTB16 #1GGGUCGAGCUUCCUGAUAAZBTB16 #2CUAGGGAGCUACACUAUGGZBTB16 #3AGAAGCAUCUGGGCAUCUAGNAS#1GCUUGCUUAGAUGUUCCAAAUGNAS#2GCCAAGUACUUCAUUCGAGAUGNAQGACACCGAGAAUAUCCGCUUUGNA11#1GCUCAAGAUCCUCUACAAGUAGNA11#2GCUCAACCUCAAGGAGUACAAGNA12:CGUCAACAACAAGCUCUUCUUGNA13:GCUCGAGAGAAGCUUCAUAUUGSK3 β#1CCCAAAUGUCAAACUACCAGSK3 β#2AGUUGGUAGAAAUAAUCAAGSK3 β#3GCUAGAUCACUGUAACAUAAtg14L#1CCGGGAGAGGUUUAUCGACAAGAAtg14L#2AUCUUCGACGAUCCCAUAUAUUACUL3#1GUCGUAGACAGAGGCGCAACUL3#2GAAGGAAUGUUUAGGGAUA N171-82Q mice ( B6C3F1/J-Tg ( HD82Gln ) 81Dbo/J , Jackson Laboratory , Bar Harbour , ME ) were maintained in the animal facility in accordance with the institutional guidelines . Mice received daily intraperitoneal injection of AMD3100 ( 10 mg/kg of body weight ) or saline only . Mice were sacrificed rapidly by cervical dislocation , and the brains were harvested and immediately frozen in liquid nitrogen . Samples were stored at −80°C until use . Frozen tissues were pulverized in liquid nitrogen . All of the subsequent steps were performed at 4°C . Powdered tissue samples were homogenized in 10 vol ( wt/vol ) of buffer containing 50 mM Tris-HCl pH 7 . 6 , 10 mM EDTA , 100 mM NaF , Protease inhibitor cocktail 1 mM PMSF , 1 mM Na3VO4 , 20 mM beta-glycerophosphate , 1 mM sodium pyrophosphate . The samples were then used for western blotting analysis . Mice were trained on rotarod for 10 min each day for 3 days before examination . For testing , the speed of the rod was set to 5 rpm and increased by 0 . 5 rpm/s . Mice received 3 trials daily over 4 consecutive days . Mice were allowed to rest for 20 min between trials . Data from each group were averaged and charted using SPSS software . Mice were anaesthetized with 5% chloral hydrate and perfused intracardially with phosphate-buffered saline ( PBS , pH 7 . 2 ) followed by 4% paraformaldehyde in PBS at pH 7 . 2 . Brains were removed , cryoprotected in 30% sucrose at 4°C for 3 days , and then sectioned at 10 μm using a cryostat ( Leica ) at −20°C . The slices were examined by immunofluorescent staining with rabbit EM48 and secondary antibodies conjugated with FITC ( Jackson ImmunoResearch Laboratories ) or Alexa Fluor 555 ( Invitrogen ) . Statistical analysis was performed on biological repeats of three independent sets of experiments using ImageJ . The protein levels are determined as a ratio between protein of interest and tubulin . Results were analyzed for statistical significance using two-tailed , unpaired Student's t-test and are expressed as mean ± SD for western blots and mean ± sem for microscopic images . p < 0 . 05 ( * ) ; p < 0 . 01 ( ) ; p < 0 . 001 ( * ) . ns , no significance .	Proteins need to be folded into specific three-dimensional shapes for them to work properly . However , the folding process does not always work perfectly , and proteins are sometimes misfolded . If left to accumulate , these misfolded proteins can damage cells , and most long-term human neurodegenerative diseases , such as Huntington's disease , Parkinson's disease , and Alzheimer's disease , are caused by the build-up of misfolded proteins in the brain . Autophagy helps to clean up misfolded proteins ( and other damaged cell components ) by first wrapping them in membrane vesicles . The membrane-wrapped vesicles—known as autophagosomes—then move to fuse with lysosomes , a different kind of membrane compartment in the cell , which breaks down misfolded proteins and recycles the degradation products . In mammalian cells , a protein called Atg14L is critical in the process of autophagosome formation . The levels of autophagosome formation are regulated by signals that originate from outside the cell . However , it is not clear if and how cells respond to external signals to control the levels of autophagy by regulating the amount of Atg14L . The G-protein-coupled receptors ( GPCRs ) are the largest class of membrane proteins that our cells have that are involved in sensing and responding to external signals . The activation of GPCRs has been shown to lead to diverse physiological responses . Zhang et al . now show that when any of a wide range of different signaling molecules bind to the GPCRs , the receptors activate a protein called ZBTB16 that leads to the degradation of Atg14L to inhibit autophagy . Furthermore , Zhang et al . found that blocking the activity of the GPCRs with a drug can activate autophagy and reduce the amount of misfolded proteins in the cell . In mice that have a version of a gene that causes Huntington's disease , this inhibition also protects against the symptoms of the disease . The challenge now is to identify appropriate GPCRs that can be safely manipulated to control the levels of autophagy in the brain in order to reduce the levels of the misfolded proteins that cause neurodegeneration .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2015	G-protein-coupled receptors regulate autophagy by ZBTB16-mediated ubiquitination and proteasomal degradation of Atg14L
Regulation of organ growth is a poorly understood process . In the long bones , the growth plates ( GPs ) drive elongation by generating a scaffold progressively replaced by bone . Although studies have focused on intrinsic GP regulation , classic and recent experiments suggest that local signals also modulate GP function . We devised a genetic mouse model to study extrinsic long bone growth modulation , in which injury is specifically induced in the left hindlimb , such that the right hindlimb serves as an internal control . Remarkably , when only mesenchyme cells surrounding postnatal GPs were killed , left bone growth was nevertheless reduced . GP signaling was impaired by altered paracrine signals from the knee joint , including activation of the injury response and , in neonates , dampened IGF1 production . Importantly , only the combined prevention of both responses rescued neonatal growth . Thus , we identified signals from the knee joint that modulate bone growth and could underlie establishment of body proportions . The signaling pathways that underlie the control of organ size during development , and recovery from injury , have important implications for the treatment of growth disorders . The same pathways likely also underlie evolutionary differences in body sizes and proportions ( Haldane , 1926 ) . Inter-species transplantation experiments have shown that growth of most organs , including the limbs , eyes ( Twitty and Schwind , 1931 ) and jaws ( Schneider , 2015 ) , to a great extent follows an intrinsic genetic program , as the final size of the organ graft is close to that of the donor . However , extrinsic factors such as nutrient availability , inflammation , mechanical forces or the presence of additional organ grafts can modulate , to varying extents , growth of many organs ( reviewed in [Roselló-Díez and Joyner , 2015] ) . Therefore , interactions between organ-intrinsic and extrinsic cues must be critical to the regulation of individual organ size , relative body proportions and responses to developmental insults ( Stanger , 2008; Parker , 2011; Mirth and Shingleton , 2012 ) . The developing vertebrate limb is an excellent model to study pathways controlling organ growth , as it is amenable to non-lethal manipulation , and it is composed of multiple tissues ( muscles , tendons , bone , nerves , vessels , etc . ) that grow coordinately and are exposed to biochemical and mechanical extrinsic signals that interact with their intrinsic genetic growth programs . Long bones within the limbs have been popular models for studies of patterning , growth and evolution since at least the nineteenth century ( Owen , 1849 ) , mostly as isolated entities , although lately also as part of the musculo-skeletal unit ( Berendsen and Olsen , 2015; Shwartz et al . , 2013 ) . Long bone growth occurs through endochondral ossification , driven by a transient cartilage structure , the growth plate ( GP ) , located at both ends ( epiphyses ) of the bone ( reviewed in [Kronenberg , 2003] ) . Once mesenchymal cells condense into the anlage of the skeletal elements , they differentiate into collagen II-expressing chondrocytes that undergo sequential differentiation from bone ends to center . Resting ( quiescent ) chondrocytes produce proliferative cells that after a few rounds of duplication cease proliferation and start to differentiate into hypertrophic chondrocytes ( HTCs ) . HTCs increase their volume as they progress towards the center of the bone , laying down a collagen X-rich extracellular matrix ( ECM ) and secreting factors that recruit vasculature and bone precursors ( osteoblasts ) into the skeletal element . Some HTCs die by apoptosis , while others transdifferentiate into osteoblasts ( Yang et al . , 2014; Zhou et al . , 2014 ) , which also derive from a fibrous layer that wraps the cartilage ( perichondrium ) ( Maes et al . , 2010; Kronenberg , 2007 ) , and both osteoblast pools form the primary ossification center by replacing the cartilaginous ECM with bone . This invasion of osteoblasts , blood vessels and ossification process is later recapitulated in the center of both epiphyseal regions , giving rise to the secondary ossification centers . The GP remains as a cartilage disc between the primary and secondary ossification centers , and responds to both intrinsic and extrinsic factors that regulate bone length . Within the GP , a negative feedback loop between indian hedgehog ( IHH ) , secreted by pre-hypertrophic chondrocytes , and parathyroid hormone-like peptide ( PTHLH or PTHrP ) secreted by resting chondrocytes , couples chondrocyte proliferation and differentiation ( Karp et al . , 2000; Vortkamp et al . , 1996; Lee et al . , 1996; Long et al . , 2001 ) , and is the main conduit through which other local signals , such as fibroblast growth factors ( FGFs ) and bone morphogenetic proteins ( BMPs ) exert their function ( Roselló-Díez and Joyner , 2015 ) . A number of systemic signals ( hormones ) have long been known to impact on bone growth ( Roselló-Díez and Joyner , 2015 ) , but the contribution of local signals extrinsic to the GP has been explored less . Interestingly , the GPs at each end of the long bones contribute differentially to final bone size ( Digby , 1916; Payton , 1932; Moss-Salentijn , 1974 ) , raising the possibility that each has a different intrinsic growth potential , and/or that each is exposed to distinct local environmental cues . Indeed , while there has been an emphasis on intrinsic regulators of bone growth potential ( Nilsson and Baron , 2004 ) , classic GP transplantation experiments revealed that the local environment can modify the amount of growth produced by a GP ( Moss-Salentijn , 1974; Röhlig , 1969; Hert , 1964 ) . The identities of such local extrinsic signal ( s ) , however , have only recently begun to emerge . For example , TGFβR2 signaling from the interzone ( the embryonic precursor of the joint [Pacifici et al . , 2006] ) affects maturation and signaling in the hypertrophic zone ( HZ ) ( Longobardi et al . , 2012 ) ; while some human variants located near the gene GDF5 , expressed in the interzone , are associated with decreased height and increased osteoarthritic risk ( Sanna et al . , 2008; Wu et al . , 2012 ) . Another important player is WNT signaling from the interzone , as its balance with BMP signaling from the GP controls the differentiation of chondrogenic precursors towards either GP ( transient cartilage ) or articular ( permanent cartilage ) fate ( Ray et al . , 2015 ) . In addition , a well-known modulator of bone growth is the Growth Hormone/Insulin-like growth factor ( GH/IGF ) axis , whereby GH induces Igf1 expression in several tissues , including the liver and the GP , but also plays an IGF1-independent role in postnatal growth ( Isgaard et al . , 1988; Lupu et al . , 2001 ) . Importantly , the accumulated evidence suggests that local rather than systemic ( i . e . liver-derived ) IGF1 is the main factor controlling appendicular ( limb ) skeletal growth at early postnatal stages . For example , while genetic deletion of Igf1 in the whole embryo affects skeletal growth from late gestation onwards ( Baker et al . , 1993 ) , mice mutant for growth hormone receptor ( Ghr ) do not show a body weight or a limb length phenotype until postnatal day ( P ) 10–15 ( Lupu et al . , 2001; Zhou et al . , 1997 ) , and GH-induced , liver-specific IGF1 only begins exerting postnatal hormonal action shortly after 3 weeks of age in mice ( Stratikopoulos et al . , 2008 ) . Similarly , deletion of Igf1 in the liver ( which diminishes circulating IGF1 levels by >75% ) does not affect appendicular skeletal growth ( Sjögren et al . , 1999; Yakar et al . , 1999 ) , whereas chronic overexpression of Igf1 in liver , brain and other organs ( leading to increased serum IGF1 levels ) results in overgrowth of only a subset of organs that does not include the skeleton ( Mathews et al . , 1988 ) . Regarding local IGF1 sources , the expression of Igf1 is much lower ( if at all present ) in the GP than in the surrounding tissues ( [Parker et al . , 2007] and our data ) , raising the question of whether the main local role of IGF1 is autocrine or paracrine . The fact that deletion of Igf1 in the whole limb mesenchyme greatly diminishes chondrocyte hypertrophy by P7 ( Cooper et al . , 2013 ) , whereas specific Igf1 deletion in the chondrocyte lineage ( including osteoblasts ) affects bone mass accretion but not bone length by two weeks of age ( Govoni et al . , 2007 ) , strongly suggests that the main source of IGF1 affecting the perinatal GP is the local soft tissues . Interestingly , one of the main downstream targets of insulin/IGF signaling , mechanistic target of rapamycin ( mTOR ) , also plays a key role in chondrocyte hypertrophy , but it is not clear what are the sources and identities of the upstream molecules that activate this pathway in chondrocytes of the GP ( Chen and Long , 2014; Lai et al . , 2013; Phornphutkul et al . , 2008; Srinivas et al . , 2009 ) . Probably even more complex than growth regulation during normal development is growth regulation after an injury in a growing organ , as the deleterious effects of the injury have to be resolved while the rest of the body is growing . Tissue injury triggers activation of a specific group of genes characterized by a very quick response to signals such as cellular stress ( Bahrami and Drabløs , 2016 ) . Such genes are often referred to as immediate-early genes . Tissue injury often also triggers an inflammatory response characterized by immune cell recruitment and release of inflammatory cytokines , which have been shown to impair organ growth , including bones ( MacRae et al . , 2006; Mårtensson et al . , 2004 ) . Elucidating the mechanisms by which local signals extrinsic to the GP modulate bone growth will be relevant to understanding regional differences in bone growth , both within and between species , and also for developing non-invasive therapies for congenital or acquired growth disorders . A major obstacle for exploring local bone growth regulation is a lack of animal models in which only cells outside the GP are manipulated . We show here that transient unilateral induction of cell death in the soft tissues of the postnatal limb impairs GP function and bone growth , leading to left-right limb length inequality . The growth defect is associated with reduced chondrocyte proliferation and hypertrophy , due to two paracrine signaling branches from the knee joint: ( 1 ) Injury-induced local inflammation reduces IGF1 in the knee fat pad , impairing mTOR and IHH signaling in the GP . Interestingly , this branch is active in neonates but not older mice with a secondary ossification center . ( 2 ) The injury response cascade is activated in the joint and prospective articular cartilage , affecting multiple signaling pathways in the adjacent resting zone ( RZ ) of the GP . Both IGF1 loss and the injury response pathway have a causative role in the reduction of bone growth , as combined maintenance of IGF1 expression and inhibition of the injury response rescues bone growth following cell death outside the neonatal GP , whereas either treatment alone does not . Our study reveals that neonatal bone growth is modulated by extrinsic signals from the fat pad and the articular cartilage , opening new avenues for developmental and evolutionary studies , and for the correction of growth defects in humans . As a means to study local bone grow control mechanisms , we developed a genetic model in mice in which transient cell death is induced specifically in the mesenchyme of the left limb , allowing the right limb to be an internal control ( Figure 1A ) . A transgenic mouse line that drives Cre expression in the left lateral plate mesoderm under the control of the Pitx2 asymmetric enhancer ( Shiratori et al . , 2006 ) was combined with an R26LSL-DTR strain that expresses diphtheria toxin receptor ( DTR ) following Cre-induced recombination ( Buch et al . , 2005 ) ( Pit::DTR mice ) . As expected , Pit::DTR mice expressed DTR primarily in the left but not right limb mesenchyme , excluding muscles and blood vessels ( Figure 1A’; Figure 1—figure supplement 1A–B ) . To model an acute injury at birth , a single systemic injection of DT was administered at postnatal day ( P ) 1 ( P1-Pit::DTR model ) , resulting in cell death in the left hindlimb mesenchyme between 1 and 3 days post-injection ( dpi ) , with little in the right limb ( Figure 1B and not shown ) . Unexpectedly , however , cell death was not detected in the left GPs of the tibia and femur of P1-Pit::DTR mice , and the front of apoptotic cells only reached the outermost layers of the prospective articular cartilage and perichondrium from P2 to P5 ( Figure 1B–B’’; the proportion of cells undergoing cell death in the perichondrial groove of Ranvier was 18 ± 7% at P2 , n = 6 ) . It is likely that DT ( a ~60 kDa molecule ) cannot diffuse efficiently into the postnatal GP , which is avascular ( Figure 1—figure supplement 1D ) and a very effective barrier to the diffusion of molecules larger than 10 kDa ( Williams et al . , 2007 ) . To rule out an abnormal response of chondrocytes to DT , we cultured Pit::DTR tibiae from P1 mice in various concentrations of DT ( 5 , 50 and 500 ng/ml ) for 24 hr , to test whether a very high concentration of DT could overcome the potential diffusion barrier and ablate GP chondrocytes . We found that at the highest concentration of DT , cell death was triggered in some cells in the core of the GP ( Figure 1—figure supplement 1E ) , confirming that postnatal chondrocytes can respond to DT . 10 . 7554/eLife . 27210 . 003Figure 1 . Postnatal growth of the left long bones is impaired following transient left-specific cell death in the adjacent mesenchyme . ( A ) Pit::DTR model for the induction of transient unilateral cell death in the limb mesenchyme . See Figure 1—figure supplement 1 for a characterization of the Pitx2-Cre lineage . LSL= loxSTOPlox . ( A’ ) Sagittal sections of Pit::DTR knees immunostained for DTR ( approximate region indicated by the boxes in ( A ) . RZ , PZ , HZ= resting , proliferative , hypertrophic zone . IPFP= infrapatellar fat pad . ( B–B’’ ) Time course of cell death induction ( TUNEL , arrowheads ) after DT injection at P1 . Insets show details at the knee joint ( pAC= prospective articular cartilage ) . The cytoplasmic signal in ( B’ ) is unspecific background . ( C ) Skeletal preparation ( P1-Pit::DTR tibiae ) and plot of left/right length ratio for P1-Pit::DTR or control femur and tibia from mice collected as indicated . Analysis was done by 2-way ANOVA ( alpha = 0 . 05 , Bone Identity and Genotype as variables , p-value for Genotype was 0 . 0177 at 3dpi , < 0 . 0001 at 63-69dpi ) followed by Sidak’s posthoc multiple comparisons test ( p-values shown in Figure ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 00310 . 7554/eLife . 27210 . 004Figure 1—figure supplement 1 . Characterization of the Pitx2-Cre derived cell lineage . ( A–C ) Immunohistochemical detection of DTR in left ( A ) and right ( B ) hindlimbs ( LHL and RHL ) of Pitx2-Cre/+; R26LSL-DTR/+ ( Pit::DTR ) mice at E13 . 5 and in distal radii ( LDR and RDR ) at P4 ( C ) . Note that DTR expression in the forelimb ( arrowheads ) is much more sparse than in the hindlimbs . n = 2 for each stage . ( D , D’ ) CD31 and DTR double staining in the knee area at P1 reveals that blood vessels and pericytes do not express DTR [n = 3 , inset in ( D ) is magnified and channel-split in ( D’ ) ] . ( E ) Histochemical analysis of cell death ( TUNEL ) in frontal sections of P1-Pit::DTR tibiae cultured 24 hr with the indicated concentrations of DT ( n = 3 for each concentration ) . Arrowheads= apoptotic chondrocytes in the core of the GP . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 004 Interestingly , despite sparing of the GP in the P1-Pit::DTR model , cell death in the surrounding tissues resulted in ~5% ( 0 . 3mm ) length difference between the left and right femora and tibiae at three dpi that was not observed in control animals ( either DT-injected Pitx2-Cre; R26+/+ or PBS-injected Pit::DTR mice ) , and that plateaued at ~11% ( ~2 mm ) after 9 weeks ( Figure 1C , n = 4 control and 9 experimental mice at 3dpi , n = 16 and 7 respectively at 63-69dpi ) . The forelimbs were not affected , as they express little DTR ( Figure 1—figure supplement 1C ) . The left specific reduction in bone growth took place on top of a transient systemic growth reduction with respect to control littermates , both in body weight and length of the right long bones ( Figure 2—figure supplement 1A–B ) . The systemic growth reduction was likely caused by cell death in a region of the heart where Pitx2-Cre drives DTR expression ( Shiratori et al . , 2006 ) ( Figure 2—figure supplement 1C–C’ ) , as we observed that P1-Pit::DTR animals were lethargic for ~4 days after DT injection , including severe ( sometimes lethal ) hypophagia . As a consequence , the GPs of P1-Pit::DTR mice at P5 were smaller than in control animals ( Figure 2—figure supplement 1D–E’ ) . At the molecular level , one difference found between control and P1-Pit::DTR mice at P5 was a decrease in the synthesis of ECM in both the left and right GPs of P1-Pit::DTR mice , as Agcn1 and Col2a1 expression was downregulated compared to controls ( Figure 2—figure supplement 1F ) . After 4dpi , however , experimental mice resumed normal feeding and started catching up with the controls . Importantly , by the end of the longitudinal growth period , the lengths of the right but not the left long bones had normalized in P1-Pit::DTR mice ( Figure 2—figure supplement 1B ) , demonstrating the left-specific growth reduction is independent of the transient systemic effect . It is worth noting that unwanted recombination , as observed in the Pit::DTR heart , is a common caveat of most limb-targeting Cre lines , such as Prrx1-Cre and Hoxb6-Cre , which target regions of the head and gut ( see JAX005584 , 017981 and references therein ) . Unlike the bilateral approaches , however , our unilateral strategy allowed us to distinguish local and systemic effects by using the right limb as an internal control . In conclusion , the specific growth defect in the left hindlimb bones of P1-Pit::DTR mice raised the interesting possibility that the injured tissues adjacent to the GP negatively modulate bone growth via a paracrine ( i . e . extrinsic ) mechanism . As differential bone growth depends to a great extent on changes in chondrocyte proliferation and hypertrophy ( Wilsman et al . , 1996 ) , we first tested whether these processes were altered in the left GP of P1-Pit::DTR animals . Indeed , the left/right ratio of EdU incorporation in the proliferative zone ( PZ ) of the proximal tibial GPs was reduced in the P1-Pit::DTR model compared to control animals at P3 ( Figure 2A , n = 4 control and 7 experimental animals ) . Moreover , at P4-P5 we observed an obvious decrease in the height of the HZ but not the PZ ( Figure 2B ) , as further revealed by in situ hybridization for the chondrocyte maturation marker Col10a1 ( Figure 2C , n = 3 ) and hematoxylin and eosin staining ( Figure 2D , n = 4 control , 4 experimental ) . In addition , the formation and expansion of the secondary ossification centers in the left bones of P1-Pit::DTR mice at P4-5 was slightly delayed and progressed more slowly as compared to the right bones ( Figure 2E–F , n = 5 ) . Thus , not only proliferation and differentiation of chondrocytes , but also maturation of the whole skeletal unit , are impaired in the left hindlimb skeletal elements of P1-Pit::DTR animals as a consequence of transient cell death in the surrounding tissues . 10 . 7554/eLife . 27210 . 005Figure 2 . Reduced chondrocyte proliferation and maturation underlie the growth defect of the left bones of P1-Pit::DTR mice . ( A ) The fraction of EdU+ nuclei in the PZ was calculated for left and right proximal tibia GPs 2dpi and represented as L/R ratio ( mean ± SD ) . p-value for unpaired two-tailed Mann-Whitney test between control and experimental ratios is shown . ( B , D ) The PZ and HZ of left and right GPs were measured on hematoxylin and eosin-stained sections of P1-Pit::DTR proximal tibia GPs , 4dpi ( D ) , and represented as L/R ratios ( B ) 3–4 sections per GP , n = 4 mice , mean ± SD are shown ) . 2-way ANOVA ( variables: Genotype and GP region , alpha = 0 . 05 , p-value=0 . 0118 for Genotype ) followed by Sidak’s posthoc multiple comparisons test ( p-values shown ) was used . Insets in ( D ) show magnifications of the boxed regions . ( C ) RNA in situ hybridization for Col10a1 in the proximal tibia GP at 3dpi . ( E , F ) The secondary ossification centers ( SOCs , arrowheads in E ) appear later and their subsequent area ( quantified in F ) is reduced in the left P1-Pit::DTR skeletal elements . Analysis was done by 2-way ANOVA ( variables: SOC location and Side , alpha = 0 . 05 , p-value=0 . 0003 for Side ) followed by Sidak’s posthoc multiple comparisons test ( p-values shown in Figure ) . Asterisk= reduced expression . See also associated Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 00510 . 7554/eLife . 27210 . 006Figure 2—figure supplement 1 . Bone growth impairment in Pit::DTR animals takes place on top of a systemic growth delay likely caused by injury-induced hypophagia . ( A ) Comparison of body weight ( mean ± SD ) between control and P1-Pit::DTR pups at P5 ( n = 18 controls , five experimental ) . ( B ) Relative length of the right tibia from control ( black ) , DT-injected ( red ) and PBS-injected ( green ) Pit::DTR animals ( mean ± SD , n = 4 , 2 and 7 , respectively , at early stages; 9 , 7 and 6 long-term ) , normalized against the average length of the right tibia in control animals . ( C and C’ ) DTR expression and TUNEL staining ( arrowheads ) in a heart section of a P1-Pit::DTR pup , 1dpi ( n = 3 ) . A , V= atrium , ventricle . Inset ( C’ ) shows magnification of the boxed area in ( C ) . ( D–E’ ) Hematoxylin and eosin staining of the right GPs ( internal control ) from control and experimental animals at P5 ( n = 4 control and four experimental mice ) . Note the striking difference in size , including the HZ ( D’ and E’ ) . ( F ) In situ hybridization for the indicated ECM components in control and experimental animals ( n = 2 animals for each marker and genotype ) , showing the systemic delay affects ECM synthesis equally in left and right proximal tibia GPs of experimental animals . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 006 We first confirmed that changes in intrinsic growth regulation mechanisms were not the origin of the bone growth impairment in the P1-Pit::DTR model , by examining well-known signaling pathways . Importantly , HTCs did not display increased EGFR signaling ( Figure 3—figure supplement 1A ) , which can be activated by DTR shedding ( Xu et al . , 2004 ) and impairs terminal chondrocyte differentiation and cartilage remodeling ( Zhang et al . , 2011; Hall et al . , 2013 ) . In addition , the potential ablation of osteoprogenitors in the perichondrium of P1-Pit::DTR animals was an unlikely cause of the reduced bone growth , as most of these cells were not TUNEL+ in our model ( Figure 3—figure supplement 1C–C’ ) . Moreover , a near-complete ablation of osteoprogenitors using an Sp7 ( Osx ) -driven Cre ( Osx::DTR model ) failed to impair longitudinal bone growth , although it did reduce cortical thickness ( Figure 3—figure supplement 1D–F ) . Finally , we did not detect increased cell death in the osteochondral junction that could explain the observed HZ shortening ( Figure 3—figure supplement 1G ) . To probe the link between soft-tissue signals and GP function , we next tested whether the signaling pathways known to affect chondrocyte proliferation and hypertrophy were altered in the left hindlimb GPs of P1-Pit::DTR mice . Since cell death receded by P5 , we focused our analysis on the first days post-injection for the rest of the study . Significantly , we found that Ihh expression was reduced in the left compared to right GPs at P3 , as was the expression of the HH target gene Gli1 ( Figure 3A–A’ , n = 14/16 and 8/10 , respectively ) . We did not , however , observe a change in expression of the IHH target Pthlh ( Figure 3—figure supplement 2A ) . Given that the phosphatase SHP2 negatively modulates Ihh levels in cultured chondrocytes ( Guan et al . , 2014 ) , we treated P1-Pit::DTR pups with an SHP2 inhibitor from P1 to P3 . While the treatment rescued Ihh expression , it did not rescue bone growth ( Figure 3—figure supplement 2B–C’ ) , suggesting that other parallel signaling pathways were disturbed . Since mTOR complex 1 ( mTORC1 ) is implicated in chondrocyte maturation and hypertrophy ( Lai et al . , 2013; Phornphutkul et al . , 2008; Srinivas et al . , 2009 ) , we next performed immunohistochemical detection of p-S6 , a readout of mTORC1 activity , and found that indeed p-S6 levels were reduced in the left PZ and pre-HZ from P3 until at least P5 ( Figures 3B and 4 , n = 19/23 , 7/8 and 3/3 at P3 , P4 and P5 , respectively ) . We further found that this downregulation could be mimicked in the right GP of P1-Pit::DTR mice and also in both hindlimbs of WT pups by treatment with the mTORC1 inhibitor rapamycin , which also stunts overall growth ( Figure 3—figure supplement 2D–E ) . Since FGFs secreted by the perichondrium negatively regulate chondrocyte proliferation and early hypertrophy ( Liu et al . , 2002; Karuppaiah et al . , 2016 ) , we tested whether FGF signaling was altered , but did not detect an increase in FGF signaling in any region of the left GP of P1-Pit::DTR mice at P3 ( Figure 3—figure supplement 1B–B’ ) . On the contrary , Fgfr3 expression was downregulated in the region of the left GP where the secondary ossification center forms , suggesting Fgfr3 could be involved in the observed delayed formation and expansion of this structure . Moreover , decreased TGFβR2 signaling in the joint region has been shown to impair chondrocyte hypertrophy at embryonic stages via increased expression of the cytokine MCP5 ( Longobardi et al . , 2012 ) , but we were not able to detect MCP5 mRNA or protein in our model at P3 . On the other hand , RNA and protein levels of the WNT target Lef1 were upregulated in the RZ of the left P1-Pit::DTR GP compared to right ( Figure 3C; see also Figure 4—figure supplement 2 , n = 5 at P3 , 4 at P4 , 3 at P5 ) , which might impair PTHLH signaling and chondrocyte differentiation , as described ( Ray et al . , 2015; Guo et al . , 2009 ) . In summary , our results show that damage in the tissues adjacent to the early postnatal GPs results in alterations in multiple GP signaling pathways known to regulate bone growth . 10 . 7554/eLife . 27210 . 007Figure 3 . Multiple signaling pathways related to chondrocyte proliferation and maturation are altered in distinct regions of the left P1-Pit::DTR growth plate . ( A–A’ ) In situ hybridization for Ihh and the HH target Gli1 , in the left and right proximal tibia of Pit::DTR mice , 2dpi . ( B , C ) Immunohistochemical staining for phosphorylated ribosomal protein S6 [ ( B ) , readout of mTORC1 activity] or for the canonical WNT target LEF1 ( C ) , in the left and right proximal tibia of Pit::DTR mice , 2dpi . Note left-specific downregulation of p-S6 in the same region where Ihh is reduced . Arrowheads= ectopic expression , asterisks= reduced expression . See also associated Figure 3—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 00710 . 7554/eLife . 27210 . 008Figure 3—source data 1 . Cortical bone parameters measured by µCT in P5 femora from Osx::DTR mice injected with PBS ( Ctl ) or DT ( Exp ) at P1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 00810 . 7554/eLife . 27210 . 009Figure 3—figure supplement 1 . Bone growth impairment in P1-Pit::DTR animals is not caused by increased FGF or EGF signaling in the HZ , nor by ablation of osteoprogenitors in the perichondrium or increased cell death in the osteochondral junction . ( A ) Immunohistochemistry for p-EGFR did not reveal increased EGF signaling in the HZ at P3 ( n = 3 PBS- and 3 DT-injected mice ) . Arrowheads= sparse activation of EGF signaling in the left articular cartilage . ( B , B’ ) Immunohistochemistry for the FGF-signal transducer FRS2 and in situ hybridization for Fgfr3 did not reveal increased FGF signaling in the left HZ ( n = 8 , 1 , and two for P3 , P4 and P5 ) . ( C ) Co-staining for the osteoblast lineage markers COLI or SP7 ( OSX ) ( Maes et al . , 2010 ) and TUNEL were done to determine whether cell death is increased in osteoprogenitors of Pit::DTR long bones . DT-mediated ablation of osteoprogenitors ( also derived from the lateral plate mesoderm ) could contribute to the Pit::DTR bone growth defect by impairing the generation of new bone , but no increase in osteoblast death was detected . The approximate regions of ( C ) are indicated in ( A ) as a reference . ( D–F ) To further rule out a possible contribution of early osteoblast precursor ablation to the phenotype , we crossed Sp7-tTA , tetO-EGFP/Cre mice ( Rodda and McMahon , 2006 ) with R26LSL-DTR animals to generate Sp7-tTA , tetO-EGFP/Cre; R26LSL-DTR mice ( Osx::DTR ) . These animals were injected with either DT or PBS at P1 . DT-injected Osx::DTR mice showed an almost complete depletion of osteoprogenitors 2dpi ( D ) and reduced thickness of the cortical bone as compared to PBS-injected animals ( E , insets show a 2 . 5x magnification of the bracketed area; E’ shows µCT images and cortical thickness quantification of the midshaft femoral region , color-coded by litter of origin , n = 6 PBS-treated and 6 DT-treated mice , p-value for unpaired Mann Whitney test is shown . See associated Source Data 1 ) . However , no consistent reduction of bone length was observed at P5 ( F n = 8 PBS- and 7 DT-injected pups ) , ruling out a major contribution of osteoprogenitor ablation to the Pit::DTR phenotype . Note that although DT-injected mice weighed significantly less than PBS-injected ones , their bones were not significantly shorter . ( G ) Quantification of the density of apoptotic cells at the osteochondral junction of Pit::DTR pups 2-3dpi . The aim of the experiment was to test whether an increase in cell death at the osteochondral junction ( where hypertrophic chondrocytes die by apoptosis and the cartilage is replaced by bone ) could contribute to the observed reduction in the height of the HZ due to chondrocytes were being eliminated faster than produced . Note , no significant differences were detected between left and right GPs ( n = 5 pups , 2–3 sections per limb ) . PS= primary spongiosa . The graph shows left and right paired data ( TUNEL+ cells per area ) for each animal , which were compared by a ratio paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 00910 . 7554/eLife . 27210 . 010Figure 3—figure supplement 2 . Characterization of the role of IHH and mTOR in the signaling and growth defects of P1-Pit::DTR mice . ( A ) X-gal staining to detect Pthlh expression , in the left and right proximal tibia of Pit::DTR; PthlhlacZ/+ mice ( Chen et al . , 2006 ) , 2dpi ( n = 3 at P3 and 2 at P4 ) . ( B–C’ ) Treatment of P1-Pit::DTR pups with the SHP2 inhibitor NSC-87877 ( 5 mg/kg s . c . daily from P1 ) was found to rescue Ihh expression in the left GP at P3 ( B and B’ , n = 2 and 6 with partial and full rescue , respectively ) but it did not rescue bone growth by P5 ( C , frontal view of the tibiae , and C’ , n = 5 control and three experimental animals ) . p-values for unpaired Mann Whitney test ( SHP2i vs . vehicle for each bone ) are shown . ( D , D’ ) p-S6 expression in proximal tibiae from vehicle-treated ( D ) or rapamycin-treated ( D’ ) P1-Pit::DTR pups , showing dependence on mTORC1 activity . n = 2 at P3 , two at P4 . ( E ) Length of the right ( control ) bones of P4 Pit::DTR pups , treated with vehicle ( n = 3 ) or rapamycin ( n = 3 ) from P1 to P4 . Analysis was done by 2-way ANOVA ( alpha = 0 . 05 , Treatment and Bone as variables , p=0 . 0079 for Treatment ) followed by Sidak’s posthoc multiple comparisons test ( p-values shown in Figure ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01010 . 7554/eLife . 27210 . 011Figure 4 . Injury-induced inflammation in the left infrapatellar fat pad reduces local Igf1 expression in the DT-impaired growth plate . ( A , A’ ) In situ hybridization for Igf1 in the infrapatellar fat pad ( IPFP , dotted lines ) of left and right P1-Pit::DTR knees , at P3 ( A ) and P5 ( A’ ) . ( B–B’’ ) Immunohistochemistry for the mTORC1 readout p-S6 ( top ) , and RNA in situ hybridization for Ihh ( bottom ) in left P1-Pit::DTR GPs at P3 , following intraarticular injection of PBS ( B ) , IGF1 ( B’ ) or IGF1 combined with i . p . injection of the mTORC1 inhibitor rapamycin ( B’’ ) . ( C–C’’’ ) Immunohistochemistry for the neutrophil marker LY6B in the IPFP ( dotted lines ) of left and right P1-Pit::DTR knees , at P3 ( C ) , P5 ( C’ ) and P2 ( C’’’ ) . The inset in ( C’’’ ) is a 2x magnification showing that some neutrophils express TNFα in a cellular compartment . Turquoise signal= autofluorescent cells . The graph in ( C’’ ) represents the density of LY6B+ cells in left and right P1-Pit::DTR IPFP at P3 ( n = 3 mice , 3–4 sections per animal ) . A ratio paired t-test was used to offset the variability between absolute measurements . ( D–D’’ ) Immunoblockade of neutrophil infiltration with NIMP-R14 antibody after DT injection ( D ) rescues Igf1 expression in the left IPFP ( D’ ) , as well as mTORC1 signaling in the GP , 2dpi ( D’’ ) . ( E , F ) Quantification of bone length at P4 , expressed as Left/Right ratio , for vehicle ( Veh , either PBS or IgG ) -treated and IGF1- ( E ) or NIMP-R14-treated ( F ) mice ( unpaired two-tailed Mann-Whitney test ) . See also associated Figure 4—figure supplement 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01110 . 7554/eLife . 27210 . 012Figure 4—figure supplement 1 . Characterization of the IGF signaling axis in P1-Pit::DTR mice . ( A ) In situ hybridization for Igf1 in the indicated tissues of P1-Pit::DTR mice , 2dpi ( metaphysis from femora , rest from tibiae ) . To properly compare expression domains that depend on the section level , whole slide imaging was performed on sections from five animals with a slide scanner , and representative images are shown . ( B–C ) Representative examples of Igf2 ( B , n = 3 , tibiae are shown ) and Igf1r expression ( C , n = 4 , femora are shown ) in left and right GPs and surrounding tissues of P1-Pit::DTR mice . Arrowheads in ( B ) point to damaged regions of the left prospective articular cartilage where Igf2 expression was somewhat diminished . ( D ) Immunohistochemistry for SOCS3 in the GPs and surrounding tissues of P1-Pit::DTR mice ( n = 4 at P2 , three at P3 , tibiae are shown ) . Magnified views of the boxed regions are shown below the overviews . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01210 . 7554/eLife . 27210 . 013Figure 4—figure supplement 2 . IGF1 supplementation or neutrophil immunoblockade do not rescue all the signaling changes in the left RZ of P1-Pit::DTR mice . ( A , B ) In situ hybridization for Fgfr3 and Lef1 in left and right P1-Pit::DTR tibiae from P3 mice treated either with IGF1 ( A , n = 3 ) or an anti-neutrophil antibody ( B , n = 4 ) . Note that only Lef1 expression in the left GP remains consistently increased in both conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01310 . 7554/eLife . 27210 . 014Figure 4—figure supplement 3 . Multiple immune cells are recruited to the infrapatellar fat pad of Pit::DTR mice , which correlates with local Igf1 downregulation . ( A–B ) IHC for general macrophage markers in the IPFP of Pit::DTR mice at P4 . Quantifications are shown in ( A’ ) , a ratio paired t-test was used to compare left and right macrophage density in the IPFP of Pit::DTR mice . ( C ) Immunostaining for the neutrophil marker LY6B and in situ hybridization for Igf1 in the IPFP of LPS-injected ( left knee ) WT mice . ( D ) Immunostaining for the neutrophil marker LY6B and in situ cell death detection ( TUNEL ) in the IPFP of LPS-injected ( left knee ) WT mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 014 We next undertook a candidate approach to explore the possible extrinsic influences that were responsible for the signaling changes in the experimental GPs . Since mTORC1 activity was impaired in P1-Pit::DTR mice and insulin/IGF signaling is the main bone-growth related pathway known to activate mTORC1 signaling , we examined the GH/IGF signaling axis . As the right limb serves as an internal control for the insulted left limb in our model , any left-specific defect could not be due to changes in a systemic factor ( s ) ( be it GH or IGF1 ) , unless there was a local change in the signal transduction machinery that responds to those systemic factors . Furthermore , given that systemic GH/IGF1 does not play an important role in early postnatal skeletal growth ( see Introduction ) , we decided to focus on local IGF signaling . Interestingly , we found that while Igf1 was highly expressed in the right infrapatellar fat pad ( IPFP ) , an extra-synovial adipose tissue located between the femur and the tibia , it was extremely and specifically downregulated in the left IPFP of P1-Pit::DTR mice at 2-3dpi ( Figure 4A , n = 10/12 ) , and recovered by 4dpi ( Figure 4A’ , n = 2 ) . Furthermore , Igf1 expression was not consistently altered in other tissues , such as the perichondrium , the muscle bundles or cells in the metaphyseal ( shaft ) region ( Figure 4—figure supplement 1A ) . Similarly , we only detected minor to no changes in the expression of Igf2 or the main IGF receptor gene Igf1r in the GP or surrounding tissues ( Figure 4—figure supplement 1B–C ) , indicating that the IPFP was specifically affected by the insult . We further tested whether downstream signaling components were altered . In particular , since injury-induced inflammation ( see below ) could conceivably increase expression of suppressor of cytokine signaling 3 ( SOCS3 ) , which is known to exert a negative influence on IGF signaling ( Ahmed and Farquharson , 2010 ) , we performed SOCS3 immunohistochemistry on P1-Pit::DTR knee sections . However , we could not detect differential expression between the left and right GPs at P2 or P3 ( Figure 4—figure supplement 1D ) , suggesting that it is the reduced availability of IGF1 from the IPFP that causes some of the signaling changes observed in the left GP . To test this possibility , we provided exogenous IGF1 to the left knee region by intraarticular injections in P1-Pit::DTR pups ( P1-P3 ) , and found that indeed mTORC1 signaling was consistently rescued in the PZ and pre-HZ of the GPs ( Figure 4B’ top , n = 5 P3-P5 pups ) . Curiously , Ihh and Fgfr3 expression were also rescued in these animals ( Figure 4B’ bottom , n = 3 at P3 , and Figure 4—figure supplement 2A ) , suggesting that both genes are downstream of IGF signaling . To determine the cause of Igf1 downregulation , we analyzed the cellular and molecular response to cell death in the left IPFP of P1-Pit::DTR mice . We observed left-specific neutrophil recruitment at P2-P4 that had mostly receded by P5 , thus precisely correlating with transient Igf1 downregulation ( Figure 4C–C’ n = 5 , 5 , 4 , 5 at P2 , P3 , P4 , P5 ) . Moreover , macrophages , already very abundant in the IPFP in resting conditions , were additionally recruited to the left IPFP ( Figure 4—figure supplement 3A–A’ ) . As neutrophils are an important feature of joint inflammatory diseases ( Wright et al . , 2010; Wipke and Allen , 2001 ) , we tested whether neutrophils were responsible for Igf1 downregulation by blocking their recruitment with the NIMP-R14 neutralizing antibody ( Lopez et al . , 1984 ) during P1-P3 in P1-Pit::DTR mice . Neutrophil blockade was effective for at least 14 hr ( Figure 4D , n = 3 ) , and significantly correlated with partial recovery of Igf1 expression in the IPFP ( Figure 4D’ , n = 4 ) , without inhibiting cell death in this region ( Figure 4—figure supplement 3B ) . The partial recovery of Igf1 after neutrophil inhibition indicates that additional cell types inhibit its expression . Importantly , the recovery of Igf1 expression correlated with increased levels of p-S6 and Ihh in the pre-HZ ( Figure 4D’’ , n = 3 , and not shown ) , demonstrating that proper signaling in the pre-HZ requires IGF1 production from the IPFP . To further probe the role of immune-cell recruitment in Igf1 downregulation , we injected lipopolysaccharide ( LPS ) into the joint of WT animals , which caused a strong ( but patchy ) local recruitment of neutrophils to the IPFP . Of significance , this recruitment led to the predicted downregulation of Igf1 in the areas with highest accumulation of neutrophils ( Figure 4—figure supplement 3C ) . This effect took place in the absence of cell death ( Figure 4—figure supplement 3D ) , suggesting that immune cells ( probably neutrophils ) secrete factors that reduce Igf1 expression . Notably , we found that TNFα-expressing cells ( some of them neutrophils ) were present in the left and not the right IPFP in P1-Pit::DTR mice at P2 , and that neutrophils accumulated TNFα in what seemed to be the Golgi apparatus ( Figure 4C’’’ ) . TNFα has been shown to reduce Igf1 expression in vascular smooth muscle ( Anwar et al . , 2002 ) , raising the possibility that this cytokine is in part responsible for the observed Igf1 downregulation in the left IPFP of P1-Pit::DTR mice . Interestingly , partial restoration of Igf1 expression by neutrophil blockade did not rescue Fgfr3 expression ( Figure 4—figure supplement 2B ) , suggesting that full restoration of Igf1 is necessary for normal Fgfr3 expression . Despite IGF1 supplementation rescuing several critical pathways , neither IGF1 injection nor neutrophil blockade were sufficient to rescue the decreased left bone growth at P4 ( Figure 4E–F , n = 5 experimental and 5 control animals for each treatment ) , indicating that an IGF1-independent pathway also contributes to the growth defect . Indeed , expression of Lef1 in the left GP remained upregulated after IGF1 injection or neutrophil blockade ( Figure 4—figure supplement 2A–B ) , raising the possibility that increased LEF1 activity contributes to the growth defect . Given the importance of the immediate-early damage response upon injury , we tested its activation and indeed detected it in the left joint region of P1-Pit::DTR mice , not only in the IPFP but also in the prospective articular cartilage ( pAC ) . Specifically , we observed that expression of Early growth response gene 1 ( Egr1 ) was increased in the left IPFP and pAC of P1-Pit::DTR mice at P2 ( 1dpi ) ( Figure 5A , n = 2 ) . Egr1 expression in the left IPFP receded by P3 , but remained highly upregulated in the left pAC through P4 , receding by P5 ( Figure 5B–C , Figure 5—figure supplement 1A–B , n = 3 , 2 , 2 , respectively ) . Interestingly , p-S6 expression , while downregulated in the pre-HZ ( Figures 3 and 4 ) , was upregulated in the left pAC at P3 and P4 and receded by P5 ( Figure 5B–C , n = 33 , 14 and Figure 5—figure supplement 1B’ , n = 2 ) , suggesting transiently increased mTORC1 signaling in this region , as observed in osteoarthritis ( Pal et al . , 2015 ) . Other injury response markers followed a dynamic similar to Egr1 in the left pAC of P1-Pit::DTR mice . For example , we detected ectopic expression of Hif2a ( Figure 5—figure supplement 1A ) , which has a prominent role in cartilage destruction during osteoarthritis and rheumatoid arthritis ( Ryu et al . , 2014; Husa et al . , 2010 ) . In addition , Il6 , one of the main mediators of Hif2a-dependent cartilage destruction ( Ryu et al . , 2014 ) , was upregulated in the left pAC ( Figure 5—figure supplement 1A ) , and was the likely cause of the ectopic JAK1/2-dependent expression of p-STAT3 we detected in the pAC at P3 and P4 ( Figure 5B–C , n = 13 and 3; Figure 5—figure supplement 1C ) , and that receded by P5 ( not shown ) . Importantly , expression of Egr1 , p-S6 and p-STAT3 was not due to inflammation in the IPFP , as neutrophil blockade or injection of IGF1 did not preclude activation of the injury response in the pAC of P1-Pit::DTR mice ( Figure 5—figure supplement 1C and not shown ) . Associated with the injury response , histological analysis revealed extensive and persistent cell loss in the pAC/AC from P4 to at least P27 ( Figure 5D–E , n = 3 and 4 ) . Of note , most signaling changes extended beyond the area of cell death ( i . e . the pAC ) , almost reaching into the RZ of the GP ( compare TUNEL and p-STAT3 in Figure 5B–C ) . Since cells in the pAC/AC do not normally give rise to GP chondrocytes ( Kozhemyakina et al . , 2015 ) , this set of results raised the possibility that the injury response in the pAC triggers a paracrine-signaling cascade that contributes to the signaling defects observed in the RZ . 10 . 7554/eLife . 27210 . 015Figure 5 . The injury response in the articular cartilage of P1-Pit::DTR mice correlates with signaling changes in the resting zone and impaired bone growth . ( A ) Expression of the immediate-early marker Egr1 1dpi . Dotted area= IPFP . ( B , C ) Expression of Egr1 and the indicated signaling effectors ( white arrows ) and cell death ( pink arrows ) in the prospective articular cartilage ( pAC ) , two ( B ) and three dpi ( C ) . Dotted area= extent of p-STAT3 signal . ( D , E ) Hematoxylin and eosin ( H&E ) and hematoxylin and safranin O ( H&SO ) analysis of the pAC/AC , showing extensive damage at P4 ( D ) , persistent at P27 ( E ) . Magnifications of the boxes in ( B ) and ( E ) are shown in ( B’ ) and ( E’ ) , respectively . See also associated Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01510 . 7554/eLife . 27210 . 016Figure 5—figure supplement 1 . Characterization of the injury response in P1-Pit::DTR mice . ( A–B’’ ) Features of the injury response in P1-Pit::DTR knee joints and proximal tibiae at P3 ( A ) in situ hybridization ) and P5 ( B , in situ hybridization , B’ , immunohistochemistry , B’’ , hematoxylin and eosin ) . n = 2 , 6 , 2 , two at P2 , P3 , P4 , P5 . ( C ) Immunohistochemical staining for p-S6 and p-STAT3 in left and right P1-Pit::DTR tibiae from P3 mice treated with NIMP-R14 . Dotted lines delimit the prospective articular cartilage ( pAC ) . IPFP= infrapatellar fat pad . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 016 To test the role of the injury response in the altered GP signaling in P1-Pit::DTR mice , we performed pharmacological rescue experiments . Significantly , in vivo treatment with the JAK1/2 inhibitor ruxolitinib ( Fridman et al . , 2010 ) precluded p-STAT3 and LEF1 upregulation in the articular region of P1-Pit::DTR mice ( Figure 6—figure supplement 1A–A’ ) . However , like IGF1 injection , ruxolitinib failed to rescue bone growth ( Figure 6A–B , n = 5 ) , perhaps because mTORC1 signaling remained downregulated in the pre-HZ ( Figure 6—figure supplement 1C’ ) . Finally , we tested whether redundancies exist between the fat pad-HZ axis and the pAC-RZ axis by administering ruxolitinib and intraarticular IGF1 either individually or in combination in the same litters . Notably , unlike single treatments , the combined treatment had a significant rescue effect ( Figure 6A–B ) , although only some femora ( 3 out of 11 ) and none of the tibiae showed complete rescue ( see Discussion ) . 10 . 7554/eLife . 27210 . 017Figure 6 . Combined inhibition of the injury response cascade and intraarticular IGF1 injection rescues DT-impaired femur growth . ( A , B ) Representative femur preparations ( A ) and quantification of left/right length ratio of P1-Pit::DTR bones at P4 ( B ) after in vivo treatment with the indicated substances ( n = 8 , 7 , 5 , 11 for PBS , IGF1 , Ruxo , IGF1+Ruxo , respectively ) . The parallel treatments were compared by two-way ANOVA with Bone and Treatment as variables ( alpha = 0 . 05 , p=0 . 0066 for Treatment , 0 . 0678 for Bone ) , followed by Sidak’s posthoc multiple comparisons test ( only the p-values lower or close to 0 . 05 are shown in the Figure ) . ( C , D ) Summary of signaling changes in the IPFP and pAC after mesenchymal cell death outside the GP , and their interaction with GP signaling and bone growth . In ( D ) , the postnatal stages at which each pathway operates are indicated . ( E ) Speculative model for the regulation of Lef1 and Ihh in the experimental GP . Green/Red lettering indicates , respectively , pathways up/downregulated following inflammation and injury response . See also associated Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01710 . 7554/eLife . 27210 . 018Figure 6—figure supplement 1 . Inhibition of JAK1/2 or mTORC1 activity prevents the upregulation of LEF1 in P1-Pit::DTR mice . ( A–A’’ ) STAT3 signaling in the AC/RZ of P1-Pit::DTR tibiae ( A ) is inhibited in mice treated with the JAK1/2 inhibitor ruxolitinib ( A’ ) but not in mice treated with the mTORC1 inhibitor rapamycin ( A’’ ) . n = 2 vehicle-treated , two ruxolitinib-treated and three rapamycin-treated Pit::DTR pups . Dotted lines= IPFP . ( B ) TUNEL staining is not rescued in the left IPFP ( dotted lines ) of IGF1- and ruxolitinib-injected P1-Pit::DTR pups ( n = 3 at P2 ) . ( C–C’’ ) LEF1 but not p-S6 levels are normalized in the left RZ of ruxolitinib-treated P1-Pit::DTR pups ( C’ , n = 2 ) , as compared to the vehicle-treated animals ( C , n = 2 ) . Similarly , LEF1 levels are not elevated in the left RZ of rapamycin-treated P1-Pit::DTR pups ( C’’ , n = 3 ) . Arrowheads= regions of elevated LEF1 levels . Asterisks= loss of expression as compared with vehicle-treated . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 018 Given that during the first two postnatal weeks the local extrinsic environment of the murine GP changes from having one to two adjacent ossification centers ( Pannier et al . , 2010 ) , and therefore is located farther from the IPFP and potentially more exposed to diffusing molecules from the vasculature , we tested the effect of DT injection at P14 in Pit::DTR mice . Similar to the P1 injections , analysis of cell death 3-4dpi in P14-Pit::DTR mice revealed apoptotic cells only in the surrounding soft tissues , and not the GP ( Figure 7A , n = 4 ) . As in the P1-Pit::DTR model , the height of the left HZ zone at 4dpi was reduced in P14-Pit::DTR mice compared to controls ( Figure 7B , n = 2 ) , and the hindlimbs developed a progressive left-right asymmetry over time , although it was non-significant at P20 ( ~7–13% by P57 , Figure 7F–G , n = 3–4 per genotype at P20 , n = 7 per genotype at P57 ) . Furthermore , although neutrophils infiltrated the knee joint , mTORC1 activity and Ihh signaling were not reduced at 3-4dpi in the left HZ as compared to the right in P14-Pit::DTR mice , unlike in the P1-Pit::DTR experiment ( Figure 7C , n = 4 and not shown ) . Interestingly , we found that Igf1 is no longer expressed in the right nor the left IPFP at P17-18 ( Figure 7D , inset , n = 3 ) , suggesting that mTORC1 activity in the HZ eventually becomes independent of this extrinsic source of IGF1 . On the other hand , similar to P1-Pit::DTR mice , Egr1 and LEF1 were ectopically activated in the left cartilage of P14-Pit::DTR mice ( Figure 7D–E , n = 4 ) . In conclusion , we observed a long-term impact on bone growth of a mesenchymal injury at P1 and P14 , although the short-term mechanisms were only partially shared between both stages . 10 . 7554/eLife . 27210 . 019Figure 7 . Induction of soft-tissue cell death at P14 also impairs bone growth . ( A ) TUNEL staining ( arrowheads ) in sagittal section of P17 proximal tibia from Pit::DTR mice injected DT at P14 ( P14-Pit::DTR ) . SOC= secondary ossification center ( outlined ) . ( B ) Hematoxylin and eosin-stained sagittal sections of P18 distal femur from P14-Pit::DTR mice . Vertical bars indicate the length of the HZ . ( C ) Immunostaining for p-S6 in the GP of P14-Pit::DTR mice at P17 . ( D , E ) In situ hybridization for Egr1 ( D ) and Igf1 ( D , insets ) and LEF1 immunostaining ( E ) in knees from P14-Pit::DTR mice . ( F ) Alizarin red staining of the skeletal elements from a representative P14-Pit::DTR mouse at P57 . ( G ) Quantification of the left/right length ratio for femora and tibiae of P20 and P57 Control and P14-Pit::DTR mice . Analysis was done by 2-way ANOVA ( alpha = 0 . 05 , Bone Identity and Genotype as variables , p-value for Genotype was 0 . 1970 at P20 and < 0 . 0001 at P57 ) followed by Sidak’s posthoc multiple comparisons test ( p-values shown in Figure ) . Arrowheads denote ectopic expression . AC= articular cartilage . RZ , PZ , HZ= resting , proliferative , hypertrophic zones . IPFP= infrapatellar fat pad . See also associated Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 01910 . 7554/eLife . 27210 . 020Figure 7—figure supplement 1 . Developmental description of infrapatellar fat pad formation and correlation with limb allometry . ( A–A’ ) Developmental progression of the knee interzone/infrapatellar fat pad region ( A ) at the indicated stages , and the elbow region at E17 . 5 ( A’ ) . Hematoxylin and eosin ( H&E ) , Igf1 expression and FABP4 ( adipocyte marker ) and CD44 plus CD90 ( mesenchymal cell markers ) immunohistochemistry are shown in ( A ) . At least two specimens per stage and technique were examined . Color-coded arrowheads point to cells expressing the marker of the corresponding color . Yellow and white dotted lines delineate , respectively , the fat pads and the cartilage anlagen . Asterisks denote lack of detectable expression . F , T , H , R , U= femur , tibia , humerus , radius , ulna . ( B ) Radius/Tibia length ratio in WT mice at the indicated stages ( at least four animals per stage ) . Data were analyzed by one-way ANOVA with Stage as variable ( alpha = 0 . 05 , p<0 . 0001 ) . p-values for Sidak’s multiple comparisons test ( comparing each stage with the preceding one ) are shown . Note that the ratio stabilizes between P8 and P14 , coinciding with the cessation of Igf1 expression in the infrapatellar fat pad . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 020 Given the new trophic role that we ascribe to the IPFP , the temporal differences we observed in Igf1 expression in this tissue prompted us to characterize the appearance , cellular composition and expression of Igf1 in the IPFP during mouse hindlimb development ( Figure 7—figure supplement 1A ) . The IPFP appeared as a defined region at around E17 . 5 , and this coincided with the accumulation of a high level of Igf1 expression in the area as compared to the surrounding mesenchymal tissues . While perivascular stem cells ( CD44/CD90+ ) ( Hindle et al . , 2017 ) were detected in the interzone area even before the IPFP was formed , adipocytes ( detected either by hematoxylin and eosin staining or expression of FABP4 ) were not clearly seen until P3-P4 , indicating that adipocytes are not the main source of Igf1 . As expected , Igf1 expression levels gradually decreased postnatally until they were barely detectable at P8 and not detected at P12 . Intrigued by the dynamic expression of Igf1 in the IPFP , we finally explored whether the dynamics correlate with any change in the ratio of the lengths of the radius ( far from any fat pad , see Figure 7—figure supplement 1A’ ) and the tibia during development . Interestingly , we observed that the radius grows more slowly than the tibia at the stages when Igf1 levels are high in the IPFP , whereas the ratio stabilizes after P8 , roughly coinciding with the cessation of Igf1 expression ( Figure 7—figure supplement 1B ) . This correlation is consistent with the hypothesis that paracrine signals from the IPFP play a role in the establishment of body proportions , a possibility with clinical and evolutionary implications ( see Discussion ) . Our study demonstrates that transient cell death in the early postnatal left hindlimb mesenchyme surrounding the GPs leads to reduced long bone growth due to altered signaling from two tissues in the adjacent knee joint . Only a few recent studies have addressed the regulation of longitudinal bone growth by the surrounding tissues , and showed that TGFβR2 signaling from the joint interzone stimulates chondrocyte hypertrophy in the GP ( Longobardi et al . , 2012 ) , and FGFs produced by the perichondrium negatively modulate bone growth in a GP non-autonomous fashion ( Liu et al . , 2002; Karuppaiah et al . , 2016 ) . Of significance , our study identifies two neonatal joint components that can modulate bone growth , at least following injury: the IPFP and the pAC/AC . Taken together , our data support a model whereby extrinsic signals from the IPFP and/or pAC/AC modulate distinct GP-intrinsic signaling pathways , leading to reduced bone growth , even in the absence of intrinsic damage to the GP ( Figure 6C–D ) . While we think this model can be generally applied , the apparently milder rescue of the tibia compared to the femur ( Figure 6B ) could mean that each bone’s response to the same growth-modulating cues varies with the developmental stage in a different manner . Indeed , the specific effect of limb immobilization on the growth of different bones has recently been shown to depend on the time of treatment in chicken and crocodile embryos ( Pollard et al . , 2017 ) . By inducing cell death outside the GP , we found a prolonged long bone growth defect in vivo that was associated with inflammation in the knee joint . This result is reminiscent of in vitro studies showing that pro-inflammatory cytokines can exert a negative influence on bone growth that continues well beyond the cytokine exposure period ( MacRae et al . , 2006 ) . As in our in vivo study , the inflammation-dependent in vitro growth defect was only partially rescued by IGF1 treatment ( Mårtensson et al . , 2004 ) , suggesting that inflammation impairs bone growth via both IGF-dependent and -independent pathways . Using our P1-Pit::DTR mice , we also uncovered that loss of mTORC1 signaling in the pre-HZ correlates with loss of Ihh expression , and that restoration of IGF1 levels in the joint rescues both pathways . At the same time , SHP2 inhibition in P1-Pit::DTR mice rescues Ihh expression but not mTORC1 activity , and mTORC1 inhibition does not reduce Ihh expression in WT animals ( Sanchez et al . , 2009 ) , nor does it impair IGF1-mediated rescue of Ihh expression in Pit::DTR mice ( Figure 4B’’ ) . Therefore , we speculate that IGF1 activates Ihh expression via an mTORC1-independent pathway , possibly by inhibition of SHP2 ( Figure 6E ) . Notably , the requirement of IGF signaling for Ihh expression seems to be restricted to a narrow perinatal window , as Igf1r deletion in the cartilage only reduces Ihh expression when triggered at P5 but not prenatally ( Wang et al . , 2011 ) , and we show that Igf1 is no longer expressed in the IPFP by P12 . Unlike in the pre-HZ , we found mTORC1 activity was increased in the left pAC and RZ of P1-Pit::DTR mice . Since mTORC1 was induced despite reduction of local IGF1 levels , it follows that another signal must be responsible for mTORC1 activation in the AC/RZ . One candidate is IL6 expressed ectopically in the AC , as this cytokine can activate mTORC1 in some cell types ( Kim et al . , 2008 ) . Interestingly , inhibition of mTORC1 activity dampened LEF1 levels in the RZ without impairing STAT3 phosphorylation , ( Figure 6—figure supplement 1A’’ , C’’ ) , while JAK1/2 inhibition also precluded LEF1 upregulation , without impeding p-S6 activation ( Figure 6—figure supplement 1A’ , C’ ) . These results suggest that both mTORC1 and JAK/STAT signaling pathways are independently required to activate Lef1 expression ( Figure 6E ) . We speculate that the local influence exerted by the joint region could at least in part explain poorly understood phenomena observed in the field of bone growth , such as the fact that the proximal tibia and distal femur GPs ( i . e . closest to the fat pad ) grow faster than the distal tibia or proximal femur ( Digby , 1916; Payton , 1932; Moss-Salentijn , 1974 ) . Our proposal is biologically relevant because differential GP growth has been recently suggested to be evolutionary optimized to achieve energy-efficient scaling of the growing bones by minimizing the remodeling of ossified cortical structures ( Stern et al . , 2015 ) . While intrinsic differences in the GP likely exist and account in part for differential growth of proximal and distal GPs , classic experiments showed a change in growth rate upon proximo-distal transposition of the GPs within a bone ( Moss-Salentijn , 1974; Hert , 1964 ) , strongly suggesting the involvement of local extrinsic factors . Fat pads are probably one of the main participants in this process , as suggested by our study and those of others showing that adipocyte-secreted signals can stimulate long bone growth in the absence of growth hormone ( Shtaif et al . , 2015 ) . In this regard , our finding that Igf1 is almost absent from the IPFP by P8 correlates with the fact that the femur and tibia grow faster than the rest of the body ( including forelimb bones ) until roughly the same stage ( Figure 7—figure supplement 1B and [Roselló-Díez and Joyner , 2015] ) . The GP likely becomes independent from the IPFP after ~P8 ( due both to the intervening presence of the secondary ossification center as well as to cessation of Igf1 expression in the fat pad ) , which could also explain why it takes longer for bone growth to be significantly impaired when the injury is induced at P14 vs . P1 . Finally , we cannot exclude the possibility that the IPFP and/or other local tissues hosting cells with osteochondrogenic potential , such as the synovium , bone marrow or the perichondrial groove of Ranvier ( Hindle et al . , 2017; Yang et al . , 2013; Karlsson et al . , 2009; Chung and Xian , 2014 ) , can contribute to bone growth not only with paracrine signals , but also with progenitor cells ( see for example Fig . S9 in [Yang et al . , 2013] ) , such that cell death within these tissues in Pit::DTR mice depletes a pool of cells that participates in recovery of the injured bones . In conclusion , further studies using mouse models should confirm and expand the repertoire of local extrinsic regulators of bone growth . This repertoire would be a valuable resource for evolutionary studies addressing the change of body proportion across phyla , and it could potentially be harnessed to develop improved therapies to correct local long bone growth defects . The Pitx2-Cre ( Shiratori et al . , 2006 ) ( RRID:IMSR_RBRC03487 , kind gift of Dr . H . Hamada ) , Sp7-tTA , tetO-EGFP/Cre ( Rodda and McMahon , 2006 ) ( RRID:IMSR_JAX:006361 ) , PthlhlacZ ( Chen et al . , 2006 ) ( RRID:MGI:5519222 , provided by Dr . Chitra Dahia ) and R26LSL-DTR ( Buch et al . , 2005 ) ( RRID:IMSR_JAX:007900 ) mouse lines were maintained in an outbred Swiss Webster background and genotyped as previously described . Noon of the day of vaginal plug detection was considered E0 . 5 . The equivalent of E19 . 5 is referred to as P0 . All animal studies were performed under an approved Institutional Animal Care and Use Committee mouse protocol according to MSKCC institutional guidelines . A stock solution of DT ( Sigma ) was prepared at 40 ng/µl in sterile PBS , aliquoted and stored at −80°C until used . A working solution ( 5 ng/µl in PBS for P1 and 15 ng/µl for P14 ) was prepared and 15 ng/g injected subcutaneously ( scruff of the neck ) , using a 25 µl syringe ( Hamilton ) . Note: For all rescue experiments , it was important to determine that any putative rescue was not actually due to defective DT injection . Therefore , only the specimens that showed obvious signs of mesenchymal death ( e . g . a characteristic deformity of the left foot and/or AC damage on histological examination ) were used to assess the effect of drugs on the P1-Pit::DTR phenotype . Staining of cartilage and bone was performed as described ( Rigueur and Lyons , 2014 ) . For young mouse pups ( ≤P5 ) , ossified bone length was measured on digital microphotographs using the line tool in Adobe Photoshop . For adolescent and adult mice , the limbs were dissected out , skinned and incubated in 2% KOH to remove the soft tissues . Individual bones were then measured using digital calipers ( EZCal from iGaging ) . Tibiae were measured from the intercondylar eminence to the distal articular surface , while femora were measured from the trochanteric fossa to the intercondylar fossa . Femora from Osx::DTR mice , injected either with PBS or DT at P1 , were collected at P5 and processed for skeletal staining as described above . After photographing , they were progressively transferred to 70% EtOH and stored at 4°C until scanning . Before scanning , the samples were allowed to reach room temperature . A 0 . 6 mm region of the mid-diaphysis ( right below minor trochanter ) was scanned on a Scanco µCT 35 system ( Scanco Medical , Brüttisellen , Switzerland ) using a 3 . 5 µm voxel size , 55KVp , 0 . 36 degrees rotation step ( 180 degrees angular range ) and a 400 ms exposure time per view ( performed in 70% EtOH ) . The Scanco μCT software ( HP , DECwindows Motif 1 . 6 ) was used for 3D reconstruction and viewing of images . After 3D reconstruction , volumes were segmented using a global threshold of 0 . 3 g/cc . Tissue mineral density ( TMD ) , cortical area fraction ( Ct . Ar . /Tt . Ar . ) , and thickness of the cortex ( Ct . Th . ) were calculated for the cortical bone . NIMP-R14 ( Adipogene #AG-20B-0043PF-C500 ) was solubilized at 2 mg/ml . 6 µg/g were injected i . p . 2xdaily , using a 25 µl syringe ( Hamilton ) . An equivalent volume of rat IgG or PBS was used as injection control . Mouse pups were euthanized by decapitation after hypothermia-induced analgesia . Knees were dissected out , skinned and fixed by immersion in either 4% paraformaldehyde ( PFA , Electron Microscopy Sciences ) in PBS for 2 days at 4°C ( for immunohistochemistry and in situ hybridization ) or 0 . 25% glutaraldehyde in PBS for 90 min at room temperature ( for X-Gal staining ) . After several washes with PBS , the tissue was then cryoprotected first by brief incubation with a solution of 15% sucrose and then 30% sucrose in PBS for at least 4 hr at 4°C , and then embedded in Cryomatrix ( Thermo ) using dry-ice-cold isopentane ( Sigma ) . The knees were oriented sagittally and facing each other , with the tibiae on the bottom of the block ( i . e . closest to the blade when sectioning ) . Serial 8-micron sections were collected with a Leica Cryostat on Superfrost slides , allowed to dry for at least 30 min and stored at −80°C until used . For paraffin embedding , the fixative step was followed by 1 week decalcification with EDTA 0 . 5M in PBS ( pH 7 . 4 ) at 4°C , dehydration by 30 min incubations with graded ethanol series and xylene at room temperature , and paraffin incubations at 65°C . For all histological techniques , frozen slides were allowed to reach room temperature in a closed box , and Cryomatrix was washed away for 15 min with warm PBS ( 37°C ) . Paraffin sections were deparaffinized and rehydrated prior to the staining protocol . The protocol described in ( Nomura and Hirota , 2003 ) was followed . For young pups ( P1-P5 ) , samples were not decalcified . Except for Col2a1 , Col10a1 and Ihh ( provided by Dr . Licia Selleri ) , the templates for most riboprobes were generated by PCR from embryonic cDNA , using primers containing the SP6 or T7 RNA polymerase promoters . Primer sequences are shown in Table 1 . After purification of the PCR product ( Qiagen PCR purification kit ) , DIG-labeled probes were transcribed following manufacturer instructions ( Roche ) , treated with DNAase for 30 min and purified by LiCl-mediated precipitation in alcoholic solvent . Probes were kept at −80°C in 50% formamide ( Fluka ) . 10 . 7554/eLife . 27210 . 021Table 1 . Sequence of the primers used to amplify template for riboprobe synthesis from cDNA . DOI: http://dx . doi . org/10 . 7554/eLife . 27210 . 021Primer nameSequence Igf1 F SP6GCCGATTTAGGTGACACTATAGAAGTGGATGCTCTTCAGTTCGTG Igf1 R T7GAAATTAATACGACTCACTATAGGGTGTTTTGCAGGTTGCTCAAG Fgf18 F SP6GCCGATTTAGGTGACACTATAGAAGCCGCCTGCACTTGCCTGTG Fgf18 R T7GAAATTAATACGACTCACTATAGGGTGGTTTCTCGCAGTTTCCTC Egr1 FGTCTTTCAGACATGACAGCGAC Egr1 R SP6GCGATTTAGGTGACACTATAGGTGTCACACAAAAGGCACCAA Lef1 FTGAAGCCTCAACACGAACAG Lef1 R SP6GCGATTTAGGTGACACTATAGTTTCCGAAACAACCGTTTTC Hif2a FCACTGAGACACCTGCCACCTC Hif2a R SP6CATTTAGGTGACACTATAGGAGGCACCAGCCACCATG Agc1 FCCAGCCTGACAACTTCTTTG Agc1 R T7GTAATACGACTCACTATAGGGGGGCACATTATGGAAGCTC Fgf18 coding FGCCGAGGAGAATGTGGACTTCCG Fgf18 coding R SP6GCGATTTAGGTGACACTATAGCTAGCCGGGGTGAGTGGGG IL6 FCTCTGGTCTTCTGGAGTACC IL6 R T7CGATGTTAATACGACTCACTATAGGGACCATCTGGCTAGGTAACAG Mcp5 FGCTTACTCTTCATCTGCTGC Mcp5 R T7CGATGTTAATACGACTCACTATAGGGCTGGTGAAGTGTTTGCAGG For enzymatic detection of β-galactosidase activity , the frozen sections were postfixed 5 min with 4% paraformaldehyde ( PFA , Electron Microscopy Sciences ) in PBS at RT . After PBS washes , the sections were incubated 2 × 5 min with X-gal buffer ( 2 mM MgCl2 , 0 . 02% NP40 and 0 . 05% deoxycholate in PBS 0 . 1 M pH 7 . 4 ) and then overnight at 37°C in X-gal reaction buffer ( 20 mg/ml X-gal , 5 mM K4Fe ( CN ) 6 and 5 mM K3Fe ( CN ) 6 in X-gal wash buffer ) . After PBS rinses , the sections were postfixed 10 min in 4% PFA and PBS-rinsed again . The sections were then counterstained with Nuclear Fast Red 0 . 005% for 15 min , serially dehydrated , incubated 3 × 1 min with xylene , and cover-slipped using DPX mountant ( Fisher ) . Sections were incubated in citrate buffer ( 10 mM citric acid , 0 . 05% Tween 20 , pH 6 . 0 ) for 15 min at 90°C , allowed to cool down , washed with PBSTx ( PBS containing 0 . 1% Triton X-100 ) , blocked with 5% BSA in PBSTx 30 min at RT , and incubated with the primary antibody o/n at 4°C ( see list of antibodies below ) . After PBSTx washes , incubation with Alexa647- and/or Alexa555-conjugated secondary antibodies ( Molecular Probes , 1/500 in PBSTx with DAPI ) was performed for 1 hr at RT . After PBSTx washes , the slides were mounted with Fluoro-Gel ( Electron Microscopy Sciences ) . When TUNEL staining was included , it was performed after the citrate step , before the BSA blocking . Endogenous biotin was blocked with the Avidin-Biotin blocking kit ( Vector ) , and the in situ cell death detection kit ( Roche ) was subsequently used following manufacturer instructions . Biotin-tagged DNA nicks were revealed with Alexa488- or Alexa647-conjugated streptavidin ( Molecular Probes , 1/1000 ) during the incubation with the secondary antibody . The antibodies used were ( description , vendor catalog# , dilution ) : p-S6 ( Ser235/236 , Cell Signaling #2211S , 1/200 ) , p-STAT3 ( Tyr705 , Cell Signaling #9145P , 1/200 ) , LEF1 ( Cell Signaling #2230P , 1/200 ) , HB-EGF ( Diphtheria Toxin Receptor , R and D #AF259NA , 1/300 ) , CD31 ( clone MEC 13 . 3 , BD Pharmingen #550274 , 1/300 ) , LY6B ( clone 7/4 , Cedarlane #CL8993AP , 1/100 ) , Collagen Type I ( Rockland #600-401-103-0 . 1 , 1/200 ) , OSX ( Abcam #ab22552 , 1/500 ) , p-FRS2 ( Tyr436 , R and D #AF5126SP ) , p-EGFR ( Tyr1068 , Abcam #ab40815 , 1/200 ) , SOCS3 ( Abcam #ab16030 , 1/100 ) , FABP4 ( clone D25B3 , Cell Signaling #3544 , 1/200 ) , CD44 ( clone IM7 , BD Pharmingen #550538 , 1/500 ) , CD90 ( clone G7 , eBioscience #14-0901-81 , 1/100 ) . 5 mg/ml EdU in PBS was injected subcutaneously ( 50 µg/g body weight ) 1 . 5 hr before euthanizing the mice . EdU was detected using the Click-iT Alexa488 Imaging Kit ( Thermo Fisher Scientific , Waltham , MA , C10337 ) , once the immuno-histochemistry and/or TUNEL staining were finished on the same slides . The fraction of nuclei that were positive for EdU in the proliferative zone of the GP was determined semi-automatically , using Cell Profiler . The number of TUNEL+ cells in the osteochondral junction and the perichondrial groove of Ranvier were counted manually using Fiji/ImageJ . Bright-field and fluorescence images were taken on a Zeiss inverted microscope ( Observer . Z1 ) using Axiovision software ( Zeiss ) . Mosaic pictures were automatically reconstructed from individual 10x ( brightfield ) or 20x ( fluorescence ) tiles . In some cases , whole slide imaging ( WSI ) was performed using a Nanozoomer S210 slide scanner ( Hamamatsu , Japan ) . When data were available for control and experimental animals , a left/right ratio was calculated for both and compared by an unpaired Mann-Whitney test ( one variable and two conditions ) , or by one-way ANOVA ( one variable and ≥3 conditions ) or by two-way ANOVA ( two variables and two or more conditions ) . When left and right parameters were compared within experimental animals only , a paired two-tailed t-test was used . The data met the assumptions of the tests ( e . g . normal distribution by Shapiro-Wilk test ) . F-test was used to test that the variance was similar between the groups compared . For rescue experiments , animals were assigned to control and experimental groups such that both groups had similar distributions of initial body weight . For each experiment , the minimum sample size was estimated using an online tool ( http://powerandsamplesize . com/Calculators ) , based on the average SD observed in pilot experiments , to achieve an effect size of 3% ( left/right bone length ratio ) , or 25% ( rest of parameters ) , with a power of 0 . 8 and a 95% confidence interval . For comparison of qualitative expression , a minimum of two specimens per stage and five across several stages were used . The investigator measuring bone length was blinded to the treatment/genotype of the specimens . No blinding was done for other measurements . Most analyses were done with Prism 7 . 0 software . A previously described protocol ( Agoston et al . , 2007 ) was slightly adapted . Briefly , tibiae were obtained from P1 pups , dissected free of soft tissues and allowed to recover from dissection for 6 hr in 24-well plates with serum-free DMEM ( Gibco ) containing 0 . 2% Bovine Serum Albumin ( BSA ) , 0 . 5 mM L-glutamine , 40 U/ml penicillin/streptomycin ( Gibco ) , 0 . 05 mg/ml ascorbic acid ( Sigma ) and 1 mM betaglycerophosphate ( Sigma ) . Different quantities of DT in PBS were then added , followed by 24 hr incubation . Tibiae were then fixed in PFA and processed for histological analysis . We thank the Joyner lab for scientific discussions . We are grateful to Juanma González-Rosa for comments on the manuscript and suggesting the use of Pitx2-Cre ( kind gift of Hiroshi Hamada ) , and Chitra Dahia for insightful discussions , critical reading of the manuscript , providing access to the µCT scanner at the Hospital for Special Surgery and for interpretation of the data . This work was supported by grant R21HD083860 ( NIH-NICHD ) to ALJ , a National Cancer Institute Cancer Center Support Grant [P30 CA008748] to MSKCC , and by postdoctoral fellowships from HFSP and the Revson Foundation to ARD .	As bones grow , their size is carefully controlled and coordinated with the growth of the other organs in the body . The mechanisms that control organ size also help the body to recover from injury , and play a key role in controlling body size and proportions . Over the course of evolution , these mechanisms have likely changed to produce the distinct body sizes and proportions seen in humans and other animals . Despite their importance , it is not well understood how signals from both inside and outside an organ work together to regulate its size . In growth disorders this signaling goes wrong , which can lead to a person having unusual proportions such as a very short stature or having one leg shorter than the other . Currently , most growth disorders that affect leg proportions are treated with painful surgical procedures . Researchers would like to know how bone growth is affected by signals from the surrounding tissues because this could help them to develop new non-invasive treatments for these conditions . Long bones , for example those in the leg , grow from structures near their ends called growth plates . Roselló-Díez et al . have now engineered mice in which an injury shortly after birth caused cells in the knee in the rear left leg to die off . At the same time , the rear right leg of the mice developed as normal , allowing the growth of the two legs to be compared . Roselló-Díez et al . found that the left leg of these mice grew more slowly than the right leg , even though none of the cells in the growth plate of the left leg bone had been damaged . Further investigation revealed that this was because the injury caused an imbalance between the growth-promoting and growth-restricting signals that are produced by the fat pad and articular cartilage in the knee joint . Restoring the lost balance allowed the left leg bone to grow to a more normal length . In the future , boosting bone growth signals might provide a way to treat conditions like dwarfism or leg-length discrepancies . Understanding how different tissues influence body proportions could also help researchers to investigate how different animals evolved different body proportions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods", "Acknowledgments" ]	[ "developmental", "biology" ]	2017	Altered paracrine signaling from the injured knee joint impairs postnatal long bone growth
The neural circuits responsible for animal behavior remain largely unknown . We summarize new methods and present the circuitry of a large fraction of the brain of the fruit fly Drosophila melanogaster . Improved methods include new procedures to prepare , image , align , segment , find synapses in , and proofread such large data sets . We define cell types , refine computational compartments , and provide an exhaustive atlas of cell examples and types , many of them novel . We provide detailed circuits consisting of neurons and their chemical synapses for most of the central brain . We make the data public and simplify access , reducing the effort needed to answer circuit questions , and provide procedures linking the neurons defined by our analysis with genetic reagents . Biologically , we examine distributions of connection strengths , neural motifs on different scales , electrical consequences of compartmentalization , and evidence that maximizing packing density is an important criterion in the evolution of the fly’s brain . Table 1 shows the hierarchy of the named brain regions that are included in the hemibrain . Table 2 shows the primary regions that are at least 50% included in the hemibrain sample , their approximate size , and their completion percentage . Our names for brain regions follow the conventions of Ito et al . , 2014 with the addition of ‘ ( L ) ’ or ‘ ( R ) ’ to indicate whether the region ( most of which occur on both sides of the fly ) has its cell bodies in the left or right , respectively . The mushroom body ( Tanaka et al . , 2008; Aso et al . , 2014 ) and central complex ( Wolff et al . , 2015; Wolff and Rubin , 2018 ) are further divided into finer compartments . Appendix 1—table 6 provide the list of identified neuron types and their naming schemes . These include newly identified sensory inputs and motor outputs . The nature of the proofreading process allows us to improve the data even after their initial publication . Our initial data release was version v1 . 0 ( Xu et al . , 2020c ) . Version v1 . 1 is now available , including improvements such as better accuracy , more consistent cell naming and typing , and inclusion of anatomical names for central complex neurons . The old version ( s ) remain online and available , to allow reproducibility of older analyses , but we strongly recommend all new analyses use the latest version . The analyses in this article , and in the corresponding articles on the mushroom body and central complex , are based on version v1 . 1 , unless otherwise noted . This research focused on the neurons of the brain and the chemical synapses between them . Every step in our process , from staining and sample preparation through segmentation and proofreading , has been optimized with this goal in mind . While neurons and their chemical synapses are critical to brain operation , they are far from the full story . Other contributors , known to be important , could not be included in our study , largely for technical reasons . Among these are gap junctions , glia , and structures internal to the cell such as mitochondria . Gap junctions , or electrical connections between neurons , are difficult to reliably detect by FIB-SEM under the best of circumstances and not detectable at the low ( for EM ) resolution needed to complete this study in a reasonable amount of time . Their contribution to the connectome will need to be established through other means - see the section on future research . Glial cells were difficult to segment , due to both staining differences and convoluted morphologies . We identified the volumes where they exist ( a glia ’mask’ , which allows these regions to be color-coded when viewed in NeuroGlancer ) but did not separate them into cells . Structures internal to the neurons , except for synapses , are not considered here even though many are visible in our EM preparation . The most obvious example is mitochondria . Again , we have identified many of them so we could evaluate their effect on segmentation , but they are not included in our connectome . Finally , autapses ( synapses from a neuron onto itself ) are known to exist in Drosophila , but are sufficiently rare that they fall well below the rate of false positives in our automated synapse detection . Therefore most of the putative autapses are false positives , and we do not include them in our connectivity data . Most accounts of neurobiology define the operation of the mammalian nervous system with , at most , only passing reference to invertebrate brains . Fly ( or other insect ) nervous systems differ from those of vertebrates in several aspects ( Meinertzhagen , 2016b ) . Some main differences include: Producing a connectome comprising reconstructed neurons and the chemical synapses between them required several steps . The first step , preparing a fly brain and imaging half of its center , produced a dataset consisting of 26 teravoxels of data , each with 8 bits of grayscale information . We applied numerous machine-learning algorithms and over 50 person-years of proofreading effort over ≈2 calendar years to extract a variety of more compact and useful representations , such as neuron skeletons , synapse locations , and connectivity graphs . These are both more useful and much smaller than the raw grayscale data . For example , the connectivity could be reasonably summarized by a graph with ≈25 , 000 nodes and ≈3 million edges . Even when the connections were assigned to different brain regions , such a graph took only 26 MB , still large but roughly a million fold reduction in data size . Many of the supporting methods for this reconstruction have been recently published . Here , we briefly survey each major area , with more details reported in the companion papers . Major advances include: Each of these is explained in more detail in the following sections and , where necessary , in the appendix . The companion papers are ‘The connectome of the Drosophila melanogaster mushroom body: implications for function’ ( Li et al . , 2020 ) and ‘A complete synaptic-resolution connectome of the Drosophila melanogaster central complex’ by Jayaraman , et al . The first steps , fixing and staining the specimen , have been accomplished taking advantage of three new developments . These improved methods allow us to fix and stain a full fly’s brain but nevertheless recover neurons as round profiles with darkly stained synapses , suitable for machine segmentation and automatic synapse detection . We started with a 5-day-old female of wild-type Canton S strain G1 x w1118 , raised on a 12 hr day/night cycle . 1 . 5 hr after lights-on , we used a custom-made jig to microdissect the brain , which was then fixed and embedded in Epon , an epoxy resin . We then enhanced the electron contrast by staining with heavy metals , and progressively lowered the temperature during dehydration of the sample . Collectively , these methods optimize morphological preservation , allow full-brain preparation without distortion ( unlike fast freezing methods ) , and provide increased staining intensity that speeds the rate of FIB-SEM imaging ( Lu et al . , 2019 ) . The hemibrain sample is roughly 250 × 250 × 250 μm , larger than we can FIB-SEM without introducing milling artifacts . Therefore , we subdivided our epoxy-embedded samples into 20-μm-thick slabs , both to avoid artifacts and allow imaging in parallel ( each slab can be imaged in a different FIB machine ) for increased throughput . To be effective , the cut surfaces of the slabs must be smooth at the ultrastructural level and have only minimal material loss . Specifically , for connectomic research , all long-distance processes must remain traceable across sequential slabs . We used an improved version of our previously published ‘hot-knife’ ultrathick sectioning procedure ( Hayworth et al . , 2015 ) which uses a heated , oil-lubricated diamond knife , to section the Drosophila brain into 37 sagittal slabs of 20 μm thickness with an estimated material loss between consecutive slabs of only ∼30 nm – sufficiently small to allow tracing of long-distance neurites . Each slab was re-embedded , mounted , and trimmed , then examined in 3D with X-ray tomography to check for sample quality and establish a scale factor for Z-axis cutting by FIB . The resulting slabs were FIB-SEM imaged separately ( often in parallel , in different FIB-SEM machines ) , and the resulting volume datasets were stitched together computationally . Connectome studies come with clearly defined resolution requirements – the finest neurites must be traceable by humans and should be reliably segmented by automated algorithms ( Januszewski et al . , 2018 ) . In Drosophila , the very finest neural processes are usually 50 nm but can be as little as 15 nm ( Meinertzhagen , 2016a ) . This fundamental biological dimension determines the minimum isotropic resolution requirements for tracing neural circuits . To meet the demand for high isotropic resolution and large volume imaging , we chose the FIB-SEM imaging platform , which offers high isotropic resolution ( <10 nm in x , y , and z ) , minimal artifacts , and robust image alignment . The high-resolution and isotropic dataset possible with FIB-SEM has substantially expedited the Drosophila connectome pipeline . Compared to serial-section imaging , with its sectioning artifacts and inferior Z-axis resolution , FIB-SEM offers high-quality image alignment , a smaller number of artifacts , and isotropic resolution . This allows higher quality automated segmentation and makes manual proofreading and correction easier and faster . At the beginning , deficiencies in imaging speed and system reliability of any commercial FIB-SEM system capped the maximum possible image volume to less than 0 . 01% of a full fly brain , problems that persist even now . To remedy them , we redesigned the entire control system , improved the imaging speed more than 10x , and created innovative solutions addressing all known failure modes , which thereby expanded the practical imaging volume of conventional FIB-SEM by more than four orders of magnitude from 103μm3 to 3⋅107 μm3 , while maintaining an isotropic resolution of 8 × 8 × 8 nm voxels ( Xu et al . , 2017; Xu et al . , 2020a ) . In order to overcome the aberration of a large field of view ( up to 300 μm wide ) , we developed a novel tiling approach without sample stage movement , in which the imaging parameters of each tile are individually optimized through an in-line auto focus routine without overhead ( Xu et al . , 2020b ) . After numerous improvements , we have transformed the conventional FIB-SEM from a laboratory tool that is unreliable for more than a few days of imaging to a robust volume EM platform with effective long-term reliability , able to perform years of continuous imaging without defects in the final image stack . Imaging time , rather than FIB-SEM reliability , is now the main impediment to obtaining even larger volumes . In our study here , the Drosophila 'hemibrain' , 13 consecutive hot-knifed slabs were imaged using two customized enhanced FIB-SEM systems , in which an FEI Magnum FIB column was mounted at 90° upon a Zeiss Merlin SEM . After data collection , streaking artifacts generated by secondary electrons along the FIB milling direction were computationally removed using a mask in the frequency domain . The image stacks were then aligned using a customized version of the software platform developed for serial section transmission electron microscopy ( Zheng et al . , 2018; Khairy et al . , 2018 ) , followed by binning along the z-axis to form the final 8 × 8 × 8 nm3 voxel datasets . Milling thickness variations in the aligned series were compensated using a modified version of the method described by Hanslovsky et al . , 2017 , with the absolute scale calibrated by reference to the MicroCT images . The 20 μm slabs generated by the hot-knife sectioning were re-embedded in larger plastic tabs prior to FIB-SEM imaging . To correct for the warping of the slab that can occur in this process , methods adapted from Kainmueller ( Kainmueller et al . , 2008 ) were used to find the tissue-plastic interface and flatten each slab’s image stack . The series of flattened slabs was then stitched using a custom method for large-scale deformable registration to account for deformations introduced during sectioning , imaging , embedding , and alignment ( Saalfeld et al . in prep ) . These volumes were then contrast adjusted using slice-wise contrast limited adaptive histogram equalization ( CLAHE ) ( Pizer et al . , 1987 ) , and converted into a versioned database ( Distributed , Versioned , Image-oriented Database , or DVID ) ( Katz and Plaza , 2019 ) , which formed the raw data for the reconstruction , as illustrated in Figure 2 . Computational reconstruction of the image data was performed using flood-filling networks ( FFNs ) trained on roughly five billion voxels of volumetric ground truth contained in two tabs of the hemibrain dataset ( Januszewski et al . , 2018 ) . Initially , the FFNs generalized poorly to other tabs of the hemibrain , whose image content had different appearances . Therefore , we adjusted the image content to be more uniform using cycle-consistent generative adversarial networks ( CycleGANs ) ( Zhu et al . , 2017 ) . Specifically , ‘generator’ networks were trained to alter image content such that a second ‘discriminator’ network was unable to distinguish between image patches sampled from , for example , a tab that contained volumetric training data versus a tab that did not . A cycle-consistency constraint was used to ensure that the image transformations preserved ultrastructural detail . The improvement is illustrated in Figure 3 . Overall , this allowed us to use the training data from just two slabs , as opposed to needing training data for each slab . FFNs were applied to the CycleGAN-normalized data in a coarse-to-fine manner at 32 × 32 × 32 nm3 and 16 × 16 × 16 nm3 , and to the CLAHE-normalized data at the native 8 × 8 × 8 nm3 resolution , in order to generate a base segmentation that was largely over-segmented . We then agglomerated the base segmentation , also using FFNs . We aggressively agglomerated segments despite introducing a substantial number of erroneous mergers . This differs from previous algorithms , which studiously avoided merge errors since they were so difficult to fix . Here , advances in proofreading methodology described later in this report enabled efficient detection and correction of such mergers . We evaluated the accuracy of the FFN segmentation of the hemibrain using metrics for expected run length ( ERL ) and false merge rate ( Januszewski et al . , 2018 ) . The base segmentation ( i . e . the automated reconstruction prior to agglomeration ) achieved an ERL of 163 μm with a false merge rate of 0 . 25% . After ( automated ) agglomeration , run length increased to 585 μm but with a false merge rate of 27 . 6% ( i . e . nearly 30% of the path length was contained in segments with at least one merge error ) . We also evaluated a subset of neurons in the volume , ∼500 olfactory PNs and mushroom body KCs chosen to roughly match the evaluation performed in Li et al . , 2019 which yielded an ERL of 825 μm at a 15 . 9% false merge rate . Accurate synapse identification is central to our analysis , given that synapses form both a critical component of a connectome and are required for prioritizing and guiding the proofreading effort . Synapses in Drosophila are typically polyadic , with a single presynaptic site ( a T-bar ) contacted by multiple receiving dendrites ( most with PSDs ) as shown in Figure 4A . Initial synapse prediction revealed that there are over 9 million T-bars and 60 million PSDs in the hemibrain . Manually validating each one , assuming a rate of 1000 connections annotated per trained person , per day , would have taken more than 230 working years . Given this infeasibility , we developed machine learning approaches to predict synapses as detailed below . The results of our prediction are shown in Figure 4B , where the predicted synapse sites clearly delineate many of the fly brain regions . Given the size of the hemibrain image volume , a major challenge from a machine learning perspective is the range of varying image statistics across the volume . In particular , model performance can quickly degrade in regions of the data set with statistics that are not well-captured by the training set ( Buhmann et al . , 2019 ) . To address this challenge , we took an iterative approach to synapse prediction , interleaving model re-training with manual proofreading , all based on previously reported methods ( Huang et al . , 2018 ) . Initial prediction , followed by proofreading , revealed a number of false positive predictions from structures such as dense core vesicles which were not well-represented in the original training set . A second filtering network was trained on regions causing such false positives , and used to prune back the original set of predictions . We denote this pruned output as the ‘initial’ set of synapse predictions . Based on this initial set , we began collecting human-annotated dense ground-truth cubes throughout the various brain regions of the hemibrain , to assess variation in classifier performance by brain region . From these cubes , we determined that although many regions had acceptable precision , there were some regions in which recall was lower than desired . Consequently , a subset of cubes available at that time was used to train a new classifier focused on addressing recall in the problematic regions . This new classifier was used in an incremental ( cascaded ) fashion , primarily by adding additional predictions to the existing initial set . This gave better performance than complete replacement using only the new classifier , with the resulting predictions able to improve recall while largely maintaining precision . As an independent check on synapse quality , we also trained a separate classifier ( Buhmann et al . , 2019 ) , using a modified version of the ‘synful’ software package . Both synapse predictors give a confidence value associated with each synapse , a measure of how firmly the classifier believes the prediction to be a true synapse . We found that we were able to improve recall by taking the union of the two predictor’s most confident synapses , and similarly improve precision by removing synapses that were low confidence in both predictions . Figure 5A and B show the results , illustrating the precision and recall obtained in each brain region . Since machine segmentation is not perfect , we made a concerted effort to fix the errors remaining at this stage by several passes of human proofreading . Segmentation errors can be roughly grouped into two classes - ‘false merges’ , in which two separate neurons are mistakenly merged together , and ‘false splits’ , in which a single neuron is mistakenly broken into several segments . Enabled by advances in visualization and semi-automated proofreading using our Neu3 tool ( Hubbard et al . , 2020 ) , we first addressed large false mergers . A human examined each putative neuron and determined if it had an unusual morphology suggesting that a merge might have occurred , a task still much easier for humans than machines . If judged to be a false merger , the operator identified discrete points that should be on separate neurons . The shape was then resegmented in real time allowing users to explore other potential corrections . Neurons with more complex problems were then scheduled to be re-checked , and the process repeated until few false mergers remained . In the next phase , the largest remaining pieces were merged into neuron shapes using a combination of machine-suggested edits ( Plaza , 2014 ) and manual intuition , until the main shape of each neuron emerged . This requires relatively few proofreading decisions and has the advantage of producing an almost complete neuron catalog early in the process . As discussed below , in the section on validation , emerging shapes were compared against genetic/optical image libraries ( where available ) and against other neurons of the same putative type , to guard against large missing or superfluous branches . These procedures ( which focused on higher-level proofreading ) produced a reasonably accurate library of the main branches of each neuron , and a connectome of the stronger neuronal pathways . At this point , there was still considerable variations among the brain regions , with greater completeness achieved in regions where the initial segmentation performed better . Finally , to achieve the highest reconstruction completeness possible in the time allotted , and to enable confidence in weaker neuronal pathways , proofreaders connected remaining isolated fragments ( segments ) to already constructed neurons , using NeuTu ( Zhao et al . , 2018 ) and Neu3 ( Hubbard et al . , 2020 ) . The fragments that would result in largest connectivity changes were considered first , exploiting automatic guesses through focused proofreading where possible . Since proofreading every small segment is still prohibitive , we tried to ensure a basic level of completeness throughout the brain with special focus in regions of particular biological interest such as the central complex and mushroom body . In a parallel effort to proofreading , the sample was annotated with discrete brain regions . Our progression in mapping the cells and circuits of the fly’s brain bears formal parallels to the history of mapping the earth , with many territories that are named and with known circuits , and others that still lack all or most of these . For the hemibrain dataset , the regions are based on the brain atlas in Ito et al . , 2014 . The dataset covers most of the right hemisphere of the brain , except the optic lobe ( OL ) , periesophageal neuropils ( PENP ) and gnathal ganglia ( GNG ) , as well as part of the left hemisphere ( Table 2 ) . It covers about 36% of all synaptic neuropils by volume , and 54% of the central brain neuropils . We examined innervation patterns , synapse distribution , and connectivity of reconstructed neurons to define the neuropils as well as their boundaries on the dataset . We also made necessary , but relatively minor , revisions to some boundaries by considering anatomical features that had not been known during the creation of previous brain maps , while following the existing structural definitions ( Ito et al . , 2014 ) . We also used information from synapse point clouds , a predicted glial mask , and a predicted fiber bundle mask to determine boundaries of the neuropils ( Figure 6A ) . The brain regions of the fruit fly ( Figure 6 , B and C ) include synaptic neuropils and non-synaptic fiber bundles . The non-synaptic cell body layer on the brain surface , which contains cell bodies of the neurons and some glia , surrounds these structures . The synaptic neuropils can be further categorized into two groups: delineated and diffuse neuropils . The delineated neuropils have distinct boundaries throughout their surfaces , often accompanied by glial processes , and have clear internal structures in many cases . They include the antennal lobe ( AL ) , bulb ( BU ) , as well as the neuropils in the optic lobe ( OL ) , mushroom body ( MB ) , and central complex ( CX ) . Remaining are the diffuse neuropils , sometimes referred to as terra incognita , since most have been less investigated than the delineated neuropils . Since many of the terra incognita neuropils are not clearly partitioned from each other by solid boundaries such as glial walls , it is important to evaluate if the current boundaries reflect anatomical and functional compartments of the brain . To check our definitions , which are mostly based on morphology , we compute metrics for each boundary between any two adjacent neuropil regions . The first is the area of each boundary , in square microns , as shown in Figure 8A . The map shows results for brain regions that are over 75% in the hemibrain region , restricted to right regions with exception to the asymmetric AB ( L ) . By restricting our analysis to the right part of the hemibrain , we hopefully minimize the effect of smaller , traced-but-truncated neuron fragments on our metric . Next , for each boundary , we compute the number of ‘excess’ neuron crossings by traced neurons , where excess crossings are defined as 0 for a neuron that does not cross the boundary , and n-1 for a neuron crosses the same boundary n times . There is no contribution to the metric from neurons that cross a boundary once , since most such crossings are inevitable no matter where the boundary is placed . Figure 8B shows the number of excess crossings normalized by the area of boundary . A bigger dot indicates a potentially less well-defined boundary . We spot checked many of the instances and in general note that the brain regions with high excess crossings per area , such as those in SNP , INP and VLNP , tend to have less well-defined boundaries . In particular , the boundaries at SMP/CRE , CRE/LAL , SMP/SIP , and SIP/SLP have worse scores , indicating these boundaries may not reflect actual anatomical and functional segregation of the neuropils . These brain regions were defined based on the arborization patterns of characteristic neuron types , but because neurons in the terra incognita neuropils tend to be rather heterogeneous , there are many other neuron types that do not follow these boundaries . The boundary between the FB and the AB also has a high excess crossing score , suggesting the AB is tightly linked to the neighboring FB . Defining cell types for groups of similar neurons is a time-honored means to help to understand the anatomical and functional properties of a circuit . Presumably , neurons of the same type have similar circuit roles . However , the definition of what is a distinct cell type and the exact delineation between one cell type and another remains inherently subjective and represents a classic taxonomic challenge , pitting ‘lumpers’ against ‘splitters’ . Therefore , despite our best efforts , we recognize that our typing of cells may not be identical to that proposed by other experts . We expect future revisions to cell type classification , especially as additional dense connectome data become available . One common method of cell type classification , used in flies , exploits the GAL4 system to highlight the morphology of neurons having similar gene expression ( Jenett et al . , 2012 ) . Since these genetic lines are imaged using fluorescence and confocal microscopy , we refer to them as ‘light lines’ . Where they exist and are sufficiently sparse , light lines provide a key method for identifying types by grouping morphologically similar neurons together . However , there are no guarantees of coverage , and it is difficult to distinguish between neurons of very similar morphology but different connectivity . We enhanced the classic view of morphologically distinct cell types by defining distinct cell types ( or sub-types ) based on both morphology and connectivity . Connectivity-based clustering often reveals clear cell type distinctions , even when genetic markers have yet to be found , or when the neuronal morphologies of different types are hardly distinguishable in optical images . For example , the two PEN ( protocerebral bridge - ellipsoid body - noduli ) neurons have very similar forms but quite distinct inputs ( Figure 9; Turner-Evans et al . , 2019 ) Confirming their differences , PEN1 and PEN2 neurons , in fact , have been shown to have different functional activity ( Green et al . , 2017 ) . Based on our previous definition of cell type , many neurons exhibit a unique morphology or connectivity pattern at least within one hemisphere of the brain ( with a matching type in the other hemisphere in most cases ) . Because our hemibrain volume covers only the right-side examples of ipsilaterally-projecting neurons , and the contralateral arborizations of bilaterally-projecting neurons arising from the left side of the brain were in practice very difficult to match to neurons in the right side , many partial neurons were therefore left uncategorized . As a result , many neuron types consisting of a distinct morphology and connectivity have only a single example in our reconstruction . It is possible to provide coarser groupings of neurons . For instance , most cell types are grouped by their cell body fiber representing a distinct clonal unit , which we discuss in more detail below . Furthermore , each neuron can be grouped with neurons that innervate similar brain regions . In this paper , we do not explicitly formalize this higher level grouping , but data on the innervating brain regions can be readily mined from the dataset . Assigning types and names to the more than 20 , 000 reconstructed cells was a difficult undertaking . Less than 20% of neuron types found in our data have been described in the literature , and half of our neurons have no previously annotated type . Adding to the complexity , prior work focused on morphological similarities and differences , but here we have , for the first time , connectivity information to assist in cell typing as well . Many cell types in well-explored regions have already been described and named in the literature , but existing names can be both inconsistent and ambiguous . The same cell type is often given differing names in different publications , and conversely , the same name , such as PN for projection neuron , is used for many different cell types . Nonetheless , for cell types already named in the literature ( which we designate as published cell types , many indexed , with their synonyms , at http://virtualflybrain . org ) , we have tried to use existing names . In a few cases , using existing names created conflicts , which we have had to resolve . ‘R1’ , for example , has long been used both for photoreceptor neurons innervating the lamina and medulla , and ring neurons in the ellipsoid body of the central complex . Similarly , ‘LN’ has been used to refer to lateral neurons in the circadian clock system , ‘local neurons’ in the antenna lobe , and LAL-Nodulus neurons in the central complex . To resolve these conflicts , the ellipsoid body ring neurons are now named ’ER1’ instead of ‘R1’ , and the nodulus neurons are now ‘LNO’ and ’GLNO’ instead of ‘LN’ and ‘GLN’ . The names of the antennal lobe local neuron are always preceded by lowercase letters for their cell body locations to differentiate them from the clock neuron names , for example , lLN1 versus LNd . Similarly , ‘dorsal neurons’ of the circadian clock system and ‘descending neurons’ in general , both previously abbreviated as ‘DN’ , are distinguished by the following characters - numbers for the clock neurons ( e . g . DN1 ) and letters for descending neurons ( e . g . DNa01 ) . Overall , we defined a ‘type’ of neurons as either a single cell or a group of cells that have a very similar cell body location , morphology , and pattern of synaptic connectivity . We were able to trace from arborizations to the cell bodies for 15 , 912 neurons in the hemibrain volume , ≈85% of which are located in the right side of the brain while the rest are in the medialmost part of the left-side brain . We classified these neurons in several steps . The first step classified all cells by their lineage , grouping neurons according to their bundle of cell body fibers ( CBFs ) . Neuronal cell bodies are located in the cell body layer that surrounds the brain , and each neuron projects a single CBF towards synaptic neuropils . In the central brain , cell bodies of clonally related neurons deriving from a single stem cell ( called a neuroblast in the insect brain ) tend to form clusters , from each of which arises one or several bundles of CBFs . Comparing the location , trajectory , and the combined arborization patterns of all the neurons that arise from a particular CBF with the light microscopy ( LM ) image data of the neuronal progeny that derive from single neuroblasts ( Ito et al . , 2013; Yu et al . , 2013 ) , we confirmed that the neurons of each CBF group belong to a single lineage . We carefully examined the trajectory and origins of CBFs of the 15 , 752 neurons on the right central brain and identified 192 distinct CBF bundles . Neurons arising from four specific CBF bundles arborize primarily in the contralateral brain side , which is not fully covered in the hemibrain volume . We characterized these neurons using the arborization patterns in the right-side brain that are formed by the neurons arising from the left-side CBFs . The CBF bundles and associated neuronal cell body clusters were named according to their location ( split into eight sectors of the brain surface with the combination of Anterior/Posterior , Ventral/Dorsal , and Medial/Lateral ) and a number within the sector given according to the size of cell population . Thus , CBF group ADM01 is the group with the largest number of neurons in the Anterior Dorsal Medial sector of the brain’s surface ( see the cellBodyFiber field of the Neuprint database explained later ) . For the neurons of the four CBF bundles that arborize primarily in the contralateral brain side - AVM15 , 18 , 19 , and PVM10 - we indicated CBF information in the records of the left-side neurons . Among the 192 bundles , 155 matched the CBF bundles of 92 known and six newly identified clonal units ( Ito et al . , 2013; Yu et al . , 2013 ) , a population of neurons and neuronal circuits derived from a single stem cell . The remaining 37 CBF bundles are minor populations and most likely of embryonic origin . In addition , we found 80 segregated cell body fiber bundles ( SCB001-080 , totalling 112 cells ) with only one or two neurons per bundle . Many of them are also likely of embryonic origin . We were able to identify another 6682 neurons that were not traced up to their cell bodies . For the neurons that arise from the contralateral side , we gave matching neuron names and associated CBF information , provided their specific arborization patterns gave us convincing identity information by comparison with cells that we identified in the right side of the brain . For the neurons arising from the ventralmost part of the brain outside of the hemibrain volume , we identified and gave them names if we could find convincingly specific arborization patterns , even if the CBF and cell body location data were missing . Sensory neurons that project to the specific primary sensory centers were also identified insofar as possible . In total , we typed and named 22 , 594 neurons . Different stem cells sometimes give rise to neurons with very similar morphologies . We classified these as different types because of their distinct developmental origin and slightly different locations of their cell bodies and CBFs . Thus , the next step in neuron typing was to cluster neurons within each CBF group . This process consisted of three further steps , as shown in Figure 10 . First , we used NBLAST ( Costa et al . , 2016 ) to subject all the neurons of a particular CBF group to morphology-based clustering . Next , we used CBLAST , a new tool to cluster neurons based on synaptic connectivity ( see the next section ) . This step is an iterative process , using neuron morphology as a template , regrouping neurons after more careful examination of neuron projection patterns and their connections . Neurons with similar connectivity characteristics but with distinguishable shapes were categorized into different morphology types . Those with practically indistinguishable shapes but with different connectivity characteristics were categorized into connectivity types within a morphology type . Finally , we validated the cell typing with extensive manual review and visual inspection . This review allowed us both to confirm cell type identity and help ensure neuron reconstruction accuracy . In total we identified 5229 morphology types and 5609 connectivity types in the hemibrain dataset . ( See Table 3 for the detailed numbers and Appendix 1—table 6 for naming schemes for various neuron categories . ) In spite of this general rule , we assigned the same neuron type name for the neurons of different lineages in the following four cases . ‘Lumping’ versus ‘splitting’ is a difficult problem for classification . Following the experiences of taxonomy , we opted for splitting when we could not obtain convincing identity information , a decision designed to ease the task of future researchers . If we split two similar neuron types into Type 1 and Type 2 , then there is a chance future studies might conclude that they are actually subsets of a common cell type . If so , then at that time we can simply merge the two types as Type 1 , and leave the other type name unused , and publish a lookup table of the lumping process to keep track of the names that have been merged . The preceding studies can then be re-interpreted as the analyses on the particular subsets of a common neuron type . If , on the contrary , we lump the two similar neurons into a common type , then a later study finds they are actually a mixture of two neuron types , then it would not be possible to determine which of the two neuron types , or a mixture of them , was analyzed in preceding studies . In the hemibrain , using the defined brain regions ( neuropils ) and reference to known expression driver strains , we were able to assign a cell type to many cells . Where possible , we matched previously defined cell types with those labeled in light data using a combination of Neuprint , an interactive analysis tool ( described later ) , Color_MIP_mask search ( Otsuna et al . , 2018 ) , and human recognition to find the matching cell types , especially in well-explored neuropils such as the mushroom body and central complex , where abundant cell type information was already available and where we are more confident in our anatomical expertise ( see the accompanying MB and CX papers ) . Even though most of the cell types in the MB and CX were already known , we still found new cell types in these regions , an important vindication of our methods . In these cases , we tried to name them using the existing schemes for these regions , and further refined these morphological groupings with relevant information on connectivity . To give names to neuron types , we categorized neurons that share certain characteristics into groups and distinguished individual types by adding identifiers ( IDs ) with numbers , uppercase letters , or combinations of these . ( See Appendix 1—table 6 for the summary of the naming schemes of all the neuron types ) . For example , the tangential neurons of the fan-shaped body ( FB ) of the central complex were grouped as ‘FB’ , and an ID of their primary innervating FB layer was added with numbers 1–9 . Different types of neurons that arborize in each layer were further distinguished by uppercase letters . Thus , for example the FB7B neurons are the second type of tangential neurons that arborize in the seventh layer of FB . We also used uppercase letters to subdivide the neuron types that have previously been reported as a single type to keep naming consistency . For example , a population of antennal lobe local neurons that has been known as LN2L was divided into five morphology subtypes as lLN2F , 2P , 2R , 2S , and 2T for their full , patchy , regional , star-like and tortuous arborization patterns while still indicating that they are part of the LN2 population . The letter ‘L’ at the end of the previous name , which referred to the cell body location on the lateral side of the AL , was moved in front of LN to keep consistency with the established naming scheme for the olfactory projection neurons ( e . g . , DA1_lPN ) . Neuron types that are known to exist were sometimes not identified in the particular brain sample used for the hemibrain EM dataset . In such cases , the corresponding ID numbers were kept blank . For example , the MBON08 neurons were not identified in the current sample and the number was therefore skipped . Although the morphology type names generally end with either numbers or uppercase letters , in a few cases lower case letters were used for distinguishing morphological subtypes to keep the naming convention of that cell group consistent . For example , subtypes of the neurons in the optic lobes were distinguished as , for example LC28a and LC28b , because such subtypes of the optic lobe neurons have historically been distinguished by lowercase letters . If neurons of near-identical morphology could be further subdivided into different connectivity types , they were suffixed with an underscore and a lowercase letter , for example FB2F_a , FB2F_b , and FB2F_c . A neuron type without such a suffix consists of a single connectivity type . The cell type names are indicated in the ‘type’ field of the NeuPrint database . In the ‘instance’ field , information about the side of the neuronal cell body , when it is known , is added as _R and _L after the cell type name . The name of the CBF group is indicated in the ‘cellBodyFiber’ field of the right-side neurons except for those that belong to AVM15 , 18 , 19 , and PVM10 groups , and in the same field of the left-side neurons for those four CBF groups . For the rest of the neurons , the CBF information is shown in the ‘instance’ field in parentheses when it is known . Across the brain , we looked for neurons that correspond to already known cell types , and as far as possible gave them consistent names . These include: olfactory projection neurons and local neurons associated with the antennal lobe ( Tanaka et al . , 2012; Bates et al . , 2020; Marin et al . , 2020 ) , neurons associated with the lateral horn ( Dolan et al . , 2019; Frechter et al . , 2019; Bates et al . , 2020 ) , aminergic and peptidergic neurons ( Bergland et al . , 2012; Busch et al . , 2009; Mao and Davis , 2009; Martelli et al . , 2017; Pech et al . , 2013; Pooryasin and Fiala , 2015; Shao et al . , 2017; White et al . , 2010 ) , neurons associated with the circadian clock ( Helfrich-Förster et al . , 2007 ) , and neurons that express the fruitless gene ( Cachero et al . , 2010; Yu et al . , 2010; Zhou et al . , 2014; Wang et al . , 2020 ) . In some cases , we found candidate neurons that do not precisely match previously identified neurons . For example , in addition to the three cell types that match the octopaminergic ( OA ) neurons OA-ASM1 , 2 and 3 ( Busch et al . , 2009 ) , we found two neuron types in the same location that appear to match some of the tdc2-Gal4 expressing neurons in the FlyCircuit database of single-cell labeling images ( Chiang et al . , 2011 ) . Because of the remaining uncertainty we gave them the canonical names SMP143 and SMP149 , but added ‘Tdc2 ( OA ) -ASM candidates’ in the Notes field . We also found that the FB2B neurons share the same cell body location and appear to match another type of tdc2-Gal4 expressing neurons in the FlyCircuit database . Although OA-immunoreactive neurites have been observed in the FB ( Sinakevitch et al . , 2005 ) , it is not known from where they are derived . Considering that the particular neurons may produce only tyramine ( TA ) but not OA , we added ‘Tdc2 ( TA ) -ASM candidates’ in the Notes . Due to similar considerations , the number of candidate neurons may not match the actual known numbers for many neuron types . For the multiglomerular olfactory projection neurons and local interneurons of the antennal lobe , we devised new naming schemes by expanding the naming scheme of uniglomerular projection neurons , which consists of the contributing antennal lobe glomerulus and the location of the cell body cluster ( Bates et al . , 2020; Marin et al . , 2020 ) . Because the list of contributing glomeruli is not a useful designator for the multiglomerular projection neurons , we used information about the antennal lobe tract ( ALT ) projection pathways instead . Unique type ID numbers were then added at the end of the names of the multiglomerular projection neurons ( 1-92 ) and local neurons ( 1-50 ) . For the local neurons LN1-6 the numbers were kept consistent with the published neuron names ( Tanaka et al . , 2012 ) ; for the newly identified local neurons and for the multiglomerular projection neurons , ID numbers were sorted according to the cell body location from dorsal to ventral . For the neurons associated with the lateral horn , we expanded the existing naming scheme ( names such as PV5a1 ) based on the cell body cluster location ( uppercase letters and first number ) , anatomically associated groups ( lower case letter ) , and individual neuron type ( last number ) , which has previously been applied for ≈30% of the lateral horn neurons ( Frechter et al . , 2019; Bates et al . , 2020 ) . The neuron types that have been defined in the lateral horn sometimes contain slightly larger morphological varieties of neurons than would be categorized as different types in the hemibrain volume . To reconcile this slight discrepancy while keeping the published neuron type names as consistent as possible , in some cases we used suffices _a , _b , etc . , for distinguishing not only the neurons that are different in their connectivity but also those that have minute but distinct morphological differences . Because of this technical issue more neurons are distinguished by suffices in the lateral horn than in other brain regions . In cases where we gave new neuron names to the already known ones , or slightly modified the existing names for the sake of naming scheme consistency , we indicated the most commonly used previous names in the notes field , from where users can look for further synonyms using the Virtual Fly Brain database ( http://virtualflybrain . org ) . For the optic lobe neurons , we categorized only the VPNs based primarily on the specific projection patterns of their axon terminals in the central brain . Newly identified neuron types were given higher numbers than those already used ( Fischbach and Dittrich , 1989; Panser et al . , 2016; Otsuna and Ito , 2006; Hausen , 1984 ) . Neurons that arborize only in the optic lobe are not classified , except for several intrinsic neurons in the lobula , because the hemibrain dataset does not provide enough information about their projection patterns in the optic lobe for conclusive cell typing . Olfactory- , thermo- , and hygro-receptor ( sensory ) neurons were named according to their target glomeruli in the antennal lobe ( Fishilevich and Vosshall , 2005; Couto et al . , 2005; Gallio et al . , 2011; Enjin et al . , 2016; Frank et al . , 2017; Marin et al . , 2020 ) . Some of the auditory receptor neurons ( Johnston’s organ neurons ) were also identified , but their precise target zones in the antennal mechanosensory and motor center ( Kamikouchi et al . , 2006 ) were not determined because of the insufficient information in the hemibrain image volume . The neurons associated with the ocellar ganglion ( OCG ) , a detached ganglion just beneath the ocelli , were categorized into eight types based on the morphology of their terminals in the central brain . Precise classification of OCG neurons is not possible without the projection pattern information in the OCG . To remedy this problem the neurons that share the common projection patterns within the brain were classified as OCG1 , OCG2 , etc . , and when the projection pattern information in the OCG is available they will be classified in more detail as OCG1A , OCG1B , etc . Outside the heavily studied regions , and the neuron types explained above , the fly’s circuits are largely composed of cells of so-far unknown type . Because such neurons , in what is called the terra incognita of the fly brain , account for nearly 70% of the total neuron types , it was necessary to devise a systematic naming scheme to give them names that annotate reasonable morphological characteristics and are easy to pronounce . About 40% of these neurons extend their projections to regions outside of the imaged volume of the hemibrain EM dataset , such as the contralateral brain side , the ventralmost parts of the brain , and the optic lobes . Since whole brain reconstructions of such neurons will soon become available , the naming scheme should provide reasonable names for the neurons that are not fully traceable within the hemibrain image volume . To address this problem , we tested various naming schemes using single-cell LM images of about 500 neuron types in these regions . LM images have much lower spatial resolution but visualize entire projection patterns across the brain compared to the EM data . We found the regions ( neuropils ) of the central brain with the most extensive arborization by counting the voxel numbers of the three-dimensional LM data . We also simulated the numbers of output and input synapses available in the EM data by assessing the number of boutons and spines - characteristic morphology of output and input synaptic sites - in the LM images . Regions with the largest number of output synapses tend to lie on the contralateral side of the brain , out of the hemibrain volume , making it difficult to use EM information as a primary determining factor . Regions with the largest number of input synapses often showed discrepancies between EM and LM images , mainly due to the varying completeness of fine dendritic fragments in the EM data . We found the names based on the neuropils with the largest number of voxels gave the most consistent names , regardless of whether we used the information of the entire brain or only the image area that corresponds to the hemibrain volume . Because the still unmapped fragments of input dendritic arborizations are thin and tiny , with much smaller volumes compared to the already mapped major branches , we found the voxel counts of dendrites are much less affected by potential incompleteness than the counts of input synapses . We then applied the above LM-based naming scheme to the EM data of terra incognita neurons , and found that naming based on EM voxel count matched with either the neuropils with the largest or second-largest number of output or input synapses for more than 95% of the neuron types . For the remaining types , we took the neuropil names with the second largest voxel numbers , which resulted in near-perfect match with the neuron type name and either the region of the most major or second major output/input synapses , making the names reasonable for connectivity analysis . There is one more factor we had to consider . Certain groups of neuron types tend to share common core projection patterns and differ slightly only in the extent of arbors in each neuropil . For functional interpretation it would be more convenient if such neurons were classified into the same category of neuropils . If we gave names simply to individual neuron types , however , such neurons tend to be scattered into various neuropil categories affected by the slight differences of arborization patterns . To address this problem , we performed NBLAST morphological clustering with a higher threshold than used for individual neuron typing , to group the neurons that share the same CBF bundle and rather similar morphology into a common neuropil category . This additional process , however , sometimes caused mismatches between the resulting neuropil name and the most major or second major output/input synapses if the arborization pattern of that neuron type deviates too much from the rest of the group . In these cases we split such neuron types from the group and assigned them into more appropriate neuropil categories . Between 45 and 630 neuron types were assigned into each neuropil category and distinguished with three-digit ID numbers , for example SLP153 and WED048 , using the standard nomenclature abbreviations of the neuropils ( Ito et al . , 2014 ) . We gave sequential numbers to the neuron types that share the same CBF bundles and common core projection patterns so that neurons with similar appearance would be assigned similar names , as far as possible . Within each CBF group , neurons are sorted from the ones with broader and more extensive projections to the ones with restricted local arborizations . Because of this numbering scheme , broadly arborizing neurons have scattered numbers within the number range of each neuropil category , depending on the CBF groups they belong to . Using the workflow of Figure 10 , we identified 22 , 594 neurons with 5229 morphological types and 5609 connectivity types ( Table 3 ) . Over 2000 of these are types with only a single instance , although presumably , for a whole brain reconstruction , most of these types would have partners on the opposite side of the brain . Figure 11 shows the number of distinct neuron types found in different brain regions . Figure 12 shows the distribution of the number of neurons in each cell type . In spite of our extensive efforts , the assignment of type names to neurons is still ongoing . Because we opted for splitting rather than lumping of hard to differentiate cell types , it is possible that some of the neuron types may be merged with others in the future . In such cases , the number that is unused after the merger should not be re-used for other later-discovered neuron types , in order to avoid confusion . There may also be cases where neuron types could be split , or that neuron types that are missing in the current brain sample might be identified in EM or LM images of other brain samples . In such cases the newly identified neurons are expected to be given numbers above the current number range . Although cell types and names may change , and indeed have already changed between versions v1 . 0 and v1 . 1 of our reconstruction , what will not change are the unique body ID numbers given in the database that refer to a particular ( traced ) cell in this particular image dataset . We strongly advise that such body IDs be included in any publications based on our data to avoid confusion as cell type names evolve . As part of our effort to assign cell types , we built a tool for cell type clustering based on neuron connectivity , called CBLAST ( by analogy with the existing NBLAST [Costa et al . , 2016] , which forms clusters based on the shapes of neurons ) . The overall flow of the tool is described in Figure 13 , and the code and instructions on how to install and run it can be found at https://github . com/connectome-neuprint/CBLAST ( Plaza and Dreher , 2020; copy archived at https://github . com/elifesciences-publications/CBLAST ) . Partitioning a network into clusters of nodes that exhibit similar connectivity is known as community detection or graph clustering ( Fortunato and Hric , 2016 ) . Numerous methods have been proposed for selecting such partitions , the best known being the stochastic block model . To non-theoreticians , the process by which most methods choose a partitioning is not intuitive , and the results are not easily interpretable . Furthermore , most approaches do not readily permit a domain expert to guide the partitioning based on their intuition or on other features of the nodes that are not evident in the network structure itself . In contrast , CBLAST is based on traditional data clustering concepts , leading to more intuitive results . Additionally , users can apply their domain expertise by manually refining the partitioning during successive iterations of the procedure . This is especially useful in the case of a network like ours , in which noise and missing data make it difficult to rely solely on connectivity to find a good partitioning automatically . Additionally , other graph clustering methods do not accommodate the notion of left-right symmetry amongst communities , a feature that is critical for assigning cell types in a connectome . CBLAST clusters neurons together using a similarity feature score defined by how the neuron distributes inputs and outputs to different neuron types . However , this is a circular requirement since neuron types must already be defined to use this technique . CBLAST therefore uses an iterative approach , refining cell type definitions successively . Initial cell type groups are putatively defined using an initial set of features based on morphological overlap as in NBLAST and/or based on the distribution of inputs and outputs in defined brain regions . These initial groups are fed into CBLAST in which the user can visualize and analyze the results using plots such as that in Figure 14 . Given the straightforward similarity measure , the user can look at the input and output connections for each neuron to better understand the decision made by the clustering algorithm . As the definitions of cell type definitions are improved , the clustering becomes more reliable . In some cases , this readily exposes incompleteness ( e . g . , due to the boundary of the hemibrain sample ) in some neurons which would complicate clustering even for more computationally intensive strategies such as a stochastic block model . Based on these interactions , the user makes decisions and refines the clusters manually , iterating until further changes are not observed . Our large , dense connectome is a key requirement for CBLAST . Unless a significant fraction of a neuron’s inputs and outputs is known , neurons that are in fact similar may not cluster together correctly . This requirement is not absolute , as we note that CBLAST is often able to match left and right symmetric neurons , despite some of these left side neurons being truncated by the boundaries of the dataset . Nonetheless , reconstruction incompleteness and any noise in the reconstruction can contribute to noise in clustering results . CBLAST usually generates clusters that are consistent with the morphological groupings of the neurons , with CBLAST often suggesting new sub-groupings as intended . This agreement serves as some validation of the concepts behind CBLAST . In some cases it can be preferable to NBLAST , since the algorithm is less sensitive to exact neuron location , and for many applications the connectivity is more important than the morphology . In Figure 14 , we show the results of using CBLAST on a few neuron types extracted from the ellipsoid body . The clusters are consistent with the morphology , with exception to a new sub-grouping for R3p being suggested as a more distinct group than type ExR7/ExR6 . Verifying correctness and completeness in these data is a challenging problem because no existing full brain connectome exists against which our data might be compared . We devised a number of tests to check the main features: Are the morphologies correct ? Are the regions and cell types correctly defined ? Are the synaptic connection counts representative ? Assessing completeness is much easier than assessing correctness . Since the reconstruction is dense , we believe the census of cells , types , and regions should be essentially complete . The main arbors of every cell within the volume are reconstructed , and almost every cell is assigned a cell type . Similarly , since the identified brain regions nearly tile the entire brain , these are complete as well . For checking morphologies , we searched for major missing or erroneous branches using a number of heuristics . Each neuron was reviewed by multiple proofreaders . The morphology of each neuron was compared with light microscopy data whenever it was available . When more than one cell of a given type was available ( either left and right hemisphere , or multiple cells of the same type in one hemisphere ) , a human examined and compared them . This helped us find missing or extra branches , and also served as a double check on the cell type assignment . In addition , since the reconstruction is dense , all sufficiently large ‘orphan’ neurites were examined manually until they were determined to form part of a neuron , or they left the volume . To help validate the assigned cell types , proofreaders did pairwise checks of every neuron with types that had been similarly scored . For subregions in which previous dense proofreading was available ( such as the alpha lobes of the mushroom body ) , we compared the two connectomes . We were also helped by research groups using both sparse tracing in the full fly brain TEM dataset ( Zheng et al . , 2018 ) , and our hemibrain connectome . They were happy to inform us of any inconsistencies . There are limits to this comparison , as the two samples being compared were of different ages and raised under different conditions , then prepared and imaged by different techniques , but this comparison would nevertheless have revealed any gross errors . Finally , we generated a ‘probabilistic connectome’ based on a different segmentation , and systematically visited regions where the two versions differed . As discussed in the section on finding synapses , we evaluated both precision ( the fraction of found synapses that are correct ) and recall ( fraction of true synapses that were correctly predicted ) on sample cubes in each brain region . We also double checked by comparing our findings with a different , recently published , synapse detection algorithm ( Buhmann et al . , 2019 ) . As a final check , we also evaluated the end-to-end correctness of given connections between neurons for different cell types and across brain regions . Specifically , for each neuron , we sampled 25 upstream connections ( T-bar located within the neuron ) and 25 downstream connections ( PSD located within the neuron ) , and checked whether the annotations were correct , meaning that the pre/post annotation was valid and assigned to the correct neuron . In total , we examined 1735 traced neurons spanning 1518 unique cell types ( therefore examining roughly 43 , 000 upstream connections and 43 , 000 downstream connections ) . The histogram of synapse accuracy ( end-to-end precision of predicted synapses ) is given in Figure 15 . Median precision for upstream connections , as well as for downstream connections , is 88% . Additionally , 90% of cell types have an accuracy of at least 70% . For the few worst cases , we manually refined the synapse predictions afterwards . We note that the worst outlier , having an upstream connection accuracy of 12% , is both a case involving few total connections ( 17 T-bars ) , and some ambiguity in the ground-truth decisions ( whether the annotated location is an actual T-bar ) . We also evaluated single-connection pathways across each brain region . In the fly , functionally important connections are thought typically to have many synapses , with the possible exception of cases where many neurons of the same type synapse onto the same downstream partner . However , the presence of connections represented by few synapses is also well known , even if the biological importance of these is less clear . Regardless , we wanted to ensure that even single connection pathways were mostly correct . We sampled over 5500 single-connection pathways , distributed across 57 brain regions . Mean synapse precision per brain region was 76 . 1% , suggesting that single-connection accuracy is consistent with overall synapse prediction accuracy . We also undertook a preliminary evaluation of two-connection pathways ( two synapses between a single pair of neurons ) . We sampled 100 such two-connection pathways within the FB . Overall synapse precision ( over the 200 synapses ) is 79% , consistent with the single-edge accuracy . Moreover , the results also suggest that synapse-level accuracy is largely uncorrelated with pathway/bodies , implying that the probability that both synapses in a two-connection pathway were incorrect is 4 . 4% ( 1-0 . 792 ) , close to the observed empirical value of 3% . ( Applying a χ2 goodness of fit test with a null hypothesis of independence gives a p value of 0 . 7 . ) A synapse in the fly’s brain consists of a presynaptic density ( with a characteristic T-bar ) and typically several postsynaptic partners ( PSDs ) . The T-bars are contained in larger neurites , and most ( >90% ) of the T-bars in our dataset were contained in identified neurons . The postsynaptic densities are typically in smaller neurites , and it is these that are difficult for both machine and human to connect with certainty . With current technology , tracing all fine branches in our EM images is impractical , so we sampled among them ( at completeness levels typically ranging from 20% to 85% ) and traced as many as practical in the allotted time . The goal is to provide synapse counts that are representative , since completeness is beyond reach and largely superfluous . Assuming the missing PSDs are independent ( which we try to verify ) , then the overall circuit emerges even if a substantial fraction of the connections are missing . If a connection has a synapse count of 10 , for example , then it will be found in the final circuit with more than 99 . 9% probability , provided at least half the individual synapses are traced . If unconnected small twigs are the main source of uncertainty in our data ( as we believe to be the case ) , then as the proofreading proceeds the synapse counts of existing connections should only increase . Of course corrections resulting in lower synapse counts , such as correcting a false connection or removing an incorrect synapse , are also possible , but are considerably less likely . To see if our proofreading process worked as expected , we took a region that had been read to a lower percentage completion and then spent the manual effort to reach a higher percentage , and compared the two circuits . ( A versioned database such as DVID is enormously helpful here . ) If our efforts were successful , ideally what we see is that almost all connections that changed had more synapses , very few connections got fewer synapses , and no new strong ( many synapse ) connections appeared ( since all strong connections should already be present even in low coverage proofreading ) . If this is the behavior we find , we could be reasonably certain that the circuits found are representative for all many-synapse connections . Figure 16 shows such an analysis . The results support our view that the circuits we report reflect what would be observed if we extrapolated to assign all pre- and postsynaptic elements . Given the complexity of the reconstruction process , and the many different errors that could occur , how confident should the user be that the returned synapse counts are valid ? This section gives a quick guide in the absence of detailed investigation . The number of synapses we return is the number we found . The true number could range from slightly less , largely due to false synapse predictions , to considerably more , in the regions with low percentage reconstructed . For connections known to be in a specific brain region , the reciprocal of the completion percentage ( as shown in Table 1 ) gives a reasonable estimate of the undercount . If we return a count of 0 ( the neurons are not connected ) , there are two cases . If the neurons do not share any brain regions , then the lack of connections is real . If they do share a brain region or regions , then a count of 0 is suspect . It is possible that there might be a weak connection ( count 1–2 ) and less likely there is a connection of medium strength ( 3–9 synapses ) . Strong connections can be confidently ruled out , minus the small chance of a mis- or un-assigned branch with many synapses . If we report a weak connection ( 1–2 synapses ) , then the true strength might range from 0 ( the connection does not exist ) through a weak connection ( 3–9 synapses ) . If your model or analysis relies on the strength of these weak connections , it is a good idea to manually check our reconstruction . If your analysis does not depend on knowledge of weak connections , we recommend ignoring connections based on three or fewer synapses . If we report a medium strength connection ( 3–9 synapses ) then the connection is real . The true strength could range from weak to the lower end of a strong connection . If we report a strong connection ( 10 or more synapses ) , the connection not only exists , but is strong . It may well be considerably stronger than we report . The representation of connectomics data is a significant problem for all connectomics efforts . The raw image data on which our connectome is based is larger than 20 TB , and takes 2 full days to download even at a rate of 1 gigabit/second . Looking forward , this problem will only get worse . Recent similar projects are generating petabytes worth of data ( Yin et al . , 2019 ) , and a mouse brain of 500 mm3 , at a typical FIB-SEM resolution of 8 nm isotropic , would require almost 1000 petabytes . In contrast , most users of connectivity information want a far smaller amount of much more specific information . For example , a common query is ‘what neurons are downstream ( or upstream ) of a given target neuron ? ’ . This question can be expressed in a few tens of characters , and the desired answer , the top few partners , fits on a single page of text . Managing this wide range of data , from the raw gray-scale through the connectivity graph , requires a variety of technologies . An overview of the data representations we used to address these needs is shown in Figure 17 . This organization offers several advantages . In most cases , instead of transferring files , the user submits queries for the portion of data desired . If the user needs only a subset of the data ( as almost all users do ) then they need not cope with the full size of the data set . Different versions of the data can be managed efficiently behind the scenes with a versioned database such as DVID ( Katz and Plaza , 2019 ) that keeps track of changes and can deliver data corresponding to any previous version . The use of existing software infrastructure , such as Google buckets or the graph package neo4j , which are already optimized for large data , helps with both performance and ease of development . The advanced user is not limited to these interfaces - for those who may wish to validate or extend our results; we have provided procedures whereby the user can make personal copies of each representation , including the grayscale , the DVID data storage , and our editing and proofreading software . These allow other researchers to establish an entirely independent version of all we have done , completely under their control . Contact the authors for the details of how to copy all the underlying data and software . Grayscale data correspond to traditional electron microscope images . This is written only once , after alignment , but often read , because it is required for segmentation , synapse finding , and proofreading . We store the grayscale data , eight bits per voxel , in Google buckets , which facilitates access from geographically distributed sites . Segmentation , synapses , and identifying regions annotate and give biological meaning to the grayscale data . For segmentation , we assign a 64 bit neuron ID to each voxel . Despite the larger size per voxel ( 64 vs 8 bits ) compared with the grayscale , the storage required is much smaller ( by a factor of more than 20 ) since segmentation compresses well . Although the voxel level segmentation is not needed for connectivity queries , it may be useful for tasks such as computing areas and cross-sections at the full resolution available , or calculating the distance between a feature and the boundary . Synapses are stored as point annotations - one point for a presynaptic T-bar , and one point for each of its postsynaptic densities ( or PSDs ) . The segmentation can then be consulted to find the identity of the neurons containing their connecting synapses . The compartment map of the brain is stored as a volume specified at a lower resolution , typically a 32 × 32 × 32 voxel grid . At 8 nm voxels , this gives a 256 nm resolution for brain regions , comparable to the resolution of confocal laser scanning microscopy . Unlike the grayscale data , segmentation , synapses , and regions are all modified during proofreading . This requires a representation that must cope with many users modifying the data simultaneously , log all changes , and be versioned . We use DVID ( Katz and Plaza , 2019 ) , developed internally , to meet these requirements . Neuron skeletons are computed from the segmentation ( Zhao and Plaza , 2014 ) , and not entered or edited directly . A skeleton representation describes each neuron with ( branching ) centerlines and diameters , typically in the SWC format popularized by the simulator Neuron ( Carnevale and Hines , 2006 ) . These are necessarily approximations , since it is normally not possible ( for example ) to match both the cross-sectional area and the surface area of each point along a neurite with such a representation . But SWC skeletons are a good representation for human viewing , adequate for automatic morphology classification , and serve as input to neural simulation programs such as ‘Neuron’ . SWC files are also well accepted as an interchange format , used by projects such as NeuroMorpho ( Ascoli et al . , 2007 ) and FlyBrain ( Shinomiya et al . , 2011 ) . The connectivity graph is also derived from the data and is yet more abstract , describing only the identity of neurons and a summary of how they connect - for example , Neuron ID1 connects to neuron ID2 through a certain number of synapses . In our case , it also retains the brain region information and the location of each synapse . Such a connectivity graph is both smaller and faster than the geometric data , but sufficient for most queries of interest to biologists , such as finding the upstream or downstream partners of a neuron . A simple connectivity graph is often desired by theorists , particularly within brain regions , or when considering neural circuits in which each neuron can be represented as a single node . A final , even more abstract form is the adjacency matrix: This compresses the connectivity between each pair of neurons to a single number . Even this most economical form requires careful treatment in connectomics . As our brain sample contains more than 25K traced neurons as well as many unconnected fragments , the adjacency matrix has more than a billion entries ( most of which are zero ) . Sparse matrix techniques , which report only the non-zero coefficients , are necessary for practical use of such matrices . For the hemibrain project , we provide access to the data through a combination of a software interface ( Clements et al . , 2020 ) and a server ( https://neuprint . janelia . org , also accessible through https://doi . org/10 . 25378/janelia . 12818645 ) . Login is via any Google account; users who wish to remain anonymous can create a separate account for access purposes only . Data are available in the form of gray-scale , pixel-level segmentation , skeletons , and a graph representation . Two previous connectomics efforts are available as well ( a seven-column optic lobe reconstruction [Takemura et al . , 2015] and the alpha lobe of the mushroom body [Takemura et al . , 2017] ) . These can be found at https://neuprint-examples . janelia . org . The most straightforward way to access the hemibrain data is through the Neuprint ( Clements et al . , 2020 ) interactive browser . This is a web-based application that is intended to be usable by biologists with minimal or no training . It allows the selection of neurons by name , type , or brain region , displays neurons , their partners , and the synapses between these in a variety of forms , and provides many of the graphs and summary statistics that users commonly want . Neuprint also supports queries from languages such as Python ( Sanner , 1999 ) and R , as used by the neuroanatomy tool NatVerse ( Manton et al . , 2019 ) . Various formats are supported , including SWC format for the skeletons . In particular , the graph data can be queried through an existing graph query language , Cypher ( Francis et al . , 2018 ) , as seen in the example below . The schema for the graph data is shown in Figure 18 . This query looks for all neurons that are presynaptic to any neuron of type ‘MBON18’ . For each such neuron it returns the types and internal identities of the presynaptic neuron , and the count of synapses between them . The whole list is ordered in order of decreasing synapse count . This is just an illustration for a particular query that is quite common and supported in Neuprint without the need for any programming language . Adjacency matrices , if needed , can be derived from the graph representation . We provide a small demonstration program that queries the API and generates such matrices , either with or without the brain regions . The two matrices themselves are available in gzipped Python format . The raw greyscale images , with overlays of segmentation and feature masks ( such as glia and mitochondria ) , can be viewed in the publicly available tool NeuroGlancer ( Perlman , 2019 ) . This viewer can be selected from the Neuprint browser . For more information on accessing data and other hemibrain updates , please see https://www . janelia . org/project-teams/flyem/hemibrain . No two flies are identical , and brain samples differ in size and orientation . Furthermore , different preparation methods cause tissues to swell and shrink by varying amounts . Therefore , the first step when comparing the features of different brains is registration to a common reference frame . Some of these differences are illustrated in Figure 19 . Compared to the hemibrain EM data ( Figure 19 ( a ) ) , the confocal laser scanning microscopy images of the previous brain atlas ( Ito et al . , 2014 ) are about 17% smaller ( Figure 19 ( b ) ) , and the JRC2018 unisex template brain used for the registration of EM and light microscopy brain images ( Bogovic et al . , 2020 ) is about 30% smaller ( Figure 19 ( c ) ) . Since unfixed brains right after dissection in saline are 15–20% larger than the antibody-labeled brains mounted in 80% glycerol – similar to Figure 19 ( b ) – a raw female brain will be nearly the same size as the hemibrain EM stack . The orientation of the brain samples may also vary . There is about 18 . 5° of tilt between the hemibrain EM stack and the 2014 brain atlas , and about 14° of tilt between hemibrain EM and the JRC2018 template . To create matching vertical or horizontal sections , therefore , each image stack should be re-sliced after applying the corresponding rotation . The raw EM data , segmentation , and skeletons ( as displayed in Neuprint ) were all computed in a reference frame corresponding to Figure 19 ( a ) , whereas the light lines and tools such as Color_MIP_mask search ( Otsuna et al . , 2018 ) use the reference frame of Figure 19 ( c ) . Therefore , registration is required to map between the EM and light representations . We registered the hemibrain EM data to the JRC2018 Drosophila template brain using an automatic registration algorithm followed by manual correction . We began by using the automated T-bar predictions ( described in section on synapse prediction ) to generate a T-bar density volume rendered at a resolution comparable to those from light microscopic images . This hemibrain synapse density volume was automatically registered to the template brain using elastix ( Klein et al . , 2010 ) . The resulting registration was manually fine-tuned using BigWarp ( Bogovic et al . , 2016 ) . The total transform is the composition of the elastix and BigWarp transformations , and can be found at https://www . janelia . org/open-science/jrc-2018-brain-templates . We estimated a corresponding inverse transformation and make that available as well . Using these transformations , an implementation that matches EM to light lines , and vice versa , is publicly accessible at https://neuronbridge . janelia . org/ . This matching software is accessible directly from the Neuprint browser , where it can be launched from the tabular display of selected neurons . For those not familiar with NeuronBridge , an explanatory video explains the matching process . The details of the underlying algorithm will be covered in a separate paper by Otsuna et al . , but are briefly sketched here . If starting from an EM neuron of interest , researchers can use NeuronBridge to identify GAL4 lines labeling that neuron . First , the EM representation of the neuron is spatially transformed into the JRC 2018 unisex template space where GAL4 driver line images are registered . The EM neuron is then used to create a mask ( Otsuna et al . , 2018 ) that narrows the search space considerably , making it easier to find corresponding neurons even in crowded GAL4 driver line images . The opposite direction , finding an EM neuron that corresponds to a light neuron , is also supported . In this case the scoring of a potential match must be modified , since the light image contains the entire neuron , but many EM neurons are trunctated by the limits of our reconstructed volume . Both of these cases are discussed in the upcoming paper , with examples . As another option , since hemibrain neurons are skeletonized , users can query GAL4 neuronal skeleton databases using NBLAST ( Costa et al . , 2016 ) . Historically , archival data from biology data have been expressed as files that are included with supplementary data . However , for connectivity data this practice has two main problems . First , the data are large , and hard to store . Journals , for example , typically limit supplemental data to a few 10s of megabytes . The data here are about 6 orders of magnitude larger . Second , connectome data are not static , during proofreading and even after initial publication . As proofreading proceeds , the data improve in their completeness and quality . The question then is how to refer to the data as they existed at some point in time , required for reproducibility of scientific results . If represented as files , this would require many copies , checkpointed at various times - the ‘as submitted’ version , the ‘as published’ version , the ‘current best version’ , and so on . We resolve this , at least for now , by hosting the data ourselves and making them available through query mechanisms . Underlying our connectome data is a versioned database ( DVID ) so it is technically possible to access every version of the data as it is revised . However , as it requires effort to host and format this data for the Neuprint browser and API , only selected versions ( called named versions ) are available by default from the website , starting with the initial versions , which are ‘hemibrain:v1 . 0’ and the much improved ‘hemibrain:v1 . 1’ . Since multiple versions are available , when reproducibility is required ( such as when referencing the data in a paper ) it is best to refer explicitly to the version used by name ( such as ‘hemibrain:v1 . 1’ ) because we expect new milestone versions every few months , at least at first . We will supply a DOI for each of these versions , and each is archived , can be viewed and queried through the web browser and APIs at any time , and will not change . The goal of multiple versions is that later versions should be of higher quality . Towards this end we have implemented several systems for reporting errors so we can correct them . Users can add annotations in NeuroGlancer ( Perlman , 2019 ) , the application used in conjunction with Neuprint to view image data , where they believe there are such errors . To make this process easier , we provide a video explaining it . We will review these annotations and amend those that we agree are problems . Users can also contact us via email about problems they find . Archival storage is an issue since , unlike genetic data , there is not yet an institutional repository for connectomics data and the data are too large for journals to archive . We pledge to keep our data available for at least the next 10 years . Of necessity , most previous analyses have concentrated on particular circuits , cell types , or brain regions with relevance to specific functions or behaviors . For example , a classic paper about motifs ( Song et al . , 2005 ) sampled the connections between one cell type ( layer five pyramidal neurons ) in one brain region ( rat visual cortex ) , and found a number of non-random features , such as over-represented reciprocal connections and a log-normal strength distribution . However , it has never been clear which of these observations generalize to other cell types , other brain regions , and the brain as a whole . We are now in a position to make much stronger statements , ranging over all brain regions and cell types . In addition , many analyses are best performed ( or can only be performed ) on dense connectomes . Type-wide observations depend on a complete census of that cell type , and depending on the observation , a complete census of upstream and downstream partners as well . Some analyses , such as null observations about motifs ( where certain motifs do not occur in all or portions of the fly’s brain ) can only be undertaken on dense connectomes . One analysis enabled by a dense whole-brain reconstruction involves the comparison between the circuit architectures of different brain areas within a single individual . The compartments vary considerably . Table 4 shows the connectivity statistics of compartments that are completely contained within the volume , have at least 100 neurons , and have the largest or smallest value of various statistics . Across regions , the number of neurons varies by a factor of 74 , the average number of partners of each neuron by a factor of 36 , the network diameter ( defined as the maximum length of the shortest path between any two neurons ) by a factor of 4 , the average strength of connection between partner neurons by a factor of 5 , and the fraction of reciprocal connections by a factor of 5 . The average graph distance between neurons is more conserved , differing by a factor of only 2 . Neurons in the fly brain are tightly interconnected , as shown in Figure 20 , which plots what fraction of neuron pairs are connected as a function of the number of interneurons between them . Three quarters of all possible pairs are connected by a path with fewer than three interneurons , even when only connections with ≥5 synapses are included . If weaker connections are allowed , the paths become shorter yet . These short paths and tight coupling are very different from human designed systems , which have much longer path lengths connecting node pairs . As an example , a standard electrical engineering benchmark ( S38584 from Brglez et al . , 1989 ) is shown alongside the hemibrain data in Figure 20A–B . The connection graph for this example has roughly the same number of nodes as the graph of the fly brain , but pair-to-pair connections involve paths more than an order of magnitude longer – a typical node pair is separated by 60 intervening nodes . This is because a typical computational element in a human designed circuit ( a gate ) connects only to a few other elements , whereas a typical neuron receives input from , and sends outputs to , hundreds of other neurons . The distribution of connection strengths has been studied in mammalian tissue , looking at specific cell types in specific brain areas . These findings , such as the log-normal distribution of connection strengths in rat cortex , do not appear to generalize to flies . Assuming the strength of a connection is proportional to the number of synapses in parallel , we can plot the distribution of connection strengths , summing over the whole central brain , as shown in Figure 21 . We find a nearly pure power law with an exponential cutoff , very different from the log-normal distribution of strengths found by Song et al . , 2005 in pyramidal cells in the rat cortex , or the bimodal distribution found for pyramidal cells in the mouse by Dorkenwald et al . , 2019 . However , we caution that these analyses are not strictly comparable . Even aside from the very different species examined , the three analyses differ . Both Song and Dorkenwald looked at only one cell type , with excitatory connections only , but one looked at electrical strength while the other looked at synapse area as a proxy for strength . In our analysis , we use synapse count as a proxy for connection strength , and look at all cell types , including both excitatory and inhibitory synapses . As mentioned earlier , there have been many studies of small motifs , usually involving limited circuits , cell types , and brain regions . We emphatically confirm some traditional findings , such as the over-representation of reciprocal connections . We observe this in all brain regions and among all cell types , confirming similar findings in the antennal lobe ( Horne et al . , 2018 ) . This can now be assumed to be a general feature of the fly’s brain , and possibly all brains . In the fly , the incidence varies somewhat by compartment , however , as shown in Table 4 . We define a large motif as a graph structure that involves every cell of an abundant type ( N ≥ 20 ) . The most tightly bound motif is a clique , in which every cell of a given type is connected to every other cell of that type , with synapses in both directions . Such connections , as illustrated in Figure 22 ( a ) , are extremely unlikely in a random wiring model . Consider , for example , the clique of ER4d cells found in the ellipsoid body , as shown in Table 5 . In the ellipsoid body , two cells are connected with an average probability of 0 . 19 . Therefore , the odds of finding all 600 possible connections between ER4d cells , assuming a random wiring model , is 0 . 19600≈10-432 . In the fly’s brain , only a few cell types form large cliques , as shown in Table 5 . All true cliques are among the ring neurons in the central complex , with a near-clique among the KCab-p cells of the mushroom body . The cell types PFNa and PFNd are included although they do not form a clique as shown in Figure 22 ( a ) . However , these neurons are part of symmetrical structures , the noduli , that occur on both sides of the brain . Within each side , the cells form a clique , as shown in Figure 22 ( d ) . The cliques within the central complex , and their potential operation , are discussed in detail in the companion paper on the central complex by Jayaraman et al . The next most tightly bound motifs are individual cells that connect both to and from all cells of a given type , but are themselves of a different type . This is illustrated in Figure 22 ( b ) . Such a motif is often speculated to be a gain or sparseness controlling circuit , where the single neuron reads the collective activation of a population and then controls their collective behavior . A well-known example is the APL neuron in the mushroom body , which connects both to and from all the Kenyon cells , and is thought to regulate the sparseness of the Kenyon cell activation ( Lin et al . , 2014 ) . We search for this motif by looking at cells with few instances ( one or two ) connecting bidirectionally to almost all cells ( at least 90% ) of an abundant type ( N ≥ 20 ) . We find this motif in three regions of the brain – it is common in the CX ( 73 different cells overseeing 22 cell types ) , the optic lobe circuits ( 19 cells overseeing 14 types ) , and somewhat in the MB ( 12 types overseeing nine types ) . A spreadsheet containing these cell types , who they connect to , and the numbers and strengths of their connections is described in the appendix and included as supplementary data . We only analyze the optical circuits here , since the mushroom body and central complex are the subjects of companion papers . We observe three variations on this motif - a single cell connected to all of a type ( Figure 23 ( a ) , found five times ) , a single cell with bidirectional connections to many types ( Figure 23 ( b ) , found once ) , and multiple cells all connected bidirectionally to a single type ( Figure 23 ( c ) ) , found three times . We find one circuit that is a combination: There is one cell that connects bidirectionally to all the LC17 neurons , and then a higher order cell that connects bidirectionally to a larger set ( LPLC1 , LPLC2 , LLP1 , LPC1 , and LC17 ) . In this case , these are all looming-sensitive cells and hence these circuits may regulate the features of the overall looming responses . It is tempting to speculate that the more complex structures of Figure 23 ( b ) and ( c ) arose from the simpler structures of ( a ) through cell type duplication followed by divergence , but the connectomes of many more related species will be needed before this argument could be made quantitative . The least tightly bound large motif is a cell that connects either to or from ( but not both ) all cells of a given type , as shown in Figure 22 ( c ) . Examples include the mushroom body output neurons ( Takemura et al . , 2017 ) . This is a very common motif , found in many regions . We find more than 500 examples of this in the fly’s brain . How does the compartmentalization of the fly brain affect neural computation ? In a few cases this has been established . For example , the CT1 neuron performs largely independent computations in each branch ( Meier and Borst , 2019 ) , whereas estimates show that within the medulla , the delays within each neuron are likely not significant for single column optic lobe neurons , and hence the neurons likely perform only a single computation ( Takemura et al . , 2013 ) . Similarly , compartments of PEN2 neurons in the protocerebral bridge have been shown to respond entirely differently from their compartments in the ellipsoid body ( Green et al . , 2017; Turner-Evans et al . , 2019 ) . Our detailed skeleton models allow us to construct electrical models of neurons . ( In what follows , we use the word ‘compartment’ to mean a named physical region of the brain , as shown in Table 1 , as opposed to the electrical sub-divisions used in simulation . ) In particular , to look more generally at the issues of intra– vs inter–compartment delays and amplitudes , we can construct a linear passive model for each neuron . Our method is similar to that elsewhere ( Segev et al . , 1985 ) , except that instead of using right cylinders , we represent each segment of the skeleton as a truncated cone . This is then used to derive the axonic resistance , the membrane resistance , and membrane capacitance for each segment . To analyze the effect of compartment structure on neuron operation , we inject the neuron at a postsynaptic density ( input ) with a signal corresponding to a typical synaptic input ( 1 nS conductance , 1 ms width , 0 . 1 ms rise time constant , 1 ms fall time constant , 60 mV reversal potential ) . We then compute the response at each of the T-bar sites ( outputs ) . Since the synapses , both input and output , are annotated by the brain region that contains them , this allows us to calculate the amplitudes and delays from each synapse ( or a sample of synapses ) in each compartment to each output synapse in all other compartments . In general , we find the compartment structure of the neuron is clearly reflected in the electrical response . Consider , for example , the EPG neuron ( Figure 24 ( a ) ) with arbors in the ellipsoid body , the protocerebral bridge , and the gall ( the gall is a sub-compartment of the LAL , the lateral accessory lobe ) . Figure 25 ( a ) shows the responses to synaptic input in the gall . Within the gall , the delays are very short , and the amplitude relatively high and variable , depending somewhat on the input and output synapse within the gall . From the gall to other regions , the delays are longer ( typically a few milliseconds ) and the amplitudes much smaller and nearly constant , largely independent of the exact transmitting and receiving synapse . There is a very clean separation between the within-compartment and across-compartment delays and amplitudes , as shown in Figure 25 ( a ) . The same overall behavior is true for inputs into the other regions - short delays and strong responses within the compartment , with longer delays and smaller amplitudes to other compartments . This simple pattern motivates a model that describes delays and amplitudes not as a single number , but as an N×N matrix , where N is the number of compartments . Each row contains the estimated amplitude and delay , measured in each compartment , for a synaptic input in the given compartment . This gives a much improved estimate of the linear response . For the example EPG neuron above , with nominal values for Ra , Rm , and Cm , if we represent all delays by a single number then the standard deviation of the error is 0 . 446 ms . If instead we represent the delays as a 3 × 3 matrix indexed by the compartment , the average error is 0 . 045 ms , for 10x greater accuracy . Similarly , the average error in amplitude drops from 0 . 168 mv to 0 . 021 mv , an eightfold improvement . While the improvement in error will depend on the neuron topology , in all cases it will be more accurate than a point model , for relatively little increase in complexity . The absolute values of delay and amplitude are strongly dependent on the electrical parameters of the cell , however . A wide range of electrical properties has been reported in the fly literature ( see Table 6 ) and it is plausible that these vary on a cell-to-cell basis . In addition gap junctions , which are not included in our model , could affect the apparent value of Rm . In light of these uncertainties , we simulate with minimum , medium , and maximal values of Ra and Rm , for a total of 9 cases , as shown in Figure 25 ( b ) . All are needed since the resistance parameters interact non-linearly . We fix the value of Cm at 0 . 01 F/m2 since this value is determined by the membrane thickness and is not expected to vary from cell to cell ( Kandel et al . , 2000 ) . The results over the parameter range are shown in Figure 25 ( b ) for the case of the EPG neuron above for delay from the gall to the PB . The intra-compartment and between-compartment values are well separated for any value of the parameters ( not shown ) . Programs that deduce synaptic strength and sign by fitting a computed response to a connectome and measured electrical or calcium imaging data ( Tschopp et al . , 2018 ) may at some point require estimates of the delays within cells . If this is required , the above results suggest this could be accomplished with reasonable accuracy with a compartment-to-compartment delay table and two additional parameters per neuron , RA and RM . This is relatively few new parameters in addition to the many synaptic strengths already fitted . A number of neurons have parallel connections in separate compartments ( see Figure 24 ( b ) ) . This motif is common in the fly’s brain – about 5% of all connections having a strength ≥6 are spread across two or more non-adjacent compartments . Given the increased delays and lower amplitudes of cross-compartment responses , this type of interaction differs electrically from those in which all connections are contained in a single compartment . A point neuron model cannot generate an accurate response for such connections – a synapse in region A will result in a fast response in A and a slower , smaller response in B , and vice versa , even though both of these events involve communication between the same two neurons . It is not known if this configuration has a significant influence on the neurons’ operation . From these models , we conclude ( a ) the compartment structure of the fly brain shows up directly in the electrical response of the neurons , and ( b ) the compartment structure , although defined anatomically , matches that of the electrical response . From the clear separation in Figure 25 , it is likely that the same compartment definitions could be found starting with the electrical response , although we have not tried this . ( c ) These results suggest a low dimensional model for neural operation , at least in the linear region . A small region-to-region matrix can represent the delays and amplitudes well . ( d ) Absolute delays depend strongly ( but in a very predictable manner ) on the values of axial and membrane resistance , which can vary both from animal to animal and from cell to cell . ( e ) Neurons that have parallel connections in separate compartments have a different electrical response than they would have with the same total number of synapses in a single compartment . Rent’s rule ( Lanzerotti et al . , 2005 ) is an empirical observation that in human designed computing systems , when the system is packed as tightly as possible , at every level of the hierarchy the required communication ( the number of pins ) scales as a power law of the amount of contained computation , measured in gates . Rent’s rule is an observed relationship , not derived from underlying theory , and the relationship is not exact and still contains scatter . A biological equivalent might be the observation that brain size tends to vary as a power law of body size ( Harvey and Krebs , 1990 ) , across a wide range of species occupying very different ecological and behavioral niches . Rent’s rule is roughly true over many orders of magnitude in scale , and for almost every system in which it has been measured . Somewhat surprisingly , Rent’s rule applies almost independently of the function performed by the computation being performed , and at every level of a hierarchical system . It also applies whether the compactness criterion is minimization of communication ( partitioning ) or physical close packing . Rent’s rule is expressed asP⁢i⁢n⁢s=a* ( c⁢o⁢m⁢p⁢u⁢t⁢a⁢t⁢i⁢o⁢n ) bwhere a is a scale factor ( typically in the range 1–4 ) , and b is the ‘Rent exponent’ describing how the number of connections to the compartment varies as a function of the amount of computation performed in the compartment . The Rent exponent has a theoretical range of 0 . 0 to 1 . 0 , where 0 represents a constant number of connections , with no dependence on the amount of computation performed , and 1 . 0 represents a circuit in which every computation is visible on a connection . Human designed computational systems occupy almost the full range , from spreadsheets in which every computation is visible , to largely serial systems in which minimizing communication ( pins ) is critical . This relationship is shown in Figure 26 . However , when the overriding criterion is that the system must be packed as tightly as possible , Rent observed that the exponent of the power law falls in a close range of roughly 0 . 5–0 . 7 . For electrical circuits , the computation is measured in gates , and the connections are measured by pin count . These ranges are shown in Figure 26 for circuits that are roughly the size of the fly’s brain , packed in either two ( Yang et al . , 2001 ) or three ( Das et al . , 2004 ) dimensions . Also shown in this plot are the values for the fly’s brain computational regions . In this case , the computation is measured as the number of contained T-bars , and the connection count is the number of neurons that have at least one synapse both inside and outside the compartment . ( Very similar results are obtained if the computation is measured as the number of PSDs , or the number of unique connection pairs ) . Almost all the fly brain compartments fall well within the range of exponents expected for packing-dominated systems , while the ellipsoid body ( EB ) falls just outside the expected area . This is perhaps due to the large number of strongly connected clique-containing circuits in the ellipsoid body ( see Table 5 ) , since such circuits have relatively few connections for the amount of synapses they contain . Both human designed and biological systems have huge incentives to pack their computation as tightly as possible . A tighter packing of the same computation yields faster operation , lower energy consumption , less material cost , and lower mass . A natural speculation , therefore , is that both the human-designed and evolved systems are dominated by packing considerations , and that both have found similar solutions . In this work , we have achieved a dream of anatomists that is more than a century old . For at least the central brain of at least one animal with a complex brain and sophisticated behavior , we have a complete census of all the neurons and all the cell types that constitute the brain , a definitive atlas of the regions in which they reside , and a graph representing how they are connected . To achieve this , we have made improvements to every stage of the reconstruction process . Better means of sample preparation , imaging , alignment , segmentation , synapse finding , and proofreading are all summarized in this work and will form the basis of yet larger and faster reconstructions in the future . We have provided the data for all the circuits of the central brain , at least as defined by nerve cells and chemical synapses . This includes not only circuits of regions that are already the subject of extensive study , but also a trove of circuits whose structure and function are yet unknown . We have provided a public resource that should be a huge help to all who study fly neural circuits . Finding upstream and downstream partners , a task that until now has typically taken months of challenging experiments , is now replaced by a lookup on a publicly available web site . Detailed circuits , which used to require considerable patience , expertise , and expertise to acquire , are now available for the cost of an internet query . More widely , a dense connectome is a valuable resource for all neuroscientists , enabling novel , system-wide analyses , as well as suggesting roles for specific pathways . A surprising revelation is the richness of anatomical synaptic engagements , which far exceeds pathways required to support identified fly behaviors , and suggests that most behaviors have yet to be identified . Finally , we have started the process of analyzing the connectome , though much remains to be done . We have quantified the difference between computational compartments , determined that the distribution of strengths is different from that reported in mammals , discovered cliques and other structures and where these occur , examined the effect of compartmentalization on electrical properties , and provided evidence that the wiring of the brain is consistent with optimizing packing . Many of the extensions of this work are obvious and already underway . Not all regions of the hemibrain have been read to the highest accuracy possible , insofar as we have concentrated first on the regions overlapping with other projects , such as the central complex and the mushroom body . We will continue to update other sections of the brain , and distributed circuits such as clock and modulatory neurons that are not confined to one region , but spread throughout the brain . There is much more to be learned about the graph properties of the brain , and how these relate to its function . The two sexes of the Drosophila brain are known to differ ( Auer and Benton , 2016 ) . so that reconstructing a male fly is critical to compare the circuits of the two sexes . A ventral nerve cord ( VNC ) should be reconstructed , preferably attached to the brain of the same individual , since the circuits in the VNC are known to be crucial for fly motor behavior ( Yellman et al . , 1997 ) . At least one optic lobe should be included to simplify analysis of visual inputs to the central brain . A whole brain connectome is preferable to the hemibrain , since then most cell types would have at least two examples , left and right , which would lend increased confidence to our reconstructions . It would also provide complete reconstruction to the many neurons that span the brain , especially the clock and modulatory neurons , and are incomplete in the hemibrain . These four goals are combined in a project that is currently underway , to image and reconstruct an entire male central nervous system ( CNS ) including the VNC and optic lobes . We continue to improve sample preparation , imaging , and reconstruction both to decrease the efforts expended on reconstruction and to speed reconstruction of more specimens . Improvements include multi-beam imaging , etching methods ( Hayworth et al . , 2020 ) that can handle larger areas , and yet better reconstruction techniques . These improvements , however , will still rely on FIB-SEM technology , and additional methods will likely be required to fill in other information . Gap junctions will continue to be difficult to see in FIB-SEM , and other methods such as optical labeling , expansion microscopy , and RNA-SEQ ( to find which neurons express gap junction proteins ) will be required . Methods for estimating the extent of diffusion of the secreted modulatory transmitters and gaseous signal molecules such as NO remain to be established . Different staining methods ( and expression driver lines ) may be needed to study glia to the same extent we currently study neurons . A wide variety of techniques will be needed to understand the subcellular architecture of the neurons we have reconstructed . Finally , larger animal brains beckon , such as the brain of a mouse and eventually a human . The data we present here is only a start .	Animal brains of all sizes , from the smallest to the largest , work in broadly similar ways . Studying the brain of any one animal in depth can thus reveal the general principles behind the workings of all brains . The fruit fly Drosophila is a popular choice for such research . With about 100 , 000 neurons – compared to some 86 billion in humans – the fly brain is small enough to study at the level of individual cells . But it nevertheless supports a range of complex behaviors , including navigation , courtship and learning . Thanks to decades of research , scientists now have a good understanding of which parts of the fruit fly brain support particular behaviors . But exactly how they do this is often unclear . This is because previous studies showing the connections between cells only covered small areas of the brain . This is like trying to understand a novel when all you can see is a few isolated paragraphs . To solve this problem , Scheffer , Xu , Januszewski , Lu , Takemura , Hayworth , Huang , Shinomiya et al . prepared the first complete map of the entire central region of the fruit fly brain . The central brain consists of approximately 25 , 000 neurons and around 20 million connections . To prepare the map – or connectome – the brain was cut into very thin 8nm slices and photographed with an electron microscope . A three-dimensional map of the neurons and connections in the brain was then reconstructed from these images using machine learning algorithms . Finally , Scheffer et al . used the new connectome to obtain further insights into the circuits that support specific fruit fly behaviors . The central brain connectome is freely available online for anyone to access . When used in combination with existing methods , the map will make it easier to understand how the fly brain works , and how and why it can fail to work correctly . Many of these findings will likely apply to larger brains , including our own . In the long run , studying the fly connectome may therefore lead to a better understanding of the human brain and its disorders . Performing a similar analysis on the brain of a small mammal , by scaling up the methods here , will be a likely next step along this path .	[ "Abstract", "Introduction" ]	[ "computational", "and", "systems", "biology", "neuroscience" ]	2020	A connectome and analysis of the adult Drosophila central brain
Mutations in the Trypanosoma brucei aquaporin AQP2 are associated with resistance to pentamidine and melarsoprol . We show that TbAQP2 but not TbAQP3 was positively selected for increased pore size from a common ancestor aquaporin . We demonstrate that TbAQP2’s unique architecture permits pentamidine permeation through its central pore and show how specific mutations in highly conserved motifs affect drug permeation . Introduction of key TbAQP2 amino acids into TbAQP3 renders the latter permeable to pentamidine . Molecular dynamics demonstrates that permeation by dicationic pentamidine is energetically favourable in TbAQP2 , driven by the membrane potential , although aquaporins are normally strictly impermeable for ionic species . We also identify the structural determinants that make pentamidine a permeant although most other diamidine drugs are excluded . Our results have wide-ranging implications for optimising antitrypanosomal drugs and averting cross-resistance . Moreover , these new insights in aquaporin permeation may allow the pharmacological exploitation of other members of this ubiquitous gene family . The Trypanosoma brucei-group species are protozoan parasites that cause severe and fatal infections in humans ( sleeping sickness ) and animals ( nagana , surra , dourine ) ( Giordani et al . , 2016; Büscher et al . , 2017 ) . The treatment is dependent on the sub-species of trypanosome , on the host , and on the stage of the disease ( Giordani et al . , 2016; P . De Koning , 2020 ) . Many anti-protozoal drugs are inherently cytotoxic but derive their selectivity from preferential uptake by the pathogen rather than the host cell ( Munday et al . , 2015a; P . De Koning , 2020 ) . Conversely , loss of the specific drug transporters is a main cause for drug resistance ( Barrett et al . , 2011; Baker et al . , 2013; Munday et al . , 2015a; P . De Koning , 2020 ) . This is the case for almost all clinically used trypanocides , including diamidines such as pentamidine and diminazene ( Carter et al . , 1995; de Koning , 2001a; de Koning et al . , 2004; Bridges et al . , 2007 ) , melaminophenyl arsenicals such as melarsoprol and cymelarsan for cerebral stage human and animal trypanosomiasis , respectively ( Carter and Fairlamb , 1993; Bridges et al . , 2007 ) , and the fluorinated amino acid analogue eflornithine for human cerebral trypanosomiasis ( Vincent et al . , 2010 ) . The study of transporters is thus important for anti-protozoal drug discovery programmes as well as for the study of drug resistance ( Lüscher et al . , 2007; Munday et al . , 2015a ) . In Trypanosoma brucei , the phenomenon of melarsoprol-pentamidine cross-resistance ( MPXR ) was first described shortly after their introduction ( Rollo and Williamson , 1951 ) , and was linked to reduced uptake rather than shared intracellular target ( s ) ( Frommel and Balber , 1987 ) . The first transporter to be implicated in MPXR was the aminopurine transporter TbAT1/P2 ( Carter and Fairlamb , 1993; Mäser et al . , 1999; Munday et al . , 2015b ) but two additional transport entities , named High Affinity Pentamidine Transporter ( HAPT1 ) and Low Affinity Pentamidine Transporter ( LAPT1 ) , have been described ( de Koning , 2001a; de Koning and Jarvis , 2001; Bridges et al . , 2007 ) . HAPT1 was identified as Aquaglyceroporin 2 ( TbAQP2 ) via an RNAi library screen , and found to be the main determinant of MPXR ( Baker et al . , 2012 , Baker et al . , 2013; Munday et al . , 2014 ) . The apparent permissibility for high molecular weight substrates by TbAQP2 was attributed to the highly unusual selectivity filter of TbAQP2 , which lacks the canonical aromatic/arginine ( ar/R ) and full NPA/NPA motifs , resulting in a much wider pore ( Baker et al . , 2012; Munday et al . , 2014; Munday et al . , 2015a ) . Importantly , the introduction of TbAQP2 into Leishmania promastigotes greatly sensitised these cells to pentamidine and melarsen oxide ( Munday et al . , 2014 ) . Moreover , in several MPXR laboratory strains of T . brucei the AQP2 gene was either deleted or chimeric after cross-over with the adjacent TbAQP3 gene , which , unlike AQP2 , contains the full , classical ar/R and NPA/NPA selectivity filter motifs and is unable to transport either pentamidine or melaminophenyl arsenicals ( Munday et al . , 2014 ) . Similar chimeric genes and deletions were subsequently isolated from sleeping sickness patients unresponsive to melarsoprol treatment ( Graf et al . , 2013; Pyana Pati et al . , 2014 ) and failed to confer pentamidine sensitivity when expressed in a tbaqp2-tbaqp3 null T . brucei cell line whereas wild-type TbAQP2 did ( Munday et al . , 2014; Graf et al . , 2015 ) . The model of drug uptake through a uniquely permissive aquaglyceroporin ( Munday et al . , 2015a ) was challenged by a study arguing that instead of traversing the TbAQP2 pore , pentamidine merely binds to an aspartate residue ( Asp265 ) near the extracellular end of the pore , above the selectivity filter , followed by endocytosis ( Song et al . , 2016 ) . This alternative , ‘porin-receptor’ hypothesis deserves careful consideration given that ( i ) it is an exceptional assertion that drug-like molecules with molecular weights grossly exceeding those of the natural substrates , can be taken up by an aquaglyceroporin and ( ii ) the fact that bloodstream form trypanosomes do have , in fact , a remarkably high rate of endocytosis ( Field and Carrington , 2009; Zoltner et al . , 2016 ) . The question is also important because aquaporins are found in almost all cell types and the mechanism by which they convey therapeutic agents and/or toxins into cells is of high pharmacological and toxicological interest . While TbAQP2 is the first aquaporin described with the potential to transport drug-like molecules , this ability might not be unique , and the mechanism by which the transport occurs should be carefully investigated . We therefore conducted a mutational analysis was undertaken , swapping TbAQP2 and TbAQP3 selectivity filter residues and altering pore width at its cytoplasmic end . This was complemented with a thorough structure-activity relationship study of the interactions between pentamidine and TbAQP2 , using numerous chemical analogues for which inhibition constants were determined and interaction energy calculated . The pentamidine-TbAQP2 interactions were further modelled by running a molecular dynamics simulation on a protein-ligand complex , and in addition , we investigated a potential correlation between the T . brucei endocytosis rate and the rate of pentamidine uptake . Our results provide strong evidence for pentamidine permeating directly through the central pore of TbAQP2 . Having identified the essential characteristics that allow the transport of large , flexible molecules through TbAQP2 , this should now allow the evaluation of aquaporins in other species for similar adaptations . One highly conserved motif of aquaporins , believed to be essential for permeant selectivity , is NPA/NPA which is present in TbAQP3 but not in TbAQP2 , where , uniquely , it is NS131A/NPS263 instead . We therefore constructed a TbAQP2 variant with the classical NPA/NPA motif ( TbAQP2S131P/S263A ) and expressed it in the aqp2/aqp3 null cell line ( Baker et al . , 2012; Munday et al . , 2014 ) . In this cell line , uptake of 30 nM [3H]-pentamidine was reduced to 4 . 40 ± 0 . 71% ( n = 4 ) of the rate in the control line expressing TbAQP2WT ( p<0 . 05 , Student’s unpaired t-test ) , as well as significantly different from the rate measured in parallel in the tbaqp2/tbaqp3 null cells ( p<0 . 01 ) ( Figure 2A ) . The remaining pentamidine uptake was sufficient to strongly sensitise the TbAQP2S131P/S263A cells to pentamidine in a standard protocol of 48 hr incubation with the drug followed by a further 24 hr in the presence of the resazurin indicator dye ( p<0 . 0001 vs tbaqp2/tbaqp3 null ) but the EC50 was still significantly higher than the TbAQP2WT control ( p<0 . 05 ) ( Figure 2B ) . A similar effect was observed for the melaminophenyl arsenical drug cymelarsan , but there was no change in sensitivity to diminazene or the control drug phenylarsine oxide ( PAO ) , which is believed to diffuse directly across the membrane ( Fairlamb et al . , 1992; Figure 2B ) . The mutant L258Y , which has the AQP3 Tyr-250 half of the highly conserved aromatic/arginine ( ar/R ) motif , responsible for pore restriction and proton exclusion ( Wu et al . , 2009 ) , introduced into the TbAQP2 pore , yielded a drug transport phenotype similar to TbAQP2S131P/S263A . The [3H]-pentamidine transport rate was reduced to 6 . 6 ± 1 . 4% of TbAQP2WT ( p<0 . 05 ) but remained above the rate in the tbaqp2/tbaqp3 null cells ( p<0 . 01 ) ( Figure 2A ) . Pentamidine and cymelarsan EC50 values were also significantly different from both the TbAQP2WT and the tbaqp2/tbaqp3 null controls ( Figure 2C ) . The ar/R motif is part of the larger selectivity filter , usually WGYR , present in both TbAQP1 and TbAQP3 but uniquely consisting of I110VL258L264 in TbAQP2 ( Baker et al . , 2013 ) , all non-polar , open chained residues . Cell lines expressing mutations AQP2I110W and AQP2L264R , either alone or in combination , displayed pentamidine transport rates , and pentamidine and cymelarsan EC50 values that were not significantly different from the tbaqp2/tbaqp3 null controls but highly significantly different from the TbAQP2WT drug-sensitive controls , showing that their capacity for pentamidine and cymelarsan uptake had been reduced to zero ( Figure 2A , D–F ) . We conclude that the unique TbAQP2 replacement of the NPA/NPA motif and all of the WGYR selectivity filter mutations are necessary for the observed pentamidine and melaminophenyl arsenical sensitivity observed in cells expressing wild-type TbAQP2 . The knockdown of the CRK12 kinase in T . brucei causes a highly reproducible defect in endocytosis that affects an estimated one third of cells 12 hr after RNAi induction and is ultimately lethal ( Monnerat et al . , 2013 ) . We utilized this system to investigate whether a link between endocytosis and pentamidine transport exists . At 12 hr of CRK12 RNAi induction with tetracycline , CRK12 mRNA levels were reduced by 42% ( p<0 . 001 ) relative to uninduced controls as determined by qRT-PCR ( Figure 5A ) . Samples from the culture taken at this time point showed an increased abundance of cells with swelling characteristic of endocytosis defects , although this was hard to quantify as a minority of cells were affected , and to various degrees , as the 12 hr time point was deliberately taken at as early a point as possible , so as not to affect cell viability ( Figure 5—figure supplement 1 ) or cause excessive cellular pathology . We thus performed parallel uptake experiments with [3H]-pentamidine and [3H]-suramin , with suramin acting as positive control as it is known to enter T . brucei bloodstream forms through endocytosis after binding to surface protein ISG75 ( Zoltner et al . , 2016; Zoltner et al . , 2020 ) . After 12 hr of CRK12 RNAi induction , pentamidine uptake was not significantly less than in the T . brucei 2T1 parental cells , whereas uptake of [3H]-suramin was ( p=0 . 019 , n = 5; Figure 5B , C ) . As this approach is necessarily limited to a partial inhibition of endocytosis in BSF T . brucei , it was also investigated whether pentamidine induces the internalisation and turnover of TbAQP2 , as could be expected if the protein acts to internalise substrate by receptor-mediated endocytosis . Cells pre-treated with the protein synthesis inhibitor cycloheximide ( 100 µg/ml ) were incubated in the presence or absence of 25 nM pentamidine and the abundance of 3 × HA tagged TbAQP2 was followed over a period of 6 hr by western blot ( Figure 6—figure supplement 1 ) . Quantification of the bands showed identical turnover rates with or without pentamidine present in the medium ( Figure 6 ) . It has been reported that knock-down of the HA1–three plasma membrane proton pumps of T . brucei ( which are essential for maintaining the plasma membrane potential ) , confers pentamidine resistance ( Alsford et al . , 2012; Baker et al . , 2013 ) . Interestingly , this locus only conferred resistance to ( dicationic ) pentamidine , not to the ( neutral ) melaminophenyl arsenicals , unlike knockdown of the TbAQP2/TbAQP3 locus ( Alsford et al . , 2012 ) . We have previously reported that the HAPT-mediated pentamidine uptake in T . brucei procyclics correlates strongly with the proton-motive force ( PMF ) ( de Koning , 2001a ) . However , it is not clear whether this dependency indicates that pentamidine uptake is mediated by a proton symporter , as known for many T . brucei nutrient transporters ( de Koning and Jarvis , 1997a; de Koning and Jarvis , 1997b; de Koning and Jarvis , 1998; de Koning et al . , 1998 ) , or reflects the energetics of uptake of cationic pentamidine being driven by the strong inside-negative membrane potential Vm . The absence of an effect of HA1–three knockdown on sensitivity to the neutral melaminophenyl arsenicals strongly argues against a mechanism of proton symport for HAPT1/AQP2 but a ( partial ) dependency of HAPT1/AQP2-mediated uptake of dicationic pentamidine on PMF or Vm would be expected if the substrate traverses the channel , as opposed to binding a single Asp residue on the extracellular side of the protein , as suggested in the endocytosis model ( Song et al . , 2016 ) . Here we show that the same ionophores that inhibit HAPT1-mediated pentamidine transport in procyclic cells , and inhibit hypoxanthine uptake in both bloodstream form ( BSF ) ( de Koning and Jarvis , 1997b ) and procyclic ( de Koning and Jarvis , 1997a ) T . brucei , also dose-dependently inhibit [3H]-pentamidine uptake in BSF ( Figure 7A ) . This confirms that pentamidine needs the membrane potential for rapid uptake , as predicted by the dependence on the HA1–three proton pumps . Using [3H]-suramin as an endocytosed substrate ( Zoltner et al . , 2016 ) , we found that 20 µM CCCP also inhibits endocytosis in T . brucei , by 32 . 6% ( p=0 . 029; pre-incubation 3 min , plus suramin accumulation over 10 min ) ( Figure 7B ) . While that means that the ionophore experiments do not perfectly discriminate between endocytosis and trans-channel transport for di-cationic pentamidine , they do for neutral melaminophenyl arsenicals: the non-dependence of these neutral TbAQP2 substrates on the proton gradient ( Alsford et al . , 2012 ) indicates that , unlike suramin , they are not endocytosed . Although there is a good correlation between the proton-motive force and TbAQP2-mediated pentamidine transport ( Figure 7C ) , the effect of CCCP was stronger than expected , and stronger than previously observed for [3H]-hypoxanthine uptake in T . brucei bloodstream forms ( de Koning and Jarvis , 1997b ) and we thus investigated whether CCCP might have a direct effect on TbAQP2 . Indeed , CCCP inhibited uptake of ( neutral ) [3H]-glycerol in tbaqp1-2-3 null cells expressing TbAQP2-WT , with an IC50 of 20 . 7 ± 2 . 6 µM ( n = 3 ) and inhibited [3H]-pentamidine uptake in the same cells with a similar IC50 ( Figure 7D , E ) , showing CCCP to inhibit TbAQP2 directly , irrespective of effects on the membrane potential . Figure 7D also shows that pentamidine , used as a control , inhibits [3H]-glycerol uptake with an EC50 value ( Mean of 27 . 5 nM , n = 2 ) similar to the EC50 of pentamidine inhibiting uptake of [3H]-pentamidine . To further investigate pentamidine binding and permeation in TbAQP2 , we used the coordinates of the TbAQP2-pentamidine complex that was modelled in our previous study ( Munday et al . , 2015a ) . The stability of the protein model was first confirmed by unbiased atomistic molecular dynamics simulations ( Figure 8—figure supplement 1 ) . We then conducted force-probe simulations , in which a moving spring potential was used to enforce unbinding of pentamidine from its docked binding position and subsequently reconstructed the free-energy profile of pentamidine association-dissociation along the pore axis by employing Jarzynski’s equality ( Park et al . , 2003 ) . Figure 8A shows that the docked position of pentamidine correctly identified its minimum free-energy binding site inside the TbAQP2 pore . Pentamidine adopts an extended state inside the TbAQP2 pore , adapting its molecular shape to the narrow permeation channel; pentamidine binding poses display inter-amidine lengths in the range 16 . 5–17 Å . Importantly , our steered simulations reveal that pentamidine can exit the channel in either direction , and that unbinding on the route towards the cytoplasm occurs on a free-energy surface roughly symmetric to that towards the extracellular side . Apart from overcoming the strongly attractive binding interaction in the centre , there are no major further free-energy barriers in either direction . The computed free-energy profile of pentamidine binding to the TbAQP2 structural model slightly overestimates its experimentally recorded binding affinity . However , the pentamidine conformation binding the narrow pore may not be the lowest-energy internal conformation of the small molecule , a factor that may be underrepresented in the profile as simulations were started from the protein-bound state . A further source of uncertainty stems from the protein model , which is expected to be somewhat less accurate than a crystal structure . Due to the dicationic character of pentamidine , the free-energy profile of the molecule within TbAQP2 strongly depends on the membrane voltage . The voltage drop of −125 mV across the cytoplasmic membrane of T . b . brucei ( de Koning and Jarvis , 1997b ) , with a negative potential inside the cell , results in an overall inward attraction of ~22 kJ/mol ( Figure 8A , arrow ) , that is exit from TbAQP2 into the cytoplasm is substantially more favourable for pentamidine than towards the extracellular side . Taken together , the free-energy profile under membrane voltage explains the strong coupling between pentamidine uptake and Vm observed in the experiments . The high affinity of the binding interaction leads to slow off-rates and a relatively low Vmax ( 0 . 0044 ± 0 . 0004 pmol ( 107 cells ) −1 s−1 ) ( de Koning , 2001a ) . We further investigated the bound positions of pentamidine and melarsoprol in AQP2 by docking , in order to rationalise the differential behaviour of pentamidine and melaminophenyl arsenicals observed in the studied AQP2 mutants . The binding modes in the central pore of AQP2 obtained by docking are shown in Figure 8B . They reveal that both drugs are likely to bind to the same general region within the central pore in spite of their different sizes and charge states . Conversely , the shorter arsenical agent is more affected by introducing the NPA/NPA motif since its terminal polar function intensely interacts with the NSA/NPS motif , whereas the major interactions of pentamidine are seen outside this motif . In the case of the L258Y mutant , the difference can be attributed to the extended flexible linker region in the centre of pentamidine , which is likely to enable it to bend around the added bulky side chain in the mutant , while cymelarsan lacks this level of flexibility in its centre . Finally , the single and double mutations I190T and W192G have broadly similar effects on pentamidine and cymelarsan permeation across AQP2 ( Figure 2 ) as both positions at the entrance to the pore exhibit similar interaction patterns to both pentamidine and the arsenical agent . In order to study substrate binding and selectivity by the T . b . brucei High Affinity Pentamidine Transporter ( HAPT1/TbAQP2 ) , competition assays were performed with a series of pentamidine analogues and other potential inhibitors , in the presence of 1 mM unlabelled adenosine to block diamidine uptake by the TbAT1 aminopurine transporter ( de Koning , 2001a; Bridges et al . , 2007 ) . High specific activity [3H]-pentamidine was used at 30 nM , below the Km value ( de Koning , 2001a ) . Uptake was linear for at least 3 min ( de Koning , 2001a ) and we utilized 60 s incubations for the determination of inhibition constants ( Ki ) . At 30 nM [3H]-pentamidine there is virtually no uptake through LAPT1 ( Bridges et al . , 2007 ) ( Km value ~1000 fold higher than HAPT1 ) ( de Koning , 2001a ) . The full dataset of 71 compounds is presented in Supplementary file 1 , featuring Kis spanning five log units . There is overwhelming consensus that expression of TbAQP2 is associated with the extraordinary sensitivity of T . brucei to pentamidine and melaminophenyl arsenicals , and that mutations and deletions in this locus cause resistance ( Baker et al . , 2012 , Baker et al . , 2013; Graf et al . , 2013; Graf et al . , 2015; Graf et al . , 2016; Pyana Pati et al . , 2014; Munday et al . , 2014; Munday et al . , 2015a; Unciti-Broceta et al . , 2015 ) . What has remained unclear , however , is the mechanism underpinning these phenomena – there are currently no documented other examples of aquaporins transporting such large molecules . Yet , considering how ubiquitous aquaporins are to almost all cell types , this question is of wide pharmacological importance: if large cationic and neutral drugs ( pentamidine and melarsoprol , respectively ) can be taken up via an aquaglyceroporin of T . brucei , what other pharmacological or toxicological roles may these channels be capable of in other cell types ? This manuscript clearly shows that changes in the TbAQP2 WGYR and NPA/NPA motifs , which collectively enlarge the pore and remove the cation filter , allow the passage of these drugs into the cell , and thereby underpin the very high sensitivity of the parasite to these drugs . TbAQP2 has evolved , apparently by positive selection given the high dN/dS ratio , to remove all main constriction points , including the aromatic amino acids and the cationic arginine of the ar/R selectivity filter , and the NPA/NPA motif , resulting in an unprecedentedly enlarged pore size ( Baker et al . , 2012; Munday et al . , 2015a ) . Whereas the advantage of this to T . b . brucei is yet unknown , the adaptation is stable within the brucei group of trypanosomes , and found in T . b . rhodesiense ( Munday et al . , 2014; Graf et al . , 2016 ) , T . b . gambiense ( Graf et al . , 2013; Graf et al . , 2015; Munday et al . , 2014; Pyana Pati et al . , 2014 ) , T . equiperdum and T . evansi ( Philippe Büscher and Nick Van Reet , unpublished ) . As such , it is not inappropriate to speculate that the wider pore of TbAQP2 ( i ) allows the passage of something not transported by TbAQP1 and TbAQP3; ( ii ) that this confers an a yet unknown advantage to the cell; and ( iii ) that uptake of pentamidine is a by-product of this adaptation . It is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis . For instance , the rate of endocytosis in bloodstream trypanosomes is much higher than in the procyclic lifecycle forms ( Langreth and Balber , 1975; Zoltner et al . , 2016 ) , yet the rate of HAPT-mediated [3H]-pentamidine uptake in procyclics is ~10 fold higher than in bloodstream forms ( de Koning , 2001a; Teka et al . , 2011 ) , despite the level of TbAQP2 expression being similar in both cases ( Siegel et al . , 2010; Jensen et al . , 2014 ) . Moreover , in procyclic cells TbAQP2 is spread out over the cell surface ( Baker et al . , 2012 ) but endocytosis happens exclusively in the flagellar pocket ( Field and Carrington , 2009 ) ( which is 3-fold smaller in procyclic than in bloodstream forms [Demmel et al . , 2014] ) , as the pellicular microtubule networks below the plasma membrane prevent endocytosis ( Zoltner et al . , 2016 ) . Thus , TbAQP2-mediated pentamidine uptake should be all but impossible in procyclic T . brucei , if dependent on endocytosis . Similarly , the expression of TbAQP2 in Leishmania mexicana promastigotes produced a rate of [3H]-pentamidine uptake more than 10-fold higher than observed in T . brucei BSF ( Munday et al . , 2014 ) , despite these cells also having a low endocytosis rate ( Langreth and Balber , 1975 ) . The Km and inhibitor profile of the TbAQP2-mediated pentamidine transport in these promastigotes was indistinguishable from HAPT in procyclic or bloodstream form T . brucei ( de Koning , 2001a ) . The experimental Vmax for HAPT-mediated pentamidine uptake in T . brucei BSF and procyclics ( de Koning , 2001a ) can be expressed as 9 . 5 × 105 and 8 . 5 × 106 molecules/cell/h , respectively; given a 1:1 stoichiometry for AQP2:pentamidine the endocytosis model would require the internalisation and recycling of as many units of TbAQP2 and this seems unlikely , especially in procyclic cells , as even in BSF the half-life time for TbAQP2 turnover is >4 hr ( Quintana et al . , 2020 ) and procyclic cells have a lower endocytosis rate and cannot easily internalise the aquaporins spread over the cell surface , as discussed above . Given the observed rate of uptake and turnover rate , this would require the presence of ~4 × 106 TbAQP2 units per BSF cell in the flagellar pocket . These observations are all inconsistent with the contention that pentamidine uptake by trypanosomes is principally dependent on endocytosis . Although it is likely that AQP2-bound pentamidine is internalised as part of the natural turnover rate of the protein , this is not likely to contribute very significantly to the overall rate of uptake of this drug . Furthermore , the Gibbs free energy of −42 kJ/mol for the pentamidine/AQP2 interaction ( de Koning , 2001a; Zoltner et al . , 2016 ) is highly unlikely to be the result of the one interaction between one terminal amidine and Asp265 as required in the endocytosis model ( Song et al . , 2016 ) . For the TbAT1 transporter , a double H-bond interaction of Asp140 with the N1 ( H ) /C ( 6 ) NH2 motif of adenosine or with one amidine of pentamidine ( Munday et al . , 2015b ) is estimated to contribute only ~16 kJ/mol to the total ΔG0 of −34 . 5 kJ/mol for adenosine ( −36 . 7 kJ/mol , pentamidine ) ( de Koning and Jarvis , 1999 ) . The endocytosis model also does not address the internalisation of melaminophenyl arsenicals , which presumably would equally need access to Asp265 , or address why most diamidines including furamidines and diminazene aceturate are at best extremely poor substrates for TbAQP2 ( Teka et al . , 2011; Ward et al . , 2011 ) . Here we systematically mapped the interactions between the aquaporin and pentamidine ( ΔG0 for 71 compounds ) , yielding a completely consistent SAR with multiple substrate-transporter interactions , summarised in Figure 9D . The evidence strongly supports the notion that pentamidine engages TbAQP2 with both benzamidine groups and most probably with at least one of the linker oxygens , and that its flexibility and small width are both required to optimally interact with the protein . This is completely corroborated by molecular dynamics modelling , which shows minimal energy to be associated with a near-elongated pentamidine centrally in the TbAQP2 pore , without major energy barriers to exiting in either direction , but driven to the cytoplasmic side by the membrane potential . This contrasts with the contention ( Song et al . , 2016 ) that pentamidine could not be a permeant for TbAQP2 because it did not transport some small cations and that this proves that the larger pentamidine cannot be a substrate either . There is scant rationale for that assertion: out of many possible examples: there are 5 orders of magnitude difference in affinity for pentamidine and para-hydroxybenzamidine ( 35 nM vs 2 . 9 mM; Supplementary file 1 ) ; adenine is not a substrate for the T . brucei P1 adenosine transporter ( de Koning and Jarvis , 1999 ) , the SLC1A4 and SLC1A5 neutral amino acid transporters transport Ala , Ser , Cys and Thr but not Gly ( Kanai et al . , 2013 ) , Na+ is not a permeant of K+ channels ( Zhorov and Tikhonov , 2013 ) and some NAT family transporters from bacterial , plant and fungal species display much higher affinity for xanthine and uric acid but not for hypoxanthine ( Gournas et al . , 2008 ) . The endocytosis model identifies only two key residues for pentamidine access ( Leu264 ) and binding ( Asp265 ) in TbAQP2 ( Song et al . , 2016 ) . Yet , multiple clinical isolates and laboratory strains contain chimeric AQP2/3 genes associated with resistance and/or non-cure that have retained those residues and should thus allow binding and internalisation of pentamidine ( Graf et al . , 2013; Pyana Pati et al . , 2014; Unciti-Broceta et al . , 2015; Munday et al . , 2014 ) . Although we find that introduction of the AQP3 Arg residue in position 264 ( TbAQP2L264R ) disables pentamidine transport , we would argue that this is because the positively charged arginine , in the middle of the pore , is blocking the traversing of all cations through the pore , as is its common function in aquaporins ( Beitz et al . , 2006; Wu et al . , 2009 ) . Indeed , the W ( G ) YR filter residues appear to be key determinants for pentamidine transport by AQPs and the introduction of all three TbAQP2 residues into TbAQP3 ( AQP3W102I/R256L/Y250L ) was required to create an AQP3 that at least mildly sensitised to pentamidine , and facilitated a detectable level of pentamidine uptake . Conversely , any one of the mutations I110W , L258Y or L264R was sufficient to all but abolish pentamidine transport by TbAQP2 . Similarly , the conserved NPA/NPA motif , and particularly the Asp residues , present in TbAQP3 but NSA/NPS in TbAQP2 , is also associated with blocking the passage of cations ( Wree et al . , 2011 ) . The unique serine residues in this TbAQP2 motif , halfway down the pore , might be able to make hydrogen bonds with pentamidine . Reinstating the NPA/NPA motif resulted in a TbAQP2 variant with a 93 . 5% reduced rate of [3H]-pentamidine transport . Tryptophan residues were introduced towards the cytoplasmic end of the TbAQP2 pore ( L84W , L118W , L218W ) to test the hypothesis that introducing bulky amino acids in that position would block the passage of pentamidine . Each of these mutants was associated with reduced sensitivity to pentamidine and cymelarsan and a > 90% reduction in [3H]-pentamidine uptake . This effect was size-dependent as the pentamidine transport rate of L84M and L218M was statistically identical to that of control TbAQP2 cells , and L118M also displayed a higher transport rate than L118W ( p<0 . 0001 ) . These mutant AQPs were still functional aquaglyceroporins as their expression in tbaqp1-2-3 null cells made those cells less sensitive to the TAO inhibitor SHAM , and increased glycerol uptake . Independence from endocytosis was investigated by employing the tetracycline-inducible CRK12 RNAi cell line previously described to give a highly reproducible and progressive endocytosis defect in T . brucei ( Monnerat et al . , 2013 ) , with the aim to distinguish between uptake via endocytosis and transporters , as current evidence suggests that in T . brucei all endocytosis , taking place exclusively in the flagellar pocket , is clathrin-dependent and AP-2 independent ( Morgan et al . , 2002; Allen et al . , 2003 ) . This means that the endocytotic mechanisms of TbAQP2 and suramin receptor ISG75 , which are both directed to the lysosome after ubiquitylation ( Quintana et al . , 2020; Zoltner et al . , 2015 ) , are likely to be similar enough for a direct comparison . Twelve hours after CRK12 RNAi induction pentamidine transport was not significantly reduced although uptake of [3H]-suramin , which is accumulated by endocytosis through the T . brucei flagellar pocket ( Zoltner et al . , 2016 ) , was reduced by 33% ( p=0 . 0027 ) , indicating successful timing of the experiment to the early stage of endocytosis slow-down . We also show that pentamidine , at approximately its half-maximal occupancy concentration , did not influence the half-life time of TbAQP2 turnover , as is often the case in receptor-mediated endocytosis . Nor should TbAQP2 , which exists as a rigid tetramer of tetramers in T . brucei bloodstream forms ( Quintana et al . , 2020 ) , be able to undergo the type of conformational change necessary to signal receptor occupancy and internalisation as observed in well-documented examples of ligand-triggered internalisation of transporters ( e . g . Gournas et al . , 2010; Gournas et al . , 2017; Ghaddar et al . , 2014; Keener and Babst , 2013 ) . Thus , the combined evidence , taken together , strongly suggests that pentamidine is not taken up by endocytosis , not induce endocytosis of TbAQP2 . Although several ionophores , including CCCP , nigericin and gramicidin strongly inhibited pentamidine uptake , similar to what has been previously reported for transport processes in T . brucei that are linked to the protonmotive force ( de Koning and Jarvis , 1997a; de Koning and Jarvis , 1997b; de Koning and Jarvis , 1998; de Koning et al . , 1998 ) , this is probably due to the inside-negative membrane potential of −125 mV ( de Koning and Jarvis , 1997b ) attracting the dicationic pentamidine . This is consistent with the prediction of the molecular dynamics modelling , and the reported role of the HA1–three proton pumps in pentamidine but not melarsoprol resistance ( Alsford et al . , 2012; Baker et al . , 2013 ) . Although CCCP does inhibit pentamidine through direct , competitive inhibition of TbAQP2 as well , this only starts to have a measurable impact above ~5 µM ( IC50 of 20 . 7 µM ) , whereas its effects after preincubation , i . e . the combination of competitive inhibition and reducing the protonmotive force , shows ~63% inhibition of pentamidine transport at 1 µM and ~90% inhibition at 5 µM , showing that the more important effect of CCCP is via reduction of the PMF . This is consistent with the conclusion from the molecular dynamics analysis that inward pentamidine flux is dependent on the inside-negative membrane potential . Altogether , we conclude that the primary entry of the sleeping sickness drugs pentamidine and melarsoprol into T . brucei spp . is through the unusually large pore of TbAQP2 , rendering the parasite extraordinarily sensitive to the drugs ( compare Leishmania mexicana [Munday et al . , 2014] ) . This is the first report providing detailed mechanistic evidence of the uptake of organic drugs ( of MW 340 and 398 , respectively ) by an aquaporin . We show that this porin has evolved through positive selection and identify the adaptations in the constriction motifs that enabled it . We consider that other pore-opening adaptations may have evolved in other organisms , including pathogens , which could initiate the pharmacological exploitation of aquaporins and lead to the design of new drug delivery strategies . The drug-sensitive clonal T . b . brucei strain 427 ( MiTat 1 . 2/BS221 ) ( de Koning et al . , 2000 ) was used for all the work on the SAR of pentamidine transport . The tbaqp2/tbaqp3 null cells ( Baker et al . , 2012 ) and tbaqp1-2-3 null cells ( Jeacock et al . , 2017 ) ( both obtained from David Horn , University of Dundee , UK ) are derived from the 2T1 strain of T . b . brucei ( Alsford and Horn , 2008 ) . The CRK12 RNAi cell line28 was obtained from Dr Tansy Hammarton ( University of Glasgow , UK ) and is also based on the 2T1 cell line; RNAi expression was induced with 1 µg/ml tetracycline in the medium . All experiments were performed with bloodstream form trypanosomes grown in vitro in HMI-11 medium as described ( Wallace et al . , 2002 ) at 37°C in a 5% CO2 atmosphere . Cultures were routinely maintained in 10 ml of this medium , being seeded at 5 × 104 cells/ml and passed to fresh medium at reaching approximately 3 × 106 cells/ml after 48 hr . For transport experiments 150 or 200 ml of culture was seeded at the same density in large flasks and incubated until the culture reached late-log phase . A complete list of diamidine analogues and other chemicals used for the SAR study is given as a table with their sources ( Supplementary file 1 ) . Ionophores and uncouplers nigericin , gramicidin , carbonyl cyanide m-chlorophenyl hydrazone ( CCCP ) and valinomycin , as well as the T . brucei proton pump inhibitor N-ethylmaleimide ( NEM ) were all purchased from Sigma-Aldrich . New compounds synthesised for this study are listed and described in Supplementary file 3 . Transport assays - Transport of [3H]-pentamidine was performed exactly as previously described for various permeants ( Wallace et al . , 2002; Bridges et al . , 2007; Teka et al . , 2011 ) in a defined assay buffer ( AB; 33 mM HEPES , 98 mM NaCl , 4 . 6 mM KCl , 0 . 55 mM CaCl2 , 0 . 07 mM MgSO4 , 5 . 8 mM NaH2PO4 , 0 . 3 mM MgCl2 , 23 mM NaHCO3 , 14 mM glucose , pH 7 . 3 ) . [3H]-pentamidine was custom-made by GE Healthcare Life Sciences ( Cardiff , UK ) with a specific activity of 88 Ci/mmol . Incubations of bloodstream form trypanosomes with 30 nM of this label ( unless otherwise indicated ) were performed in AB at room temperature for 60 s ( unless otherwise indicated ) and terminated by addition of 1 ml ice-cold ‘stop’ solution ( 1 mM unlabelled pentamidine ( Sigma ) in AB ) and immediate centrifugation through oil ( 7:1 dibutylphthalate:mineral oil v/v ( both from Sigma ) ) . Transport was assessed in the presence of 1 mM adenosine to block uptake through the P2 aminopurine transporter; adenosine does not affect HAPT1-mediated transport ( de Koning , 2001a; Bridges et al . , 2007 ) . Inhibition assays were performed routinely with 6–10 different concentrations of inhibitor over the relevant range , diluting stepwise by one third each time , in order to obtain a well-defined and accurate sigmoid plot and IC50 value ( inhibitor concentration giving 50% inhibition of pentamidine transport; calculated by non-linear regression using Prism 6 . 0 ( GraphPad ) , using the equation for a sigmoid curve with variable slope ) . Highest concentration was usually 1 mM unless this was shown to be insufficient for good inhibition , or when limited by solubility . Ki values were obtained from IC50 values using ( 1 ) ki=IC50/[1+ ( L+km ) ]in which L is the [3H]-pentamidine concentration and Km the Michaelis-Menten constant for pentamidine uptake by HAPT1 ( Wallace et al . , 2002 ) . The Gibbs Free energy of interaction ΔG0 was calculated from ( 2 ) ΔG0=−RTlnKiin which R is the gas constant and T is the absolute temperature ( Wallace et al . , 2002 ) . Transport of [3H]-glycerol and [3H]-suramin was performed essentially as for [3H]-pentamidine . For [3H]-glycerol ( American Radiolabeled Chemicals , 40 . 0 Ci/mmol ) , 107 BSF T . brucei were incubated with radiolabel at a final concentration of 0 . 25 µM , for one minute . When the effect of CCP was studied , CCCP was added 3 min prior to the addition of the radiolabel . [3H]-suramin ( American Radiolabeled Chemicals , 20 . 0 Ci/mmol ) was also used at 0 . 25 µM final concentration , using 15 min incubations in the presence and absence of 100 µM unlabelled suramin ( used as saturation ) control . All mutations in the TbAQP2 and TbAQP3 genes were introduced to the relevant backbone WT vector , either pRPaGFP-AQP2 or pRPaGFP-AQP3 ( Baker et al . , 2012 ) , by site-directed mutagenesis . Use of the pRPa vector for transfection of 2T1-derived T . brucei ensures integration in a prepared locus in the ribosomal rRNA spacer region and a high level of stable expression ( Alsford et al . , 2005 ) . For mutations S131P , S263A , I110W , L264R , L258Y , I190T and W192G in AQP2 , and W102I , R256L and Y250L in AQP3 mutations were inserted using the QuikChange II kit ( Agilent , Santa Clara , CA , USA ) , following the manufacturer's instructions . For mutations L84W , L118W , L218W , L84M , L118M and L218M were introduced using the Q5 Site-Directed Mutagenesis Kit ( E0554S ) , ( New England BioLabs ) according to manufacturer’s instructions . The following primer pairs ( itemised in Supplementary file 4 ) were used to insert the named TbAQP2 mutations: for S131P , primers HDK1062 and HDK1063; in combination with mutation S263A , using primers HDK1064 and HDK1065 to produce plasmid pHDK166 . For I110W , primers HDK607 and HDK608 , to produce pHDK84; for L264R , primers HDK609 and HDK610 to produce pHDK167; the combination I110W/L264R was produced using primers HDK609 and HDK610 on plasmid pHDK84 to give plasmid pHDK78; for L258Y , primers HDK1109 and HDK1110 to produce pHDK168; for I190T , primers HDK1056 and HDK1057 to produce pHDK163; for W192G , primers HDK1058 and HDK1059 to produce pHDK164; the combination I190T/W192G was produced using primers HDK1060 and HDK1061 on plasmid pHDK163 to give plasmid pHDK165; for L84W , primers HDK1276 and HDK1277 , producing pHDK210; for L118W , primers HDK1274 and HDK1275 , producing pHDK208; for L218W , primers HDK1272 and HDK1273 , producing pHDK209; for the combination L84W/L118W , primers HDK1276 and HDK1277 on template pHDK208 , producing pHDK227; for L84M , primers HDK1364 and HDK1367 , producing pHDK234; for L118M , primers HDK1365 and HDK1367 , producing pHDK235; and for L218M , primers HDK1366 and HDK1367 , producing pHDK236 . To insert the named mutations into TbAQP3 , the following primers were used ( Supplementary file 5 ) : for W102I , primers HDK511 and HDK512 , in combination with mutation R256L , with primers HDK513 and HDK514 , to produce plasmid pHDK71; and to add mutation Y250L to this combination , primers HDK795 and HDK796 , to produce pHDK121 . All plasmids were checked by Sanger Sequencing ( Source BioScience , Nottingham , UK ) for the presence of the correct mutation ( s ) and the cassette for integration digested out with AscI ( NEB , Hitchin , UK ) prior to transfection . For transfection , 10 µg of digested plasmid and 1–2 × 107 parasites of the desired cell line ( either aqp2/aqp3 null or aqp1/aqp2/aqp3 null ) were resuspended in transfection buffer and transfected using an Amaxa Nucleofector , with program X-001 . After a recovery period ( 8–16 hr ) in HMI-11 at 37°C and 5% CO2 , the parasites were cloned out by limiting dilution with the selection antibiotic ( 2 . 5 µg/ml hygromycin ) . In all cases the presence of the construct and its correct integration into the designed rRNA locus was verified by three PCR reactions , one using primers for the amplification of the full-length aquaporin ( primers HDK529 and HDK209 ) . The second PCR was performed to amplify the gene with surrounding parts of the expression cassette using ( primers HDK1011 and HDK430 ) . The third PCR was to assess whether the expression cassette had linearized and integrated into the T . brucei genome using ( primers HDK991 and HDK713 ) . Drug sensitivity assays for T . b . brucei bloodstream forms used the cell viability dye resazurin ( Sigma ) and were performed exactly as described ( Wallace et al . , 2002; Bridges et al . , 2007 ) in 96-well plates with doubling dilutions of test compound , starting at 100 µM , over 2 rows of the plate ( 23 dilutions plus no-drug control ) . Incubation time with test compound was 48 hr ( 37°C/5% CO2 ) , followed by an additional 24 hr in the presence of the dye . T . brucei 2T1 containing 3×HAAQP2 were incubated with 1 µg/ml tetracycline for 24 hr at 37°C/5% CO2 to induce expression . Tetracycline was then washed away by four consecutive washes with fresh supplemented HMI-9 , and cells were then incubated with 25 nM of pentamidine ( 5 × EC50 ) for 1 hr at 37°C/5% CO2 prior to harvest cells for pulse chase experiments . To determine protein half-life , translation was blocked by addition of cycloheximide ( 100 µg/ml ) and cells were harvested at various times by centrifugation ( 800 × g for 10 min at 4°C ) . Cells were washed with ice-cold PBS , then resuspended in 1 × SDS sample buffer ( Thermo ) and incubated at 70°C for 10 min . Proteins were subjected to electrophoresis using 4–12% precast acrylamide Bis-Tris gels ( Thermo ) and transferred to polyvinylidene difluoride ( PVDF; Sigma-Aldrich ) membranes with a iBlot2 system ( Thermo ) at 23 V for 6 min , exactly as described ( Quintana et al . , 2020 ) , blocking non-specific binding with 5% ( w/v ) bovine serum albumin ( BSA; Sigma ) in Tris-buffered saline ( pH 7 . 4 ) with 0 . 2% ( v/v ) Tween-20 ( TBST ) and using rat anti-HA IgG1 ( Sigma ) or anti-mouse β-tubulin ( clone KMX-1; Millipore ) at 1:5000 or 1:10 , 000 dilution in TBST , respectively . Membranes were washed five times with TBST and then incubated in TBST/1% BSA with the appropriate horseradish peroxidase ( HRP ) -coupled secondary antibody ( Sigma ) , at 1: 10 , 000 . Bands were visualised using the ECL Prime Western Blotting System ( Sigma ) and GE healthcare Amersham Hyperfilm ECL ( GE ) . Densitometry quantification was conducted using ImageJ software ( NIH ) . Molecular dynamics simulations were performed using the GROMACS software package , version 5 . 1 . 1 ( Abraham et al . , 2015 ) . We used the coordinates from the homology model of TbAQP2 published in Figure 2A in Munday et al . , 2015a , which was inserted into POPC/POPE ( 4:1 ) membranes , approximately reflecting the membrane composition of T . b . brucei ( Smith and Bütikofer , 2010 ) . The membrane models were constructed using the CHARMM-GUI webserver ( Jo et al . , 2008 ) . Subsequently , extended stability tests of the modelled structure and the bound pentamidine were carried out using unbiased simulations of 100 ns length . The root-mean-square deviation ( RMSD ) of the protein remained relatively low with a backbone RMSD converging to ~3 Å after 100 ns simulated time ( Figure 8—figure supplement 1 ) bound to the binding site defined previously using molecular docking ( Figure 8A; Munday et al . , 2015a ) . For these and all following simulations , we used the CHARMM36 force field ( Klauda et al . , 2010 ) pentamidine was parameterised using the CHARMM generalized force field approach ( CHGenFF [Vanommeslaeghe et al . , 2010] ) . All simulations employed a time step of 2 fs under an NPT ensemble at p=1 bar and T = 310 K . To obtain non-equilibrium work values for removing pentamidine from the internal AQP2 binding site , we then conducted steered MD simulations with a probe speed of 0 . 005 nm/ns and a harmonic force constant of 300 kJ/mol nm2 , pulling pentamidine in both directions along the pore axis . The free energy profile of pentamidine binding to the AQP2 pore was reconstructed by using the Jarzynski , 1997 equality . All transport experiments were performed in triplicate and all values such as rate of uptake , percent inhibition , Ki , Km , Vmax etc were performed at least three times completely independently . For drug sensitivity tests , all EC50 values were based on serial dilutions over two rows of a 96-well plate ( 23 doubling dilutions plus no-drug control ) , which were obtained independently at least three times . EC50 and IC50 values were determined by non-linear regression using the equation for a sigmoid curve with variable slope and are presented as average ± SEM . Statistical significance between any two data points was determined using Student’s t-test ( unpaired , two-tailed ) .	African sleeping sickness is a potentially deadly illness caused by the parasite Trypanosoma brucei . The disease is treatable , but many of the current treatments are old and are becoming increasingly ineffective . For instance , resistance is growing against pentamidine , a drug used in the early stages in the disease , as well as against melarsoprol , which is deployed when the infection has progressed to the brain . Usually , cases resistant to pentamidine are also resistant to melarsoprol , but it is still unclear why , as the drugs are chemically unrelated . Studies have shown that changes in a water channel called aquaglyceroporin 2 ( TbAQP2 ) contribute to drug resistance in African sleeping sickness; this suggests that it plays a role in allowing drugs to kill the parasite . This molecular ‘drain pipe’ extends through the surface of T . brucei , and should allow only water and a molecule called glycerol in and out of the cell . In particular , the channel should be too narrow to allow pentamidine or melarsoprol to pass through . One possibility is that , in T . brucei , the TbAQP2 channel is abnormally wide compared to other members of its family . Alternatively , pentamidine and melarsoprol may only bind to TbAQP2 , and then ‘hitch a ride’ when the protein is taken into the parasite as part of the natural cycle of surface protein replacement . Alghamdi et al . aimed to tease out these hypotheses . Computer models of the structure of the protein were paired with engineered changes in the key areas of the channel to show that , in T . brucei , TbAQP2 provides a much broader gateway into the cell than observed for similar proteins . In addition , genetic analysis showed that this version of TbAQP2 has been actively selected for during the evolution process of T . brucei . This suggests that the parasite somehow benefits from this wider aquaglyceroporin variant . This is a new resistance mechanism , and it is possible that aquaglyceroporins are also larger than expected in other infectious microbes . The work by Alghamdi et al . therefore provides insight into how other germs may become resistant to drugs .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "microbiology", "and", "infectious", "disease" ]	2020	Positively selected modifications in the pore of TbAQP2 allow pentamidine to enter Trypanosoma brucei
The NuRD complex is a multi-protein transcriptional corepressor that couples histone deacetylase and ATP-dependent chromatin remodelling activities . The complex regulates the higher-order structure of chromatin , and has important roles in the regulation of gene expression , DNA damage repair and cell differentiation . HDACs 1 and 2 are recruited by the MTA1 corepressor to form the catalytic core of the complex . The histone chaperone protein RBBP4 , has previously been shown to bind to the carboxy-terminal tail of MTA1 . We show that MTA1 recruits a second copy of RBBP4 . The crystal structure reveals an extensive interface between MTA1 and RBBP4 . An EM structure , supported by SAXS and crosslinking , reveals the architecture of the dimeric HDAC1:MTA1:RBBP4 assembly which forms the core of the NuRD complex . We find evidence that in this complex RBBP4 mediates interaction with histone H3 tails , but not histone H4 , suggesting a mechanism for recruitment of the NuRD complex to chromatin . The nucleosome remodelling and deacetylase ( NuRD ) complex is one of at least five corepressor complexes that recruits and activates class I histone deacetylase ( HDACs ) enzymes and directs their activity to chromatin ( Bantscheff et al . , 2011; Guenther et al . , 2000; Humphrey et al . , 2001; Itoh et al . , 2015; Kelly and Cowley , 2013; Laherty et al . , 1997; Oberoi et al . , 2011; Vermaak et al . , 1999; Watson et al . , 2012; Wen et al . , 2000; Xue et al . , 1998; Zhang et al . , 1999 ) . The NuRD complex plays a major role in the regulation of gene expression as well as being involved in chromatin organization , DNA damage repair , and genomic stability ( Denslow and Wade , 2007; Lai and Wade , 2011 ) . The NuRD complex is composed of six proteins , each with several isoforms: HDAC1/2 , MTA1/2/3 , RBBP4/7 , P66α/β , MBD2/3 , and CHD3/4 ( Allen et al . , 2013 ) . The complex is special in its ability to exhibit dual enzymatic functionality through its HDAC ( protein deacetylation ) and ATPase ( chromatin remodelling ) domains . Histone Deacetylases ( HDACs ) are important regulators of transcription that are involved in silencing genes and/or priming them for successive rounds of transcription ( Micelli and Rastelli , 2015; Verdin and Ott , 2014; Wang et al . , 2009 ) . HDACs act by removing the acetyl groups from modified lysine residues on histone tails and other non-histone proteins . These enzymes have been shown to be overexpressed in various cancer tissues and several inhibitors against these enzymes are currently used in the clinic as anti-cancer therapeutics ( Falkenberg and Johnstone , 2014; Marks and Xu , 2009 ) . Metastasis-associated protein 1 ( MTA1 ) ( and its closely related homologues MTA2 and 3 ) are up-regulated in various cancer tissues ( Fujita et al . , 2003; Li et al . , 2012; Sen et al . , 2014; Toh et al . , 1994 ) . They serve as scaffold proteins for the assembly of the NuRD complex . The amino-terminal region contains four established domains: an amino-terminal BAH domain , followed by ELM2 and SANT domains , and a GATA-like zinc finger domain towards the centre of the protein ( Manavathi and Kumar , 2007; Millard et al . , 2014 ) . The ELM2 domain has been shown by X-ray crystallography to mediate dimerization of the complex and , in conjunction with the SANT domain , recruits HDAC1 & 2 ( Millard et al . , 2013 ) . The structure and function of the MTA-BAH and zinc finger domains remain to be determined . The carboxy-termini of the MTA proteins are much more diverse and structure predictions suggest that much of the protein is intrinsically disordered . The exception is at the very carboxy-terminus of MTA1 and MTA2 where a short helix has been shown to mediate interaction with RBBP4/7 ( Alqarni et al . , 2014 ) . A binding site for CHD4 has also been identified in this region , but has not been structurally characterised ( Nair et al . , 2013 ) . RBBP4 and RBBP7 are homologous ( 92% identical ) histone binding proteins ( also known as RbAp48 and RbAp46 ) that have been shown to be core components of the NuRD complex . These proteins are unusual in that they are present in multiple complexes including other class I HDAC corepressors such as the Sin3A and PRC2 complexes ( Ciferri et al . , 2012; Kuzmichev , 2002; Zhang et al . , 1997 ) . These proteins have also been reported to have roles in nucleosome assembly: RBBP4 , along with sNASP and Asf1 forms a multi chaperone complex whose function is to maintain a supply of newly synthesized histone H3 and H4 proteins under replicative stress ( Groth et al . , 2005 ) . RBBP4 has also been found to be a subunit of the chromatin-assembly factor-1 ( CAF-1 ) complex whose function is to initiate nucleosome assembly by assembling histones H3 and H4 onto newly synthesized DNA ( Parthun et al . , 1996; Verreault et al . , 1996 ) . In addition , RBBP7 has been found to be an essential subunit of the HAT1 histone-acetyltransferase complex , which acetylates newly synthesized histone proteins ( Verreault et al . , 1998 ) . RBBP4 and RBBP7 have been shown to interact directly with histone H3 and H4 tails through two distinct binding sites and therefore have the potential to act as chromatin recruitment modules ( Murzina et al . , 2008; Nowak et al . , 2011; Schmitges et al . , 2011; Song et al . , 2008; Zhang et al . , 2013 ) . RBBP4 and RBBP7 are WD40 domain proteins with a characteristic seven bladed β-propeller structure ( Xu and Min , 2011 ) . RBBP4/7 interacts with histone H3 through a binding site on the ‘top’ face of the protein ( Schmitges et al . , 2011 ) . A particular feature of the RBBP4/7 WD40 domain is an amino-terminal helix of 30 residues that forms an extension to the β-propeller domain . A groove that is formed adjacent to this helix mediates interaction with the amino-terminus of histone H4 ( Murzina et al . , 2008 ) . Interestingly , the extreme carboxy-terminus of MTA1 ( residues 671–690 ) has been shown to interact with RBBP4/7 in this groove , making very similar contacts to those made by histone H4 ( Alqarni et al . , 2014 ) . In this study , we have identified a second RBBP4 recruitment site within MTA1 that is positioned near the centre of the corepressor . This site shows some sequence homology to the previously characterised carboxy-terminal recruitment site , but is considerably more extensive . We show that both sites within the MTA1 monomer are capable of independently recruiting RBBP4 proteins such that the HDAC1:MTA1 dimer recruits four molecules of RBBP4 to the NuRD complex . The crystal structure of this central recruitment site reveals an extensive interface that requires some structural rearrangement within RBBP4 . Small angle X-ray scattering ( SAXS ) and single-particle negative-stain electron microscopy ( EM ) reveal the architecture of the core NuRD complex . RBBP4 proteins are tethered to the HDAC1 dimer by MTA1 forming an elongated zig-zag conformation . We show that when in complex with MTA1 , RBBP4 is able to interact with histone H3 , but not histone H4 suggesting a mechanism for recruitment of the NuRD complex to chromatin . During purifications of various HDAC1:MTA1 complexes expressed in mammalian cells , we observed an endogenous protein of c . 50kDa co-purifying with these complexes . Peptide mass spectrometry identified this protein as being the WD40 protein RBBP4 and/or RBBP7 . The interaction between RBBP4/7 and the carboxy-terminus of MTA1 ( residues 670–691 ) is well characterised and an atomic resolution structure of a 22 residue MTA1 peptide bound to RBBP4 has been previously solved by X-ray crystallography ( Alqarni et al . , 2014 ) . However , we were intrigued to find that RBBP4/7 was recruited to constructs that lacked this well-characterised interaction domain . This appears to fit with previous findings that indicated that RBBP4/7 may interact with a region containing the MTA1 zinc finger domain and adjacent sequences ( Roche et al . , 2008 ) . These findings suggested that two regions of MTA1 are able to mediate interaction with RBBP4 . What remained unclear was whether these two regions collaborate to bind to a single RBBP4 protein or whether there are two separate binding sites that result in two RBBP4 proteins binding to MTA1 . To define the biologically relevant regions of MTA1 that recruit RBBP4/7 , various fragments of MTA1 , and full-length HDAC1 , were co-transfected into mammalian cells ( Figures 1a and 1b ) . These complexes were purified utilising a FLAG tag on MTA1 and we examined the apparent stoichiometry for the recruitment of endogenous RBBP4/7 to these complexes using gel densitometry ( Figure 1c ) . Importantly , the MTA1-B construct spanning residues 162–546 appears to pull down a stoichiometric 1:1 ratio of HDAC1:RBBP4/7 suggesting that this region is sufficient to form a bona fide interface with RBBP4/7 . Interesting a construct spanning residues 162–715 ( MTA1-C ) pulls down a supra-stoichiometric ratio of 1:1 . 5 of HDAC1:RBBP4/7 . This implies that there may be two RBBP4/7 proteins associated with MTA1 ( Figure 1c ) . 10 . 7554/eLife . 13941 . 003Figure 1 . MTA1 co-purifies with endogenous RBBP4/7 in a supra-stoichiometric ratio . ( a ) Schematic representation of the domain structure of MTA1 , with the R1 and R2 RBBP4 recruitment domains shown in purple . A summary of fragments used in the interaction studies is shown below . ( b ) MTA1-B and MTA1-C co-purify with endogenous RBBP4/7 , as identified by mass spectrometry , in a stoichiometric and a supra-stoichiometric ratio respectively . ( c ) The ratio of endogenous RBBP4/7 to co-transfected HDAC1 is quantified from the SDS-PAGE gel by densitometry . ( d ) and ( e ) Co-expression of the MTA1-D:RBBP4 and MTA1-R1:RBBP4 complexes . ( f ) and ( g ) Co-expressed MTA1-D:RBBP4 and MTA1-R1:RBBP4 are shown to form complexes of 1:2 and 1:1 stoichiometry , respectively , as determined by size exclusion chromatography coupled to multi-angle light scattering ( MALS ) . See Figure 1—figure supplement 1 for information about expression , purification and crystallisation . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 00310 . 7554/eLife . 13941 . 004Figure 1—figure supplement 1 . Purification of MTA1:RBBP4 complexes . ( a ) SDS-PAGE gel showing purification of the co-transfected MTA1-D ( 390–715 ) :RBBP4 complex by gel filtration . Peak elution of the complex is in lane 9 . ( b ) and ( c ) SDS-PAGE gel and Superdex S200 gel filtration profile of the co-transfected MTA1-R1 ( 464–546 ) :RBBP4 complex . Peak elution of the complex is in lane 6 . ( d ) Crystals of the MTA1-R1:RBBP4 complex . ( e ) Electron density ( 2Fo-Fc ) contoured at 1σ showing the interface between MTA1-R1 domain ( blue ) and RBBP4 ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 004 To confirm the stoichiometry of the complex we co-expressed RBBP4 with different length MTA1 fragments extending from the zinc finger domain . The resulting complexes were purified by gel filtration and analysed with Multi-Angle Light Scattering ( MALS ) to determine the overall molecular weight . The longer fragment of MTA1-D ( residues 390–715 ) in complex with RBBP4 was shown to have a molecular weight of 130 kDa ( Figures 1d and 1f; see also Figure 1—figure supplement 1 ) . This closely fits with the predicted molecular weight ( 131 kDa ) of a 1:2 stoichiometry of one MTA1 fragment bound to two RBBP4 proteins . The shorter MTA1-R1 fragment ( residues 464–546 ) in complex with RBBP4 was determined to have a molecular weight of 54 kDa which is close to the predicted molecular weight of 58 kDa for the 1:1 complex ( Figures 1e and 1g; see also Figure 1—figure supplement 1 ) . It is clear that these experiments support a model in which two RBBP4 proteins are recruited by MTA1 to the NuRD complex . Since there are two recruitment sites for RBBP4 we have called these R1 ( residues 464–546 ) and R2 ( residues 670–691 ) . To understand the nature of the interface between the R1 domain from MTA1 and RBBP4 we co-expressed and purified a complex between MTA1-R1 ( residues 464–546 ) and full-length RBBP4 . The complex was affinity purified on Flag resin using an amino-terminal tag on MTA1 before being cleaved from the resin and gel filtrated . The complex was concentrated to 10 mg/ml and small crystals were obtained . X-ray diffraction data were collected to 3 . 2 Å and the structure solved by molecular replacement using the structure of RBBP4 ( pdb code: 4PC0 ) . The density for MTA1 was remarkably clear and the structure of the complex was built through multiple rounds of refinement ( Figure 1—figure supplement 1 ) . The seven bladed beta-propeller fold of RBBP4 is largely unchanged by the binding of MTA1-R1 . The first 10 residues of RBBP4 are unstructured , and similar to previous structures of RBBP4/7 , the 14 carboxy-terminal residues are not observed . A notable difference in this structure is the stabilisation of a loop that supports binding of MTA1 ( see below ) . The interface between MTA1-R1 and RBBP4 is extensive with a buried surface area of 2 , 837 Å2 ( Figures 2a and 2b , Table 1 and Video 1 ) . MTA1 interacts on the side of the WD40 beta-propeller domain making interactions in three adjacent grooves . Two of these grooves lie on either side of the long amino-terminal helix that forms an extension to the WD40 domain . The contacts made by MTA1 on the surface of RBBP4 can be rationalised into three regions hereafter termed interaction regions A , B and C ( Figure 2c ) . The residues involved in all three interaction regions are conserved in both MTA2 and MTA3 ( Figure 2a ) . 10 . 7554/eLife . 13941 . 005Figure 2 . The crystal structure of MTA1-R1 domain bound to RBBP4 . ( a ) Schematic representation of the domain structure of MTA1 . The secondary structure of MTA1-R1 ( 464–546 ) , corresponding to the R1 domain , is shown . Those residues of MTA1 that form an interface with RBBP4 in the crystal structure are indicated . Below is the sequence alignment of the R1 domains from MTA1 , 2 and 3 . Residues coloured red are identical and residues coloured orange are conserved . ( b ) Cartoon representation of the MTA1-R1:RBBP4 complex with MTA1 in purple and RBBP4 in grey . ( c ) The structure is rotated by 90° and the surface of RBBP4 is shown . The dotted white line indicates the part of MTA1 that is disordered . The 5G loop in RBBP4 is shown as sticks with grey coloured carbon atoms . MTA1-R1 is rationalised into three interaction regions A , B and C and these are shown in more detail in ( d ) , ( e ) and ( g ) . MTA1 is shown as a cartoon and RBBP4 as a grey surface . ( f ) Electrostatic surface view of RBBP4 with MTA1 shown in yellow . See Video 1 for a 3D view of the MTA1-R1:RBBP4 complex and Table 1 for the crystallographic data collection and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 00510 . 7554/eLife . 13941 . 006Video 1 . The crystal structure of MTA1-R1 bound to RBBP4 MTA1 ( residues 464-546 ) interacts in three adjacent grooves on the side of RBBP4 . RBBP4 is shown as surface ( grey ) and MTA1 as cartoon ( purple ) . This video relates to Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 00610 . 7554/eLife . 13941 . 007Table 1 . Crystallographic data collection and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 007Data Collection MTA1-R1:RBBP4 complex Beamline Diamond I24 Space group P1 21 1 Wavelength , Å 0 . 96861 Cell dimensions a , b , c , Å 81 . 29 , 150 . 07 , 95 . 59 α , β , γ , ° 90 , 94 . 54 , 90 Resolution range , Å* 95 . 29–3 . 2 ( 3 . 34–3 . 2 ) Rmerge* 0 . 165 ( 0 . 805 ) Rmeas* 0 . 176 ( 0 . 872 ) CC1/2* 0 . 975 ( 0 . 552 ) Mean I/σI* 5 . 3 ( 2 . 0 ) Completeness , %* 97 . 7 ( 98 . 0 ) Multiplicity* 4 . 2 ( 4 . 2 ) Refinement Resolution range , Å 95 . 29–3 . 2 No . of reflections 35 , 054 Rwork/Rfree 0 . 249/0 . 291 Number of atoms Protein 14345 Water 0 B-factors , Å2 Protein 90 . 7 Rmsd from ideal values Bond lengths , Å 0 . 007 Bond angles , ° 1 . 053 Ramachandran plot Favoured , % 94 . 7 Allowed , % 4 . 6 Outliers , % 0 . 7 Missing residues MTA1 , 464–467 , 519–528 RBBP4 , 1–10 , 90–103 , 176–179 , 412–425 *The highest resolution shell is shown in parentheses . The interaction region A of MTA1 takes the form of an extended strand followed by a short helix ( H1 ) ( Figure 2d ) . The extended strand positions a number of non-polar sidechains into a groove formed between the amino-terminal helix and adjacent strand of RBBP4 that forms part of the ‘linking’ blade of the WD40 domain . This region of MTA1 is held in place by a loop within RBBP4 that is an insert in blade one ( residues 104–113 ) of the WD40 domain . This loop contains five glycine residues and is referred to as the 5G-loop . The 5G-loop adopts an ordered structure on MTA1 binding such that several non-polar residues form an additional interface with MTA1 . Helix H1 of interaction region A is immediately carboxy-terminal to the extended strand and is oriented perpendicularly to it . The helix lies across the base of the long amino-terminal helix of RBBP4 ( Figure 2d ) . The interaction involves four arginine residues forming a zipper-like series of salt bridges with three interdigitating acidic residues on the surface of RBBP4 . The central interaction region B of MTA1 consists of a short helix H2 , a structured PYxPI loop and a longer helix H3 ( Figure 2e ) . The helix H2 and PYxPI loop lie in a deep groove between the amino-terminal helix of RBBP4 and the so-called PP-loop ( which contains two adjacent prolines ) . The groove is largely non-polar in character , but the periphery is strongly acidic ( Figure 2f ) . This complements the interacting amino acids in helix H2 and the PYxPI loop of MTA1 . Helix H3 interacts with two phenylalanines on RBBP4; F30 on the surface of the amino-terminal helix and F105 with the 5G-loop . The final interaction region C is composed primarily of helix H4 which binds in a third groove on the surface of RBBP4 on the far side of the PP-loop ( Figure 2g ) . The interaction is primarily non-polar in character . Bioinformatic analysis of MTA1 indicates that there are only short regions of predicted disorder between the ELM2-SANT domain and the R1 domain . This suggested that the MTA1:HDAC1:RBBP4 proteins might form a structural core to the NuRD complex with a fixed arrangement of subunits . We have used a combination of small angle X-ray scattering ( SAXS ) , single-particle negative-stain electron microscopy ( EM ) and chemical crosslinking to determine the architecture of this core complex . For the SAXS analysis , we compared the scattering profile and the calculated envelope of the HDAC1:MTA1-A ( residues 162–354 ) complex with that of the MTA1-B ( residues 162–546 ) :HDAC1:RBBP4 complex . Multi-Angle-Light-Scattering ( MALS ) confirmed that , as expected from the crystal structure , the HDAC1:MTA1 complex is dimeric , containing two copies of each protein ( Figure 3—figure supplement 1 ) . The SAXS data for the complex are consistent with minimal aggregation ( linear I ( q ) at low q ) . A theoretical scattering profile was calculated from the crystal structure of HDAC1:MTA1 ( Millard et al . , 2013 ) and this was in good agreement with the experimental data with a χ2 function of 1 . 54 . Furthermore , the crystal structure of the HDAC1:MTA1 dimer was a convincing fit to the averaged ab initio molecular envelope generated from the SAXS curve ( Figures 3a , 3b and 3c ) . 10 . 7554/eLife . 13941 . 008Figure 3 . The core NuRD complex has an elongated structure . ( a ) Co-expression of the HDAC1:MTA1-A ( 162–354 ) complex . ( b ) SAXS data of HDAC1:MTA1-A with the theoretical scattering curve from the HDAC1:MTA1 dimer ( pdb code: 4BKX ) superimposed in red and the fit residuals are shown . ( c ) The HDAC1:MTA1 crystal structure is fitted into the ab initio molecular envelope derived from the SAXS curve . MTA1 is coloured purple and HDAC1 is shown in grey . ( d ) Co-expression of the MTA1-B ( 162–546 ) :HDAC1:RBBP4 complex . ( e ) SAXS data of MTA1-B:HDAC1:RBBP4 with the theoretical scattering curve from the MTA1:HDAC1:RBBP4 model superimposed in red and the fit residuals are shown . ( f ) The model of MTA1:HDAC1:RBBP4 is fitted into the ab initio molecular envelope derived from the SAXS curve . MTA1 is shown in purple , HDAC1 in grey and RBBP4 in green . The MTA1-R1:RBBP4 crystal structure is orientated based on crosslinks identified on surface of HDAC1 and RBBP4 and these are coloured red . ( g ) Isotopic crosslinking of MTA1-B:HDAC1:RBBP4 . Black dotted lines indicate crosslinks that could be mapped onto the model of MTA1:HDAC1:RBBP4 and those that could not be mapped are shown in grey . The HDAC1 carboxy-terminus is predicted highly disordered and the large number of crosslinks seen to this region suggest it is highly mobile . These crosslinks are not shown for simplicity . See Figure 3—figure supplement 1 and Figure 3—figure supplement 2 for information about purification and SAXS measurements , and Figure 3—source data 1 for crosslinks between MTA1 , HDAC1 and RBBP4 . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 00810 . 7554/eLife . 13941 . 009Figure 3—source data 1 . Crosslinks within the MTA1-B ( 162–546 ) :HDAC1:RBBP4 complex . Intermolecular crosslinks are listed with the respective proteins and amino acids identified . Crosslinks with an xQuest score greater than 14 . 5 are included . Crosslinks that fit the model are numbered . Crosslinks that do not fit the model are indicated "x" and we presume result from low levels of aggregation between complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 00910 . 7554/eLife . 13941 . 010Figure 3—figure supplement 1 . Purification and SAXS analysis of the HDAC1:MTA1-A ( 162–354 ) complex . ( a ) SDS-PAGE gel showing the purification of HDAC1:MTA1-A by gel filtration . Peak elution of the complex is in lane 8 . ( b ) The light scattering profile ( MALS ) and calculated molecular weight of HDAC1:MTA1-A . ( c ) SAXS data of HDAC1:MTA1-A with a line of best fit to the scattering superimposed in black . ( d ) P ( r ) distribution for HDAC1:MTA1-A used to determine Dmax . ( e ) Guinier plot calculated from the scattering data for HDAC1:MTA1-A . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 01010 . 7554/eLife . 13941 . 011Figure 3—figure supplement 2 . Purification and SAXS analysis of the MTA1-B ( 162–546 ) :HDAC1:RBBP4 complex . ( a ) SDS-PAGE gel showing the purification of MTA1-B:HDAC1:RBBP4 by gel filtration . Peak elution of the complex is in lane 7 . ( b ) The light scattering profile ( MALS ) and calculated molecular weight of MTA1-B:HDAC1:RBBP4 . ( c ) SAXS data of MTA1-B:HDAC1:RBBP4 with a line of best fit to the scattering superimposed in black . ( d ) P ( r ) distribution for MTA1-B:HDAC1:RBBP4 used to determine Dmax . ( e ) Guinier plot calculated from the scattering data for MTA1-B:HDAC1:RBBP4 . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 011 MALS analysis of the MTA1:HDAC1:RBBP4 complex showed that this ensemble is also dimeric , with the experimental mass of 300 kDa closely matching the theoretical mass ( 294 kDa ) ( Figure 3—figure supplement 2 ) . The ab initio molecular envelope generated from the SAXS curve in the lowest symmetry mode ( P1 ) suggests that the core NuRD complex has an elongated architecture ( Figures 3d , e and f ) . Although no symmetry was imposed , the envelope has an approximate two-fold symmetry consistent with a dimeric complex . We fitted the HDAC1:MTA1 dimer in the centre of the envelope such that the two-fold axis of symmetry matched the approximate symmetry of the envelope ( Figure 3f ) . This left two 'lobes' to the envelope that were of sufficient volume to contain the structure of the MTA1-R1:RBBP4 complex and additional volume for the zinc-finger domain that forms a link between the two X-ray structures . Our crosslinking data indicated that one face of HDAC1 could readily form crosslinks with one face of RBBP4 , suggesting a particular relative orientation of the two proteins in the core complex . We orientated the MTA1-R1:RBBP4 structure within the two 'lobes' so as to satisfy the largest number of crosslinks ( Figure 3g and Figure 3—source data 1 ) . The crosslinks that could not be satisfied are likely to derive from higher order oligomers of the core complex . A theoretical scattering profile was calculated from the docked model of the MTA1:HDAC1:RBBP4 complex . This showed reasonable agreement with the experimental data with a χ2 function of 2 . 16 . To further investigate the architecture of the MTA1:HDAC1:RBBP4 ensemble we used negative stain EM to determine a low resolution structure of the core NuRD complex , crosslinked with glutaraldehyde to maintain integrity . A 60 Å Gaussian low-pass-filtered , initial model was generated based on the model derived from the SAXS , crosslinking and crystal structure data . This initial model was refined against 17 , 841 semi-automatically picked particles ( Figure 4a ) . The resulting EM envelope showed that RBBP4 and HDAC1 form a closer association than predicted in the SAXS model . The HDAC1:MTA1 and MTA1-R1:RBBP4 structures fitted well within this envelope . In their new position the RBBP4 domains are tipped by 30° such that they are approximately parallel to each other and propagate the symmetry of the HDAC1:MTA1 dimer . This position fits well with the cross-linking data . A 30 Å filtered model was then generated , based on the new position of RBBP4 , and refined further to produce a 19 Å EM structure using the FSC criterion with a cutoff of 0 . 143 ( Figure 4b , c ) . 10 . 7554/eLife . 13941 . 012Figure 4 . Visualisation of the core NuRD complex by negative-stain electron microscopy . ( a ) 60 Å filtered initial model of the MTA1-B ( 162–546 ) :HDAC1:RBBP4 complex generated from the SAXS and crosslinking data . ( b ) The resolution of the final EM model was estimated to be 19 Å using the Fourier shell correlation criterion with a cutoff of 0 . 143 ( 26 Å using a cutoff of 0 . 5 ) . ( c ) Final EM model and fitting after refinement against 17 , 841 particles . ( d ) Comparison of different views of the MTA1-B:HDAC1:RBBP4 complex , generated after EM model fitting , with the reference-free 2D class averages . MTA1 is shown in blue , HDAC1 in grey and RBBP4 in green . See Video 2 for a 3D view of the MTA1-B ( 162–546 ) :HDAC1:RBBP4 complex and Figure 4—figure supplement 1 for class averaged particles and re-projections of the EM model . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 01210 . 7554/eLife . 13941 . 013Figure 4—figure supplement 1 . Further analysis of the EM model . Subset of ( a ) class averaged particles and ( b ) matching re-projections of the MTA1-B:HDAC1:RBBP4 complex . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 013 The resulting EM envelope , along with the fitted crystal structures , clearly shows the HDAC1:MTA1 central dimer and a detailed outline of the RBBP4 lobes ( Figure 4c and Video 2 ) . The model suggests a number of contacts between RBBP4 and HDAC1 that determine the relative orientation . The active sites of both HDACs are on the convex side of the complex and are exposed to enable substrate binding . Extra density surrounding the dimerization domain of MTA1 may indicate the position of the MTA1 zinc finger domains and/or the tails of HDAC1 and RBBP4 . The reference-free 2D class averages fit well with the model of the MTA1:HDAC1:RBBP4 complex and with the re-projections of the EM class averages ( Figure 4d and Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 13941 . 014Video 2 . EM structure of the MTA1-B:HDAC1:RBBP4 complex . Low resolution EM structure ( 19 Å ) of the core NuRD complex is shown contoured at 1 . Structures of the dimeric HDAC1:MTA1 ( pdb code: 4BKX ) and MTA1:RBBP4 ( pdb code: 5FXY ) are fitted . HDAC1 is shown in grey , RBBP4 in green and MTA1 in blue . This video relates to Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 014 Since we have established that two copies of the RBBP4 protein are recruited to MTA1 , we compared the amino-terminal recruitment domain , R1 , with the previously identified carboxy-terminal domain , R2 . Sequence alignment show that a stretch of c . 35 residues is highly similar between the two recruitment regions ( Figure 5a ) . These include the amino terminal strand , helices H1 & H2 and the PYxPI loop . A previous structure of the R2 domain in complex with RBBP4 included helix H2 and the PYxPI loop ( 22 residues ) ( Alqarni et al . , 2014 ) . Comparison of this structure with that of the equivalent region of the R1 domain ( region B in the description above ) shows a highly similar mode of binding to RBBP4 ( Figure 5b ) . Indeed , the path of the mainchain from P492 to I503 is essentially identical . The sidechains A495 , A496 and R497 in helix H2 and P499 , Y500 , P502 and I503 in the PYxPI loop mediate key interactions with RBBP4 and are identical in the R2 domain . 10 . 7554/eLife . 13941 . 015Figure 5 . Comparison of the MTA1-R1:RBBP4 structure with other RBBP4 complexes . ( a ) The domain structure of MTA1 highlighting the R1 and R2 domains . Below is a sequence alignment of the R1 domains of MTA1/2/3 , R2 domains of MTA1/2 and histone H4 . Boxes highlight the residues that have been crystallised in complex with RBBP4 . Residues coloured red are identical and residues coloured orange are conserved . ( b ) Overlay of crystal structures of MTA1-R1 domain with a peptide corresponding to the MTA1-R2 domain ( 671–690 ) in complex with RBBP4 . ( c ) Overlay of crystal structures of the MTA1-R1 domain with a histone H4 peptide ( 27–41 ) in complex with RBBP4 . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 015 The sequence alignment also indicates that the 15 residues preceding helix 2 in the R2 domain have significant sequence similarity to the R1 domain suggesting that the interface between RBBP4 and the R2 domain is likely to be rather more extensive than was observed in the previously reported crystal structure ( Figure 5a ) . We would expect these residues to adopt a structure similar to that seen in interaction region A of the complex between RBBP4 and the R1 domain . An important part of this interface is the 5G-loop that wraps over the top of the amino-terminal strand in the R1 domain . Interestingly this loop adopts a very different conformation in the structure of the MTA1-R2:RBBP4 complex . This alternative orientation may be the consequence of crystal packing interactions . The surface of RBBP4 that mediates interactions with helix H2 of the R1 and R2 domains of MTA1 has also been shown to mediate interactions with the amino-terminal region of histone H4 . Sequence alignment of the MTA1-R1 and -R2 domains with histone H4 shows that there is considerable sequence similarity between these proteins . The structure of histone H4 ( residues 27–41 ) in complex with RBBP4 shows a similar mode of binding to the structure of the bound MTA1-R1 and -R2 domains ( Figure 5c ) ( Murzina et al . , 2008 ) . All three proteins bind in the same groove and make similar sidechain contacts with an Alanine-Arginine dipeptide making identical contacts in all three structures ( Figures 5b and 5c ) . Again , sequence comparison suggests that the interface between histone H4 and RBBP4 may be more extensive than seen in the previously reported structure , since the sequence similarity extends at the amino-terminus , into the region corresponding to helix H1 of MTA1-R1 . Interestingly , a fragment of the Polycomb protein Su ( z ) 12 has also been crystallised bound to the central groove of RBBP4 . There is little structural or sequence similarity to either the histone H4 , MTA1-R1 or MTA1-R2 complexes . The previous observation that RBBP4 is able to bind to both histones H3 and H4 suggests that RBBP4 may serve as a common chromatin recruitment module in several chromatin modifying complexes . However , given that the extensive interface between RBBP4 and the R1 domain of MTA1 ( and by inference the R2 domain ) occludes the binding site for histone H4 , we would predict that RBBP4 is no longer able to bind histone H4 in the NuRD complex . However , the histone H3 interaction on RBBP4 is exposed in both the MTA1:RBBP4 binary complex and in the dimeric MTA1:HDAC1:RBBP4 complex . This implies that RBBP4 is likely to play a role in recruiting the NuRD complex to histone H3 tails , but not to histone H4 . To test this , we compared RBBP4 and the MTA1-R1 ( residues 464–546 ) :RBBP4 complex in their ability to bind to a peptide array of histone tails , bearing various post-translation modifications . As predicted , we observed that both RBBP4 and the MTA1-R1:RBBP4 complex showed an identical pattern of binding to histone H3 peptides . However , there were significant differences between the binding profiles to histone H4 modified peptides ( Figure 6a ) . The free RBBP4 bound multiple histone H4 tails . In contrast , the MTA1-R1:RBBP4 complex showed only binding to two histone H4 peptides and this was not reproduced in the duplicate arrays ( Figure 6—figure supplement 1 and Figure 6—source data 1 ) . Taken together this suggests that when bound to MTA1 , RBBP4 binds to histone H3 and not histone H4 . 10 . 7554/eLife . 13941 . 016Figure 6 . Recruitment of the core NuRD complex to chromatin could be mediated by RBBP4 binding to histone H3 . ( a ) RBBP4 binds to both histone H3 and histone H4 tail peptides on the MODified histone peptide array whereas the complex of MTA1-R1:RBBP4 is unable to bind to histone H4 tail peptides . MODified histone peptides J1 to P24 are shown . The five negative control peptides ( non-histone sequences ) are boxed in the right-hand corner . ( b ) Schematic model for the recruitment of the core NuRD complex to chromatin . Two HDAC1/2 molcules ( grey ) and four RBBP4/7 molcules ( green ) are tethered by the ELM2/SANT and R1/R2 domains of MTA1 respectively . MTA1 is coloured according to the schematic representation below . The RBBP4/7 proteins recruit the complex to chromatin through histone H3 tails ( orange ) on the same or adjacent nucleosomes ( purple ) . The other histone tails would be available for deacetylation by HDAC1/2 . The BAH domain of MTA1 is omitted from the model for clarity . See Figure 6—figure supplement 1 for the MODified and duplicate histone peptide arrays and Figure 6—source data 1 for a key to the array . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 01610 . 7554/eLife . 13941 . 017Figure 6—source data 1 . Key for the MODified histone peptide array – Rows J1 to P24 . The key includes the identity and post-translational modification state of the histone peptides in the arrays presented in Figure 6a and Figure 6—figure supplement 1 . The strength of observed binding is approximated by one or two asterisks . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 01710 . 7554/eLife . 13941 . 018Figure 6—figure supplement 1 . MODified histone peptide array . Expanded view of the MODified histone peptide array and duplicate array for both ( a ) RBBP4 alone and ( b ) the MTA1-R1:RBBP4 complex . DOI: http://dx . doi . org/10 . 7554/eLife . 13941 . 018 Here we have investigated the architecture of the core NuRD complex and derived the first model of the arrangement of three major components MTA1/2 , HDAC1/2 and RBBP4/7 ( Figure 6b ) . This core complex forms an elongated zig-zag structure as visualised by SAXS and EM , centred around the HDAC1:MTA1 dimerization domain , with RBBP4 ‘lobes’ symmetrically bound to the dimer . We have established that a central section of MTA1 directly recruits RBBP4/7 and brings this subunit into close proximity to HDAC1/2 . It is possible that the relative arrangement of HDAC1 , MTA1 and RBBP4 may be somewhat different in the holo-NuRD complex . However , the relatively close interaction between HDAC1 and RBBP4 , seen in the EM structure , suggests to us that this interface is likely to be present in the full complex . We have shown that each MTA1 protein recruits two RBBP4/7 molcules to the NuRD complex through the R1 and R2 interaction domains . Since MTA1 is itself a dimer in the complex ( Millard et al . , 2013 ) this means that the NuRD complex contains four copies of the RBBP4/7 proteins , which we have confirmed experimentally . Recognisable R1 and R2 interaction domains are present in MTA1 and 2 proteins from fly to man suggesting that the stoichiometry of the NuRD complex is evolutionarily conserved . Our findings fit partially with a previous study of the NuRD complex using label-free mass spectrometry . In this study they also report a 1:2 MTA1:RBBP4/7 stoichiometry , but in contrast to our previous crystal structure of the MTA1:HDAC1 complex , they find a 1:3 HDAC:MTA stoichiometry ( Smits et al . , 2012 ) . Alqarni et al . used isothermal titration calorimetry ( ITC ) to compare the affinity of interaction of the helical region of the R2 domain from MTA1 ( 656–686 ) with that of histone H4 , which they observed bound in the same groove of RBBP4 . The similarity in affinity led them to suggest that there may be competition between MTA1 and histone H4 for binding to RBBP4 ( Alqarni et al . , 2014 ) . Our findings suggest that there is likely to be a more extensive interface between MTA1-R2 and RBBP4 with an additional helix and beta-strand contributing to the interaction . This is in fact supported by cross-linking data ( Kloet et al . , 2015 ) . Given this extended interface we suggest that histone H4 is unlikely to be able to compete with either MTA1-R1 or R2 for binding to RBBP4 . Indeed , the MTA1-R1:RBBP4 complex appeared highly stable since the resin-bound complex could be washed with 2 M NaCl and 5% triton-X-100 without significant dissociation ( data not shown ) . Taken together our data suggest that once bound to MTA1 , the two RBBP4 molcules would remain permanently assembled as a complex . Since it appears that the histone H4 binding site on RBBP4 is blocked when associated with MTA1 ( either R1 or R2 ) this raises the question as to the role of RBBP4 in the NuRD complex . Importantly , the crystal structure of RBBP4 bound to a histone H3 tail peptide shows that the interaction surface lies on the ‘top’ of the WD40 domain ( Schmitges et al . , 2011 ) . This site is not blocked by binding to MTA1-R1 . Thus RBBP4 would theoretically remain able to bind to histone H3 tails whilst assembled into the NuRD complex . This fits with our experimental data which confirm that the core NuRD complex can bind histone H3 tails but not H4 . Such a model of recruitment to histone H3 is consistent with the architecture of the MTA1:HDAC1:RBBP4 complex since the orientation of RBBP4 relative to the HDAC1:MTA1 dimer means that the histone H3 binding surface of RBBP4 is exposed , albeit close to the interface between HDAC1 and RBBP4 , and therefore able to bind the tail of histone H3 . Furthermore , the histone recruitment site is sufficiently close to the catalytic site of HDAC1 to allow other histone tails to access the active site . Interestingly , both RBBP4 and the RBBP4:MTA1 complex show enhanced binding to modified tails of histone H3 lysine 27 ( H3K27 ) ( Figure 6a ) . In particular , H3K27 acetylation and H3K27 methylation appear to promote binding . This would correlate with previous studies that have observed association of the NuRD complex with a specific subset of gene promoters enriched with H3K27me3 ( Reynolds et al . , 2012 ) . In addition , MBD3 , a component of the NuRD complex has been reported to localise at H3K27ac and H3K27me3 enriched sites ( Shimbo et al . , 2013 ) . RBBP4 may therefore have an important role in directing the NuRD complex to chromatin bearing these modifications . A number of studies have suggested that RBBP4 and RBBP7 have different functions . Indeed , it has been suggested that they are associated with different complexes . Our structural data suggest that the residues important for mediating interaction with MTA1 are 100% identical and therefore we would expect both RBBP4 and RBBP7 to be associated with the NuRD complex depending upon relative abundance . Interestingly , in MTA3 the recruitment domain R2 is not present in any species . This suggests that MTA3 only recruits a single RBBP4/7 molcule . Furthermore , a splice variant of MTA1 ( MTA1S ) does not contain either R1 or R2 domains ( Yaguchi et al . , 2005 ) . It is intriguing to speculate that these variant NuRD complexes will exhibit different specificities due to different modes of chromatin recruitment . Presumably the presence of four RBBP4 molcules in the NuRD complex increases the avidity for chromatin . Overall this study provides significant novel insights into the assembly and chromatin recruitment of the core NuRD complex . It establishes the architectural relationship between the histone deacetylase and the RBBP4 chromatin-binding module . Further studies will be required to explore the relative organisation between the HDAC and the CHD3/4 chromatin remodeller . MTA1-A , -B , -C , -D and -E ( residues 162–335 , 162–546 , 162–715 , 390–715 and 464–546 ) were cloned into pcDNA3 vectors containing an amino terminal His10-Flag3 purification tag followed by a TEV protease cleavage site . RBBP4 ( residues 1–425 ) and HDAC1 ( residues 1–482 ) were cloned without affinity tags into the same vectors . Protein constructs were co-transfected into HEK293F suspension-grown cells ( Invitrogen ) with polyethylenimine ( PEI; Sigma ) and the cells were harvested after 48 hr as previously described ( Nettleship et al . , 2014; Portolano et al . , 2014; Watson et al . , 2012 ) . Cells were lysed in 50 mM Tris-HCl ( pH 7 . 5 ) , 100 mM potassium acetate , 10% v/v glycerol , 0 . 3% v/v Triton X-100 and Roche Complete Protease Inhibitor ( buffer A ) . The clarified lysate was applied to FLAG resin ( Sigma ) for 2 hr at 4°C . The resin was washed three times with 50 mM Tris-HCl ( pH 7 . 5 ) , 100 mM potassium acetate and 5% v/v glycerol ( buffer B ) , three times with buffer B containing 200 mM potassium acetate and a further three times with buffer B . The protein was treated with RNase A in buffer B for 1 hr at 4°C before being eluted with TEV overnight into buffer B containing 50 mM potassium acetate . The protein complexes were purified further by gel filtration using a Superdex S200 column into 25 mM Tris-HCl ( pH 7 . 5 ) , 50 mM potassium acetate and 0 . 5 mM TCEP ( GE Healthcare , UK ) . The purified complex was concentrated to 10 mg/ml . The intensity of protein bands was determined with ImageJ and the ratio of HDAC1:RBBP4 was plotted using GraphPad Prism . Purified MTA1:RBBP4 , HDAC1:MTA1 and MTA1:HDAC1:RBBP4 complexes that had been gel filtrated were concentrated to 1 mg/ml and reapplied to the Superdex S200 column ( GE Healthcare , UK ) . The mass of each protein complex was calculated immediately on elution with an Optilab T-rEX differential Refractive Index detector coupled to a DAWN HELEOS MALS detector ( Wyatt Technology ) . Diffracting crystals ( 25 μm ) of MTA1-R1:RBBP4 were grown at 10°C by sitting-drop vapour diffusion against 0 . 21 M ammonium citrate and 19% PEG 3350 . Crystals were frozen in liquid nitrogen using 30% glycerol as a cryoprotectant , initial screening was performed at the Swiss Light Source PX1 ( Switzerland ) , and data were collected at the Diamond microfocus beamline I24 ( UK ) . Data from two crystals were processed with iMOSFLM ( Leslie , 2005 ) and merged using AIMLESS ( Evans and Murshudov , 2013 ) to produce a complete dataset . A unique phase solution was found with RBBP4 ( pdb code: 4PC0 ) ( Alqarni et al . , 2014 ) as a search model using PHASER ( McCoy et al . , 2007 ) and MTA1 was manually built using multiple rounds of refinement using REFMAC ( Murshudov et al . , 2011 ) and COOT ( Emsley et al . , 2010 ) . The final model contains four copies of residues 9-411 of RBBP4 ( chains A , C , E , G ) and residues 468-546 of MTA1 ( chains B , D , F , H ) . For the SAXS experiments HDAC1:MTA1-A and MTA1-B:HDAC1:RBBP4 were concentrated ( 1 to 5 mg/ml ) in buffer E immediately before analysis . Data were collected on the samples using a Pilatus2M detector ( Dectris , CH ) at a sample-detector distance of 4014 mm at 12 . 4 keV at the Diamond B21 beamline ( UK ) . The scattering data was collected to provide a q range of 0 . 0038-0 . 42 Å-1 , where q is the magnitude of the scattering vector , using a flux of c . 1011 photons per second . The samples and matching buffer solutions were exposed to X-rays 30 times in 5s bursts at 25°C . Examination of the data suggested there was no evidence of radiation damage . The lowest concentration datasets ( 1 mg/ml ) were used for analysis using the ATSAS suite of software ( Petoukhov et al . , 2012 ) and ScÅtter for basic analysis of the SAXS datasets ( BioIsis . net ) . The radius-of-gyration ( Rg ) was calculated from Guinier plots and these values were in good agreement with the Rg calculated from the second moment of the P ( r ) function . The maximum dimension of the scattering molecular species ( Dmax ) was estimated from the P ( r ) curve and again these values were favourable comparable to those calculated using the program GNOM ( Svergun , 1992 ) . Theoretical scattering profiles were calculated from the crystal structure of HDAC1:MTA1 ( Millard et al . , 2013 ) and combined crystal structures of HDAC1:MTA1 and MTA1-R1:RBBP4 and compared to the experimental data using FoXS ( Schneidman-Duhovny et al . , 2010 ) . Thirteen ab initio dummy atom models were generated using the programme DAMMIF ( Franke and Svergun , 2009 ) and were superposed , merged and filtered using the programme DAMAVER ( Volkov and Svergun , 2003 ) . Gel filtrated MTA1-B:HDAC1:RBBP4 complex was buffer exchanged into 50 mM HEPES ( pH 7 . 5 ) , 50 mM potassium acetate , diluted to 0 . 5 mg/ml , and the isotopically labelled crosslinker CBDPSS-H8/D8 ( CyanurBiotinDimercaptoPropionylSulfoSuccinimide , Creative Molecules Inc . ) ( Petrotchenko et al . , 2011 ) was added to a final concentration of 0 . 7 mM . After incubation at 30°C for 30 min the reaction was stopped by addition of 0 . 5 M NH4HCO3 to a final concentration of 40 mM . The protein was analysed by SDS-PAGE and protein bands corresponding to the crosslinked-dimeric complex were excised from the gel . The sample was reduced with 10 mM DTT , alkylated with 100 mM iodoacetamide and digested with trypsin . LC-MS/MS was carried out as described previously ( Alam et al . , 2015 ) . Raw data files were converted to . mzXML format using the ProteoWizard msconvert toolkit ( Chambers et al . , 2012 ) and cross-linked peptides were identified using the xQuest/xProphet ( Leitner et al . , 2014 ) pipeline with appropriate parameters defined for the CBDPSS isotopically labelled crosslinker . For crosslinks to be considered valid the xQuest Id-Score was required to exceed 14 . 5 , and on inspection of the MS/MS spectra a minimum of either four fragment ions per peptide or three consecutive fragment ions per peptide were required to match . Gel filtrated MTA1-B ( 162-546 ) :HDAC1:RBBP4 was prepared for EM analysis by using GraFIX ( Kastner et al . , 2007 ) . A 5-25% sucrose gradient with 0-1% glutaraldehyde in 20 mM HEPES/KOH ( pH 7 . 5 ) , 40 mM NaCl was prepared using a Gradient Master IP ( BIOCOMP ) with a SW60 rotor . The protein complex was added directly on top of the gradient and centrifuged for 15 hr at 166400g in a swing-out rotor . The gradient was manually fractionated into 175 μl fractions and fractions containing monodisperse complex were pooled . The sucrose was removed and cross linking stopped by buffer exchange in a centrifugal filter concentrator ( Millipore , Amicon , Ultra-0 . 5 , 10 kDa cut-off ) using 20 mM Tris/HCl ( pH 7 . 5 ) , 40 mM NaCl . Negative stain grids were prepared by glow-discharging carbon coated copper 300 mesh grids ( Agar Scientific ) at 10 mA for 30 s . The protein complex was diluted to 0 . 1 mg/ml and 5 μl was applied to the grid . The protein was allowed to absorb for 1 min and blotted to remove excess liquid . The grid was then stained with 2% w/v uranyl acetate . Grids were imaged using a Jeol 2010F TEM operating at 200 kV fitted with a Gatan UltraScan™ 4000 CCD camera with a 15 μm pixel size . 308 micrographs were collected at various defocus values between 0 . 5 µm and 5 µm , each micrograph contained 20 – 90 particles . 17 , 841 particles were semi-automatically picked using the EMAN2 e2boxer . py swarm function and extracted using a 224 × 224 pixel box size . Reference-free 2D class averages were generated , with c . 100 – 200 particles per class average using e2refine2d . py ( Tang et al . , 2007 ) . A 60 Å filtered model of the complex was generated based on the crystal structures of HDAC1:MTA1 and MTA1:RBBP4 fitted to the SAXS and crosslinking data , refined against the 17 , 841 particle dataset , and a model was generated with C1 symmetry using e2refine_easy . py . Resolution of refined model was 19 Å using a FSC criterion cutoff of 0 . 143 . Gel filtrated Flag-RBBP4 and Flag-MTA1-R1-:RBBP4 proteins were diluted to 200 nM in 5 ml TTBS ( 10 mM Tris-HCl ( pH 7 . 4 ) , 0 . 05% Tween 20 and 150 mM NaCl ) and applied to separate MODified histone peptide arrays ( Active Motif ) containing 384 unique histone modification combinations . The arrays were washed and protein detected using an anti-Flag antibody according to the manufacture’s protocol . Data from each array was analysed using the commercially available software ( Active Motif ) .	The correct regulation of our genes is essential for life . Genes are actively switched on or off through the action of assemblies of proteins that act together as molecular machines . Some of these machines alter the way that DNA is packaged inside cells . Packaged DNA – called chromatin – consists of DNA wrapped around proteins called histones , which together form structures called nucleosomes . Changing how tightly nucleosomes are packed together can alter whether a gene is active: tighter packing makes it harder to access the genes in that stretch of DNA and therefore inactivates them . In humans , an assembly of proteins called the NuRD complex makes chromatin more compact by removing acetyl groups from nucleosomes . This complex is important for early development and for the stability and repair of our genes . Three proteins make up its core: HDAC1 , which removes the acetyl group from the nucleosome; MTA1 , which acts as a scaffold to hold the complex together; and RBBP4 , which enables the complex to interact with nucleosomes . Understanding how protein complexes are assembled tells us a lot about how they work . Millard et al . have therefore used a number of structural techniques to investigate the three-dimensional architecture of the three core proteins in the NuRD complex . The resulting structures have revealed how the HDAC1 , MTA1 and RBBP4 proteins interact to influence how the complex is recruited to nucleosomes . The next step will be to assemble all the remaining proteins of the NuRD complex to understand its architecture as a whole .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2016	The structure of the core NuRD repression complex provides insights into its interaction with chromatin
Adhesion and morphogenesis of many non-muscle cells are guided by contractile actomyosin bundles called ventral stress fibers . While it is well established that stress fibers are mechanosensitive structures , physical mechanisms by which they assemble , align , and mature have remained elusive . Here we show that arcs , which serve as precursors for ventral stress fibers , undergo lateral fusion during their centripetal flow to form thick actomyosin bundles that apply tension to focal adhesions at their ends . Importantly , this myosin II-derived force inhibits vectorial actin polymerization at focal adhesions through AMPK-mediated phosphorylation of VASP , and thereby halts stress fiber elongation and ensures their proper contractility . Stress fiber maturation additionally requires ADF/cofilin-mediated disassembly of non-contractile stress fibers , whereas contractile fibers are protected from severing . Taken together , these data reveal that myosin-derived tension precisely controls both actin filament assembly and disassembly to ensure generation and proper alignment of contractile stress fibers in migrating cells . Cell migration is essential for embryonic development , wound healing , immunological processes and cancer metastasis . Cell migration is driven by assembly and disassembly of protrusive and contractile actin filament structures . The force in protrusive actin filament structures , including lamellipodium and filopodia at the leading edge of cell , is generated through actin polymerization against the plasma membrane . In contractile actin filament bundles , such as stress fibers , the force is generated by sliding of bipolar myosin II bundles along actin filaments . Notably , whereas the assembly-mechanisms of protrusive actin filament structures are relatively well understood , general principles underlying the assembly of contractile actomyosin bundles have remained elusive ( Pollard and Cooper , 2009; Bugyi and Carlier , 2010; Michelot and Drubin , 2011; Burridge and Wittchen , 2013 ) . The most prominent contractile actomyosin structures in most cultured non-muscle cells are stress fibers . Beyond cell migration , stress fibers guide adhesion , mechanotransduction , endothelial barrier integrity , myofibril assembly , and receptor clustering in T-lymphocytes ( Burridge and Wittchen , 2013; Wong et al . , 1983; Sanger et al . , 2005; Tojkander et al . , 2012; Yi et al . , 2012 ) . Due to their intrinsic properties , stress fibers have become an important model system for studying the general principles by which contractile actomyosin bundles are assembled in cells . Stress fibers can be divided into three main categories based on their protein compositions and interactions with focal adhesions ( Small et al . , 1998 ) . Dorsal ( radial ) stress fibers are connected to focal adhesions at their distal ends and rise towards the dorsal surface of the cell at their proximal region ( Hotulainen and Lappalainen , 2006 ) . They elongate through vectorial actin polymerization at focal adhesions ( i . e . coordinated polymerization of actin filaments , whose rapidly elongating barbed ends are facing the focal adhesion , is responsible for growth of dorsal stress fibers ) . These actin filament bundles do not contain myosin II , and dorsal stress fibers are thus unable to contract ( Hotulainen and Lappalainen , 2006; Cramer et al . , 1997; Tojkander et al . , 2011; Oakes et al . , 2012; Tee et al . , 2015 ) . However , dorsal stress fibers interact with contractile transverse arcs and link them to focal adhesions . Transverse arcs are curved actin bundles , which display periodic α-actinin – myosin II pattern and undergo retrograde flow towards the cell center in migrating cells . They are derived from α-actinin- and tropomyosin/myosin II- decorated actin filament populations nucleated at the lamellipodium of motile cells ( Hotulainen and Lappalainen , 2006; Tojkander et al . , 2011; Burnette et al . , 2011; 2014 ) . In fibroblasts and melanoma cells , filopodial actin bundles can be recycled for formation of transverse arc –like contractile actomyosin bundles ( Nemethova et al . , 2008; Anderson et al . , 2008 ) . Ventral stress fibers are defined as contractile actomyosin bundles , which are anchored to focal adhesions at their both ends . Despite their nomenclature , the central regions of ventral stress fibers can bend towards the dorsal surface of the lamellum ( Hotulainen and Lappalainen , 2006; Schulze et al . , 2014 ) . Migrating cells display thick ventral stress fibers that are typically oriented perpendicularly to the direction of migration , and thinner ventral stress fibers that are often located at the cell rear or below the nucleus . At least the thick ventral stress fibers , which constitute the major force-generating actomyosin bundles in migrating cells , are derived from the pre-existing network of dorsal stress fibers and transverse arcs . However , the underlying mechanism has remained poorly understood ( Burridge et al . , 2013; Hotulainen and Lappalainen , 2006 ) . Stress fibers and focal adhesions are mechanosensitive structures . Stress fibers are typically present only in cells grown on rigid substrata and they disassemble upon cell detachment from the matrix ( Mochitate et al . , 1991; Discher et al . , 2005 ) . Furthermore , after applying fluid shear stress , stress fibers align along the orientation of flow direction in endothelial cells ( Sato and Ohashi , 2005 ) . Also focal adhesions develop only on rigid surfaces , and applying external tensile force promotes their enlargement ( Chrzanowska-Wodnicka and Burridge , 1996; Pelham et al . , 1999; Riveline et al . , 2001 ) . Focal adhesions contain several mechano-sensitive proteins , including talin , filamin and p130Cas , whose activities and interactions with other focal adhesion components can be modulated by forces of ~∼10–50 pN range ( Sawada et al . , 2006; del Rio et al . , 2009; Ehrlicher et al . , 2011 ) . Furthermore , the protein compositions of focal adhesions are regulated by tension supplied by myosin II activity and external forces applied to the cell ( Zaidel-Bar et al . , 2007; Kuo et al . , 2011; Schiller et al . , 2011 ) . Importantly , despite wealth of information concerning mechanosensitive focal adhesion proteins , possible effects of tensile forces on actin filament assembly at focal adhesions have remained elusive . Furthermore , the mechanisms by which tension contributes to the alignment of stress fibers and actin dynamics within these actomyosin bundles have not been reported . Here we reveal that formation of mature contractile actin bundles from their precursors is a mechanosensitive process . We show that arc fusion during centripetal flow is accompanied by increased contractility that inhibits vectorial actin polymerization at focal adhesions through AMPK-mediated phosphorylation of VASP , thus insuring formation of ventral stress fibers . Conversely , activation of AMPK allows generation of contractile ventral stress fibers in cells growing on compliant matrix , where their formation is normally prevented . Furthermore , we provide evidence of mechanosensitive actin filament disassembly by ADF/cofilins during stress fiber assembly . These data provide support to a new mechanobiological model explaining the principles of assembly and alignment of ventral stress fibers in migrating cells . Transverse arcs are generated from actin filament arrays at the lamellipodium –— lamella interface ( Tojkander et al . , 2011; Shemesh et al . , 2009; Burnette et al . , 2011 ) . During their assembly , thin arcs associate with elongating dorsal stress fibers to form a spider-net -like structure ( Figure 1—figure supplement 1A and 1B; Tojkander et al . , 2011 ) . This network , consisting of several non-contractile dorsal stress fibers and multiple thin arcs , flows towards the cell center and maturates to thick , contractile ventral stress fibers through a mechanism that has remained poorly understood ( Hotulainen and Lappalainen , 2006 ) . Interestingly , proper stress fiber network does not form in cells grown on compliant matrix ( Discher et al . , 2005; Prager-Khoutorsky et al . , 2011 ) , but whether the assembly of all above-mentioned stress fiber categories , or only a specific one , is mechanosensitive has not been reported . By plating U2OS cells on soft ( 0 . 5 kPa ) and stiff ( 64 kPa ) substrata , we revealed that dorsal stress fibers and arcs are also present in cells grown on compliant matrix . In contrast , ventral stress fiber assembly is compromised under these conditions ( Figure 1A ) . While 89% of cells plated on 64 kPa matrix contained ventral stress fibers , only 10% of cells plated on 0 . 5 kPa matrix exhibited ventral stress fibers as defined by presence of straight , contractile actin bundles connected to focal adhesions at each end . Thus , generation of ventral stress fibers appears to be the mechanosensitive phase in the formation of the stress fiber network . 10 . 7554/eLife . 06126 . 003Figure 1 . Transverse arcs fuse during centripetal flow to generate a contractile ventral stress fiber . ( A ) U2OS cells have three subtypes of stress fibers: Dorsal stress fibers ( red arrowheads ) , which are attached to focal adhesion at their distal end; Transverse arcs ( yellow arrowheads ) , which are curved actomyosin bundles oriented parallel to the leading edge; Ventral stress fibers ( orange arrowheads ) , which are thick contractile bundles connected to focal adhesions at both ends . All three stress fiber categories are present in cells grown on glass or on stiff ( E = 64 kPa ) silicone matrix , whereas assembly of contractile ventral stress fibers is compromised in cells grown on soft ( 0 . 5 kPa ) matrix . ( B ) Live-imaging of U2OS cells expressing GFP-calponin-3 ( CaP3 ) revealed that transverse arcs fuse with each other during centripetal flow to form thicker actomyosin bundles . Red and yellow brackets highlight fusing arcs , and the orange dashed line indicates the thick ventral stress fiber derived from the fusing arcs . ( C ) Arc fusion often initiates at the connection points of dorsal stress fibers and transverse arcs ( indicated by red arrowheads ) . Two separate video frame series are shown in the panels . In the images , dorsal stress fibers and arcs are oriented vertically and horizontally , respectively . Stress fibers were visualized by expression of GFP-actin . Bar , 1 μm . ( D ) Live imaging of YFP-Tm4 and CFP-α-actinin-1 expressing U2OS cell reveals that homotypic coalescence of adjacent Tm4 and α-actinin foci occurs during arc fusion , thus allowing to retain the periodic pattern of transverse arcs . Yellow brackets indicate fusing arcs and yellow arrowheads highlight pairs of fusing α-actinin-1 foci . Bar , 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 00310 . 7554/eLife . 06126 . 004Figure 1—figure supplement 1 . Fusion of transverse arcs . ( A ) Dorsal stress fibers and transverse arcs have multiple connections points , thus forming a spider-net -like structure . Stress fibers were visualized by expression of GFP-actin . Red arrowheads indicate dorsal stress fibers and yellow arrowheads transverse arcs . Bar , 5 μm . ( B ) Connections between dorsal stress fibers and transverse arcs are formed early during the arc fusion process . The stress fiber network was visualized by expression of GFP-CaP3 . Red arrowhead indicates the distal end of a dorsal stress fiber , and yellow brackets highlight arc fusion . Bar , 5 μm . ( C ) Frames from a movie of representative GFP-CaP3 and CFP-zyxin expressing cell demonstrates that the spacing between individual CaP3 foci decreases as the arc flows towards the cell center and fuses with adjacent arcs . Bar , 5 μm . ( D ) Also in fixed U2OS cells expressing GFP-CaP3 , the distance between adjacent CaP3 foci correlates inversely with the distance of the arc from the cell edge . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 004 To reveal how ventral stress fibers are derived from arcs and to elucidate the mechanosensitive basis of this process , we examined the dynamics of the stress fiber network in U2OS cells , where all three stress fiber categories can be readily visualized by live-cell microscopy ( Hotulainen and Lappalainen , 2006 ) . We first followed this process by using GFP-calponin-3 ( CaP3 ) , which compared to other stress fiber components allows better visualization of thin arc precursors . Interestingly , live-imaging of GFP-CaP3 -transfected cells revealed that the thin arc precursors fused with each other to form thicker actomyosin bundles during their flow towards the cell center ( Figure 1B; Figure 1—figure supplement 1B ) . Fusion appeared to often initiate at the sites where arcs were connected to elongating dorsal stress fibers ( Figure 1C ) . Live-imaging of cells expressing CFP-α-actinin and YFP-tropomyosin-4 demonstrated that homotypic coalescence of tropomyosin-4/myosin II foci and α-actinin foci of adjacent arcs occurred during the fusion process in all observed cases ( Figure 1D ) . Thus , thin arc precursors fuse with each other during centripetal flow to generate thicker actomyosin bundles , where the periodic α-actinin — myosin II pattern is retained . Traction force microscopy was applied to examine whether arc fusion during centripetal flow is accompanied by changes in their contractility . These experiments revealed that thick ventral stress fibers exhibit stronger traction forces to focal adhesions as compared to forces applied by dorsal stress fibers ( Figure 2A and B ) , similarly to what was recently demonstrated with model-based traction force microscopy by Soine et al . ( 2015 ) . Furthermore , spacing between individual CaP-3 foci , which co-localize with α-actinin in stress fibers ( Small and Gimona , 1998 ) , decreased as the arcs flowed towards the cell center and become thicker as detected both from several fixed samples and live-cell imaging experiments ( representative examples are shown Figure 1—figure supplement 1C and D ) . This correlates well with the increased contractility of the structures ( Aratyn-Schaus et al . , 2011 ) . 10 . 7554/eLife . 06126 . 005Figure 2 . Vectorial actin polymerization at focal adhesions halts upon increased contractility and formation of ventral stress fibers . ( A ) Representative images of U2OS cells grown on 26 kPa polyacrylamide dishes with fluorescent nanobeads together with the corresponding force maps . Adhesions located at the ends of ventral stress fibers ( orange arrowheads ) apply stronger forces to their substrate compared to adhesions located at the ends of dorsal stress fibers ( red arrowheads ) . Bar , 10 um . ( B ) Quantification of traction forces at adhesions located in the ends of dorsal and ventral stress fiber adhesions . Mean +/- SEM , n = 20 cells , 4–8 adhesions per cell . ( C ) Recovery of GFP-actin signal was measured next to dorsal stress fiber adhesions ( dsf FA , red box ) and ventral stress fiber adhesions ( vsf FA , yellow box ) . Bar , 5 um . ( D ) Representative example of a fluorescence-recovery-after-photobleaching ( FRAP ) experiment performed on a GFP-actin expressing cell at a ventral stress fiber ( vsf ) region close to a focal adhesion . Yellow box indicates the photobleached region and the orange arrowhead the distal end of the ventral stress fiber . Scale , 3 um . ( E ) Quantification of the recovery speed ( μm/min ) for GFP-actin from focal adhesions located at the tips of dorsal stress fibers ( dsf ) or ventral stress fibers ( vsf ) . Means +/- SEM , n ( dorsal stress fibers ) = 11; n ( ventral stress fibers ) = 17 . ( F ) Activation of PA-GFP-actin in focal adhesions is followed by centripetal flow of photoactivated actin along the dorsal stress fiber . In contrast , PA-GFP-actin activated at a focal adhesion located at the tip of ventral stress fiber does not distribute from the adhesion to the stress fiber . Activated PA-GFP-actin is in green , focal adhesion marker mCherry-zyxin in red and the yellow dashed lines show the borders of the photoactivated region . Bar , 2 , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 00510 . 7554/eLife . 06126 . 006Figure 2—figure supplement 1 . Alignment of focal adhesions linked to the ends of transverse arcs . ( A ) Thick actomyosin bundles derived through arc fusion apply tension to focal adhesions leading to enlargement and alignment of those focal adhesions that are linked to the ends of the actomyosin bundle through dorsal stress fibers . Eventually , this results in a formation of a straight ventral stress fiber , where the focal adhesions are aligned along the direction of the actomyosin bundle . Focal adhesions and stress fibers were visualized by expression of GFP-paxillin and mCherry-actin , respectively . The two ‘terminal’ focal adhesions , whose alignments during the process are shown in the insets , are indicated by red arrowheads in the larger cell frames . Bars , 10 μm . ( B ) Fusion of transverse arcs into thicker actomyosin structures correlates with the alignment of the ‘terminal’ focal adhesions along the direction of the arc bundle . Stress fibers were visualized by expression of GFP-CaP3 . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 00610 . 7554/eLife . 06126 . 007Figure 2—figure supplement 2 . Actin dynamics in focal adhesions located at the tips of dorsal and ventral stress fibers . The dynamics of GFP-actin in focal adhesions of U2OS cells were examined by fluorescence recovery after photobleaching ( FRAP ) . Recovery curves with mean +/- SD are shown . n ( dorsal stress fiber adhesions ) = 11; n ( ventral stress fiber adhesion ) = 12 . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 00710 . 7554/eLife . 06126 . 008Figure 2—figure supplement 3 . Inhibition of myosin light chain phosphorylation results in formation of abnormally long dorsal stress fibers . ( A ) Examples of a control U2OS cell and a cell incubated for 2 h with ROCK kinase inhibitor , Y27632 . This inhibitor results in an abnormal morphology of transverse arcs or their total disapparance . Y27632 –treated cells also contain fewer dorsal stress fibers , which appear abnormally long and thin . Red arrowheads indicate dorsal stress fibers . Bar , 10 μm . ( B ) Quantification of the lengths of dorsal stress fibers in control and Y27632 -treated cells . Mean lengths ( +/- SD ) of dorsal stress fibers from 20 cells is shown . ( C ) Inhibition of myosin light chain kinase ( MLCK ) by ML-7 leads to a loss of transverse arcs and concomitant defects in formation of ventral stress fibers . Bar , 10 μm . ( D ) Dorsal stress fibers are abnormally long in ML-7 treated cells . Mean length ( +/- SD ) of 55 dorsal stress fibers measured from control and ML-7 treated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 00810 . 7554/eLife . 06126 . 009Figure 2—figure supplement 4 . Expression of dominant inactive Rif leads to abnormal elongation of dorsal stress fibers . ( A ) Video frames from a U2OS cells expressing GFP-actin alone ( right panel ) or GFP-actin and Rif-TN . In control cells , ventral stress fibers ( yellow arrowhead ) are generated through fusion of relatively short dorsal stress fibers ( red arrows ) and transverse arcs . Rif-TN expressing cells lack transverse arcs and concomitant formation of ventral stress fibers . Importantly , dorsal stress fibers in Rif-TN expressing cells are abnormally long and continue elongating through the observation period . Bar , 10 μm . ( B ) Images of phalloidin stained control and Rif-TN expressing cells . Dorsal stress fibers are indicated by red arrowheads . ( C ) Quantification of the lengths of dorsal stress fibers length from 20 control and Rif-TN expressing cells . Mean lengths of the dorsal stress fibers ( +/- SEM ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 009 Transverse arcs are typically connected to several focal adhesion-attached dorsal stress fibers along their length ( Hotulainen and Lappalainen , 2006 ) . To elucidate how increased contractility of arcs affects the associated focal adhesions , we examined possible changes in adhesion alignment during the arc maturation process . These experiments revealed that the ‘distal’ focal adhesions , linked via dorsal stress fibers to the ends of the arc , turned and aligned along the direction of arc . In contrast , focal adhesions linked to the central region of the arc did not display similar alignment during the process . Alignment of ‘distal’ focal adhesions correlated with arc fusion , and was accompanied by enlargement of adhesions ( Figure 2—figure supplement 1A and B ) . Thus , arc fusion during centripetal flow correlates with their increased contractility , consequent enlargement of distal focal adhesions and their alignment along the direction of the actomyosin bundle . Eventually , this leads to formation of a directed ventral stress fiber , containing one properly aligned large focal adhesion at its both ends . Dorsal stress fibers elongate through actin polymerization at focal adhesions . In U2OS cells , this ‘vectorial’ actin polymerization promotes elongation of the actin filament bundle with a rate of ∼0 . 25 μm/min ( Hotulainen and Lappalainen , 2006 ) . In addition , focal adhesions may contain other actin filament populations that are not directly associated with vectorial actin polymerization and consequent elongation of dorsal stress fibers . This is because several tropomyosin isoforms , which are likely to decorate distinct actin filament populations , localize to focal adhesions ( Tojkander et al . , 2011 ) and because several proteins involved in actin polymerization regulate actin dynamics at focal adhesions ( e . g . Hotulainen and Lappalainen , 2006; Skau et al . , 2015 ) . Furthermore , FRAP experiments performed at focal adhesions show rapid , uniform recovery of GFP-actin fluorescence ( Videos 1 and 2; Figure 2—figure supplement 2 ) , whereas FRAP experiments performed at dorsal stress fiber regions below focal adhesions exhibit treadmilling-like recovery that is indicative of vectorial actin polymerization ( Hotulainen and Lappalainen , 2006; Tee et al . , 2015 ) . 10 . 7554/eLife . 06126 . 010Video 1 . Fluoresence recovery after photobleaching ( FRAP ) of GFP-actin at dorsal stress fiber-associated focal adhesions ( FAs ) . GFP-actin at FAs of dorsal stress fibers were bleached with 100% laser power of 488 laser line for 1 ms . Signal of GFP-actin displayed relatively uniform recovery at the site of focal adhesion . Duration of the movie is 8 , 5 min and the display rate is 10 frames/second . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 01010 . 7554/eLife . 06126 . 011Video 2 . Fluoresence recovery after photobleaching ( FRAP ) of GFP-actin at ventral stress fiber-associated focal adhesions ( FAs ) . GFP-actin at FAs of dorsal stress fibers were bleached with 100% laser power of 488 laser line for 1 ms . Signal of GFP-actin displayed relatively uniform recovery at the site of focal adhesion . Duration of the movie is 8 , 5 min and the display rate is 10 frames/second . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 011 Because contractility promotes focal adhesion enlargement and alignment during maturation of arcs to ventral stress fibers , we examined whether this process would be accompanied by alterations in vectorial actin polymerization at focal adhesions . Fluorescence-recovery-after-photobleaching ( FRAP ) was first applied to visualize the recovery of GFP-actin signal within actin filament bundles of dorsal and ventral stress fibers . Region of interest was chosen beneath focal adhesions to exclude other focal adhesion associated actin filament populations that are not directly involved in vectorial actin polymerization and elongation of stress fibers . As previously reported , elongation of a bright actin filament bundle ( with a rate of ∼0 , 26 μm/min ) from focal adhesions located at the distal ends of dorsal stress fibers was observed ( Hotulainen and Lappalainen , 2006 ) . Importantly , when a FRAP analysis was performed on a corresponding ventral stress fiber region , only very slow elongation ( ∼0 , 02 μm/min ) of a bright actin filament bundle from the adhesion was observed . Instead , we mainly detected recovery of GFP-actin fluorescence evenly along the photobleached region ( Figure 2C-E ) . As an alternative approach , we utilized photoactivatable ( PA ) -GFP-actin to follow its incorporation into dorsal and ventral stress fibers . In both cases , significant fraction of activated PA-GFP-actin remained at/close to focal adhesions , probably corresponding to actin filament pools associated with focal adhesions ( Tojkander et al . , 2011 ) . Importantly , PA-GFP-actin displayed centripetal flow along the actin filament bundle from focal adhesions in dorsal stress fibers , while similar flow of PA-GFP-actin was not detected from focal adhesions located at the tips of ventral stress fibers ( Figure 2F ) . Therefore , in contrast to dorsal stress fibers , ventral stress fibers do not elongate through vectorial actin polymerization at focal adhesions . To elucidate whether inhibition of vectorial actin polymerization in focal adhesions at the tips of ventral stress fibers is dependent on tension applied by myosin II , we examined the morphology of the stress fiber network in cells treated with myosin light chain kinase ( MLCK ) inhibitor ML-7 . This compound induced rapid disassembly of most contractile ventral stress fibers and transverse arcs , without affecting integrity of non-contractile dorsal stress fibers ( Figure 2—figure supplement 3C ) . Importantly , dorsal stress fibers in cells treated for 2 h with ML-7 were ∼1 . 5 times longer compared to the ones in control cells ( Figure 2—figure supplement 2D ) . Similarly , disruption of contractile stress fibers by ROCK inhibitor , Y27632 , or by over-expression of dominant inactive Rif GTPase ( Rif-TN ) , which prevents assembly of contractile arcs ( Tojkander et al . , 2011 ) , led to formation of abnormally long dorsal stress fibers ( Figure 2—figure supplement 3A and B , and Figure 2—figure supplement 4 ) . Importantly , live-imaging of GFP-actin expressing cells revealed that the abnormally long dorsal stress in Rif-TN transfected cells continued to elongate throughout the entire observation period . During their uncontrolled elongation , the dorsal stress fibers of Rif-TN expressing cells occasionally bent or fused with another elongating dorsal stress fiber initiated from the opposite side of the cell ( Figure 2—figure supplement 4A; Videos 3 and 4 ) . 10 . 7554/eLife . 06126 . 012Video 3 . Control movie on stress fiber dynamics in GFP-actin expressing U2OS cell . U2OS cells were transfected with GFP-actin 24 hr before imaging . Images were acquired every 15 s . Display rate is 15 frames/second and total duration is 73 , 3 min . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 01210 . 7554/eLife . 06126 . 013Video 4 . Expression of dominant-inactive Rif causes uncontrolled dorsal stress fiber elongation . U2OS cells were transfected with Rif-TN and GFP-actin 24 hr prior to imaging . Loss of proper contractile structures due to Rif-TN expression causes abnormal elongation of the dorsal stress fibers as visualized by GFP-actin . Images were acquired every 15 s . Display rate is 15 frames/s , total video duration 53 , 5 min . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 013 To more directly test the role of myosin II-derived tension in stress fiber elongation , we examined whether local relaxation of contractile ventral stress fibers could re-induce vectorial actin polymerization at focal adhesions . Thus , we applied pointed laser ablation on ventral stress fibers ( see Figure 3—figure supplement 1 ) followed by a similar FRAP assay as shown in Figure 2D . Whereas vectorial actin polymerization in intact contractile fibers was very slow ( ∼0 , 02 μm/min ) , ablated ventral stress fibers displayed approximately 10-fold higher rate of vectorial actin polymerization ( ∼0 , 23 μm/min ) , which is comparable to the one of dorsal stress fibers ( Figure 3A and B , and data not shown ) . Importantly , also photoactivation experiments on PA-GFP-actin expressing cells demonstrated specific elongation of laser ablated ventral stress fibers and lack of vectorial actin polymerization at the focal adhesions located at the ends of non-ablated ventral stress fibers within the same cell ( Figure 3C and D ) . 10 . 7554/eLife . 06126 . 014Figure 3 . Local relaxation of ventral stress fibers induces vectorial actin polymerization at focal adhesion . ( A ) The effect of tension on actin polymerization at focal adhesions was monitored by fluorescence recovery after photobleaching ( FRAP ) in laser-ablated ( indicated by red box/arrowhead ) and intact ventral stress fibers ( indicated by yellow box/arrowhead ) . FRAP experiment was initiated 10 seconds after ablation . Symbol ( ¤ ) indicates the ablation site . ( B ) Kymographs recorded along the center of ablated and non-ablated ventral stress fiber regions ( shown in the red and yellow boxes to the left from the kymograps ) reveal that non-ablated and ablated fibers display differences in actin dynamics . In contractile control fibers , the rate of vectorial actin polymerization is slow ( 0 , 023 μm/min +/- 0 , 007 μm/min; SEM; n = 17 ) , whereas relaxation induces vectorial actin polymerization from the adhesion located at the end of ablated ventral stress fiber ( 0 , 257 μm/min +/- 0 , 035 μm/min; SEM; n = 12 ) . See also Figure 7B for a graphical representation of the data . ( C ) Photoactivation of GFP-PA-actin in contractile ( yellow box ) and relaxed ( red box ) ventral stress fibers . ( D ) Kymograph analysis performed along the center of indicated ventral stress fiber regions ( shown in yellow and red boxes to the left from the kymographs ) demonstrate induction of vectorial actin polymerization at the focal adhesion located in the end of an ablated ventral stress fiber . However , no detectable vectorial actin polymerization occurred in the non-ablated ventral stress fiber . Photoactivation was performed 10 s after ablation of the contractile stress fiber . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 01410 . 7554/eLife . 06126 . 015Figure 3—figure supplement 1 . Method for laser ablation of ventral stress fibers . Ablation of contractile stress fibers was achieved by using full laser power of 405 laser line ( 3I Marianas microscope ) for 3 x 200 ms pulses . Symbol ( ¤ ) indicates ablation sites . Breakage of the fibers was confirmed by following retraction of mCherry-actin bundle with 561 channel for 10 seconds ( A ) or by immediate recruitment of a repair protein zyxin to the ablation site ( B ) . Bar , 10 um . ( C ) Ablation of a ventral stress fiber leads to an immediate decrease in cell-mediated forces at focal adhesions located at the tips of the stress fiber . A representative example is shown in the figure , and quantification of the data from six adhesions at the ends of ablated stress fibers from three different cells revealed 27 , 3 +/- 7 , 0 ( SD ) % decrease in tension at the area around focal adhesions . Traction forces were measured both before , and 0 , 5 seconds after ablation . Bar , 10 um . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 015 These data demonstrate that vectorial actin filament assembly , which promotes elongation of dorsal stress fibers , is inhibited in focal adhesions located at the tips of contractile ventral stress fibers . Furthermore , laser ablation experiments as well as assays with MLCK and ROCK inhibitors , and dominant inactive Rif provide evidence that tension applied by myosin II–mediated contractility is important for inhibition of vectorial actin polymerization at focal adhesions located at the ends of ventral stress fibers . Two proteins promoting actin filament elongation , Dia1 formin and vasodilator-stimulated phosphoprotein ( VASP ) , have been linked to actin polymerization in focal adhesions ( Hotulainen and Lappalainen , 2006; Oakes et al . , 2012; Watanabe et al . , 1999; Gateva et al . , 2014; Figure 4—figure supplement 1A and B ) . Because from these proteins only VASP , and its family members Mena and Evl , accumulate to focal adhesions ( Reinhard et al . , 1992; Gertler et al . , 1996; Lambrechts et al . , 2000; Hoffman et al . , 2006 ) , we focused on examining the possible role of VASP in tension-controlled actin filament assembly in focal adhesions . Previous studies demonstrated zyxin-mediated recruitment of VASP to the sites of stress fiber repair and remodelling ( Smith et al . , 2010; Hoffman et al . , 2012 ) and mechanosensitive recruitment of VASP to epithelial zonula adherens ( Leerberg et al . , 2014 ) . However , whether the activity of VASP within adhesions can be regulated through tension has not been reported . Immunofluorescence microscopy revealed that VASP localizes to focal adhesions located at the tips of both dorsal and ventral stress fibers ( Figure 4A and B ) . Therefore , regulation of VASP localization does not offer an explanation for the lack of vectorial actin polymerization at the tips of ventral stress fibers . Interestingly , previous work demonstrated that phosphorylation of specific residues ( Ser239 and Thr278 ) of VASP inhibit its actin filament binding and polymerization activities ( Harbeck et al . , 2000; Benz et al . , 2009; Figure 4C ) . To study the possible role of VASP phosphorylation in actin filament assembly at focal adhesions , we first examined the localization of phosphorylated VASP in U2OS cells . From several VASP phospho-Ser239/Thr278 antibodies tested , only 16C2 ( Millipore ) worked in immunofluorescence experiments . Although the signal with this antibody was weak , it specifically stained focal adhesions located at the tips of ventral stress fibers , whereas enrichment of phospho-Ser239 VASP to adhesions at the tips of dorsal stress fibers was not detected ( Figure 4D , E and F ) . To confirm this result by an alternative approach , we examined by Western blotting phospho-Ser239 and phospho-Thr278 VASP levels in control cells and in cells where the assembly of contractile ventral stress fibers was stimulated or inhibited by over-expression of dominant active RhoA or by plating cells on compliant matrix , respectively . In line with the data presented above , both phospo-Ser239 ( Figure 4G and H ) and phospho-Thr278 ( data not shown ) levels were >2-fold elevated in the cell population transfected with a construct expressing dominant active RhoA , whereas phospo-Ser239 levels were ∼5-fold diminished in cells plated on soft matrix and unable to form contractile ventral stress fibers ( Figure 4I and J ) . Thus , VASP phosphorylation in focal adhesions correlates with increased contractility of stress fibers . 10 . 7554/eLife . 06126 . 016Figure 4 . Increased VASP phosphorylation in focal adhesions located at the tips of ventral stress fibers . ( A ) VASP localizes to focal adhesions at the tips of both dorsal and ventral stress fibers . Focal adhesions positioned at the tips of dorsal ( dsf FA ) and ventral stress fibers ( vsf FA ) are indicated by red and orange arrowheads , respectively . Bar , 10 μm . ( B ) Line profile intensity graphs of the adhesions ( highlighted in panel A ) and adjacent stress fiber regions show similar localizations of VASP and vinculin in focal adhesions positioned at the tips of dorsal and ventral stress fibers . ( C ) The domain structure of VASP . Phosphorylation of VASP at Ser239 and Thr278 inhibits its actin polymerization activity . ( D ) Localization of phospho-Ser239 VASP in focal adhesions at the tips of dorsal ( red arrowheads ) and ventral stress fibers ( orange arrowheads ) . Bar , 5 μm . ( E ) Line profile intensities along the yellow lines ( indicated in panel D ) demonstrating that phospho-Ser239 VASP is enriched at the tips of ventral stress fibers , but not at the tips of dorsal stress fibers . phospho-Ser239-VASP–red; actin-green ( F ) Quantification of the relative fluorescence intensity ratio of phospho-Ser239-VASP: total VASP in focal adhesions located at the tips of dorsal ( dsf FAs ) and ventral ( vsf FAs ) stress fibers . The obtained intensity value from ventral stress fibers was set to 1 . Mean intensity values ( +/- SEM ) of 21 adhesions are shown . ( G ) Western blot analysis demonstrating that expression of dominant active Rho-14V expression leads to an increase in the total Ser239 phosphorylation levels of VASP . ( H ) Quantification of the relative phospho-Ser239-VASP: total VASP ratios in control and Rho-14V transfected cells . The ratio in control cells was set to 1 and the mean values ( +/- SEM ) from three separate experiments are shown . ( I ) Western blot analysis demonstrating increased phospho-Ser239-VASP levels in cells grown on stiff ( 64 kPa ) martix compared to cells grown on compliant ( 0 . 5 kPa ) matrix . ( J ) Quantification of the relative phospho-Ser239-VASP: total VASP ratios in cells grown on soft ( 0 . 5 kPa ) or rigid ( 64 kPa ) matrices . The phospho-Ser239-VASP: total VASP ratio in cells from stiff matrix was set to 1 and the mean values ( +/- SEM ) from three separate experiments are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 01610 . 7554/eLife . 06126 . 017Figure 4—figure supplement 1 . VASP regulates the elongation of dorsal stress fibers . ( A ) Depletion of VASP from U2OS cells by siRNA leads to loss of dorsal stress fibers ( examples indicated by red arrowheads in the control cell ) , but these cells still contain focal adhesions and ventral stress fibers ( examples are indicated by orange arrowheads in the VASP-depleted and control cell ) . However , the ventral stress fibers are abnormally aligned in VASP knockdown cells . The cells were stained with VASP antibody , vinculin antibody ( to visualize focal adhesions ) and phalloidin ( to visualize F-actin ) . The VASP depleted cell is surrounded by a dashed line . Bar , 10 μm . ( B ) VASP depleted cells still exhibit ventral stress fibers , but they are predominantly poorly aligned and form abnormal clusters ( orange arrowheads ) . Time-lapse frames from a movie of VASP-depleted U2OS cell expressing GFP-actin also demonstrate that in the absence of VASP ( and visible dorsal stress fibers ) , ventral stress fibers ( orange arrowheads ) form from transverse arcs very close to the lamellipodium . Bar , 10 μm . ( C ) Over-expression of constitutively active ( Ser239Ala;Thr278Ala ) VASP affects elongation of stress fibers and often leads to a formation of abnormally curly actin filament bundles ( visualized by phalloidin staining ) . Bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 017 VASP phosphorylation at Ser239 and Thr278 is regulated by cAMP- and cGMP dependent protein kinases PKA and PKG as well as by AMP-activated Protein Kinase ( AMPK ) ( Butt et al . , 1994; Blume et al . , 2007 ) . To elucidate the possible role of VASP phosphorylation in controlling actin polymerization in focal adhesions , we examined the effects of PKA , PKG and AMPK inhibitors ( KT5720 , DT-2 , compound C and KT5823 ) on the organization of the stress fiber network . As AMPK inhibitors , compound C and KT5823 , had most pronounced effects on the elongation of stress fiber precursors , we decided to focus on AMPK rather than PKA and PKG in this study . Both of compound C and KT5823 inhibited VASP phosphorylation at Ser239 and Thr278 ( Figure 5C ) . Importantly , incubation of U2OS cells for 4 hr in the presence of these inhibitors resulted in a nearly complete lack of contractile ventral stress fibers and defects in arc fusion . Furthermore , these inhibitors promoted formation of abnormally long dorsal stress fibers , which often bent at their proximal regions ( Figure 5A and B ) . It is important to note that these inhibitor treatments also led to an increase in the total cell area that may result from the lack of contractile ventral stress fibers , which are important regulators cell morphogenesis . 10 . 7554/eLife . 06126 . 018Figure 5 . Elongation of dorsal stress fibers is regulated by VASP phosphorylation . ( A ) U2OS cells treated with cAMPK and PKA inhibitors , compound C and KT5823 , display abnormally long dorsal stress fibers ( red arrows ) . Also fusion of arcs appears defective in compound C and KT5823 treated cells , because thick actin bundles ( yellow arrowhead in the control cell ) are largely absent from these cells . Actin filaments were visualized by phalloidin . Bar , 10 μm . ( B ) Quantification of dorsal stress fiber lengths ( μm ) from control cells and cells treated with compound C or KT5823 . Mean lengths ( +/- SEM ) of 40 dorsal stress fibers from each sample are shown . ( C ) Western blot demonstrating that lysates of compound C or KT5823 -treated cells display decreased phosphorylation of VASP at Ser239 and Thr278 . GADPH and total VASP were probed as loading controls . ( D ) Expression of constitutively active VASP mutant ( Ser239Ala;Thr278Ala ) induces formation of abnormally long dorsal stress fibers , whereas similar phenotype was not observed in wild-type VASP expressing cells . Bar , 10 μm . ( E ) Quantification of dorsal stress fiber lengths from cells expressing wild-type and Ser239Ala;Thr278Ala mutant VASP . Mean lengths ( +/- SEM ) of 55 dorsal stress fibers measured from both samples are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 018 To confirm that the stress fiber phenotype in compound C and KT5823 –treated cells was specific to VASP , and did not result from diminished phosphorylation of other AMPK targets , morphology of the stress fiber network of U2OS cells expressing a ‘constitutively active’ Ser239Ala;Thr278Ala VASP mutant was examined ( Benz et al . , 2009 ) . Also Ser239Ala;Thr278Ala mutant VASP expressing cells displayed significantly longer dorsal stress fibers as compared to wild-type VASP expressing cells ( Figure 5D and E ) . Furthermore , over-expression of Ser239Ala;Thr278Ala mutant VASP occasionally resulted in formation of ‘curly’ actin filament bundles ( Figure 4—figure supplement 1C ) . Thus , inhibition of VASP phosphorylation at Ser239 and Thr278 results in a similar phenotype compared to the one resulting from the inhibition of contractile arc assembly through over-expression of Rif-TN ( see Figure 2—figure supplement 4 ) , suggesting that VASP phosphorylation has a key role in tension-controlled actin filament assembly at focal adhesions . Conversely , activation of AMPK by AICAR ( Carling et al . , 2008 ) leads to an increased Ser239 phosphorylation of VASP and early maturation of ventral stress fibers ( Figure 6A and B ) . Cells exposed for 16 h to AICAR displayed shorter dorsal stress fibers compared to control cells and their ventral stress fibers were typically located close to cell perimeter ( Figure 6A ) . This phenotype was similar to the one resulting from depletion of VASP ( Figure 4—figure supplement 1 ) , suggesting that AICAR indeed affects stress fibers mainly through inducing VASP phosphorylation . Importantly , activation of AMPK and VASP phosphorylation by AICAR treatment was sufficient to induce the formation of ventral stress fibers on soft ( 0 . 5 kPa ) matrix , where their formation is normally inhibited ( Figure 6C and D ) . 10 . 7554/eLife . 06126 . 019Figure 6 . AMPK activation leads to maturation of contractile actomyosin bundles . ( A ) U2OS cells grown on glass and treated with AMPK activator AICAR exhibited short dorsal stress fibers . Ventral stress fibers were typically localized close to cell edges , indicating early maturation of the contractile actomyosin bundles compared to control cells . ( B ) Western blot analysis of the corresponding samples showed elevation of Ser239-P-VASP in AICAR treated cells . A representative Western blot and quantification of three separate experiments ( mean +/- SEM ) are shown . ( C ) Activation of AMPK bypasses the need for stiff substrata in ventral stress fiber formation . U2OS cells grown on soft ( 0 . 5 kPa ) matrix do not typically contain contractile ventral stress fibers , but their formation on soft matrix can be induced by AICAR . ( D ) Quantification of the amount of cells ( % ) containing ventral stress fibers . For each condition , 50–70 cells were analysed and the data are presented as mean +/- SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 019 To directly test the role of VASP in mechanosensitive actin filament assembly at focal adhesions , we performed photoablation experiments on VASP knockdown cells , and applied FRAP to compare the vectorial actin polymerization rates of ventral stress fibers in control vs . VASP-depleted cells . These experiments revealed approximately 3-fold decrease in ablation-induced vectorial actin polymerization at the tips of ventral stress fibers in VASP knockdown cells compared to control cells ( Figure 7 ) , demonstrating that VASP is indeed important for mechanosensitive actin filament assembly in focal adhesion . Together , these results provide evidence that AMPK-mediated phosphorylation of VASP is essential for tension-sensitive inhibition of vectorial actin filament assembly at focal adhesions , and consequent formation and stabilization of ventral stress fibers . 10 . 7554/eLife . 06126 . 020Figure 7 . VASP-depletion leads to a decrease in tension-sensitive actin polymerization at focal adhesions . ( A ) Actin polymerization at focal adhesions was monitored by fluorescence recovery after photobleaching ( FRAP ) in laser-ablated ventral stress fibers of control cells and VASP knockdown cells . FRAP experiments were initiated 10 seconds after ablation . Yellow arrowheads indicate the ablation sites and red boxes the regions of stress fibers that were followed for vectorial actin polymerization . Kymographs on the right were recorded along the center of the ablated ventral stress fibers . Release of tension induces vectorial actin polymerization from the adhesion located at the end of an ablated ventral stress fiber in a control cell , whereas the rate of vectorial actin polymerization was slower in an ablated ventral stress fiber in a VASP-depleted cell . Bar , 10 um . ( B ) Quantification of vectorial actin polymerization rates in intact ventral stress fibers , in ablated ventral stress fibers of control cells , and in ablated ventral stress fibers of VASP-depleted cells . Mean +/- SEM is shown; n ( intact ventral stress fibers ) = 17; n ( ablated ventral stress fibers ) = 12; n ( ablated ventral stress fibers from VASP knockdown cells ) = 17 . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 020 Maturing arcs are typically connected to several dorsal stress fibers and focal adhesions , but only the ones located at the ends of the arc bundle are used for the formation of a ventral stress fiber ( Figure 2—figure supplement 1 ) . To gain insight into the fate of other arc-associated dorsal stress fibers and focal adhesions , we performed live-imaging of cells expressing GFP-zyxin and mCherry-actin . Focal adhesions connected to the ends of arcs elongated during maturation of arcs into a ventral stress fiber , whereas adhesions associated with the central region of the arc through dorsal stress fibers diminished in size and eventually disappeared ( Figure 8—figure supplement 1 ) . To reveal what happens to those dorsal stress fibers , which are located at the ‘unstable zone’ at the central region of the leading edge ( see Figure 8D ) , we followed the stress fiber network in cells expressing GFP-actin . These experiments revealed that dorsal stress fibers , oriented perpendicularly to the contractile arc , sense weaker myosin II -generated tension and disassemble during stress fiber maturation process ( Figure 8A and C ) . Furthermore , those dorsal stress fiber regions , which reach beyond the contractile arc/ventral stress fiber , disassemble during the process ( Figure 8B ) . 10 . 7554/eLife . 06126 . 021Figure 8 . Dorsal stress fibers exhibit different lifespans depending on their interactions with the actomyosin network . ( A ) Individual frames from a representative movie of GFP-actin expressing cell demonstrating the disassembly of non-contractile dorsal stress fibers located at the ‘unstable’ zone . Bar , 5 μm . ( B ) Frames from a movie of GFP-actin expressing cell displaying the disassembly of the non-contractile dorsal stress fiber region extending beyond the contractile transverse arc . Bar , 5 μm . ( C ) Quantification of the stability of dorsal stress fibers at different cell regions revealed that these actin bundles are more stable at the sides of the leading edge as compared to the central region of the leading edge . Amount of dorsal stress fibers ( % ) , remaining after 30 min follow-up , is shown ( mean +/- SD ) , n = 5 cells , 8–20 fibers per cell were analysed . ( D ) Representation of the ‘unstable’ and ‘stable’ dorsal stress fiber zones in U2OS cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 02110 . 7554/eLife . 06126 . 022Figure 8—figure supplement 1 . Enlargement and lifespan of focal adhesions depends on their location in migrating cells . ( A ) Dorsal stress fiber-attached focal adhesions on the sides of a migrating cell ( orange arrowheads ) become more prominent in size and participate in formation of contractile ventral stress fibers . In contrast , dorsal stress fibers and focal adhesions in the central region of the leading edge ( red arrowheads ) , which do not participate in the formation of ventral stress fibers , diminish in size and eventually disappear . Image frames are from movies of U2OS cells expressing GFP-actin and CFP-zyxin . Bar , 10 um . ( B ) Quantification of the areas of individual focal adhesions ( fold change in comparison to the initial area of the adhesion ) located at the tips of dorsal stress fibers and maturing ventral stress fibers . Cells were imaged for one hour and the mean values ( +/- SD ) of focal adhesions from 7 cells are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 022 Actin depolymerizing factor ( ADF ) /cofilin proteins are essential regulators of F-actin disassembly in all eukaryotic cells ( Poukkula et al . , 2011 ) . Interestingly , ADF/cofilins were recently shown to preferentially bind and disassemble flexible actin filaments in vitro . ADF/cofilins did not localize to contractile stress fibers in intact cells , but translocated to stress fibers when pre-stretched elastic substratum was relaxed ( Hayakawa et al . , 2011 ) . Furthermore , ADF/cofilins affect the actin filament bending mechanics ( McCullough et al . , 2008; Elam et al . , 2013 ) . Thus , we examined whether ADF/cofilins could be responsible for specific disassembly of those dorsal stress fibers that are not under tension in migrating U2OS cells . The major ADF/cofilin isoform , cofilin-1 , is highly abundant protein in non-muscle cells where it displays mainly diffuse cytoplasmic localization . Interestingly , endogenous cofilin-1 and flag-tagged cofilin-1 also localized to dorsal stress fibers in U2OS cells , whereas contractile ventral stress fibers and thick arcs did not exhibit detectable enrichment of cofilin-1 ( Figure 9A and B; Figure 9—figure supplement 1A , B and C ) . Furthermore , cofilin-1 localized to the ‘curly’ actin filament bundles that were occasionally present in cells over-expressing Ser239Ala;Thr278Ala mutant VASP ( Figure 9—figure supplement 1D ) . These bundles contain myosin II , but exert defective contractile properties as detected by live cell imaging ( data not shown ) . Thus , cofilin-1 appears to localize specifically to dorsal stress fibers and other non-contractile actin bundles in U2OS cells . 10 . 7554/eLife . 06126 . 023Figure 9 . Cofilin-1 promotes disassembly of non-contractile dorsal stress fibers . ( A ) Endogenous cofilin-1 localizes to dorsal stress fibers ( red arrowheads ) but it is absent from contractile arcs ( yellow arrowheads ) as shown by phalloidin and anti-cofilin-1 staining of a U2OS cell treated with Triton-X 100 prior to PFA fixation . Bar , 10 μm . ( B ) Line intensity profiles show incorporation of cofilin-1 into dorsal stress fibers ( dsf ) but not to the contractile ventral stress fibers ( vsf ) . Cofilin ( green ) ; Actin ( red ) . ( C ) Depletion of cofilin-1 leads to an appearance of abnormally long dorsal stress fibers and defects in the fusion of transverse arcs . Bar , 10 μm . ( D ) Quantification of the lengths of dorsal stress fibers ( μm ) in control and cofilin-1-depleted cells . Mean lengths ( +/- SEM ) of 50 dorsal stress fibers from control and cofilin-1 RNAi cells are displayed in the graph . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 02310 . 7554/eLife . 06126 . 024Figure 9—figure supplement 1 . Cofilin-1 localizes to dorsal stress fibers and affects their turnover . ( A ) Flag-tagged cofilin-1 localizes to non-contractile dorsal stress fibers , whereas it is absent from contractile ventral stress fibers and transverse arcs . Cofilin-1 ( red ) was visualized by anti-flag antibody and F-actin ( green ) by phalloidin staining . Bar , 10 μm . ( B ) Magnifications of a flag-tagged cofilin-1 expressing cell . The regions selected for the line intensity profiles from dorsal and ventral stress fibers are indicated by red and orange lines , respectively . ( C ) Line intensity profiles show incorporation of cofilin-1 into dorsal stress fibers ( dsf ) but not to the contractile ventral stress fibers ( vsf ) . ( D ) Expression of constitutively active Ser239Ala;Thr278Ala mutant VASP leads to formation of curly actomyosin structures . Cofilin-1 localizes to these structures . ( E ) Cofilin-1-depleted cells show abnormally long and thick dorsal stress fibers as well as impaired fusion of transverse arcs into thick contractile ventral stress fibers . Stress fibers were visualized with phalloidin staining . Bar , 10 μm . ( F ) Time lapse frames from a cofilin-1-depleted U2OS cell , transfected with GFP-actin . Depletion of cofilin-1 leads to formation of abnormally stable dorsal stress fibers , which appear to prevent arc fusion , but continue to grow between adjacent arcs . Red arrowhead indicates a dorsal stress fiber and yellow brackets indicate two arcs that are not able to fuse . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 024 Depletion of cofilin-1 leads to defects in multiple actin-based structures such as lamellipodia and sites of endocytosis due to diminished filament disassembly as well as to cortical F-actin accumulation due to excessive myosin II activity ( e . g . Hotulainen et al . , 2005; Sidani et al . , 2007; Kiuchi et al . , 2007; Wiggan et al . , 2012 ) . Importantly , depletion of cofilin-1 from U2OS cells resulted also in an appearance of abnormally thick and long dorsal stress fibers ( Figure 9C and D; Figure 9—figure supplement 1E ) . Furthermore , fusion of arcs during their centripetal flow was diminished , leading to problems in the formation of proper ventral stress fibers ( Figure 9—figure supplement 1F; Video 5 ) . Defects in arcs fusion suggest that proper ADF/cofilin-mediated turnover of dorsal stress fibers is required for proper coalescence of dorsal stress fiber–associated arcs . Together , these data provide evidence that ADF/cofilins are important for turnover of non-contractile dorsal stress fibers , whereas contractile arcs and ventral stress fibers appear to be protected from ADF/cofilin-mediated F-actin disassembly . 10 . 7554/eLife . 06126 . 025Video 5 . Effects of cofilin-1 depletion on stress fiber dynamics . Cofilin-1 depleted U2OS cells were transfected with GFP-actin 24 h prior to imaging . Lack of cofilin-1 leads to elongated dorsal stress fibers and impairs transverse arc fusion . Images were acquired every 15 s . Display rate is 15 frames/s , total video duration 62 , 5 min . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 025 Ventral stress fibers play an important role in cell adhesion , morphogenesis and migration , but how these and other contractile actomyosin bundles are generated has remained elusive . Here we have revealed several new aspects concerning the mechanisms underlying the assembly of contractile ventral stress fibers . We provide evidence that: ( 1 ) Formation of ventral stress fibers from their precursors ( arcs and dorsal stress fibers ) is a mechanosensitive process . ( 2 ) Arcs fuse with each other during centripetal flow to form thicker and more contractile actomyosin bundles , which apply tension to focal adhesions located at their ends . ( 3 ) This tension activates AMPK-mediated phosphorylation of VASP that leads to inhibition of vectorial actin polymerization at focal adhesions . ( 4 ) AMPK-mediated VASP phosphorylation is necessary for assembly and proper alignment of contractile ventral stress fibers . Conversely , activation of AMPK can bypass the need of stiff matrix for ventral stress fiber assembly . ( 5 ) ADF/cofilin–mediated disassembly of non-contractile dorsal stress fibers is important for the proper maturation of the stress fiber network . We propose that similar mechanosensitive actin filament assembly and disassembly may have general role in formation and alignment of diverse contractile actomyosin bundles in different cell-types . A working model for mechanosensitive assembly of contractile ventral stress fibers is presented in Figure 10 . Nascent adhesions appear at the lamellipodium of migrating cell and a fraction of them matures to focal adhesions ( Burnette et al . , 2011; Choi et al . , 2008 ) . Dorsal stress fibers are initiated from focal adhesions located at the leading edge of cells , and elongate through ‘vectorial’ actin polymerization at focal adhesions , mediated at least by VASP and Dia1 formin ( Hotulainen and Lappalainen , 2006; Watanabe et al . , 1999; Gateva et al . , 2014 ) . Similarly to filopodia , where VASP localizes at the tip-complex and promotes assembly of unipolar actin filament bundles , VASP at focal adhesions is expected to catalyse polymerization of an unipolar actin filament bundle towards the cell center . Elongating dorsal stress fibers associate with multiple myosin II containing arcs , which are derived from the lamellipodial actin structures ( Figure 10A ) . During centripetal flow of this spider-net like structure , arcs fuse with each other to generate a thicker , more contractile bundle ( Figure 10B and C ) . As a result , those focal adhesions that are attached via dorsal stress fibers to the ends of the contractile arc sense strong myosin II-generated tension , leading to their enlargement and turning along the direction of the arc . In support to this , ventral stress fibers apply strong traction forces to the substrate through their terminally located focal adhesions ( Figure 2 and Möhl et al . , 2012 ) . Moreover , tension-mediated maturation of terminal focal adhesions leads to inhibition of vectorial actin polymerization , which is at least partially mediated by phosphorylation of VASP ( Figure 10C ) . However , because VASP-depletion did not result in a compete inhibition of vectorial actin polymerization after releasing the tension in ventral stress fibers ( Figure 7 ) , other proteins are also likely to contribute to vectorial actin polymerization at focal adhesions . It is also important to note that , although the actin polymerization activity of VASP is inhibited at focal adhesions under high tension , VASP protein is still present in that location . Thus , VASP may contribute to integrity of the adhesions at the tips of ventral stress fibers through its other activities , including actin filament bundling ( Bear and Gertler , 2009 ) . Furthermore , VASP phosphorylation is not expected to inhibit all actin dynamics in focal adhesions , because VASP appears to specifically contribute to vectorial actin polymerization at focal adhesions , whereas other proteins such as formins may promote the turnover of other focal adhesion-associated actin filament populations . Importantly , inhibition of vectorial actin polymerization is essential for proper alignment and contractility of the ventral stress fiber , because continuous elongation of the actin filament bundle from focal adhesions would counteract myosin II-driven shortening of the actomyosin bundle . Finally , we propose that mechanosensitive binding of cofilin-1 to those dorsal stress fibers and dorsal stress fiber regions , which are not under myosin II-applied tension , leads to disassembly of these ‘non-productive’ regions of the stress fiber network . Consequently , the focal adhesions located at the distal ends of ‘central’ dorsal stress fibers , which are oriented perpendicularly to the contractile arcs and hence do not sense strong myosin II–derived tension , diminish in size and disappear . On the other hand , contractile arcs and mature ventral stress fibers are protected from ADF/cofilin-mediated actin filament disassembly ( Figure 10D and E ) . Consistent with this model , dorsal stress fibers and arcs form in cells grown on soft substrata , whereas contractile ventral stress fibers fail to assemble in compliant matrix ( Figure 1A ) . Focal adhesions are mechanosensitive structures ( e . g . Iskratsch et al . , 2014 ) . Their maturation from nascent adhesions and maintenance require relatively small forces that can be generated by retrograde actin flow without myosin II-driven contractility ( Oakes et al . , 2012; Stricker et al . , 2013 ) . However , focal adhesions mature into larger , elongated adhesions under stronger , myosin II-derived tension ( Geiger et al . , 2009 ) . Myosin II -generated force also affects the protein composition and dynamics in focal adhesions ( Kuo et al . , 2011; Schiller et al . , 2011; Wolfenson et al . , 2011 ) . Here , we provide evidence that vectorial actin polymerization , which drives elongation of stress fibers , is strictly controlled in focal adhesions . Thus , different force regimes seem to have distinct effects on actin dynamics and molecular composition of focal adhesions . While weak forces exerted by retrograde actin flow appear to be required for VASP recruitment and to promote vectorial actin polymerization at focal adhesions , stronger force applied by contractility of the myosin II-containing ventral stress fiber efficiently inhibits actin polymerization at focal adhesions . Importantly , during this process VASP is phosphorylated by AMPK , whose activity at least in muscle cells can be controlled by tension through a currently uncharacterized mechanism ( Blair et al . , 2009 ) . It is likely that activities or localizations of additional actin-polymerization associated proteins in focal adhesions are regulated by contractility . Indeed , an interaction partner of VASP , palladin ( Gateva et al . , 2014 ) , as well as tropomyosins Tm1 and Tm5NM1 ( Figure 10—figure supplement 1 ) are enriched in focal adhesions located at the tips of dorsal stress fibers , but absent from adhesions at the tips of ventral stress fibers . Furthermore , formins contribute to actin filament nucleation and/or processive polymerization in focal adhesions , and it is likely that also their activities are regulated during this process . Recent studies demonstrated that low-regime forces ( <3 pN ) increase the actin polymerization activity of formins ( Courtemanche , et al . , 2013; Jégou et al . , 2013 ) . Furthermore , the activity of formins can be controlled by a local increase in G-actin levels ( Higashida et al . , 2013 ) . In the future , it will be interesting to examine the effects of imposed external forces , comparable in magnitude to tension applied by a myosin II-containing ventral stress fiber , on formins . However , similar to shown here for VASP , we propose that possible regulation of formin activity in focal adhesions is more likely controlled through biochemical signalling cascades than direct mechanical regulation of the formin molecule . This is because culturing cells on hyaluronic acid containing soft gels can produce a similar formation of ventral stress fibers that is otherwise observed only in cells cultured on rigid substrates ( Chopra et al . , 2014 ) . Furthermore , several tyrosine kinases play an important role in focal adhesion mechanosensing , and their inactivation can shift the stiffness regime for assembly of large focal adhesions and ventral stress fibers ( Prager-Khoutorsky et al . , 2011 ) . 10 . 7554/eLife . 06126 . 026Figure 10 . A working model for mechanosensitive generation of ventral stress fibers in U2OS cells . ( A ) Dorsal stress fibers elongate through vectorial actin polymerization from focal adhesions located at the leading edge of the cell , and form a spider net -like structure with multiple transverse arcs . At least Dia1 formin and VASP are involved in vectorial actin polymerization and consequent dorsal stress fiber elongation from focal adhesions . ( B ) Arcs flow along the elongating dorsal stress fibers towards the cell center , and fuse with each other to form thicker and more contractile actomyosin bundles . ( C ) Tension provided by the contraction of arcs is mediated through dorsal stress fibers to those focal adhesions that are linked to the end of the arc . This leads to enlargement of ‘terminal’ adhesions and their alignment along the direction of the contractile arc bundle . Tension provided by myosin II –driven contractility of the arc inhibits vectorial actin polymerization in ‘terminal’ focal adhesions , and this is at least partially mediated by VASP phosphorylation . Consequently , elongation of the actomyosin bundle ceases , thus allowing its efficient contractility . Please note that focal adhesions are likely to be composed of many actin filament populations , and for simplicity only the one undergoing vectorial actin polymerization and thus promoting stress fiber elongation is shown in the model . ( D ) Cofilin-1 specifically binds to and promotes the disassembly of non-contractile dorsal stress fibers , which are connected to the central regions of the arc and thus do not participate in the formation of the ventral stress fiber . ( E ) Whereas non-contractile stress fibers are disassembled by cofilin-1 , contractile stress fibers are protected from tension-sensitive cofilin-1–induced severing . Eventually , this leads to the formation of a contractile ventral stress fiber , which is connected to one large focal adhesion at its each end and aligned perpendicularly to the direction of lamellipodium extension . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 02610 . 7554/eLife . 06126 . 027Figure 10—figure supplement 1 . Differential localizations of Tm1 and Tm5NM1 in focal adhesions at the tips of dorsal and ventral stress fibers . Tm1 and Tm5 localize to focal adhesion at the tips of dorsal stress fiber sites ( dsf FAs ) , while they are not enriched in focal adhesions located at the tips of contractile ventral stress fiber adhesions ( vsf FAs ) . Cells expressing GFP-tagged tropomyosins were stained with a focal adhesion marker vinculin . Red and orange arrowheads indicate focal adhesions at the tips of dorsal and ventral stress fibers , respectively . Bars 10 um . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 027 In addition to vectorial actin polymerization at focal adhesions , actin filaments within the stress fiber network undergo turnover with a half-life of an approximately one minute ( Hotulainen and Lappalainen , 2006 ) . We propose that maintenance or disappearance of individual stress fibers depends on a balance between assembly and tension-sensitive disassembly of actin filaments . In addition to its role in actin polymerization in focal adhesions , VASP contributes to actin filament assembly within the stress fiber network ( Gateva et al . , 2014; Smith et al . , 2010; Hoffman et al . , 2012 ) . Our data propose that stress fibers are maintained or become thicker under tension , whereas mechanosensitive binding of ADF/cofilins to stress fibers that are not under tensions shifts the balance from steady state ( or net assembly ) to net disassembly . Eventually , this leads to disappearance of the stress fiber and the focal adhesion associated with its end . What are the functions of dorsal stress fibers ? Our data demonstrate that arc fusion during centripetal flow occurs preferentially at the intersections with dorsal stress fibers . This suggests that dorsal stress fibers may functions as ‘rails’ to facilitate coalescence of adjacent arcs in the 3D-environment inside lamellum . However , arc fusion and formation of focal adhesion–attached ventral stress fibers can occur also in VASP-depleted cells , which either do not contain dorsal stress fibers or where these actin filament bundles are very thin and fragile ( Figure 4—figure supplement 1A and B; Video 6 ) . Furthermore , many cell-types including epithelial cells can assemble peripheral actomyosin bundles resembling ventral stress fibers in the apparent absence of dorsal stress fibers . Thus , focal adhesion–attached contractile actomyosin bundles can be generated at least in non-motile cells without prominent dorsal stress fibers . It is , however , important to note that the stress fiber network is typically poorly organized in VASP-depleted U2OS cells ( Figure 4—figure supplement 1B ) , suggesting that dorsal stress fibers are required for proper alignment of ventral stress fibers in migrating cells . Furthermore , dorsal stress fibers play an important role in directional cell migration ( Kovac et al . , 2013 ) . 10 . 7554/eLife . 06126 . 028Video 6 . VASP-depleted cells lack dorsal stress fibers and have poorly aligned contractile actomyosin bundles . VASP-depleted U2OS cells were transfected with GFP-actin one day prior to imaging . Images were acquired every 15 s . Display rate is 15 frames/s , total video duration 21 , 3 min . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 028 Collectively , our findings reveal that mechanosensitive actin filament assembly and disassembly are essential for generation of contractile ventral stress fibers , and function as selection processes to ensure proper alignment of ventral stress fibers perpendicularly to the direction of cell migration . In the future , it will be important to identify the signalling pathway regulating mechanosensitive phosphorylation of VASP in focal adhesions . Here it will be especially interesting to examine the possible contribution of mechanosensitive Ca2+ channels and Ca2+ -activated CaMKK family kinases , because the latter can activate AMPK kinase ( Carling et al . , 2008 ) . Furthermore , it will be important to reveal how the activities of other proteins contributing to actin polymerization at focal adhesions are regulated by myosin II -dependent tension . Finally , it will be interesting to examine how mechanosensitive actin filament assembly and disassembly contribute to generation and proper alignment of stress fibers and other contractile actomyosin bundles in the tissue environment . Human osteosarcoma ( U2OS ) cells were maintained as described in Hotulainen and Lappalainen ( 2006 ) . Transient transfections were performed with LipofectamineTM2000 ( Invitrogen ) according to manufacturer’s instructions . Cells were subsequently incubated for 24 hr and further fixed with 4% PFA or detached with trypsin-EDTA and plated on fibronectin-coated ( 10 μg/ml fibronectin ) glass-bottomed dishes ( MatTek ) for live cell imaging . Fibronectin-coated CytoSoftTM 35 mm biocompatible silicone dishes ( Advanced BioMatrix ) with elastic modulus of 0 . 5 and 64 kPa were used for studying the effect of matrix rigidity on stress fiber composition . For siRNA silencing , 2100 ng of pre-annealed 3′ Alexa Fluor 488–labelled oligonucleotide duplexes were transfected into cells on 35 mm plates by using GeneSilencer's siRNA transfection reagent ( Gene Therapy Systems ) according to the manufacturer's instructions . Cells were incubated for 72–96 hr for efficient depletion of the target protein . For inhibition of VASP phosphorylation , cells were treated with AMPK inhibitor , Compound C ( final concentration of 5 uM for 5 hr ) or PKA/PKG inhibitor , KT5823 ( 1 uM , 4 hr ) and for AMPK activation , 25 uM AICAR was used for 16 hr . For disruption of contractile structures , cells were treated with either myosin light chain kinase ( MLCK ) inhibitor ML-7 ( 1 uM , 2 hr ) and ROCK inhibitor , Y27632 ( 1 uM , 2 hr ) . All chemical compounds were purchased from Sigma-Aldrich . Cells were transfected , incubated for 24 hr , and re-plated prior to imaging on 10 μg/ml fibronectin–coated glass-bottomed dishes ( MatTek Corporation ) . The time-lapse images were acquired with 3I Marianas imaging system ( 3I intelligent Imaging Innovations ) , with an inverted spinning disk confocal microscope Zeiss Axio Observer Z1 ( Zeiss ) and a Yokogawa CSU-X1 M1 confocal scanner , or with an inverted microscope ( IX-71; Olympus ) equipped with a Polychrome IV monochromator ( TILL Photonics ) . Both systems have appropriate filters , heated sample chamber ( +37°C ) , and controlled CO2 . With 3I Marianas , a 63x/1 . 2 W C-Apochromat Corr WD = 0 . 28 M27 objective was used . SlideBook 5 . 0 software ( 3I intelligent Imaging Innovations ) and sCMOS ( Andor ) Neo camera were used for the image acquirement and recording . With Olympus , a 60x water objective with 1 . 6× magnification was used . TILL Vision 4 software ( TILL Photonics ) and Imago QE ( TILL Photonics ) and Andor iXon ( Andor ) cameras were used for the image acquirement and recording . Deconvolution of the time-lapse videos was performed with AutoQuant AutoDeblur 2D non-blind Deconvolution ( AutoQuant Imaging , Inc . ) . Further analyses of the video frames were performed with Image Pro Plus 6 . 0 . U2OS cells , transfected with Cherry-Actin , were cultured for 3–8 hr on collagen-1-coated polyacrylamide ( PAA ) gel substrates ( elastic modulus = 26 kPa ) that were coated with sulfate fluorescent microspheres ( diameter = 100 or 200 nm , Life Technologies ) ( Marinkovic et al . , 2012 ) . Using an inverted fluorescence microscope ( Leica DMI6000 ) , images of cells and of the fluorescent microspheres directly underneath the cells were imaged during the experiment and after cell detachment with trypsin . By comparing the fluorescent microsphere images before and after cell detachment , we computed spatial maps of cell-exerted displacement . With knowledge of the displacement field and that of the substrate stiffness , we computed the traction field using the well-established method of constrained fourier transform traction microscopy ( Butler et al . , 2002; Krishnan et al . , 2009 ) . From the cell traction map , we computed local force within an ∼13µm2 area around pre-selected points corresponding to tips of focal adhesions at either dorsal ( red ) or ventral ( orange ) stress fibers ( Figure 2A ) . DNA transfections were performed as described in Tojkander et al . ( 2011 ) . The following constructs were used in experiments: wild-type GFP-VASP , GFP-VASPser239ala , thr278ala , , which was generated from the triple mutant AAA-GFP-VASP construct ( Benz et al . , 2009 ) , GFP-CaP3 ( Burgstaller et al . , 2002 ) , PA-GFP-actin ( a kind gift from Maria Vartiainen ) , cofilin-1-Flag ( Hotulainen et al . , 2005 ) , Rif-TN , YFP-Tm4 , CFP-and YFP-α-actinin , CFP-and mCherry-Zyxin ( Hotulainen and Lappalainen , 2006 ) , GFP-and Cherry-actin ( Tojkander et al . , 2011 ) , dominant active RhoA ( Vartiainen et al . , 2000 ) . For depletion of VASP , Dharmacon ON-TARGETplus Smartpool cat# L-019763-01 , Lot# 121105 was used . For depletion of cofilin-1 target sequence “AAG GAG GAT CTG GTG TTT ATC” was used for a 5´-Alexa Fluor 488 labelled siRNA , which was purchased from Qiagen . Cells were fixed with 4% PFA , washed 3 x with 0 . 2% Dulbecco/BSA and permeabilized with 0 . 1% Triton X-100 in TBS . Immunofluorescence stainings were performed as in ( Tojkander et al . , 2011 ) . Images were acquired with a charge-coupled device camera ( AxioCam HRm; Zeiss ) on a microscope ( Axio Imager . M2; Zeiss ) . AxioVision Rel . 4 . 8 ( Zeiss ) and PlanApo 63x/1 . 40 ( oil ) objective ( Zeiss ) was used for the image acquirement . The following reagents and antibodies were used for the stainings: Alexa phalloidin 488 , 568 , 594 and 647 ( 1:200–400 dilutions ) ( Life TechnologiesTM ) , anti-cofilin-1 antibody ( Abcam , ab11062 ) , anti-VASP antibodies ( 1:50–100 ) ( Sigma , HPA005724 and Enzo , IE273 ) , VASP-phospho-T278 antibody ( 1:50 ) ( ImmunoWay ) , VASP-phospho-S239 antibody ( 1:50 ) ( Millipore , 16C2 ) , anti-vinculin antibody ( 1:50 ) ( Sigma , hVin-1 ) . DAPI and secondary antibodies , which were conjugated to Alexa Fluor 488 , Alexa Fluor 568/594 , or Cy5 were from Life Technologies . Cells were washed with cold PBS , scraped , and lysed in PBS , 1% Triton X-100 ( with 0 . 3 mM PMSF and protease and phosphatase inhibitor cocktail ( Pierce ) . Protein concentrations were measured using Bradford reagent ( Sigma-Aldrich ) . Alternatively , cells were lysed after washes into 4x LSB-DTT buffer for obtaining total cell lysates . Lysates were briefly sonicated prior to boiling . Mixture of 5% milk/BSA was used for blocking . Following antibodies were used for detection with dilutions recommended by the manufacturers: rabbit polyclonal anti-VASP antibodies ( Sigma , HPA005724 and Enzo , IE273 ) , VASP-phospho-T278 antibodies ( ImmunoWay and ECM Biosciences , VP2781 ) , VASP-phospho-S239 antibodies ( Millipore , 16C2; Sigma SAB4504565; Abcam , 16C2 ) , anti-GAPDH ( Sigma , G8795 ) . Appropriate HRP-linked secondary antibodies ( Promega ) and ECL reagent ( AmershamTM , GE Healthcare ) were applied for chemiluminescence detection of the blots . Quantity One 4 . 1 . 1 program ( Bio-Rad ) was used to quantify the band intensities of blots . Live cell imaging with PA-GFP-actin/Cherry-Zyxin-transfected U2OS cells was performed as above with 3I Marianas imaging system . Three captures were taken before activation of PA-GFP-actin with 405 lasers . Activation was performed at the adhesion sites at the tips of dorsal and ventral stress fibers . 488 and 561 lasers were used to visualize activated protein and focal adhesion marker Zyxin , respectively . Images were captured 5x every 2 ms , after which the signal was recorded every 20 s . For measuring vectorial actin polymerization as well as actin dynamics within focal adhesions by fluorescence recovery after photobleaching ( FRAP ) , cells were transfected with GFP-actin construct and incubated for 24 hr . Prior to imaging , the cells were moved to fibronectin-coated ( 10 μg/ml ) glass-bottomed dishes ( MatTek Corporation ) and 3I Marianas imaging system ( 3I intelligent Imaging Innovations ) with 63x/1 . 2 water objective ( C-Apochromat Corr WD = 0 . 28 M27 ) was used . Five pre-bleach images were acquired before bleaching with 100% intensity of 488 ( 50 mW ) for 1 x 1 ms . First post-bleach images were acquired 10x every 500 ms and after that every 10 s . In laser ablation experiments , five pre-ablation images were acquired and bleaching was performed 10 s after ablation . The rate of vectorial actin polymerization at focal adhesions was determined by a blind analysis ( performed from randomly ordered samples by a different person to the one that carried out the experiments and prepared the kymographs ) by ImagePro Plus 6 . 0 software . Here , the speed of stress fiber elongation was quantified by measuring the advancement rate of proximal ends of the photobleached stress fiber regions form the kymographs prepared from the movies . Ablation of single ventral stress fibers was performed with 100% intensity of 405 nm laser ( 100mW , in 3I Marianas imaging system ) using 3 x 200 ms pulses . Five captures were taken before the ablation , after which retraction of the fibers was followed for 10 s before recording the changes in actin dynamics in relaxed fibers . Growth rate of actin filaments from the adhesions of ablated- or non-ablated ventral stress fibers were followed by either FRAP or photoactivatable-GFP-actin with 3I Marianas as explained above . It is important to note that experiments on cells co-expressing mCherry-Actin and GFP-Zyxin demonstrated that focal adhesions at the ends of ventral stress fibers are immobile after ablation ( see Video 7 ) . In few cases , laser ablation , however , led to retraction of the cell edge and accompanied disruption of focal adhesions . All such cases were discarded from further analysis . 10 . 7554/eLife . 06126 . 029Video 7 . Ablation of contractile ventral stress fibers does not affect the position of focal adhesions . An example of a laser-ablated ventral stress fiber from a U2OS cell co-expressing mCherry-Actin and GFP-Zyxin . Five captures , every 1 s , were taken before the ablation , and retraction of the ventral stress fiber was followed for 10 x 1 s before bleaching the region below the focal adhesion . Recovery of mCherry-Actin signal was detected at the bleached area of the relaxed fiber . Display rate is 10 frames/s , and the total duration of the video is 9 min . Please note , that position of the focal adhesion ( indicated by GFP-Zyxin ) is not significantly affected by ablation of the associated stress fiber . DOI: http://dx . doi . org/10 . 7554/eLife . 06126 . 029 Image Pro Plus 6 . 0 program was used for the quantifications of focal adhesion and stress fiber properties . Dorsal stress fiber lengths from fixed ctrl or ML-7- , Y27632- , KT5823- or Compound C-treated as well as Rif-TN-transfected cells were analysed from at least 12 cells and 3 fibers per each cell ( exact numbers for each experiment are indicated in the figure legends ) . Focal adhesion sizes and angles were quantified from frames of live cell imaging captures . Angles were calculated as the change between the major axis of the focal adhesion and the vertical . Intensity of phospho-VASP antibody stainings as well as co-localizations in the adhesion sites were analysed with line profiles . Intensities of phospho-VASP ( ser239 and thr 278 ) were divided with the intensity of total VASP antibody staining and values for ventral stress fiber adhesions were normalized to 1 . Distances of periodic spacing of the transverse arc structures were also measured with line profiles in Image Pro . Differences between groups were compared using the unpaired student t-test assuming unequal variances . All data were reported as mean +/- SEM or SD as indicated in the figure legends .	Muscle cells are the best-known example of a cell in the human body that can contract . These cells contain bundles of filaments made of proteins called actin and myosin , which can generate pulling forces . However , many other cells in the human body also rely on similar “contractile actomyosin bundles” to help them stick to each other , to maintain the correct shape or to migrate from one location to another . These bundles in the non-muscle cells are often called “ventral stress fibers” . Ventral stress fibers develop from structures commonly referred to as “arcs” . Previous work has clearly established that ventral stress fibers are sensitive to mechanical forces . However , the underlying mechanism behind this process was not known , and it remained unclear how external forces could promote these actomyosin bundles to assemble , align and mature . Tojkander et al . documented the formation of ventral stress fibers in migrating human cells grown in the laboratory . This revealed that pre-existing arcs fuse with each other to form thicker and more contractile actomyosin bundles . The formation of these bundles then pulls on the two ends of the stress fibers that are attached to sites on the edges of the cell . Tojkander et al . also showed that this tension inactivates a protein called VASP , which is also found at these sites . Inactivating VASP inhibits the construction of actin filaments , which in turn stops the stress fibers from elongating and allows them to contract . Further experiments then revealed that ventral stress fibers are maintained and can even become thicker under a sustained pulling force . Conversely , stress fibers that were not under tension were decorated by proteins that promote the disassembly of actin filaments . This subsequently led to the disappearance of these fibers . Future studies could now examine whether the newly identified pathway , which allows mechanical forces to control the assembly and alignment of stress fibers , is conserved in other cell-types . Furthermore , and because the assembly of such mechanosensitive actomyosin bundles is often defective in cancer cells , it will also be important to study this pathway’s significance in the context of cancer progression .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2015	Generation of contractile actomyosin bundles depends on mechanosensitive actin filament assembly and disassembly
Unlike dogs and cats , people do not point their ears as they focus attention on novel , salient , or task-relevant stimuli . Our species may nevertheless have retained a vestigial pinna-orienting system that has persisted as a 'neural fossil’ within in the brain for about 25 million years . Consistent with this hypothesis , we demonstrate that the direction of auditory attention is reflected in sustained electrical activity of muscles within the vestigial auriculomotor system . Surface electromyograms ( EMGs ) were taken from muscles that either move the pinna or alter its shape . To assess reflexive , stimulus-driven attention we presented novel sounds from speakers at four different lateral locations while the participants silently read a boring text in front of them . To test voluntary , goal-directed attention we instructed participants to listen to a short story coming from one of these speakers , while ignoring a competing story from the corresponding speaker on the opposite side . In both experiments , EMG recordings showed larger activity at the ear on the side of the attended stimulus , but with slightly different patterns . Upward movement ( perking ) differed according to the lateral focus of attention only during voluntary orienting; rearward folding of the pinna’s upper-lateral edge exhibited such differences only during reflexive orienting . The existence of a pinna-orienting system in humans , one that is experimentally accessible , offers opportunities for basic as well as applied science . Watching the ears allows an equestrian to gauge their mount’s shifting attention . Ear movements are not a useful cue in humans or apes because higher primates have lost the ability to orient by adjusting pinna shape and focal direction . Instead we judge a person’s attention by their gaze direction . In thousands of research reports each year , though , casual observation of ocular orienting is replaced by sophisticated recording techniques . We show in the present paper that similar electrical and optical techniques allow us to extract muscular correlates of pinna-orienting in our species and even render subtle pinna-orienting movements visible . Activation of the ear muscles is directionally specific and it occurs during voluntary as well as reflexive attention . A review of research in Hackley , 2015 on pinna-orienting in humans identified three relevant findings scattered across the preceding 100-or-so years . The first was Wilson’s oculo-auricular phenomenon ( Wilson , 1908 ) , in which shifting the gaze hard to one side elicits a 1 to 4 mm deflection of the lateral rim of both ears . The relevance to spatial attention is uncertain , though , with diverging results across studies , for example see Gerstle and Wilkinson , 1929; Urban et al . , 1993; O'Beirne and Patuzzi , 1999 . Additional evidence comes from a 1987 study ( Hackley et al . , 1987 ) of the bilateral postauricular muscle ( PAM ) reflex ( onset latency = 10 ms ) to acoustic onset transients . Increased amplitudes were observed when subjects directed their attention to a stream of tones on the same side as the recorded muscle while ignoring a competing , contralateral stream . Comparisons across left/right stimulus , attention , and PAM combinations localized modulation to the motor limb of the reflex arc . This pattern could indicate that the muscle behind an ear is primed when attention is directed toward that side . Finally , an experiment in Stekelenburg and van Boxtel , 2002 found that the automatic capture of attention by unexpected sounds coming from a speaker hidden to the left of the participant elicited greater activity in the left than right PAM . Apart from the research just described , functional studies of the human auriculomotor system have been mainly limited to the PAM reflex , in the context of audiometry or affective psychophysiology . The auriculomotor system lies essentially untouched in the literature . Here we present evidence that our brains retain vestigial circuitry for orienting the pinnae during both exogenous , stimulus-driven attention to brief , novel sounds and endogenous , goal-directed attention to sustained speech . We also demonstrate a complex interplay of different auricular muscles which may be causally linked to subtle movements of the pinnae . To examine automatic , stimulus-driven attention we used novel sounds similar to those in the Stekelenburg and van Boxtel , 2002 study , for example traffic jam , baby crying , footsteps . However , we presented them randomly from four different speakers ( at ± 30° , ± 120°; Figure 1 ) rather than just one , while the subject read a boring essay . As we were interested in the interactive role of distinct muscles in attempting to shape and point the pinnae , we recorded EMG from posterior , anterior , superior , and transverse auricular muscles ( PAM , AAM , SAM , and TAM ) . Visual evidence had previously been limited to still photos of Wilson’s oculo-auricular phenomenon ( Wilson , 1908 ) , so we supplemented our EMG data with videos from four high-definition cameras , see Methods and Video 1 , Video 2 , and Video 3 . To confirm that our findings would generalize to different age groups , older ( 62 . 7 ± 5 . 9 y ) as well as younger ( 24 . 1 ± 3 . 1 y ) adults were tested . The signal-averaged EMG waveforms of Figure 2 show well-defined responses with an onset latency of about 70 ms , responses that vary in amplitude , duration , and morphology according to the relative direction of the sound source . The inter- and intra-subject variability for the analyzed auricular muscles is shown in Figure 3 and Figure 4 , respectively . These plots portray the consistency of the PAM , AAM , and TAM responses across stimuli and subjects , especially for stimulation from the back . For the statistical analysis , mean amplitudes were subjected to a mixed , repeated-measures analysis of variance , with factors of age group , stimulus-muscle correspondence ( ipsi-/contralateral ) , and anterior/posterior stimulus direction . EMG amplitudes were larger for stimulus sources on the same side as the recorded ear for PAM , AAM , and TAM [F⁢ ( 1 , 26 ) = 47 . 44 , 17 . 01 , and 47 . 53 , respectively; p-values < 0 . 001; ηp2 = 0 . 65 , 0 . 40 , and 0 . 65] but not SAM . Responses were also larger to sounds emanating from the back than the front speakers for PAM , AAM , and TAM [F⁢ ( 1 , 26 ) = 32 . 1 , 12 . 0 , and 19 . 9 , respectively; p-values < 0 . 003 , ηp2 = 0 . 55 , 0 . 32 , and 0 . 43] . Posterior , ipsilateral stimulation elicited the most vigorous responses from these three muscles . In particular , the interaction between the factors ipsi/contralateral and anterior/posterior for PAM , AAM , and TAM yields F⁢ ( 1 , 26 ) = 40 . 4 , 14 . 9 , and 23 . 6 , respectively; p-values < 0 . 002; ηp2 = 0 . 61 , 0 . 36 , and 0 . 48 . There were no interactions involving age group , but a main effect indicated that older participants had smaller AAM responses [F⁢ ( 1 , 26 ) = 6 . 0 , p < 0 . 03 , ηp2 = 0 . 19] . These results support the hypothesis that the human brain retains circuits that attempt to point the ears in the direction of unexpected , potentially relevant sounds . The corresponding vestigial auriculomotor drive appears to be causally linked to very small ear displacements , see Figure 2—figure supplement 1 , Video 1 , and Video 2 . Having documented the existence of directionally-appropriate responses of the ear muscles to brief novel sounds , we turn now to a qualitatively distinct type of attention . To examine voluntary , goal-directed attention we used the classic , dichotic-listening paradigm , see Hillyard et al . , 1973; Cherry , 1953 . Two competing short stories were played either over the two front speakers or the two back speakers . To increase motivation participants were allowed to choose , after a brief introduction , which of the two stories ( podcasts ) they would like to listen to . They were then told which speaker that story would be presented from . Our subjects were instructed to listen carefully while looking at a fixation cross and , as in the immediately preceding study , holding their head still on a chin rest . Upon completion , a new story was picked and the listening direction was switched to one of the other speakers . Recording methods were identical to those of the exogenous experiment . Muscle activity was quantified as the mean of the absolute EMG energy over the entire course of each 5 min listening trial in Figure 5 and for consecutive segments of 10 s duration in a temporal analysis shown in Figure 6 . As in the exogenous study , EMG energy at PAM and AAM was largest on the side to which attention was focused [analysis corresponding to Figure 5: F ( 1 , 19 ) = 15 . 2 and 4 . 6 , respectively; p= 0 . 001 and 0 . 04; ηp2 = 0 . 44 and 0 . 20]; an effect that is particularly strong in narrow–band middle frequency components of the signal , see Figure 5—figure supplements 1 and 2 and the tables in Supplementary file 1 . A different pattern emerged for the other two muscles . Whereas TAM but not SAM activity had reflected lateralization of transient , exogenous attention , the reverse was true for sustained , goal-directed attention . That is to say , mean EMG energy at SAM was larger at the ipsi– than contralateral ear [F ( 1 , 19 ) = 16 . 3; p=0 . 001; ηp2 = 0 . 46] in Experiment 2 , but there was no such difference for TAM . Another main effect indicated that activation of all four muscles was generally enhanced when participants listened to one of the two speakers that were slightly behind as opposed to in front of them [PAM , AAM , TAM , SAM: F ( 1 , 19 ) = 5 . 7 , 3 . 1 , 8 . 1 , and 12 . 0 , respectively; p= 0 . 03 , 0 . 09 , 0 . 01 , and 0 . 003; ηp2 = 0 . 23 , 0 . 14 , 0 . 30 , and 0 . 39] . These effects did not interact with each other or with age . Although PAM activity declines over time , EMG energy of all three muscles is clearly sustained across the 5-min sessions , see Figure 6 . A corresponding sustained deflection of the pinna is also noticeable in the co-registered Video 3 . An alternative to the account we have been developing is that participants in Experiments 1 and 2 may have shifted their gaze toward the attended source . This would have then triggered Wilson , 1908 phenomenon , that is auriculomotor activity secondary to large gaze shifts . To test this hypothesis , we segmented the horizontal electrooculogram ( EOG ) in the same way as the auricular EMG . Voltages were converted to degrees of arc separately for each participant , based on findings from a cursor tracking protocol ( ± 35° ) . Figure 7 and Figure 8 document a complete absence of eye movements that were systematically related to attention direction . A limitation of these findings is that electro-oculographic recordings have a resolution of only 1 − 2° . However , gaze shifts less than 30° are rarely accompanied by auriculomotor activity , see Urban et al . , 1993 . Representative examples of macrosaccades during reading in Experiment one with co-registered auricular muscle activity can be found in Figure 7—figure supplement 1 . There is no obvious linkage of saccades and PAM responses . Note that the mean visual angle range observed in this example generalizes across subjects , see Figure 7—figure supplement 2 . Also when considering all the macrosaccades from all the subjects in Experiment 1 , our data do not exhibit a regularity between auditory stimuli and macrosaccades , see Figure 7—figure supplement 3 . Another line of evidence that the auricular responses observed in our study were not secondary to eye movements , concerns their pattern of lateralization . Activation of TAM during Wilson’s oculo-auricular phenomenon is more vigorous on the side opposite the direction of gaze , see Gerstle and Wilkinson , 1929; Urban et al . , 1993 . By contrast , we found in Experiment one that TAM activation was relatively enhanced at the ear on the same side as the attention-engaging sounds . The PAM component of Wilson's phenomenon does exhibit enhanced activity on the ipsilateral side , but this effect appears to be reliable only for gaze shifts greater than about 40 degrees ( Patuzzi and O'Beirne , 1999 , Figure 4 ) . Another alternative interpretation is that participants oriented not with their ears or eyes , but by lifting their chin from the chin rest and rotating their head toward the attended sound . If humans have a vestibulo-auricular response as do cats ( Tollin et al . , 2009 ) , such head rotations could have indirectly triggered activity in the ear muscles . However , recent research has shown that azimuthal head rotations have little effect on auricular activity in humans , see Cook and Patuzzi , 2014 . Moreover , analysis of sternocleidomastoid EMG in our data suggest that movements of the neck were rare , small , and unsystematic , see Figure 7—figure supplements 4 and 5 . An additional statistical analysis in Supplementary file 1 also rejects an influence of the sternocleidomastoid EMG and horizontal EOG . Finally , head rotations would have been too slow to generate the rapid responses of around 70 ms onset latency observed in the ear muscles in Experiment 1 . We note with interest , though , the possibility that subtle , covert activation of head turning muscles ( Corneil et al . , 2008 ) might be correlated with ocular and auricular orienting . Note that there was also no corresponding co–activation of the other measured ( non–auricular ) facial muscles , the zygomaticus and frontalis muscle , see Figure 7—figure supplements 6–9 for Experiment one and Figure 8—figure supplement 2 for Experiment 2 . The ability to swivel and point the pinnae seems to have been lost during the transition from the primarily nocturnal lifestyles of prosimians to the diurnal ones of New World monkeys , and then , Old World monkeys ( Coleman and Ross , 2004 ) . Mobility continued to decline as the ears became shorter and more rigid , see Waller et al . , 2008; Coleman and Ross , 2004 . The musculature degenerated . For example , an inferior auricular muscle to oppose SAM still exists in lesser apes such as gibbons and siamangs ( Burrows et al . , 2011 ) , but not in chimpanzees ( Burrows et al . , 2011 ) or humans ( Cattaneo and Pavesi , 2014 ) . Given that head rotation has little effect on PAM activity ( Cook and Patuzzi , 2014 ) , it seems likely that the vestibulo–auricular reflex as documented in cats ( Tollin et al . , 2009 ) has not been conserved in our species . Also presumably lost is the ability to use proprioceptive information to adjust auditory processing in accordance with pinna position , orientation , and shape as documented in cats , see Kanold and Young , 2001 . Although the ear muscles of Old World monkeys have spindles ( Lovell et al . , 1977 ) , those of humans do not ( Cattaneo and Pavesi , 2014 ) . When pinna-orienting movements became too small to modify acoustic input substantially , possibly 25 million years ago when lesser apes branched off from Old World monkeys , see Gibbs et al . , 2007 as discussed in Hackley , 2015 , selective environmental pressure ceased . The neural system became more-or-less 'frozen’ in a form optimized for controlling taller , more flexible ears , mounted on a smaller , more spherical head . This evolutionary perspective helps us to understand the surprising finding that AAM , which pulls the base of the pinna forward , was activated in Experiment one by novel sounds coming from the rear . Co-activation of opposing muscles AAM and PAM in our remote ancestors would have reduced occlusion of the ear canal by the tragus . Note that this occlusion occurs in monkeys when contraction of the PAM homolog is unopposed , see Waller et al . , 2008 , supplementary video clip 17 . In addition , PAM-AAM co-activation would have stabilized the base of the pinna and reduced myotendinous elasticity , thereby allowing quick changes in position or orientation . This perspective also illuminates our unexpected finding of ipsilateral SAM suppression in Experiment 1 . A study of pinna orienting in cats , whose tall ears resemble those of prosimians , showed that they tend to tilt the ear downward slightly when orienting to a lateral target , see Populin and Yin , 1998 . Neurobiologists have distinguished two types of pinna-orienting movements in cats , based on onset latency , see Siegmund and Santibáñez , 1982 . The short-latency response is specific to auditory stimuli and is chronometrically uncorrelated with saccades toward the target . By contrast , the long-latency response can be elicited by visual as well as auditory stimuli and it is roughly synchronous with ocular orienting , see Populin and Yin , 1998 . Using a 4-speaker set-up similar to that of the present Experiment 1 , Siegmund and Santibáñez , 1982 found cats’ unconditioned pinna responses to have an EMG onset latency that averaged 78 ms , similar to our value of about 70 ms . The animals were then trained to make gaze shifts toward the sound sources . Onset latency of the auriculomotor responses dropped to a remarkable 29 ms and the responses were resistant to extinction over the course of 125 trials . Both findings were replicated by Populin and Yin , 1998 ( mean = 26 ms; failure to extinguish across 10 , 000 unreinforced trials ) . The latter authors obtained an even more rapid response ( mean = 21 ms ) when the sound was preceded by a visual stimulus that served as warning signal and indicated that the cat should maintain gaze at a fixation point . They argued that the short-latency pinna response is too rapid to be mediated by the brain region most centrally involved in orienting , the superior colliculus ( SC ) . This is because an earlier study , Populin and Yin , 1997 , had found the average first-spike latency in the relevant portion of this structure to be 19 ms . Comparisons with these cat studies suggest that our participants' auriculomotor responses may have been primarily also of the short-latency variety that is not mediated by the SC . Two of the conditions tested by Populin and Yin , 1998 involved brief , lateralized auditory stimuli that , as in the present Experiment 1 , were not task-relevant . The stimuli elicited short-latency ipsilateral pinna movements that were temporally uncorrelated with gaze shifts ( see their Figures 6 and 8 ) . During the delayed-saccade condition of their study , laterally presented sounds were task-relevant and forward fixation was required , as in our Experiment 2 . Ipsilateral pinna movements triggered by onset of these sounds were of the short-latency variety ( 21 ms , as noted above ) . Subsequent , smaller movements were then observed in synchrony with ocular orienting , roughly 400 ms after the fixation point was extinguished ( Figure 7 ) . It is long-latency pinna movements of this sort that Populin and Yin , 1998 suggested might be mediated by the SC . Given the major role of the SC in controlling eye fixation ( Krauzlis et al . , 2017 ) , this structure may also be responsible for sustained maintenance of pinna orientation , such as in Experiment 2 . Pinna movements can be triggered by electrical stimulation of the deep and intermediate layers of the SC in accordance with a topographical pattern that is in register with that of eye movements ( Stein and Clamann , 1981 ) . Lesions of this structure reduce the likelihood of pinna orienting as well as its accuracy , see Czihak et al . , 1983 . Although monosynaptic connections from SC to auriculomotor neurons in the facial nucleus do exist ( Vidal et al . , 1988 ) , pinna control is dominated by disynaptic pathways from the SC that include the paralemniscal , oculomotor , or pontine reticular zones , see Henkel and Edwards , 1978; Takeuchi et al . , 1979; Vidal et al . , 1988 . Among these , the paralemniscal zone appears to be the most important , and its auditory input originates in the nearby nucleus sagulum , see Henkel , 1981 . Portions of neocortex also play a role in controlling pinna movements . Lesions of auditory cortex reduce the kinematic complexity of pinna orienting and slow its habituation , see Alvarado and Santibañez , 1971 . Stimulation and recording studies in the macaque have identified a premotor ear-eye field ( area 8B ) , which is connected with both auditory cortical areas and the SC , see Lanzilotto et al . , 2013 . These animal neuroanatomy and physiology studies , coupled with Wilson , 1908 seminal report , make it clear that the eyes and pinnae work together during endogenously cued attentional orienting . A recent study in humans by Gruters et al . , 2018 , showed a close relationship between movements of the left and right eardrums and multiple parameters of task-related left- and right-directed eye movements . It will be important in future research to test whether muscular responses of the middle and outer ears are linked in a coordinated manner to ocular orienting . Furthermore , exploration of the relationship between human auriculomotor activity and subtle markers of covert attention ( see the recent review given in van Ede et al . , 2019 ) corticofugal modulation of ascending auditory pathways ( Perrot et al . , 2006 ) in endogenous attention , and the neural mechanisms of orienting discussed in the preceding section has scarcely begun . Our results have implications for applied science , as well . They suggest that patterns of auricular muscle activity might serve as an easily accessible correlate of top-down processing in endogenous modes of attention . As such , the described effects might complement electroencephalographic indices of attentional focus ( de Cheveigné et al . , 2018; Schäfer et al . , 2018 ) in that their sensitivity is exclusively spatial , rather than reflecting a context-specific mixture of modality , feature , location , and object representations . Registration of pinna-orienting might better support near real-time decoding of the attentional focus and , as compared to EEG-based stimulus reconstruction approaches , does not require the exogenous sound source , for example see the discussion in Schäfer et al . , 2018 . Thus , auricular muscle monitoring might support the decoding of auditory attention in technical applications such as attentionally controlled hearing aids that preferentially amplify sounds the user is attempting to listen to . We wish to underscore , though , that the development of such applications would benefit crucially from a better understanding of how the auditory and visual attention systems interact . We hope that the results presented here will stimulate research in this direction . Both older ( N = 12 , mean age = 62 . 7 ± 5 . 9 y , 8 F , all right-handed ) and younger adult ( N = 16 , mean age = 24 . 1 ± 3 . 1 y , 8 F , 15 right-handed , one left-handed ) volunteers in Experiment one had age-typical , pure tone audiometric thresholds ( 1 , 2 , 4 , and 8 kHz; young < 20 dB; old < 40 dB ) . All served in both studies , but after the 8th participant , Experiment two was altered in several ways ( e . g . , four stimulus directions rather than two ) . Only data from the final 21 subjects were retained for Experiment 2 . The two groups in this experiment comprised 11 older adults ( mean age = 62 . 6 ± 6 . 2 y , 8 F , all right-handed ) and 10 younger adult ( mean age = 24 . 1 ± 3 . 6 y , 5 F , nine right-handed , one left-handed ) . After a detailed explanation of the procedure , all subjects signed a consent form . The study was approved by the responsible ethics committee ( ethics commission at the Ärztekammer des Saarlandes , Saarbrücken , Germany; Identification Number: 79/16 ) . The four active loudspeakers ( KH120A , Neumann , Germany ) were positioned at head level , 115 cm . Sounds in Experiment 1 and 2 were reproduced with a soundcard ( Scarlett 18i20 , Focusrite , UK ) . The experimental paradigms were programmed using software for scientific computing ( Matlab , Mathworks , USA ) and Psychtoolbox 3 . In Experiment 1 , sounds lasted 1 . 7 – 10 . 0 s , were delivered every 15 – 40 s , and had an average intensity of 70 dBC , except for foot steps ( 65 dBC ) . Each of the nine stimuli ( lemur howling , dog barking , helicopter flying , cell phone vibrating , birds singing , baby crying , mosquito buzzing , footsteps , and traffic jam ) was repeated four times ( i . e . , once per speaker ) . In Experiment 2 , the stories were 5 min long , with an average intensity of 50 dBA for younger and 60 dBA for older participants . Participants answered content questions at the conclusion of each condition in this experiment . Surface EMGs were recorded with non-recessed , Ag/AgCl electrodes ( BME4 , BioMed Electrodes , USA ) , which were 4 mm in diameter for TAM and 6 mm ( BME6 ) in all other cases , see Figure 1 . The signals were AD-converted at 9600 Hz and 24 bit resolution per channel ( 4 × USBamp , g . tec GmbH , Austria ) . Skin temperature , skin resistance , electrocardiograms , and EOGs were also recorded . All signal processing algorithms were implemented using the scientific computing software Matlab ( Mathworks , USA , Version: 2018a ) . Because surface electrodes had not previously been used to record from intrinsic ear muscles , we conducted preliminary tests with a participant who exhibited a large , reliable Wilson’s phenomenon and who could voluntarily contract her SAM and PAM . Isolation of the corresponding responses indicated that EMG from TAM electrodes was not an artifact of volume conduction from PAM or SAM . In other words , the TAM activity was not correlated with forced SAM/PAM innervation . Sternocleidomastoid EMG signals were zero-phase bandpass filtered from 60 to 1000 Hz ( FIR , 2000th order ) , the auricular EMG signals from 10 to 1000 Hz ( FIR , 2000th order ) with a notch filter at 50 Hz ( IIR , 2nd order ) . Horizontal EOG signals were zero-phase filtered from 0 . 01 to 20 Hz ( IIR , 2nd order ) . All filter operations were performed using Matlab’s filtfilt-function for zero-phase filtering . The filtered signals were then downsampled to 2400 Hz for further processing . The statistical analysis was performed using repeated measures ANOVA ( with IBM SPSS Statistics 26 ) . Within-subjects factors were stimulus-muscle correspondence ( ipsi- vs . contralateral responses ) and anteriority ( front vs . back speakers ) . The only between-subjects factor our statistical model accounted for was age and , in association with that , also stimulus level in the endogenous experiment . Other factors like head-size , audiogram shape or small electrode placement differences were not included in the model . All main and interaction effects were tested . Videos were acquired using four Ximea MQ022CG–CM color sensors with a resolution of 1936 × 1216 at 120 frames per second and an exposure of 2 ms . Two cameras were positioned on each side of the head and focused on the ears to record pairwise stereo videos . We used hardware triggering for all four cameras and recorded each camera onto a separate m . 2 solid-state-drive to reduce frame loss . We used a KOWA 35 mm macro lens with an aperture of F0 . 4 which gave us a close-up view of the ear with acceptable depth of field to allow slight movements towards the camera and enough distance such that the cameras did not cast shadows on to the scene . We illuminated the face uniformly with flicker-free LED studio illumination . The cameras were calibrated with the stereo camera calibrator app from the Mathworks Matlab Computer Vision System Toolbox . Calibration was performed whenever camera adjustment required re-alignment of relative stereo camera positions or a change of focus of one of the cameras . For 3D reconstruction and motion visualization/quantification , we used functions from the Mathworks Matlab Computer Vision System Toolbox and custom written code . Our analysis system was able to reduce redundancies in optic flow and stereo depth estimation by exploiting the unilateral scene composition and limited degrees of freedom for ear and head movements . For 3D reconstructions , we initialized a sequence with one initial estimation of disparity and subsequently tracked points independently for the left and right image sequence . We tracked points with respect to the first frame of the sequence as reference frame with dense optical flow initialized with a rigid motion estimation . Motion was visualized with a Lagrangian motion magnification approach that had a constant magnification factor with respect to the reference frame and prior removal of affine motion with respect to manually selected stable points . The results of the motion analysis with and without magnification can be seen in the videos .	Dogs , cats , monkeys and other animals perk their ears in the direction of sounds they are interested in . Humans and their closest ape relatives , however , appear to have lost this ability . Some humans are able to wiggle their ears , suggesting that some of the brain circuits and muscles that allow automatic ear movements towards sounds are still present . This may be a ‘vestigial feature’ , an ability that is maintained even though it no longer serves its original purpose . Now , Strauss et al . show that vestigial movements of muscles around the ear indicate the direction of sounds a person is paying attention to . In the experiments , human volunteers tried to read a boring text while surprising sounds like a traffic jam , a baby crying , or footsteps played . During this exercise , Strauss et al . recorded the electrical activity in the muscles of their ears to see if they moved in response to the direction the sound came from . In a second set of experiments , the same electrical recordings were made as participants listened to a podcast while a second podcast was playing from a different direction . The individuals’ ears were also recorded using high resolution video . Both sets of experiments revealed tiny involuntary movements in muscles surrounding the ear closest to the direction of a sound the person is listening to . When the participants tried to listen to one podcast and tune out another , they also made ear ‘perking’ movements in the direction of their preferred podcast . The results suggest that movements of the vestigial muscles in the human ear indicate the direction of sounds a person is paying attention to . These tiny movements could be used to develop better hearing aids that sense the electrical activity in the ear muscles and amplify sounds the person is trying to focus on , while minimizing other sounds .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2020	Vestigial auriculomotor activity indicates the direction of auditory attention in humans
Recognition and elimination of tumor cells by the immune system is crucial for limiting tumor growth . Natural killer ( NK ) cells become activated when the receptor NKG2D is engaged by ligands that are frequently upregulated in primary tumors and on cancer cell lines . However , the molecular mechanisms driving NKG2D ligand expression on tumor cells are not well defined . Using a forward genetic screen in a tumor-derived human cell line , we identified several novel factors supporting expression of the NKG2D ligand ULBP1 . Our results show stepwise contributions of independent pathways working at multiple stages of ULBP1 biogenesis . Deeper investigation of selected hits from the screen showed that the transcription factor ATF4 drives ULBP1 gene expression in cancer cell lines , while the RNA-binding protein RBM4 supports ULBP1 expression by suppressing a novel alternatively spliced isoform of ULBP1 mRNA . These findings offer insight into the stress pathways that alert the immune system to danger . Natural killer ( NK ) cells are lymphocytes of the innate immune system that play a critical role in limiting tumor growth ( Vivier et al . , 2011; Marcus et al . , 2014; Mittal et al . , 2014 ) . NK cell activation is controlled by a balance of signals from activating and inhibitory receptors , which recognize cognate ligands expressed by potential target cells ( Vivier et al . , 2011; Shifrin et al . , 2014 ) . One of the best-studied NK-activating receptors is NKG2D , which is also expressed on certain subsets of T cells ( Raulet , 2003 ) . Engagement of NKG2D by its ligands displayed on a target cell membrane leads to NK cell activation , cytokine secretion , and lysis of target cells , such as tumor cells . NKG2D recognizes a family of ligands that are structurally similar to MHC Class I proteins . Humans express up to eight NKG2D ligands ( ULBP1-6 , MICA , and MICB ) , and mice express 5–6 different ligands , depending on the strain ( RAE-1α-ε , H60a-c , and MULT1 ) ( Raulet et al . , 2013 ) . Healthy cells typically do not display NKG2D ligands on their surface and are thus poor targets for NKG2D-mediated lysis by NK cells . However , cellular stresses associated with transformation , viral infection , or other danger to the host cause the upregulation of NKG2D ligand expression ( Raulet et al . , 2013 ) . Primary tumors and cancer cell lines frequently express one or more NKG2D ligand , and NKG2D expression is important for the control of tumors in vivo in models of spontaneous cancer ( Guerra et al . , 2008 ) . Tumors arise despite the tumor-suppressive effects of the immune system , and some tumors show evidence of adaptation to escape immune control ( Schreiber et al . , 2011 ) . In the case of NKG2D-mediated tumor recognition , published results suggest that one mechanism of tumor immune evasion is the loss or decreased expression of NKG2D ligands ( Guerra et al . , 2008; McGilvray et al . , 2009 ) . In other cases , tumors progress despite sustained expression of NKG2D ligands ( Vetter et al . , 2002; Guerra et al . , 2008; McGilvray et al . , 2009; Hilpert et al . , 2012 ) . The evidence as a whole suggests that upregulation of NKG2D ligands on early stage tumor cells is part of a host defense mechanism , but that the immune response subsequently applies selective pressure for tumors that have either extinguished expression of NKG2D ligands or have activated immune suppressive mechanisms ( Raulet and Guerra , 2009 ) . Therefore , identifying factors and pathways that regulate NKG2D ligands will improve our understanding of the cellular properties used by the immune system to define unwanted cells and will also help reveal how tumors evade the corresponding immune responses . Prior investigations have identified regulators of NKG2D ligands using a candidate approach based on the roles of these regulators in known stress pathways . Such approaches have implicated the DNA damage response pathway ( Gasser et al . , 2005 ) , heat shock ( Venkataraman et al . , 2007; Nice et al . , 2009 ) , hyperproliferation ( Jung et al . , 2012 ) , and pattern recognition receptors ( Hamerman et al . , 2004 ) , among others , in the regulation of one or more NKG2D ligands . However , these pathways do not account fully for expression of ligands in tumor cells , since inhibiting them may decrease ligand expression but typically does not abrogate it . For example , the DNA damage response is active in many cancer cells and tumor cell lines , but inhibiting that pathway only partially inhibits ligand expression ( Gasser et al . , 2005; Gasser and Raulet , 2006; Soriani et al . , 2014 ) . Similarly , hyperproliferation can drive NKG2D ligand expression , but blocking proliferation does not completely eliminate NKG2D ligand expression by tumor cell lines ( Jung et al . , 2012 ) . These findings suggest that unidentified molecular cues in tumor cells also initiate the expression of NKG2D ligands , allowing potentially dangerous tumor cells to be distinguished from normal cells . Identifying those cues , especially for human NKG2D ligands , is important for understanding the biological regulation of NKG2D ligands and devising approaches for immunotherapy based on that knowledge . To identify novel drivers of NKG2D ligand expression , we performed a genome-wide loss-of-expression mutant screen in the tumor-derived human cell line HAP1 ( Carette et al . , 2009; Carette et al . , 2011 ) and used CRISPR/Cas9 gene targeting methodology for confirmation of the hits and extension of the results . The results reveal previously unknown regulators for NKG2D ligands , provide evidence for selectivity of the regulators for specific ligands , and support the cooperation of different stress pathways in the regulation of one such ligand . Many tumors and cancer cell lines express multiple NKG2D ligands , possibly due to ongoing stress responses associated with the transformed state ( Raulet et al . , 2013 ) . To identify novel drivers of human NKG2D ligand expression in transformed cells , we employed a retroviral gene-trap mutagenesis screen using the near-haploid human cell line HAP1 ( Figure 1 ) ( Carette et al . , 2009; Carette et al . , 2011 ) . Like many cell lines , HAP1 cells express multiple NKG2D ligands ( Figure 1—figure supplement 1 ) . We chose to screen for drivers of ULBP1 expression because it showed the brightest staining on HAP1 cells , making it particularly amenable to our loss-of-expression screen . Following mutagenesis , we selected for mutants with decreased expression of ULBP1 but intact expression of the unrelated GPI-anchored protein CD55 ( Figure 1A ) . Selection of CD55+ cells was used to reduce the fraction of selected cells that had lost ULBP1 expression due to mutations that alter cell surface expression of all proteins or of all GPI-linked proteins . In the first round of selection , we depleted ULBP1high cells from the mutant cell population using magnetic bead-based depletion of cells labeled with a ULBP1 antibody . After briefly expanding the selected cells , we used flow cytometry to further select for ULBP1lowCD55+ cells . Figure 1B shows ULBP1 and CD55 expression on WT and post-selection HAP1 cells . 10 . 7554/eLife . 08474 . 003Figure 1 . A genetic screen of a haploid human cell line to identify regulators of ULBP1 expression . ( A ) HAP1 cells ( ∼108 cells ) were transduced with a retroviral gene-trap vector . To enrich for mutant cells with decreased ULBP1 expression , we initially depleted ULBP1high cells by labeling cells with an anti-ULBP1 antibody followed by magnetic bead-based cell depletion . Following a brief recovery and expansion of the cells , we used FACS to further enrich ULBP1lowCD55+ cells . Deep-sequencing of genomic DNA from pre- and post-selection cells was used to map sites of gene-trap insertions , and mutations enriched in ULBP1low cells were identified . ( B ) Flow cytometric analysis of WT and post-selection HAP1 cells . Cells were stained for ULBP1 and CD55 , an irrelevant GPI-linked protein . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 00310 . 7554/eLife . 08474 . 004Figure 1—figure supplement 1 . Expression of NKG2D ligands on HAP1 cells . HAP1 cells were stained with antibodies against NKG2D ligands and analyzed by flow cytometry . Isotype control staining is shown by the shaded gray histogram . The ULBP2/5/6 antibody cross-reacts with three NKG2D ligands , but only ULBP2 is substantially expressed by HAP1 cells . The data shown are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 004 We employed deep-sequencing to map and quantify the frequencies of independent insertion sites of the retroviral gene-trap in selected cells and compared this with the landscape of insertions in unselected control cells . Table 1 shows a selected ‘hit list’ of genes that were targeted significantly more frequently in selected ULBP1lowCD55+ cells than in unselected cells . The ULBP1 gene itself was a highly significant hit , providing a validation of this approach . Many genes encoding enzymes involved in GPI synthesis were also represented despite the selection for CD55 expression; many of these were removed from Table 1 , for simplicity . The complete list of hits ( p < 0 . 05 ) is shown in Supplementary file 1 , along with the analysis of all independent insertions mapped in the selected data set . Raw sequencing data for the screen are available under NCBI Bioproject PRJNA284536 , containing the datasets for HAP1 gene trap control cells ( Accession number SAMN03703230 ) and cells from the ULBP1 screen ( Accession number SAMN03703231 ) . We chose hits for validation and follow-up experiments based on their statistical ranking and expectations that the corresponding proteins play roles in stress responses , protein biogenesis , or gene/mRNA regulation . The genes chosen encode ATF4 ( a stress-associated transcription factor ) , RBM4 ( an RNA-binding protein ) , HSPA13 ( a protein chaperone ) , and SPCS1 and SPCS2 , which are both non-catalytic subunits of the signal peptidase complex . 10 . 7554/eLife . 08474 . 005Table 1 . Selected list of genes enriched for gene-trap insertions after selection of ULBP1lowCD55+ cellsDOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 005Gene symbolFunction/Processp-valuePIGWGPI synthesis/anchoring1 . 12E-196PIGQGPI synthesis/anchoring3 . 26E-155PIGBGPI synthesis/anchoring2 . 38E-103ULBP1NKG2D ligand2 . 65E-76PIGOGPI synthesis/anchoring1 . 49E-65RBM4RNA-binding protein1 . 29E-24PIGVGPI synthesis/anchoring4 . 43E-23SPCS1Non-catalytic subunit of signal peptidase complex1 . 25E-15PIGMGPI synthesis/anchoring6 . 06E-14C1GALT1C1Protein O-linked glycosylation2 . 22E-13SLC35A1Golgi-localized CMP-sialic acid transporter1 . 77E-12ST3GAL2Sialyltransferase1 . 64E-11SPCS2Non-catalytic subunit of signal peptidase complex6 . 20E-09HSPA13Microsome-associated protein with ATPase activity1 . 24E-05FLJ37453Non-coding RNA0 . 00115SLC17A9Vesicular nucleotide transporter0 . 00715RPS25Ribosomal protein0 . 00715ATF4Stress-induced transcription factor0 . 0206PMM2Oligosaccharide synthesis , protein glycosylation0 . 0234NCRNA00167Non-coding RNA0 . 0363CRNKL1Pre-mRNA splicing0 . 0363ICKIntestinal cell kinase , MAPK-related0 . 0363TBC1D19TBC domain-containing protein0 . 0416ZNF236Zinc-finger protein0 . 0496The gene symbols of hits ( p < 0 . 05 ) are shown with a brief description of known or predicted gene functions . A p-value of enrichment was determined using Fisher's exact test , followed by correction for the false discovery rate . The list was manually curated to remove known genes that have occurred in several unrelated screens using the same cells , perhaps indicating pleiotropic effects . For simplicity , a number of genes related to GPI biosynthesis and anchoring were removed . Bold text indicates genes confirmed in this study to impact ULBP1 expression . Blue text indicates genes involved in GPI biosynthesis and anchoring . Red text indicates genes involved in protein glycosylation . Asterisks indicate two genes ( SLC17A9 and CRNKL1 ) that , when targeted with CRISPR/Cas9 , failed to result in decreased ULBP1 expression . To confirm that selected genes from the screen regulate ULBP1 , we employed the CRISPR/Cas9 mutagenesis system , targeting sites in the 5′ coding regions of each candidate gene in HAP1 cells ( Jinek et al . , 2013; Mali et al . , 2013 ) . The ULBP1 gene was targeted for comparison . After HAP1 cells were transiently transfected with plasmids encoding Cas9 and guide RNAs ( sgRNAs ) for each candidate gene , a population of ULBP1low cells appeared that was absent in control transfected cells ( Figure 2—figure supplement 1 ) . In each case , individual ULBP1low cells were sorted into 96-well plates , and expanded clones were screened for mutations by PCR and sequencing . For further analysis , we selected clones with insertions or deletions that resulted in frameshift mutations in each targeted gene ( Figure 2—figure supplement 2 ) . Since the sites targeted were near the beginning of each coding region and the cells are haploid , the frameshift mutations are expected to result in complete loss-of-function of the corresponding proteins . Analysis of HAP1 cell lysates by Western blot confirmed the loss-of-protein expression in ATF4 , RBM4 , and SPCS2 mutant cell lines ( data not shown ) . Cells with a ULBP1 mutation lacked cell surface ULBP1 staining altogether , as expected , whereas the other mutations analyzed resulted in a partial ( twofold to threefold ) decrease in cell surface expression of ULBP1 ( Figure 2A ) . The effect of each mutation was specific to ULBP1 , as we found no change in cell surface expression of other proteins , including four other NKG2D ligands ( ULBP2 , ULBP3 , MICA , and MICB ) , HLA Class I , the unrelated GPI-anchored protein CD59 , or PVR and Nectin-2 , the ligands for DNAM-1 , another NK cell-activating receptor ( Figure 2B , C , Figure 2—figure supplement 3 ) . The minor changes in ULBP3 staining seen in Figure 2—figure supplement 3B were not consistently observed across experiments . The finding that the mutations each affect only ULBP1 among the NKG2D ligands tested supports the hypothesis that different NKG2D ligands are subject to distinct regulatory processes . It was surprising that SPCS1 and SPCS2 mutations only impacted cell surface staining of ULBP1 and not the six other membrane proteins tested , as we had expected that mutating components of the signal peptidase complex would cause a more generalized defect in cell surface protein expression ( see ‘Discussion’ ) . In all cases , ULBP1 expression on mutant lines could be restored by re-expressing the gene of interest with a doxycycline-inducible lentiviral vector ( Figure 2D ) . These findings established that ATF4 , RBM4 , HSPA13 , SPCS1 , and SPCS2 each contribute partially to cell surface display of ULBP1 in HAP1 cells in steady-state culture conditions . 10 . 7554/eLife . 08474 . 006Figure 2 . Decreased ULBP1 expression upon targeted mutation of screen hits . ( A–C ) Flow cytometric analysis of cell surface expression of ULBP1 ( A ) , the NKG2D ligand MICA ( B ) , or pan-HLA Class I ( C ) on WT and mutant HAP1 cells . WT and mutant ( KO ) cells are represented as black and red histograms , respectively . The shaded gray histogram represents isotype control staining . The blue trace in panel A shows staining of ULBP1 KO HAP1 cells and matches isotype control staining . Data are representative of at least three independent experiments . ( D ) To restore expression of ULBP1 drivers , mutant cell lines were transduced with a doxycycline-inducible lentiviral vector containing the gene of interest . Cells were treated for 24 hr with doxycycline ( Dox ) at a final concentration of 100 ng/ml for ATF4 and 1000 ng/ml for all other genes . After treatment , cells were analyzed by flow cytometry . Black histograms: WT cells transduced with control vector , +Dox . Red histograms: mutant cells transduced with Dox-inducible gene of interest , −Dox . Blue histograms: mutant cells transduced with Dox-inducible gene of interest , +Dox . The shaded gray histogram represents isotype control staining . Data are representative of three independent experiments . ( E ) RT-qPCR analysis of canonically spliced ULBP1 mRNA expression levels in WT and mutant HAP1 cells . Expression levels were normalized to ACTB , GAPDH , and HPRT1 and are shown as mean ±SE . The data were analyzed by 1-way ANOVA with Dunnet's multiple comparisons test comparing all samples to WT . **p < 0 . 001 . Data are representative of three independent experiments . In one out of three total experiments performed , the level of ULBP1 mRNA was significantly increased in the SPCS2 mutant compared to WT . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 00610 . 7554/eLife . 08474 . 007Figure 2—figure supplement 1 . Mutagenesis of screen 'hits' with the CRISPR/Cas9 system . HAP1 cells were co-transfected with plasmids encoding Cas9 , a target gene-specific sgRNA , and GFP . Successfully transfected ( GFP+ ) cells were sorted 24–72 hr post-transfection and re-plated . 5–7 days post-transfection , cells were stained for ULBP1 expression and analyzed by flow cytometry . Examples are shown for ULBP1 ( A ) , RBM4 and ATF4 ( B ) , and SPCS1 ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 00710 . 7554/eLife . 08474 . 008Figure 2—figure supplement 2 . Sequences of CRISPR/Cas9 target sites in HAP1 cells . Genomic DNA sequence surrounding Cas9:gRNA target loci as shown with protospacer adjacent motif ( PAM ) in red . Inserted bases are in bold blue text . For HSPA13 and SPCS1 , the PAM was on the non-coding strand . For ATF4 , mutation ( a ) was present in the ATF4 single-KO line . Mutation ( b ) was present in all double- and triple-KO lines . For SPCS2 , mutation ( c ) was present in the ATF4/RBM4/SPCS2 triple-KO line . Mutation ( d ) was present in the SPCS2 single-KO line . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 00810 . 7554/eLife . 08474 . 009Figure 2—figure supplement 3 . Expression of additional NK cell ligands is unchanged on mutant HAP1 cells . Flow cytometric analysis of cell surface expression of ULBP2/5/6 ( A ) , ULBP3 ( B ) , MICB ( C ) , CD112 ( D ) , or CD155 ( E ) on WT and mutant HAP1 cells . WT and mutant cells are represented as black and red histograms , respectively . The shaded gray histogram represents isotype control staining . In HAP1 cells , staining with the ULBP2/5/6 antibody is due almost exclusively to expression of ULBP2 . Data are representative of three independent experiments for panels A–C . Data are representative of two independent experiments for panels D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 009 Consistent with their presumptive roles in gene expression and/or splicing , ATF4 and RBM4 KO cells each showed decreased amounts of canonically spliced ULBP1 mRNA commensurate with the change in cell surface expression ( Figure 2E ) . Consistent with their roles as chaperones and protein-processing components , HSPA13 , SPCS1 , and SPCS2 KO cells all showed WT amounts of ULBP1 mRNA . To assess whether the genes implicated in ULBP1 expression work independently or in common pathways , we generated and analyzed double and triple-mutant cell lines . We generated ATF4/RBM4 double-mutant cells by targeting ATF4 in the RBM4 KO line , and triple-mutant cells by further targeting either HSPA13 or SPCS2 in the double-mutant line . Notably , we observed stepwise decreases in ULBP1 expression with each additional mutation , with triple-mutant cells showing up to a 20-fold reduction in ULBP1 expression compared to WT cells ( Figure 3A ) . These data suggested that the genes tested ( with the likely exception of SPCS1 vs SPCS2 ) contribute largely independently to steady-state ULBP1 expression in HAP1 cells and support a model in which constitutive NKG2D ligand expression in cell lines is due , at least in some instances , to the contribution of several pathways that act cumulatively . ATF4/RBM4 double mutants showed a larger decrease in canonically spliced ULBP1 mRNA than either single mutant , suggesting that ATF4 and RBM4 may act independently in regulating the amounts of ULBP1 mRNA ( Figure 3B ) . There were no greater decreases in ULBP1 mRNAs when mutations in either HSPA13 or SPCS2 were added to the double-mutant cells , as expected if they act co-translationally or post-translationally . 10 . 7554/eLife . 08474 . 010Figure 3 . Double and triple-mutant cell lines show stepwise decreases in ULBP1 expression . ( A ) Flow cytometric analysis of ULBP1 expression on single , double , and triple-mutant HAP1 cells . ATF4−/RBM4− double-mutant cells were generated by mutagenesis of ATF4 in RBM4− cells . Triple-mutant cells were generated by mutagenesis of HSPA13 or SPCS2 in ATF4−/RBM4− double-mutant cells . Shaded gray histograms represent isotype control staining . Data are representative of three independent experiments . ( B ) RT-qPCR analysis of canonically spliced ULBP1 mRNA expression levels in the cells described in ( A ) . Expression levels were normalized to ACTB , GAPDH , and HPRT1 and are shown as mean ±SE . The data were analyzed by 1-way ANOVA with Bonferroni's multiple comparisons test and substantive significant differences are shown . Data are representative of three independent experiments . p < 0 . 05 , p < 0 . 01 , p < 0 . 001 , n . s . : not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 010 Having validated several hits from our screen , we investigated the roles of ATF4 and RBM4 in regulating ULBP1 expression . To address whether ATF4 regulates ULBP1 in other cell types , we mutated ATF4 in the K-562 chronic myelogenous leukemia cell line and the Jurkat acute T-cell leukemia cell line . ATF4 KO K-562 and Jurkat cells were generated using Cas9 and a pair of sgRNAs that flank the ATF4 gene . The entire ATF4 gene was deleted in the resulting mutant lines , eliminating the possibility of the mutant cells expressing a functional ATF4 protein fragment . HAP1 cells carrying a similar deletion of ATF4 had an identical phenotype to the ATF4 frameshift mutant line described above ( data not shown ) . WT K-562 cells had relatively low ULBP1 expression , and mutation of ATF4 resulted in the complete disappearance of ULBP1 from the cell surface , as well as an 11-fold reduction in ULBP1 mRNA ( Figure 4A , B ) . In contrast , ATF4-deficiency in Jurkat cells resulted in a modest reduction in ULBP1 mRNA and a barely detectable reduction in ULBP1 cell surface staining . By comparison , HAP1 cells showed intermediate effects of ATF4-deficiency . The results suggest that ATF4 drives basal ULBP1 expression in multiple tumor cell lines , perhaps reflecting constitutive activation of underlying stress pathways . 10 . 7554/eLife . 08474 . 011Figure 4 . ATF4 drives basal ULBP1 expression in multiple cell lines . ( A ) Flow cytometric analysis of ULBP1 expression on WT versus ATF4 KO variants of K-562 , HAP1 , and Jurkat cells . Data are representative of three independent experiments . ( B ) RT-qPCR analysis of canonically spliced ULBP1 mRNA expression levels in the cells described in ( A ) . Expression levels were normalized to ACTB , GAPDH , and HPRT1 and are shown as mean ±SE . Expression in WT cells was set to ‘1 . 0’ for each cell type; the different cell types are not comparable in this experiment . The data were analyzed by 2-way ANOVA with Bonferroni's multiple comparisons test . Data are representative of three independent experiments , though one of the three Jurkat analyses did not show a significant difference . p < 0 . 05 , **p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 011 ATF4 is induced in a variety of stress conditions that arise in unhealthy , infected , and transformed cells; coupling these stress conditions to NKG2D ligand expression may enable destruction of undesirable cells . Such stresses , including amino acid starvation , the unfolded protein response ( UPR ) , oxidative stress , and the presence of dsRNA , induce ATF4 through phosphorylation of the translation initiation factor eIF2α ( Rutkowski and Kaufman , 2003; Suragani et al . , 2012 ) . Phosphorylation of eIF2α globally suppresses protein translation but selectively activates the translation of ATF4 ( Harding et al . , 2000 ) . The transcriptional targets of ATF4 include amino acid transporters and protein chaperones , which in combination with the overall reduction in protein synthesis help mitigate the cellular stress ( Harding et al . , 2003 ) . However , ATF4 expression also drives the expression of the pro-apoptotic transcription factor CHOP ( GADD153 ) , suggesting that ATF4 not only drives adaptation to stress but also promotes cell death if the stress cannot be overcome ( Tabas and Ron , 2011; Han et al . , 2013 ) . Increased ATF4 expression has been detected in some tumors ( Bi et al . , 2005 ) , providing a possible mechanism for coupling malignant transformation to expression of ULBP1 , and consequently to tumor immunosurveillance . To impart stress known to induce ATF4 expression , we treated cells with the drug histidinol ( HisOH ) , which competitively inhibits histidine–tRNA charging , thus mimicking amino acid starvation ( Hansen et al . , 1972; Thiaville et al . , 2008 ) , or thapsigargin ( Tg ) , which induces the UPR ( Oslowski and Urano , 2011 ) . Histidinol or thapsigargin treatment each significantly induced ULBP1 mRNA in WT K-562 , HAP1 , and Jurkat cells , but induction was abrogated or severely blunted when cells lacked ATF4 ( Figure 5A ) . Induction of ULBP1 transcription by the proteasome inhibitor MG132 ( Butler et al . , 2009 ) was not prevented in the ATF4-mutant cells ( data not shown ) , indicating that certain cellular perturbations transcriptionally activate ULBP1 independently of ATF4 . RNA-Seq analysis of HAP1 cells confirmed ATF4-dependent upregulation of ULBP1 mRNA in response to histidinol treatment ( Figure 5—figure supplement 1 ) . RT-qPCR analysis of other NKG2D ligands expressed by HAP1 cells showed that histidinol treatment caused an increase in ULBP2 mRNA that was partially ATF4-dependent , while MICA expression was slightly decreased in an ATF4-independent manner ( Figure 5—figure supplement 2 ) . ULBP3 and MICB transcripts were not affected by histidinol treatment or loss of ATF4 ( Figure 5—figure supplement 2 ) . ULBP4 , ULBP5 , and ULBP6 mRNAs were absent or barely detectable by RNA-Seq analysis of untreated HAP1 cells , and those genes showed no sign of induction by histidinol ( data not shown ) . These data suggest that the ATF4-mediated stress response induces expression of ULBP1 , and to a minor extent ULBP2 , but not most other NKG2D ligands . 10 . 7554/eLife . 08474 . 012Figure 5 . ATF4 drives increased expression of ULBP1 mRNA and surface protein in response to cell stress . ( A ) Cells were treated for 24 hr with 2 mM histidinol ( HisOH ) to mimic amino acid starvation or 300 nM thapsigargin ( Tg ) to induce the unfolded protein response . RNA was isolated from treated and control cells , and canonically spliced ULBP1 mRNA levels were determined by RT-qPCR . Expression levels were normalized to ACTB , GAPDH , and HPRT1 and are shown as mean ±SE . Expression in untreated WT cells was set to ‘1 . 0’ for each cell type; the different cell types are not comparable in this experiment . For reference , the Cq values for ULBP1 in untreated WT cells were 32 . 2 for K-562 cells , 27 . 9 for HAP1 cells , and 30 . 5 for Jurkat cells . The data were analyzed by 2-way ANOVA with Bonferroni's multiple comparisons test and are representative of three independent experiments , though in one of the three analyses of ATF4 KO Jurkat cells , histidinol-treated cells trended higher than untreated cells but did not reach significance . p < 0 . 05 , **p < 0 . 001 , n . s . : not significant . ( B , C ) Flow cytometric analysis of ULBP1 ( B ) and HLA Class I expression ( C ) on cells treated with histidinol as in ( A ) . ( D , E ) Quantification of surface staining shown in ( B ) and ( C ) . Data are plotted as the geometric mean fluorescence intensity of the specific stain minus the intensity of the isotype control ( ΔgMFI ) . Data are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01210 . 7554/eLife . 08474 . 013Figure 5—figure supplement 1 . Analysis of ULBP1 expression by RNA-Seq . ( A ) WT and ATF4 KO HAP1 cells were treated for 24 hours with 2 mM histidinol ( HisOH ) . Total RNA was isolated from histidinol treated and control cells and analyzed by RNA-Seq . RNA-Seq read densities were normalized to the total number of aligned read in each sample ( reads per million ) and plotted with IGV . ( B ) Quantification of reads aligning to ULBP1 , presented as reads per kb per million ( RPKM ) . Raw sequencing data corresponding to this figure have been deposited to GEO . Accession: GSE69308 . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01310 . 7554/eLife . 08474 . 014Figure 5—figure supplement 2 . Analysis of other NKG2D ligands in response to cell stress . ( A–D ) WT and ATF4 KO cells were treated for 24 hr with 2 mM histidinol ( HisOH ) . RNA was isolated from treated and control cells , and expression of ULBP2 ( A ) , ULBP3 ( B ) , MICA ( C ) , and MICB ( D ) mRNAs was determined by RT-qPCR . Expression levels were normalized to ACTB , GAPDH , and HPRT1 and are shown as mean ±SE . The data were analyzed by 1-way ANOVA with Bonferroni's multiple comparisons test and are representative of three independent experiments , although MICA expression did not change in response to histidinol treatment in one of the three experiments performed . p < 0 . 05 , p < 0 . 01 , p < 0 . 001 , n . s . : not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 014 Stress-induced cell surface ULBP1 protein expression was also examined . Although histidinol and thapsigargin induce transcription of ATF4-regulated genes , they also inhibit protein synthesis , both directly ( e . g . , by limiting the availability of charged histidine tRNAs ) and indirectly , by causing eIF2α phosphorylation . These opposing effects make it difficult to predict whether stress induction will result in increased amounts of any specific protein encoded by an ATF4-regulated gene . Nevertheless , in K-562 cells , we observed a dramatic ATF4-dependent induction of cell surface ULBP1 by histidinol ( Figure 5B , D ) . In HAP1 and Jurkat cells , in contrast , addition of histidinol induced ULBP1 cell surface expression slightly in WT cells , but caused decreased ULBP1 expression in ATF4-mutant cells ( Figure 5B , D ) , consistent with the aforementioned inhibition of protein synthesis resulting from histidinol treatments . Indeed , histidinol treatments resulted in substantially decreased cell surface expression of HLA proteins , which are not regulated by ATF4 ( Figure 5C , E ) . As summarized in Figure 5D , the overall effect of ATF4 induction by histidinol was to induce and/or maintain ULBP1 cell surface expression in the face of global protein synthesis inhibition associated with this stress pathway . To address whether ATF4 directly regulates ULBP1 transcription , we used ChIP-Seq analysis to determine whether it binds to ULBP1 regulatory elements in histidinol-treated HAP1 cells . ChIP-Seq with three independent ATF4 antibodies showed a strong peak of ATF4 binding associated with the ULBP1 promoter , along with a smaller peak ∼27 kb downstream in the intergenic region , which could function as an enhancer . No other notable ATF4 binding was observed in the 266 kb interval shown , which includes genes encoding five other functional NKG2D ligands ( ULBP2-6; RAET1K is a pseudogene ) , at least two of which are expressed in HAP1 cells ( ULBP2 and ULBP3 ) ( Figure 6A , data not shown ) . The MICA and MICB genes were also devoid of ATF4 binding ( data not shown ) . HOMER Motif Analysis software identified ATF4-binding motifs in both peaks ( Figure 6B , C ) ( Heinz et al . , 2010 ) . Additional ATF4-binding motifs were present in the segment shown but were not bound by ATF4 . ATF4 binding at the ULBP1 promoter was confirmed by conventional ChIP-qPCR , which also demonstrated that ATF4 binds to the ULBP1 locus in HAP1 cells under steady-state conditions ( no histidinol ) , and that ATF4 binding is sharply enhanced after histidinol treatment ( Figure 6D , E ) . A similar pattern of ATF4 binding was observed at the ASNS promoter , a known target of ATF4 , but not at a negative control site in the ASNS gene body ( Figure 6D , E , Figure 6—figure supplement 1 ) ( Chen et al . , 2004 ) . 10 . 7554/eLife . 08474 . 015Figure 6 . ATF4 is bound to the ULBP1 promoter and a potential downstream regulatory element . ( A ) ATF4 ChIP-Seq: HAP1 cells were treated with 2 mM histidinol ( HisOH ) for 24 hr , followed by formaldehyde cross-linking . ATF4-bound chromatin was immunoprecipitated using three independent anti-ATF4 antibodies , and the isolated DNA was sequenced and aligned to the human genome ( hg19 ) . Locations of consensus ATF4-binding motifs are indicated . RAET1K is a pseudogene ( ψ ) . ( B , C ) HOMER Motif Analysis software identified the ATF4 binding motif ( B ) and its reverse-complement ( C ) . ( D , E ) Conventional ChIP-qPCR of ATF4 using the antibody sc-200 ( D ) or D4B8 ( E ) . Samples were treated as in ( A ) , followed by qPCR . ChIP signal was normalized to the amount of Input DNA for each sample . Data are plotted as mean ±SE and were analyzed by 2-way ANOVA with Bonferroni's multiple comparisons test . Representative data are shown . The ChIP-qPCR experiment was performed twice using both untreated and HisOH-treated cells , and a third time using only HisOH-treated cells . p < 0 . 05 , *p < 0 . 001 , n . s . : not significant . Raw sequencing data corresponding to ATF4 ChIP-Seq have been deposited to GEO . Accession: GSE69304 . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01510 . 7554/eLife . 08474 . 016Figure 6—figure supplement 1 . ATF4 ChIP-Seq signal at the ASNS promoter , a known ATF4-binding site . ATF4 ChIP-Seq: Hap-1 cells were treated with 2 mM histidinol ( HisOH ) for 24 hr , followed by formaldehyde cross-linking . ATF4-bound chromatin was immunoprecipitated using three independent anti-ATF4 antibodies , and the isolated DNA was sequenced and aligned to the human genome ( hg19 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01610 . 7554/eLife . 08474 . 017Figure 6—figure supplement 2 . ATF4 directly activates the ULBP1 promoter . ( A ) Schematic of ULBP1 promoter constructs . WT promoter sequences are listed in blue , while mutant DNA sequences are listed in red . ( B ) ATF4-mutant HAP-1 cells were co-transfected with ULBP1 promoter constructs and ATF4 or control vectors . Mutant constructs were created in the - 288 promoter fragment . Luciferase activity was measured 24 hours post-transfection . Data were analyzed by 2-way ANOVA with Bonferroni's Multiple Comparisons Test and are representative of 3 independent experiments . P<0 . 001 , n . s . : not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 017 We used luciferase reporter assays to confirm direct activation of the ULBP1 promoter by ATF4 . We inserted previously described fragments of the ULBP1 promoter ( López-Soto et al . , 2006 ) and 5ʹ UTR upstream of a luciferase reporter and tested promoter responsiveness to ATF4 overexpression in ATF4-deficient HAP1 cells ( Figure 6—figure supplement 2 ) . ATF4 overexpression increased ULBP1 promoter activity , and the fold-induction was similar for the longer and shorter promoter fragments tested . To identify the relevant ATF4-binding site ( s ) , we employed the shorter promoter fragment and mutated either the ATF4 motif we had identified , or each of two Cyclic AMP Response Element ( CRE ) sites , which were previously shown to drive ULBP1 promoter activity and are similar to the ATF4 consensus sequence . Mutation of the ATF4 motif 138–147 bp upstream of the transcription start site completely ablated promoter activation by ATF4 ( see mutant 1 , m1 ) , while mutation of the CRE sites did not reduce induction by ATF4 . We conclude that ATF4 directly transactivates the ULBP1 promoter , and that a single ATF4-binding site drives the bulk of this response . RBM4 is an RNA-binding protein that has been implicated in several steps of gene regulation but has been most studied for its regulation of RNA splicing ( Markus and Morris , 2009 ) . We compared WT and RBM4 KO HAP1 cells by RNA-Seq to determine a potential role of RBM4 in splicing of ULBP1 transcripts . As a result , we discovered a novel isoform of ULBP1 that to our knowledge has not been previously reported ( Figure 7A ) . The sequence of the novel isoform has been deposited in GenBank ( Accession number KT591165 ) . Analysis of sequence reads that aligned to splice junctions revealed that most transcripts in WT cells exhibited the expected splicing pattern to encode the normal ULBP1 protein ( Figure 7A , B ) . In RBM4 KO cells , in contrast , approximately 60% of transcripts had undergone an alternative splicing event of the first and second exons in which the 5ʹ splice site at the 3ʹ border of the first exon was ignored in favor of a 5ʹ splice site located 1 . 3 kb downstream within the first intron ( Figure 7A ) . Splicing of the downstream exons of ULBP1 appeared normal in RBM4 KO cells . The alternatively spliced transcript in RBM4 KO cells encodes a premature stop codon shortly after the predicted ER signal peptide and is thus unlikely to encode a functional protein fragment ( Figure 7C ) . We did not observe differential splicing of the other NKG2D ligands expressed by HAP1 cells ( MICA , MICB , ULBP2 , and ULBP3 , Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 08474 . 018Figure 7 . RBM4 suppresses the alternative splicing of ULBP1 mRNA . ( A ) ‘Sashimi plots’ of RNA-Seq reads aligning to ULBP1 mRNA from WT and RBM4 KO HAP1 cells . Peak height represents the number of reads aligning to a given segment ( RPKM ) . Arcs connect splice junctions identified by reads spanning the splice junction; the number of reads aligning to each splice junction is indicated . ( B ) Quantification of splice junction-spanning reads mapping to canonical or alternatively spliced ULBP1 mRNA . Reads spanning the canonical or alternative splice junction are expressed as a percent of the total junction-spanning reads ( canonical + alternative ) . n . d . : not detected . ( C ) Diagram of the canonical and alternatively spliced isoforms of the ULBP1 mRNA . The alternative isoform contains a premature stop codon and is unlikely to produce a functional protein product . The PCR amplicons used in ( D–F ) are indicated . ( D–F ) RT-qPCR analysis of ULBP1 transcript levels in WT and RBM4 KO HAP1 cells . Cells were transduced with a doxycycline-inducible RBM4 vector or control vector and treated with 100 ng/ml Dox for 24 hr . The data were analyzed by 1-way ANOVA with Bonferroni's multiple comparisons test and are representative of three independent experiments . p < 0 . 01 , *p < 0 . 001 , n . s . : not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01810 . 7554/eLife . 08474 . 019Figure 7—figure supplement 1 . Splicing of other NKG2D ligands is unchanged in RBM4 KO HAP1 cells . ‘Sashimi’ plots of RNA-Seq reads aligning to ULBP2 ( A ) , ULBP3 ( B ) , MICA ( C ) , and MICB ( D ) in WT and RBM4 KO HAP1 cells . Peak height represents the number of reads aligning to a given segment RPKM . Arcs connect splice junctions identified by reads spanning the splice junction , with numerals indicating the number of reads aligning to the junction . The data for MICA in panel C showed more reads spanning the exon 1-exon 2 splice junction in the RBM4 KO sample than in WT cells , but as this did not affect cell surface expression , the significance of this observation is not clear . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 01910 . 7554/eLife . 08474 . 020Figure 7—figure supplement 2 . Additional analysis of ULBP1 transcript in WT and RBM4 KO HAP1 cells . ( A ) Diagram of the ULBP1 primary transcripts and canonical and alternatively spliced mRNA isoforms . The PCR amplicons used in ( B–D ) are noted . ( B–D ) RT-qPCR analysis of ULBP1 transcript levels in WT and RBM4− HAP1 cells . Cells were transduced with a doxycycline-inducible RBM4 vector or control vector and treated with 100 ng/ml Dox for 24 hr . The data were analyzed by 1-way ANOVA with Bonferroni's multiple comparisons test and are representative of three independent experiments . p < 0 . 05 , **p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 020 The alternative splicing of ULBP1 transcripts in RBM4 KO cells was confirmed by RT-qPCR analysis ( Figure 7C–E , Figure 7—figure supplement 2 ) . Canonically spliced ULBP1 transcripts were less abundant in RBM4 KO cells , while the alternatively spliced transcripts were more abundant; the overall abundance of ULBP1 transcripts was unchanged as a result of RBM4-deficiency . The reduction in canonically spliced transcripts was commensurate with the reduction in cell surface ULBP1 protein , suggesting that RBM4 impacts cell surface expression of ULBP1 primarily by facilitating the correct splicing of the first two exons of ULBP1 transcripts . A key finding was that ATF4 , a mediator of several stress responses , transcriptionally activates ULBP1 . The data show that ATF4 binds to two locations at the ULBP1 locus , including the promoter region , that an ATF4 site in the ULBP1 promoter is essential for ATF4-induced transcription , and that ATF4 drives ULBP1 transcription in wild-type cells . No additional ATF4 binding or transcriptional induction was detected for most other NKG2D ligand genes , although ULBP2 was modestly upregulated by histidinol treatment in a manner partially dependent on ATF4 . Since ATF4 was not bound to the ULBP2 promoter , we speculate that ULBP2 transcription is regulated by a transcription factor that is induced by ATF4 , or that ATF4 binding to the sites in the ULBP1 gene modestly activates the neighboring ULBP2 gene . Disruption of ATF4 resulted in reduced ULBP1 expression under steady-state conditions in all three of the cell lines tested . These data suggest that ATF4 activity caused by constitutive stress in tumor cell lines contributes to ULBP1 expression . This finding has physiological relevance , as overexpression of UPR components has been found in primary tumors and cell lines derived from breast cancers ( Fernandez et al . , 2000; Fujimoto et al . , 2003 ) , hepatocellular carcinomas ( Shuda et al . , 2003 ) , and multiple myeloma ( Reimold et al . , 1996; Munshi et al . , 2004 ) . The three cell lines we examined illustrate the different ways that ATF4 can support steady-state ULBP1 expression in different contexts . In some cases , expression of ULBP1 may be critically dependent on ATF4 . This is the case in unstimulated K-562 cells in which basal ULBP1 expression was detected ( though weakly ) in WT cells but was absent from the surface of ATF4 KO cells . In other cases , exemplified by HAP1 and Jurkat cells , numerous factors contribute to ULBP1 expression , with ATF4 being only one of them . Those cells had higher basal ULBP1 expression than K-562 cells and showed a modest decrease in ULBP1 expression when ATF4 was mutated . The impact of ATF4 on ULBP1 transcription was most notable when additional stressors were applied that induce ATF4 expression . Outcomes at the level of protein expression are difficult to predict a priori , because these stressors induce transcription of ATF4 target genes , but also cause global suppression of translation as a result of eIF2α phosphorylation . We observed a range of outcomes in different cell lines . Amino acid starvation of K-562 cells , imparted by histidinol treatment , caused a substantial increase in ULBP1 mRNA and surface protein , and the response was ablated in ATF4 KO cells . In this case , the increase in ULBP1 transcription overwhelmed the global decrease in translation . In contrast , in HAP1 and Jurkat cells , amino acid starvation of WT cells caused a fivefold to sevenfold increase in ULBP1 mRNA , but a much smaller increase in ULBP1 protein at the cell surface . Nevertheless , the impact of ATF4 was made apparent by the observation that ULBP1 protein levels decreased on the surface of similarly stressed ATF4 KO cells . In these cell lines , therefore , the induction of ULBP1 transcription by ATF4 serves to maintain ( and slightly increase ) expression of ULBP1 protein in the face of cellular stress . Maintaining ULBP1 expression that is induced by distinct regulatory pathways is important in cells where translational inhibition is likely to occur , such as in hypoxic tumors . Our finding that ULBP1 expression is regulated by ATF4 fits the existing paradigm for ATF4 target genes . ATF4 is induced by amino acid starvation , oxidative stress , hypoxia , and the UPR . ATF4 target genes include amino acid transporters , anti-oxidant response genes , angiogenic factors , and protein chaperones , and thus , ATF4 initially promotes cell survival against these insults ( Jain , 2005; Roybal et al . , 2005; Han et al . , 2013 ) . Previous studies have shown that ATF4 expression is frequently induced in primary tumors and tumor cell lines ( Fernandez et al . , 2000; Shuda et al . , 2003; Ma and Hendershot , 2004 ) . Inadequate vascularization can cause tumors to become hypoxic and nutrient-starved , two conditions known to induce ATF4 . Indeed , ATF4 expression is highest in hypoxic regions of tumors ( Ameri et al . , 2004; Bi et al . , 2005 ) . However , the pro-apoptotic targets of ATF4 , such as CHOP , provide an important safety mechanism for the host ( Tabas and Ron , 2011 ) . Without a pro-apoptotic component , the transcriptional targets of ATF4 would be ripe for exploitation by expanding tumors , which are often poorly vascularized , and consequently both hypoxic and starved . Upregulation of ULBP1 by ATF4 provides an additional layer of host defense , marking potentially dangerous cells for elimination by the immune system . Although not explored in this study , induction of ULBP1 by ATF4 may also play a role in antiviral immunity . As mentioned above , ATF4 expression is induced by phosphorylation of eIF2α . Viral infection can activate PERK and PKR , which phosphorylate eIF2α and therefore induce ATF4 expression ( Mohr and Sonenberg , 2012; Jheng et al . , 2014 ) . We identified RBM4 as a second factor responsible for steady-state ULBP1 expression in HAP1 cells . RBM4-deficient cells exhibited a twofold to threefold reduction in cell surface ULBP1 and the corresponding canonical ULBP1 mRNA isoform . Interestingly , RBM4 suppresses a novel alternative splicing event in the ULBP1 transcript . In the absence of RBM4 , total ULBP1 mRNA levels were unchanged , but the alternatively spliced isoform was upregulated while the canonically spliced isoform was downregulated . The alternative ULBP1 transcript contains a premature stop codon early in the protein-coding sequence and is unlikely to produce a functional protein . Transcripts containing premature stop codons can be targets for nonsense-mediated decay ( NMD ) ( Maquat , 2004 ) , but we could nonetheless detect the alternative ULBP1 transcript by RNA-Seq . The degree to which the alternative ULBP1 transcript is degraded by NMD is presently unclear . RBM4 has also been implicated in other forms of gene regulation , including translation and microRNA-mediated gene silencing ( Lin and Tarn , 2009; Uniacke et al . , 2012 ) . In HAP1 cells , however , the change in cell surface ULBP1 in RBM4 KO cells was commensurate with the change in canonically spliced transcripts , suggesting that these other mechanisms do not play an appreciable role . It remains possible that RBM4 influences ULBP1 translation or gene expression in other cells or contexts . It was recently reported that RBM4 regulates the alternative splicing of numerous cancer-related genes . Most interesting , in light of our findings , was the finding based on RNA-Seq analysis that RBM4 expression promoted a tumor-suppressive splicing profile ( Wang et al . , 2014a ) . This action occurred in part through the alternative splicing of the BCL2L1 gene ( Bcl-x ) , wherein RBM4 promoted splicing that generated the pro-apoptotic isoform Bcl-xS , at the expense of the anti-apoptotic isoform Bcl-xL . Our data suggest that RBM4 further promotes tumor suppression by favoring the canonical splicing of ULBP1 mRNA , marking cells for elimination by the immune system . Decreased RBM4 expression has been observed in various cancers compared to paired healthy tissues , including lung , breast , and pancreatic cancers . Furthermore , lower RBM4 expression correlated with decreased patient survival ( Wang et al . , 2014a ) . We speculate that tumor cells are under selective pressure to downregulate RBM4 expression , as this would allow tumors to evade tumor-suppressive events such as expression of Bcl-xS and ULBP1 . An interesting question is whether the splicing functions of RBM4 are subject to stress-regulation . Intriguingly , the subcellular localization of RBM4 is reportedly regulated by oxidative stress or activation of the MKK3/6-p38 MAPK pathway . In that instance , however , stress favored the cytoplasmic localization of RBM4 ( Lin et al . , 2007 ) , where it reportedly enhanced translation of a subset of mRNAs ( Lin et al . , 2007; Lin and Tarn , 2009; Uniacke et al . , 2012 ) . Additional studies will be necessary to establish whether regulated re-localization of RBM4 , or other types of regulation , impact alternative splicing of ULBP1 or other transcripts . HSPA13 ( also known as STCH ) belongs to the Hsp70 family of chaperone proteins ( Otterson et al . , 1994 ) . HSPA13 is associated with microsomes and displays ATPase activity , but little is known about its function , including whether HSPA13 participates in cellular stress responses . One report showed that HSPA13 binds to the ion transporters NBCe1 and NHE1 , which help maintain intracellular pH homeostasis ( Bae et al . , 2013 ) . HSPA13 overexpression increased NBCe1 and NHE1 protein levels and functional activity . It will be interesting to pursue the mechanisms by which HSPA13 supports expression of ULBP1 , which could be through direct or indirect effects . SPCS1 and SPCS2 are non-catalytic subunits of the signal peptidase complex , but surprisingly little is known about their function ( Mullins et al . , 1996 ) . We were surprised that HAP1 cells lacking SPCS1 or SPCS2 did not have a more general defect in cell surface protein expression; out of seven cell surface proteins tested , only ULBP1 expression was decreased . However , it should be noted that seven proteins represent a small fraction of the proteome , and we expect that a proteome-wide investigation would reveal many additional affected proteins . In yeast , SPCS1 and SPCS2 are required for maximal signal peptidase activity under certain growth conditions ( Fang et al . , 1996; Mullins et al . , 1996 ) , and in mammalian cells , SPCS1 supports expression of the HCV protein NS2 ( Suzuki et al . , 2013 ) . It will be of future interest to define the properties that determine if a given secreted or membrane protein depends on SPCS1 and SPCS2 for maximal expression . The drivers of ULBP1 expression we have identified are likely to cooperate with previously described stress pathways that regulate ULBP1 . Other transcription factors known to bind the ULBP1 gene and drive gene expression include Sp1 , Sp3 , and the tumor suppressor protein p53 ( López-Soto et al . , 2006; Textor et al . , 2011 ) . It is possible that one or more of these factors contribute to ULBP1 transcription in HAP1 and Jurkat cells , considering that residual ULBP1 expression was observed even after ATF4 was disrupted . Additional stress pathways including heat shock , E2F transcription factors , and the DNA damage response drive expression of other NKG2D ligands , and the combined expression level of all the NKG2D ligands on a given cell will contribute to its ability to activate NK cells ( Gasser et al . , 2005; Venkataraman et al . , 2007; Jung et al . , 2012 ) . HAP1 cells may be representative of a very high ‘threat level’ , as they had activated numerous ULBP1 drivers , and independent drivers for several other NKG2D ligands . Moreover , these cells possessed potent NK-activating properties distinct from NKG2D ligands , as shown by studies with NKG2D-deficient NK cells ( data not shown ) . Probably as a consequence , loss of ULBP1 drivers , and even ULBP1 itself , had little detectable effect on the capacity of HAP1 cells to stimulate a response ( data not shown ) . The layers of ULBP1 regulation we have described could allow NKG2D ligand expression to be fine-tuned to reflect the ‘threat level’ of the cell . ATF4 could upregulate ULBP1 transcription , but other factors may amplify the extent of ULBP1 transcription . Maximal cell surface expression would require RBM4 to direct splicing of the functional mRNA isoform , followed by maximal translation and trafficking of the protein to the cell surface . Ubiquitous RBM4 expression is ‘safe’ for normal , healthy cells , since functional ULBP1 cannot be spliced unless the gene is transcribed in the first place . The results of our screen highlight diverse mechanisms for the control of NKG2D ligand expression . To the best of our knowledge , none of the hits we investigated are related to previously known drivers of ULBP1 or other NKG2D ligands , and the effects found were generally specific to ULBP1 . Screens for drivers of other NKG2D ligands will be the subject of future work , and the advent of CRISPR/Cas9-based screens will enable screens to be carried out in many different cell lines ( Gilbert et al . , 2014; Shalem et al . , 2014; Wang et al . , 2014b ) . Cell cultures were performed at 37°C in humidified atmosphere containing 5% CO2 . HAP1 cells were cultured in complete IMDM , consisting of IMDM ( Life Technologies , Carlsbad , CA ) , 10% fetal calf serum ( FCS , Omega Scientific , Tarzana , CA ) , 100 U/ml penicillin ( Life Technologies ) , 100 µg/ml streptomycin ( Life Technologies ) , and GlutaMAX-I ( Life Technologies ) . K-562 and Jurkat cells were cultured in complete RPMI , consisting of RPMI ( Life Technologies ) , 10% FCS , 1 mM sodium pyruvate ( Life Technologies ) , MEM non-essential amino acids ( Life Technologies ) , 100 U/ml penicillin , 100 µg/ml streptomycin , 0 . 2 mg/ml glutamine ( Sigma–Aldrich , St . Louis , MO ) , 10 µg/ml gentamycin sulfate ( Lonza , Basel , Switzerland ) , 20 mM HEPES ( Thermo Fisher Scientific , Waltham , MA ) , and 50 µM 2-mercaptoethanol ( EMD Millipore , Billerica , MA ) . Antibodies against ULBP1 ( Clone 170818 ) , ULBP2/5/6 ( Clone 165903 ) , ULBP3 ( Clone 166510 ) , MICA ( Clone 159227 ) , and MICB ( Clone 236511 ) were purchased from R&D Systems . Antibodies against HLA-A , B , C ( Clone W6/32 ) , CD55 ( Clone JS11 ) , CD59 ( Clone p282 H19 ) , CD112 ( Clone TX31 ) , and CD155 ( Clone TX24 ) , and mouse Thy1 . 1 ( Clone OX-7 ) were purchased from BioLegend ( San Diego , CA ) . In some experiments , the antibodies against HLA-A , B , C ( Clone W6/32 ) and CD59 ( Clone OV9A2 ) were purchased from eBioscience ( San Diego , CA ) . Antibodies against ATF4 used for chromatin immunoprecipitation ( ChIP ) were sc-200 ( polyclonal ) ( Santa Cruz Biotechnology , Santa Cruz , CA ) , D4B8 ( monoclonal ) ( Cell Signaling Technologies , Danvers , MA ) , and ABE387 ( polyclonal ) ( EMD Millipore ) . Histidinol , thapsigargin , and RNase A were purchased from Sigma–Aldrich . Proteinase K was purchased from Roche ( Basel , Switzerland ) . Mutagenesis of HAP1 cells was performed as previously described ( Carette et al . , 2011 ) . Selection of ULBP1low cells was performed using sequential rounds of selection . First , 108 mutagenized HAP1 cells were labeled with ULBP1 antibody , and ULBP1high cells were depleted by running the cells over two sequential MACS LD columns ( Miltenyi Biotec , San Diego , CA ) . Depletion of ULBP1high cells on this day was less efficient than we had achieved in previous experiments , so we chose to re-plate the selected cells , expand the cells for 2 days , and repeated the magnetic depletion of ULBP1high cells . We speculate that a single round of magnetic depletion may have been sufficient if adequate depletion of ULBP1high cells had been achieved . Magnetically selected cells were re-plated and expanded for 5 days . Expanded cells were labeled with ULBP1 and CD55 antibodies , and ULBP1lowCD55+ cells were sorted by FACS , re-plated , and expanded . Genomic DNA was isolated from 4 × 107 cells . Mapping of gene-trap insertion sites and statistical analysis were performed as previously described ( Carette et al . , 2011 ) . Cells were stained with the specified antibodies in 50 μl of PBS supplemented with 2 . 5% FCS or 1% BSA ( FACS buffer ) . FACS buffer included 0 . 05% sodium azide for analytical flow cytometry , but not cell sorting . Dead cells were excluded from analysis by staining with DAPI ( BioLegend ) . Multicolor flow cytometry was performed on an LSR Fortessa cytometer ( BD Biosciences , San Jose , CA ) , and data were analyzed with FlowJo software ( Tree Star , Inc . , Ashland , Oregon ) . Mutant HAP1 cells were generated by transiently co-transfecting cells with a Cas9 expression vector ( pMJ918 , a gift from Jennifer Doudna ) , an sgRNA expression vector ( Addgene plasmid #41824 , a gift from George Church ) , and a GFP expression vector using Lipofectamine 2000 . Transfected cells were sorted based on GFP expression 24–72 hr post-transfection . To enrich cells with a mutant phenotype , cells were stained for ULBP1 expression 5–7 days post-transfection , and single ULBP1low cells were sorted into 96-well plates . Mutant cell clones were identified by sequencing PCR products surrounding the Cas9:sgRNA target site . sgRNA sequences are listed in Table 2 . PCR primers used to detect mutations are listed in Table 3 . 10 . 7554/eLife . 08474 . 021Table 2 . sgRNA sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 021NamesgRNA Sequence + PAMUsed inULBP1 sgRNAGGGTCCCGGGCAGGATGGGTCGGHAP1ATF4 sgRNA 1GGCGGGCTCCTCCGAATGGCTGGHAP1ATF4 sgRNA 2GCTCGTCACAGCTACGCCCTGGGHAP1 , K-562 , JurkatATF4 sgRNA 3GTGGCCAACTATACGGCTCCAGGGHAP1 , K-562 , JurkatRBM4 sgRNAGAGTCCCACCTGCACCAATAAGGHAP1RBM4/RBM4b sgRNAGCCCCGGGAGGCTACAGAGCAGGHAP1HSPA13 sgRNAGTGCCAAATAGCCGGCCAACAGGHAP1SPCS1 sgRNAGCATCTGTTCAGCTAGCTTCTGGHAP1SPCS2 sgRNAGAGTGGCCGTAGCGGCTTGTTGGHAP1PAM , protospacer adjacent motif . For each sgRNA , the protospacer adjacent motif ( PAM ) is indicated in bold blue text . Red text indicates that the 5ʹ G is not present in the genomic target sequence and was added to the sgRNA to allow transcription from the U6 promoter . 10 . 7554/eLife . 08474 . 022Table 3 . PCR primers used to detect mutationsDOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 022TargetPrimer directionPrimer sequenceAmplicon size ( bp ) ULBP1 sgRNA target locusForwardATAAACAGCCGTGGTGTGAG401ReverseTGTCTGGGGAGATCACGATGATF4 sgRNA 1 target locusForwardCATTCCTCGATTCCAGCAAAGC345ReverseTGAGTGATGGGGCCAAGTGAGDeletion of ATF4 with sgRNAs 2 and 3ForwardCGTCCTCGGCCTTCACAATA442ReverseTCTTCAGGATGAGGCTTCTGCRBM4 sgRNA target locusForwardTGCACATAGAAGACAAGACGGC332ReverseCCTTGTGTTCAGCCCTCTACCCHSPA13 sgRNA target locusForwardTGCTGTCTGAGAGGAGTGCT476ReverseCCTCCAACTCTTCTGCGGTASPCS1 sgRNA target locusForwardATTTAATATCTTGCCCAGGCCC394ReverseACCCACAAATTTCTTACCAAACATSPCS2 sgRNA target locusForwardAACCTCAAGTCCCAGCAAGC466ReverseCGGGTCCAGGTTTGAAGTGT ATF4 KO K-562 , Jurkat , and HAP1 cells were generated using two sgRNAs flanking ATF4 ( ATF4 sgRNA 2 and ATF4 sgRNA 3 ) in order to delete the entire ATF4 gene . Cells were transiently co-transfected with GFP and the Cas9/sgRNA expression plasmid px330 ( Addgene plasmid #42230 , a gift from Feng Zhang ) . Transfected cells were sorted based on GFP expression 24–72 hr post-transfection . 5–7 days post-transfection , single cells were sorted into 96-well plates without any additional selection . For doxycycline-inducible gene expression , constructs encoding ATF4 , RBM4 , HSPA13 , SPCS1 , and SPCS2 were cloned into the lentiviral vector pFG12-TRE-UbC-rtTA-Thy1 . 1 ( hereafter pFG12 ) . The ATF4 construct was a gift from Yihong Ye ( Addgene plasmid #26114 ) . The RBM4 construct was a gift from Lan Ko , Georgia Health Sciences University . The HSPA13 construct was a gift from Kyungpyo Park , Seoul National University . The SPCS1 construct was a gift from Tetsuro Suzuki , Hamamatsu University School of Medicine . SPCS2 was cloned from HAP1 cDNA . Lentiviral supernatants were generated by co-transfecting 293T cells with pFG12 vector , the VSV-G plasmid pMD2 . G , and the packaging plasmid psPAX2 using Lipofectamine 2000 ( Life Technologies ) . Supernatants were harvested 48 hr post-transfection , filtered with a 0 . 45-µM PES syringe filter , and added to target cell cultures with 4 µg/ml polybrene . Transduced cells were sorted based on Thy1 . 1 expression . To induce gene expression , cells were cultured in complete medium with 100 or 1000 ng/mL doxycycline ( Sigma–Aldrich ) , as noted in figure legends . Total RNA was isolated from cells using the RNeasy Mini Kit and on-column DNase digestion using the RNase-free DNase Set ( Qiagen , Hilden , Germany ) . Reverse transcription was performed with iScript Reverse Transcription Supermix ( Bio-Rad , Hercules , CA ) . Quantitative real-time PCR was performed on a CFX96 thermocycler using SsoFast EvaGreen Supermix ( Bio-Rad ) with the following cycling program: Initial denaturation at 98°C , 2ʹ , followed by 40 cycles of 98°C , 2ʹʹ , 55°C , 5ʹʹ . Data were analyzed with Bio-Rad CFX Manager Software version 3 . 1 . Gene expression was normalized to the reference genes ACTB , GAPDH , and HPRT1 , normalizing the gene of interest to the geometric mean of the three reference genes ( Vandesompele et al . , 2002 ) . qPCR primer sequences are listed in Table 4 . 10 . 7554/eLife . 08474 . 023Table 4 . Primers used for qPCRDOI: http://dx . doi . org/10 . 7554/eLife . 08474 . 023TargetPrimer directionPrimer sequence ( 5ʹ-3ʹ ) Amplicon size ( bp ) ACTBForwardTTGGCAATGAGCGGTTCC92ReverseGTTGAAGGTAGTTTCGTGGATGGAPDHForwardCAACAGCGACACCCACTCCT115ReverseCACCCTGTTGCTGTAGCCAAAHPRT1ForwardAGGATTTGGAAAGGGTGTTTATTC109ReverseCAGAGGGCTACAATGTGATGGTotal ULBP1 ( exon 3-exon 3 ) ForwardGCCAGGATGTCTTGTGAGCATGAA134ReverseTTCTTGGCTCCAGGATGAAGTGCTULBP1 canonical isoform ( exon 1-exon 2 ) ForwardATCAGCGCCTCCTGTCCAC136ReverseAAAGACAGTGTGTGTCGACCCATULBP1 alternative isoform ( extended exon 1-exon 2 ) ForwardGGAATTGCAGGAGGGTGGAG183ReverseCAAAGGCTTTGGCCTTGTGGTTAAULBP1 spliced transcript ( exon 2-exon 3 ) ForwardTAAGTCCAGACCTGAACCACA477ReverseCCATTGAAGAGGAACTGCCAAGULBP1 alternative isoform and primary transcript ( exon 1-extended exon 1 ) ForwardCCGGGCAGGATGGGTCG263ReverseTGTCTGGGGAGATCACGATGULBP1 unspliced transcript ( Intron 1-exon 2 ) ForwardCCCTCAGAGGCCTTCACTTG195ReverseAAGGCCTTTCATCCACCAGGULBP2ForwardGCCGCTACCAAGATCCTTCT161ReverseTCATCCACCTGGCCTTGAACULBP3ForwardCTCGCGATTCTTCCGTACCT127ReverseTCTGGACCTCACACCACTGTMICAForwardATGTCCTGCCTGATGGGAATGGAA189ReverseCAGCAGCAACAGCAGAAACATGGAMICBForwardTGGATCTGTGCAGTCAGGGTTTCT176ReverseTGAGGTCTTGCCCATTCTCTGTCAULBP1 promoter ChIPForwardGCTGTCAGATGACGAGCCC80ReverseATACACTGGGCGGGATCCTAASNS promoter ChIPForwardTGGTTGGTCCTCGCAGGCAT66ReverseCGCTTATACCGACCTGGCTCCTASNS exon 7 ChIPForwardGCAGCTGAAAGAAGCCCAAGT62ReverseTGTCTTCCATGCCAATTGCA Cells were re-plated in fresh media 12–24 hr prior to treatment . Media was then replaced with fresh media or fresh media containing 2 mM histidinol , and cells were incubated for 24 hr . Chromatin complexes were cross-linked by replacing the culture media with fresh complete media containing 1% formaldehyde , and cells were incubated at room temperature for 10 min . Cross-linking was quenched by adding glycine to a final concentration of 0 . 125 M . Cells were washed two times with ice-cold PBS , and cell pellets were flash-frozen in a slurry of dry ice and ethanol . Cells were resuspended in ChIP Lysis Buffer #1 ( see buffer compositions below ) , incubated for 10ʹ at 4°C with rotation , and centrifuged for 8ʹ at 800×g at 4°C . Pellets were resuspended in ChIP Lysis Buffer #2 , incubated for 10ʹ at 4°C with rotation , and centrifuged for 8ʹ at 800×g at 4°C . Pellets were resuspended in 3 volumes ( ∼200 µl ) of ChIP Lysis Buffer #3 . Chromatin was sheared to an average fragment size of 200 bp using a Covaris S2 Ultrasonicator ( Covaris , Woburn , MA ) . Sheared chromatin was diluted in additional ChIP Lysis Buffer #3 , followed by addition of Triton X-100 ( final concentration 1% ) and NaCl ( final concentration 150 mM ) and divided into 1 ml aliquots for immunoprecipitation . Antibody was added to chromatin samples , followed by overnight incubation at 4°C , with rotation . Antibody-chromatin complexes were captured with Protein G Dynabeads ( Life Technologies ) at 4°C for 2 hr , with rotation . Beads were washed 2× with low-salt wash buffer , 1× with high-salt wash buffer , 1× with LiCl wash buffer , and 2× with TE . Beads were resuspended in ChIP elution buffer , and chromatin was eluted by incubating samples at 65°C for 15ʹ . Supernatant containing eluted chromatin was removed from the beads and the cross links were reversed by incubating samples overnight at 65°C . RNase A was added to samples to a final concentration of 0 . 2 mg/ml , followed by incubation at 37°C for 1 hr . Proteinase K was then added to a final concentration of 0 . 4 mg/ml , followed by incubation at 56°C for 1 hr . DNA was then isolated using the Qiagen MinElute Kit , followed by qPCR as described above or Illumina Library prep . ChIP-Seq libraries were prepared using the NEBNext Ultra DNA Library Prep Kit for Illumina ( New England BioLabs , Ipswich , MA ) , including optional size selection of adapter-ligated DNA and 13 cycles of PCR amplification . Final libraries were analyzed with a Qubit fluorometer ( Life Technologies ) and Bioanalyzer 2100 ( Agilent Technologies , Santa Clara , CA ) to assess quantity and quality . Libraries were clustered at a density of 6 . 5 pM with a cluster station and sequenced on a GAIIx Genome Analyzer ( Illumina , San Francisco , CA ) . ChIP-Seq reads were aligned to the human genome ( hg19 ) using Bowtie2 version 2 . 1 . 0 using all default settings ( Langmead and Salzberg , 2012 ) . Reads that could not be uniquely mapped were discarded . Read densities were normalized to the total number of aligned reads in a sample using Bedtools version 2 . 17 . 0 ( Quinlan and Hall , 2010 ) . ATF4-binding peaks were identified with HOMER version 4 . 7 ( Heinz et al . , 2010 ) using the program findPeaks with the following options: -style factor -i <input_tagdir> where <input_tagdir> is a directory of ChIP-Seq reads from the input DNA fraction . ATF4 binding motifs were identified with HOMER version 4 . 7 using the program findMotifsGenome . pl with the following options: -size 50 -mask Data were visualized with the Integrative Genomics Viewer ( IGV ) version 2 . 3 . 47 ( Robinson et al . , 2011 ) . ChIP Lysis Buffer #1: 50 mM HEPES pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , 10% glycerol , 0 . 5% NP-40 , 0 . 25% Triton X-100 , Protease Inhibitors ChIP Lysis Buffer #2: 10 mM Tris pH 8 . 0 , 200 mM NaCl , 1 mM EDTA , 0 . 5 mM EGTA , Protease Inhibitors* ChIP Lysis Buffer #3: 10 mM Tris pH 8 . 0 , 100 mM NaCl , 1 mM EDTA , 0 . 5 mM EGTA , 0 . 1% sodium deoxycholate , 0 . 5% N-lauroylsarcosine , Protease Inhibitors* Low-salt wash buffer: 20 mM Tris pH 8 . 1 , 150 mM NaCl , 2 mM EDTA , 0 . 1% Sodium dodecyl sulfate ( SDS ) , 1% Triton X-100 High-salt wash buffer: 20 mM Tris pH 8 . 1 , 500 mM NaCl , 2 mM EDTA , 0 . 1% SDS , 1% Triton X-100 LiCl salt wash buffer: 20 mM Tris pH 8 . 1 , 0 . 25 M LiCl , 1 mM EDTA , 1% NP-40 , 1% sodium deoxycholate TE: 10 mM Tris pH 8 . 0 , 1 mM EDTA Elution buffer: 10 mM Tris pH 8 . 0 , 300 mM NaCl , 1 mM EDTA , 1% SDS *Complete Mini EDTA-free protease inhibitor cocktail ( Roche , Basel , Switzerland ) Antibody amounts per 1 ml IP: anti-ATF4 ( sc-200 ) : 5 µg , anti-ATF4 ( D4B8 ) : 1 . 06 µg , anti-ATF4 ( ABE387 ) : 20 µg , Rabbit IgG control: 5 µg Total RNA was purified with RNAzol RT ( Sigma–Aldrich ) according to the manufacturer's instructions . For RBM4 KO samples , equal amounts of total RNA were pooled from 4 independent RBM4 KO cell clones . For ATF4 KO samples , equal amounts of total RNA were pooled from three independent ATF4 KO cell clones . RNA-Seq libraries were prepared using the NEBNext Ultra Directional RNA Library Prep Kit ( New England BioLabs ) , including poly ( A ) selection using the NEBNext Poly ( A ) mRNA Magnetic Isolation module . Library quality assessment and clustering were performed as described for ChIP-Seq . Paired-end sequencing was performed using a GAIIx Genome Analyzer ( Illumina ) . RNA-Seq reads were aligned using Tophat version 2 . 0 . 11 ( Kim et al . , 2013 ) using the reference human genome hg19 and the reference transcriptome CRCh37 . 59 using the following options: --library-type fr-firststrand --no-coverage-search --mate-inner-dist 300 Guided transcriptome reconstruction was done with Cufflinks Version 2 . 2 . 1 ( Trapnell et al . , 2012 ) . Cufflinks was run on the Tophat output file ‘accepted_hits . bam’ for each sample , using the hg19 reference genome and transcript annotation files and the options -b hg19_ucsc . fa -M hg19 . rRNA . gtf -g hg19_ucsc . refGene . gtf -u --min-frags-per-transfrag After assembly was complete for each sample , the assembled transcripts were merged together using Cuffmerge with the command template: cuffmerge -o combined/ --min-isoform-fraction 0 -s hg19 assembly_list . txt The merged transcriptome was compared against the reference transcriptome using Cuffcompare: cuffcompare -r hg19_ucsc . refGene . gtf -s hg19 -R -M merged . gtf This produced comparative statistics , and most importantly for this pipeline , a ‘ . tmap’ file that lists the most closely matching reference transcript for each of the assembled transcripts from merged . gtf . The merged transcriptome was converted to genePred format ( . gp ) using the UCSC binary utility , gtfToGenePred using the command: gtfToGenePred merged . gtf merged . gp merged . gp was joined to the . tmap file produced by Cuffcompare above , to obtain a transcriptome file that is annotated with transcript IDs , exonic coordinates , gene IDs , and short-names of the most closely matching reference transcripts . The Tophat output files ‘accepted_hits . bam’ for each sample were sorted and converted to BED format using the BEDTools-2 . 17 . 0 utility bamToBed with the command: bamToBed -split -i accepted_hits . bam\|sort -k1 , 1 -k2 , 2n -k3 , 3n > accepted_hits . bed An executable ( written in-house ) called kcGEXM ( Escobar et al . , 2014 ) was used to obtain normalized read counts and calculate reads per kilobase per million ( RPKM ) for transcripts . kcGEXM uses the genePred and BED files as input: kcGEXM -f gp -r merged . tmap . gp accepted_hits . bed > sample_name . cnt where -f precedes the format of the reference annotation file and -r precedes the reference annotation file . In Figure 5—figure supplement 1 , read densities were normalized to the total number of aligned reads in a sample using Bedtools version 2 . 17 . 0 ( Quinlan and Hall , 2010 ) , and data were visualized with IGV version 2 . 3 . 47 ( Robinson et al . , 2011 ) . Sashimi plots were generated using MISO version 0 . 5 . 2 ( Katz et al . , 2010 ) with Python version 2 . 7 . Fragments of the ULBP1 promoter were amplified from HAP1 genomic DNA and inserted into the pGL3-Basic luciferase expression vector ( Promega , Fitchburg , WI ) between the KpnI and HindIII restriction sites . Mutant ULBP1 promoter fragments were ordered from IDT ( Coralville , IA ) . The sequences of the ULBP1 promoter constructs can be found in Supplementary files 2–7 . ATF4-mutant HAP1 cells were plated at 10 , 000 cells/well in 96-well plates . 24 hr after plating , cells were co-transfected with the indicated ULBP1 promoter constructs ( 99 ng ) , pRK5-FLAG-ATF4 or control vector ( 199 ng ) , and renilla luciferase vector pRL-SV40 ( 1 ng ) . 24 hr post-transfection , cells were washed twice in PBS and lysed with 50 µl of passive lysis buffer ( Promega #E1941 ) . 15 µl of lysate was transferred to an opaque 96-well assay plate . 100 µl of D-Luciferin reagent ( synthesized in-house ) was added to the lysates and samples were immediately assessed for luminescence over a 10-s time period using an LMAX-II luminometer ( Molecular Devices , Sunnyvale , CA ) .	Cancer is caused by a series of mutations that result in uncontrolled cell growth and division . Yet , the body's immune system can often detect and destroy abnormal cells before they cause tumors and disease . Natural killer cells are part of the immune system and have receptors on their surface that allow them to tell the difference between healthy host cells and host cells that are stressed or abnormal . Some of these receptors activate the natural killer cells when they bind to their target molecules . Other receptors have the opposite effect and inhibit the natural killer cells . Activation occurs when the signaling from the activating receptors is stronger than the signals from the inhibitory receptors . One of the well-studied activating receptors recognizes a number of proteins and molecules that are produced by abnormal or tumor cells , including a protein called ULBP1 . This protein is absent from the surface of healthy cells but is found in abundance on tumor cells . However , it is still not clear what drives tumor cells to produce ULBP1 ( or other molecules ) that are recognized by natural killer cell receptors . Now , Gowen et al . report on a genetic screen that has revealed numerous genes that regulate the levels of ULBP1 in human cells . Many of these genes had independent effects that when added together accounted for most of the ULBP1 present on the cell surface . Gowen et al . then explored some of the ‘regulators’ encoded by these genes in more detail . One called ATF4 , which had previously been linked to stress responses , was shown to increase the expression of the gene for ULBP1 in cancer cells . Another regulator called RBM4 instead acted in a different way and at a later stage in ULBP1 production . All together , these findings offer insight into the stress pathways that alert the immune system to abnormal cells . The next challenge will be investigating how these pathways might be exploited for cancer immunotherapy .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2015	A forward genetic screen reveals novel independent regulators of ULBP1, an activating ligand for natural killer cells
Myxobacteria are known for complex social behaviors including outer membrane exchange ( OME ) , in which cells exchange large amounts of outer membrane lipids and proteins upon contact . The TraA cell surface receptor selects OME partners based on a variable domain . However , traA polymorphism alone is not sufficient to precisely discriminate kin . Here , we report a novel family of OME-delivered toxins that promote kin discrimination of OME partners . These SitA lipoprotein toxins are polymorphic and widespread in myxobacteria . Each sitA is associated with a cognate sitI immunity gene , and in some cases a sitB accessory gene . Remarkably , we show that SitA is transferred serially between target cells , allowing the toxins to move cell-to-cell like an infectious agent . Consequently , SitA toxins define strong identity barriers between strains and likely contribute to population structure , maintenance of cooperation , and strain diversification . Moreover , these results highlight the diversity of systems evolved to deliver toxins between bacteria . Cooperative , social organisms benefit by resource sharing and division of labor between individuals in a population . These behaviors entail directing beneficial action toward kin , often at a fitness cost to the actor . Relatedness between individuals must be high for such cooperative action to remain evolutionarily viable ( Hamilton , 1964a , 1964b ) . This requires that social organisms recognize their kin , and direct preferential action toward them ( kin discrimination ) . The mechanisms by which social microbes recognize and direct benefits toward kin cells are not well understood . However , insights in this area will help us to understand the organization of microbes into social groups and the behaviors that maintain cooperation despite seemingly opposing evolutionary pressures to be selfish . The soil bacterium Myxococcus xanthus is a model organism for the study of social behavior and cooperation ( Cao et al . , 2015 ) . Myxobacterial populations divide labor and share resources during coordinated behaviors such as swarming , predation , and starvation-induced fruiting body development . Their social lifestyle , which includes multicellular development by an aggregation strategy , requires that they direct cooperative behavior towards their clonemates or close kin . One such cooperative behavior is outer membrane exchange ( OME ) . During OME , swarming cells in a population simultaneously donate and receive prodigious amounts of outer membrane ( OM ) material between one another during cell contact . Exchanged material includes membrane lipids , lipoproteins and lipopolysaccharide ( Nudleman et al . , 2005; Wei et al . , 2011; Vassallo et al . , 2015; Pathak et al . , 2012 ) . The mechanism for exchange is thought to involve transient OM fusion catalyzed by the OM receptor TraA and an associated protein , TraB ( Cao et al . , 2015; Pathak et al . , 2012 ) . Our model predicts that transient OM fusion between two cells enables the lateral diffusion of OM lipids and proteins between OMs until cells move apart and the membranes are again separated ( Cao et al . , 2015 ) . This process occurs constitutively on surfaces and facilitates efficient OM homogenization of populations with heterogeneous OMs ( Wei et al . , 2011 ) . Exchange of fluorescent OM lipoprotein reporters , as well as endogenous OM lipoproteins , demonstrates that nearly all recipient cells receive substantial amounts of cargo protein within two hours of co-culture ( Nudleman et al . , 2005; Wei et al . , 2011 ) . Furthermore , cells with lethal defects in lipopolysaccharide biosynthesis can be sustained in a population by OME with wild-type ( WT ) donors ( Vassallo et al . , 2015 ) . Based on this observation , OME is hypothesized to help physiologically heterogeneous populations move toward homeostasis and buffer cell damage to support synchronized and cohesive group behaviors ( Vassallo et al . , 2015; Vassallo and Wall , 2016 ) . This robust system for sharing cellular goods must be discriminately targeted to closely related cells – that is , clonemates . Otherwise , this organism risks donating private goods to competing , non-kin genotypes . In this regard , we previously showed that TraA has a variable domain that specifies recognition between cells through homotypic interactions ( Pathak et al . , 2013; Cao and Wall , 2017 ) . Thus , myxobacteria with divergent , incompatible TraA receptors do not engage in OME . traA is therefore a greenbeard gene in that it allows myxobacteria to identify cells with identical alleles and to direct beneficial treatment toward those cells ( Dawkins , 1976 ) . Greenbeard alleles do not exclusively recognize kin genotypes , but instead recognize any genotype that possesses the same allele ( kind discrimination ) ( Queller , 2011; Strassmann et al . , 2011 ) . Indeed , although TraA sequence diversity in the variable domain is high , some non-kin genotypes share compatible traA alleles ( Pathak et al . , 2013 ) . In fact , some Myxococcus isolates that antagonize one another in co-culture possess the same traA alleles ( Pathak et al . , 2013 ) . Based on this observation , we hypothesized that there are additional genetic determinants that more precisely discriminate kin during social interactions . Bacterial kin discrimination is often mediated by antagonism toward non-kin . One mechanism that bacteria use to this end is the delivery of polymorphic toxins between cells in close contact ( Zhang et al . , 2012; Ruhe et al . , 2013; Cardarelli et al . , 2015; Wenren et al . , 2013 ) . These toxins usually share homology in species-specific amino-terminal domains required for presentation and/or delivery , but vary in their carboxy-terminal toxin domains ( Zhang et al . , 2012 ) . Each toxin is associated with a cognate immunity protein , typically encoded together in an operon , which specifically neutralizes toxicity in the producing cell and in clonemates or close kin that share the locus . The presence of a polymorphic toxin/immunity pair in one strain leads to antagonism toward related strains that do not possess immunity ( Riley and Wertz , 2002 ) . Examples include contact-dependent growth inhibition ( CDI; a type Vb secretion system ) ( Aoki et al . , 2005 , Aoki et al . , 2010 ) ; modular type IV secretion system ( T4SS ) ( Souza et al . , 2015 ) , type VI secretion system ( T6SS ) ( Russell et al . , 2011; MacIntyre et al . , 2010; Schwarz et al . , 2010; Hood et al . , 2010 ) , and type VII secretion system ( T7SS ) ( Cao et al . , 2016 ) effectors; as well as the MafB toxins of Neisseria ( Jamet and Nassif , 2015a ) . The competitive advantages offered by these toxins likely drives positive selection for novel toxin/immunity pairs , which in turn helps to define kin groups through inter-strain antagonism . Mining of prokaryotic genomes revealed that polymorphic toxins are indeed quite prevalent and diverse ( Zhang et al . , 2012 ) . Additionally , many homologous C-terminal toxin domains are shared between distantly related toxin-delivery systems from diverse organisms , suggesting that these toxins have evolved from a common pool of domains ( Zhang et al . , 2012 ) . Currently there is a knowledge gap between the number of toxin domains discovered through bioinformatic analysis and the experimental characterization of their delivery mechanisms ( Jamet and Nassif , 2015b; Benz and Meinhart , 2014 ) . It seems likely that additional , uncharacterized modes of polymorphic toxin delivery remain to be discovered , with each mechanism adapted to the host’s particular lifestyle . As mechanisms that promote inter-strain and inter-species conflict , polymorphic toxins appear to play a strong role in the evolution of microbes . For instance , kin discrimination by polymorphic toxins may help maintain cooperation in social organisms such as myxobacteria by promoting local relatedness ( Hamilton , 1964a; Vos and Velicer , 2009 ) . In addition , they likely play an important role in symbiosis ( Hillman and Goodrich-Blair , 2016 ) and in population structure within ecological niches such as the soil ( Varivarn et al . , 2013 ) , rhizosphere ( Ma et al . , 2014 ) , and human gut ( Zheng et al . , 2015; Russell et al . , 2014 ) . We previously showed that the widely used DK1622 reference strain of M . xanthus is killed by ancestral strains when co-cultured on surfaces ( Dey et al . , 2016 ) . A traA mutation in either strain abolishes this behavior , indicating that OME is required for antagonism . Further , this antagonism requires a hyper-variable region of the chromosome called Mx-alpha , which is composed of roughly 100 kb of prophage and mobile genetic elements and can be found in multiple copies of imperfect repeats in M . xanthus genomes ( Dey et al . , 2016 ) . In ancestral strains that antagonize DK1622 , there are three homologous Mx-alpha units apparently arranged in tandem . However , two of these units ( ~200 kb ) were lost by spontaneous deletion during the construction of DK1622 ( Dey et al . , 2016 ) . From these observations , we hypothesized that OME-delivered toxins encoded within the Mx-alpha repeat elements are responsible for antagonism . Here , we identify the genetic determinant of this antagonism as one of several related , polymorphic , OM lipoprotein toxins that are encoded on Mx-alpha and transferred to target cells by OME . OME between strains that contain different toxins leads to mutual cell death , which establishes territorial barriers between populations . These toxins belong to a large and diverse family found in myxobacteria and display features that make them unique among polymorphic toxin systems . Strikingly , we show that these toxins are serially transferred from cell-to-cell by OME , which results in a potent killing system . Finally , we provide evidence that OME-mediated antagonism contributes to the ecology and evolution of these social microbes . M . xanthus inter-strain antagonism related to the presence of Mx-alpha was originally observed as ‘swarm inhibition’ , during which a nonmotile ancestor strain ( Mx-alpha+ ) inhibited the outward swarming of a motile strain ( missing two of three Mx-alpha units ) during co-culture on agar . This phenomenon is traA-dependent and therefore is likely an outcome of OME ( Pathak et al . , 2013 ) ( see Figure 1 ) . Swarm inhibition was further demonstrated to be caused by cell death of the motile strain ( Dey et al . , 2016 ) . We first sought to identify the specific genetic determinant on Mx-alpha that was required for antagonism and cell death of the susceptible motile strain . Sequence analysis of the two Mx-alpha units retained in the ancestral DK101 strain ( a . k . a . DZF1 ) ( Müller et al . , 2013 ) but lost in DK1622 revealed a candidate toxin gene ( MXF1DRAFT_07513 ) , which we have designated sitA1 for swarm inhibition toxin . This gene encodes a predicted lipoprotein that contains a C-terminal nuclease domain with a WHH motif ( Zhang et al . , 2011 ) ( Figure 1A ) . The absence of lysine at the +2 position or alanine at the +7 position relative to the +1 cysteine in the N-terminal lipobox suggests that this protein is localized to the OM ( Bhat et al . , 2011 ) ( see Supplementary file 1 ) . Given that OME efficiently transfers OM lipoproteins between cells ( Wei et al . , 2011 ) , this open reading frame ( ORF ) represents a promising candidate for the antagonistic determinant . Immediately downstream of sitA1 is a gene ( sitI1 ) that shows homology to the SUKH-family of immunity proteins commonly found in polymorphic toxin systems ( Zhang et al . , 2011 ) ( Figure 1A ) . Upstream of sitA1 is a hypothetical gene ( sitB1 ) of unknown function . The sitB1 ORF overlaps with sitA1 by 11 base pairs , suggesting that the genes form an operon and function together ( Figure 1A ) . 10 . 7554/eLife . 29397 . 003Figure 1 . SitA1 is the swarm inhibition determinant . ( A ) sitBAI1 operon found on one of the Mx-alpha elements that was lost from DK1622 . SS , signal sequence . ( B ) Swarm inhibition assays with indicated motile and nonmotile strains . White arrow illustrates swarm inhibition with control strains . NA , not applicable . Bar , 1 mm . ( C ) Expression of sitBAI2 in a non-antagonistic nonmotile background results in modest swarm inhibition ( indicated by * ) compared to sitBAI1+ ( shown in B ) . Expression of sitBAI3 in the non-antagonistic nonmotile background results in complete swarm inhibition of ∆sitBAI3 . Here and elsewhere see Supplementary file 2A for strain details . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 003 To test if sitA1 is the swarm inhibition toxin , we used the swarm inhibition assay as a readout for the contribution of sitA or sitI toward cell death of the susceptible motile strain during co-culture with the antagonistic , nonmotile ancestral strain . In this assay , cell death of the susceptible strain results in no cells visibly escaping the mixed culture spot . In contrast , abrogation of cell death results in the appearance of the motile strain moving outward from the colony co-culture . As shown previously ( Pathak et al . , 2013; Dey et al . , 2016; Dey and Wall , 2014 ) , nonmotile ancestral cells inhibited the motility of DK1622 , but not the DK1622 ∆traA strain ( Figure 1B , rows 1–4 ) . Importantly , nonmotile ancestors carrying a sitA1 mutation did not inhibit DK1622 ( Figure 1B , row 5 ) . Further , expression of sitI1 in motile DK1622 cells also prevented antagonism ( Figure 1B , row 6 ) , consistent with the prediction that sitI encodes an immunity protein that neutralizes SitA1 . To test whether sitBAI1 is sufficient to convert non-antagonistic cells into killers , we expressed the gene cassette in a DK1622-derived nonmotile strain , which lacks two Mx-alpha units ( see Figure 2A ) , and does not cause swarm inhibition . As predicted , ectopic sitBAI1 expression allowed the nonmotile ∆Mx-alpha cells to inhibit DK1622 , thus recapitulating the antagonistic phenotype exhibited by the nonmotile ancestor strain ( Figure 1B , rows 7–8 ) . These combined results suggest that sitBAI1 may function as a toxin/immunity system responsible for the antagonistic behaviors previously described ( Pathak et al . , 2013; Dey et al . , 2016; Dey and Wall , 2014 ) . Therefore , the loss of two Mx-alpha units , and thus the sitBAI1 operon ( Figure 2A ) , during the construction of DK1622 from an ancestral DK101 strain explains why the latter strain antagonizes the former . 10 . 7554/eLife . 29397 . 004Figure 2 . SitA polymorphic toxins found on Mx-alpha units are delivered by OME . ( A ) Strain DK101 ( the ancestor of DK1622 ) carries three Mx-alpha repeats , whereas DK1622 retains only one copy . Each Mx-alpha unit contains a unique sitBAI cassette . SitB proteins contain type I signal sequences ( white boxes ) whereas SitA proteins contain type II signal sequences ( white boxes ) with a lipobox and C-terminal toxin domains . The relative sequence identities are shown . ( B ) Competition outcomes when inhibitor strains each expressing one of three sitBAI cassettes were competed against susceptible target strains that lack the corresponding sitBAI cassette . Mock-inhibitor control is shown at left ( WT vs . WT ) . See text for the calculation of competitive index . Strain genotypes ( ‘–’ , traA deletion ) are shown below histograms and further strain details provided in Supplementary file 2A . ( C ) Cells harvested from an agar co-culture of a strain expressing a SitA1-mCherry fusion with a GFP-labeled target at 0 and 6 hr . GFP targets are traA+ in the top panel and ∆traA in the bottom panel . Yellow arrows indicate two examples of GFP cells that have acquired the mCherry reporter . Boxes represent the number of mCherry positive GFP cells out of 100 . Bar , 5 μm . ( D ) Fixed-cell immunofluorescence of C-terminal FLAG-tagged SitA1 and untagged control . Bar , 2 . 5 μm . Immunoblot of protein isolated from the same strains ( right ) . SitAFLAG predicted size is 62 . 6 kDa . ( E ) Competition outcomes when inhibitor expresses one of the three sitBAI cassettes and the target strains express one of the three sitI genes . Data points at <0 . 001 indicate that no target cells remained . ( F ) E . coli MG1655 plating efficacy when equal number of cells were 10-fold serially diluted , spotted onto arabinose-supplemented agar and incubated overnight . Strains express either SitA1 or SitA3 C-terminal toxin domain ( CTD ) from a pBAD plasmid either in the absence ( ‘–’ , empty vector ) or presence of the indicated sitI genes expressed constitutively from a separate plasmid ( pKSAT ) . This image is representative of three biological replicates . In this figure and the figures below , error bars represent standard error of the mean from at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 00410 . 7554/eLife . 29397 . 005Figure 2—figure supplement 1 . Morphology of SitA-poisoned target cells . Target cells ( red ) were competed with sitBAI inhibitor strains or a ∆sitBAI mock inhibitor control at a 20 to 1 ratio . After 24 hr on agar media , cells were harvested and placed on glass slides for microscopy . Yellow arrows indicate an example of filamentous morphology for SitA1 and SitA2-poisoned cells and indicate an example of rounded cells for SitA3-poisoned cells . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 00510 . 7554/eLife . 29397 . 006Figure 2—figure supplement 2 . SitA-CTD expression in M . xanthus is toxic . ( A ) Culture growth of strains was measured over 24 hr , in the presence or absence of IPTG . Each strain expressed a SitA-CTD or a control protein ( tdTomato ) from an IPTG-inducible promoter . Red diamonds , + IPTG ( 1 mM ) ; black circles , – IPTG . ( B ) Cell morphology and DAPI stain of the strains from ( A ) when grown with 1 mM IPTG for 30 hr . Yellow arrows highlight an example cell which was arrested during cell division and contains two nucleoids . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 00610 . 7554/eLife . 29397 . 007Figure 2—figure supplement 3 . Heterologous sitAI cassettes from M . fulvus HW-1 are active in DK1622 . sitA3Mf1 ( SitA3 homolog , LILAB_02580 ) and its associated sitI3Mf1 , or sitA1Mf1 ( SitA1 homolog , LILAB_05795 ) and its associated sitB1Mf1 and sitI1Mf1 , were expressed in DK1622 . Competitive indices against WT ( DK1622 ) and DK1622 ∆traA cells were determined at 24 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 007 The three tandem Mx-alpha units in the ancestor strain are related and contain different alleles of >80 genes . This region therefore represents a rare bacterial polyploid element – that is , it contains three Mx-alpha prophage genomes with divergent gene allele sets ( Dey et al . , 2016 ) . Inspection of the other two Mx-alpha units revealed additional putative sitBAI operons . The second Mx-alpha unit ( absent from DK1622 ) carries sitA2 ( MXF1DRAFT_07313 ) , which encodes a putative lipoprotein with clear homology to the N-terminal region of SitA1 , though the C-terminal domains are unrelated ( Figure 2A ) . The sitA2 gene is flanked by a sitB1 homolog , sitB2 , and a downstream candidate immunity gene , sitI2 ( Figure 2A ) . The third Mx-alpha unit , which is shared between the ancestral strain and DK1622 , also appears to contain a sitBAI operon . Although the sitA3 gene ( MXF1DRAFT_05864 or MXAN_1899 ) has low sequence homology with sitA1 and sitA2 , the three genes nevertheless share several key features: ( 1 ) sitA3 is preceded by sitB3 , which is homologous to sitB1 and sitB2 , ( 2 ) sitA3 occupies a similar position within its Mx-alpha unit as the other sitA genes , ( 3 ) sitA3 encodes an OM lipoprotein signal sequence , ( 4 ) sitA3 encodes a predicted C-terminal tRNase toxin domain , and ( 5 ) sitA3 is adjacent to a downstream putative immunity gene , sitI3 ( Figure 2A ) . This analysis suggests that the three Mx-alpha units each contain distinct sitBAI toxin/immunity operons . To determine whether SitA lipoproteins function as toxins , we expressed each sitBAI cassette in DK1622 and tested the competitive fitness of the resulting inhibitor strains against parental DK1622 target cells that lack the corresponding sitI gene . Target strains were labeled with fluorescent markers and co-cultured with inhibitor strains on agar for 24 hr . Competition outcomes were assessed by competitive index , which is the ratio of target cells to toxin-producing inhibitor cells at 24 hr relative to the starting ratio . For example , a competitive index of 0 . 01 indicates that the ratio of target cells to inhibitor cells decreased 100-fold . In all instances , sitBAI-expressing inhibitor cells significantly outcompeted target cells , whereas the mock-inhibitor did not ( Figure 2B ) . Delivery of SitA1 and SitA2 over a 24 hr period induced filamentation and lysis of target cells , whereas SitA3 induced rounding and lysis of target cells ( Figure 2—figure supplement 1 ) . Furthermore , a ∆traA mutation in either the target or inhibitor strain abolished the inhibitor’s competitive advantage ( Figure 2B ) . We note that assessing competitive index by microscopy gives a quick and reproducible metric of one strain’s ability to outcompete another , but does not capture the full dynamic range of competition because many enumerated target cells have severe morphological abnormalities and are likely not viable at the 24 hr time point . However , microscopy allows competitive indices to be determined for these otherwise WT strains , which are not easily amenable to enumeration as colony forming units ( CFU ) because they form extraordinarily cohesive biofilms in isogenic co-cultures ( dependent on type IV pili ) . These results show that SitA lipoproteins provide a competitive advantage , conferring the ability to kill and/or inhibit the growth of competitors in a TraA-dependent manner . To examine SitA localization , we generated an mCherry reporter that carries the N-terminal lipobox from SitA1 . Cells expressing this fusion have membrane-localized fluorescence as expected for a lipoprotein ( Figure 2C ) . The TraA-dependent function of SitA shown in Figure 2B suggests that the protein is delivered by OME . Therefore , we tested whether the reporter fusion is transferred between cells . We co-cultured the reporter strain with a target strain expressing cytoplasmic GFP ( which is not exchanged [Wei et al . , 2011] ) and microscopically assayed for transfer of the reporter . At 6 hr of co-culture , we observed the mCherry signal present in the cell envelope of the GFP target strain , indicating cell-to-cell transfer of the SitA1-mCherry reporter ( Figure 2C , upper panel ) . Deletion of traA in the GFP target strain prevented the acquisition of mCherry signal ( Figure 2C , lower panel ) , recapitulating our earlier findings that OM-localized reporters are exchanged between cells in a TraA/B-dependent manner ( Nudleman et al . , 2005; Wei et al . , 2011; Pathak et al . , 2012 ) . Because inner membrane lipoproteins are not transferred during OME ( Wei et al . , 2011 ) , these data also suggest that the lipobox directs SitA1 to the OM . We confirmed that full-length SitA1 localizes to the cell envelope using immunofluorescence microscopy to detect FLAG epitope-tagged SitA1 in formaldehyde-fixed cells ( Figure 2D ) . Taken together , these results demonstrate that SitA1 resides in the OM and is transferred cell-to-cell by OME . The fact that sitI1 expression protects WT DK1622 cells from swarm inhibition suggests that this gene encodes an immunity protein that neutralizes SitA1 toxicity . To determine whether SitI proteins block SitA-mediated growth inhibition , we expressed each sitI allele individually in DK1622 ∆sitBAI3 cells and co-cultured the resulting strains with strains that express each of the three sitBAI cassettes . For each competition , only strains that express the cognate sitI were protected from growth inhibition ( Figure 2E ) , consistent with immunity function . Immunity proteins typically interact with the C-terminal domain of polymorphic toxins ( Zhang et al . , 2012; Poole et al . , 2011 ) . To test whether this was true for SitA , we expressed the predicted C-terminal toxin domains ( CTD ) of each SitA toxin in E . coli MG1655 under the inducible PBAD promoter . Expression of SitA1-CTD and SitA3-CTD blocked growth , but co-expression of cognate sitI from a second plasmid restored cell growth ( Figure 2F ) . These results confirm that SitI proteins specifically neutralize cognate SitA toxins . In addition , because the SitA-CTD constructs lack secretion signal sequences , these data show that the domains exert their toxic effects in the cytoplasm . We also tested SitA2-CTD expression constructs , but none inhibited E . coli MG1655 growth . Because SitA2-mediated inhibition is obvious in M . xanthus competition co-cultures ( Figure 2B and E ) , we tested the SitA-CTD expression constructs in M . xanthus and found that each inhibited cell growth ( Figure 2—figure supplement 2A ) . Therefore , SitA2-CTD is indeed toxic when expressed in the cytoplasm of M . xanthus . Given that sitBAI2 expression confers a significant advantage in competition co-culture ( see Figure 2B ) , it is unclear why swarm inhibition was not observed with the sitA1– nonmotile ancestral strain ( see Figure 1B , row 5 ) , considering these cells should still deploy SitA2 and that DK1622 lacks the SitI2 immunity protein . Therefore , we tested whether nonmotile ∆Mx-alpha cells that ectopically express sitBAI2 are able to inhibit DK1622 swarming . In agreement with the prior result , we found that DK1622 motility was only partially inhibited by the sitBAI2-expressing strain ( Figure 1C , row 1 ) . This result confirms that SitA2 contributes to the swarm inhibition phenotype , but is not sufficiently potent by itself to block outward swarming of DK1622 . Together , these results indicate that SitA1 is the major swarm-inhibition toxin . We note that SitA3 does not contribute to the originally observed swarm inhibition phenotype because both ancestor and DK1622 strains contain the sitBAI3 operon ( see Figure 2A ) . However , we found that a nonmotile strain expressing sitBAI3 fully inhibits the motility of a DK1622 ∆sitBAI3 strain that lacks the sitI3 immunity gene ( Figure 1C , row 2 ) . Homologous CTDs are often associated with different toxin delivery systems from phylogenetically distant bacteria ( Zhang et al . , 2012 ) . SitA3-CTD is homologous to a previously characterized tRNase domain found at the C-terminus of CdiA from Burkholderia pseudomallei 1026b ( Morse et al . , 2012; Nikolakakis et al . , 2012 ) and an orphan CdiA-CTD encoded by Yersinia pseudotuberculosis YPIII ( Figure 3—figure supplement 1 ) . To determine whether SitA3-CTD also has tRNase activity , we expressed the toxin in E . coli and compared its activity to the CDI toxins . Induction of SitA3-CTD expression inhibited cell growth in the same manner as the CdiA-CTD toxins ( Figure 3A ) . Examination of tRNA from SitA3-CTD intoxicated cells revealed cleavage of tRNAUGCAla , similar to the specific tRNase activity of the B . pseudomallei toxin ( Figure 3A ) . 10 . 7554/eLife . 29397 . 008Figure 3 . Toxic function of SitA1 and SitA3 CTDs . ( A ) SitA3-CTD is a toxic tRNase . Expression of the indicated CTDs was induced with arabinose in E . coli , and cell growth monitored by measuring the optical density of the cultures at 600 nm ( OD600 ) . OD600 values are reported as the average ± standard error for three independent experiments ( left ) . RNA was isolated after 90 min of toxin expression and analyzed by Northern blot hybridization using probes to the indicated tRNAs ( right ) . The arrow indicates cleaved tRNAUGCAla . ( B ) SitA1-CTD has DNase activity . E . coli cells were stained with DAPI at 0 hr and after 6 hr of toxin expression . Cells expressing sitA1-CTD became filamentous and lost DAPI staining at 6 hr ( left ) . In contrast , sitA3-CTD expressing cells retained DAPI staining ( right ) , though their nucleoids became compacted ( yellow arrow ) . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 00810 . 7554/eLife . 29397 . 009Figure 3—figure supplement 1 . Alignment of SitA3-CTD with CdiA-CTD tRNase toxins . SitA3-CTD shares homology with CdiA-CTD domains from Burkholderia pseudomallei 1026b ( BP1026B_II2207 ) and Yersinia pseudotuberculosis YPIII ( Ga0077885_11584 ) . Predicted nuclease active-site residues are conserved and marked in red ( Morse et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 009 Next , we investigated the toxic activities of SitA1-CTD and SitA2-CTD . We had previously observed that SitA1 induces cell filamentation and loss of DAPI staining in M . xanthus target cells , which is consistent with DNase activity mediated by the predicted Colicin-DNase domain ( Pfam 12639 , E = 6 . 7 e-21 ) containing the WHH motif ( Zhang et al . , 2011 ) . To test this , we induced SitA1-CTD expression in E . coli and found that cells became filamentous and had reduced DAPI stain signal ( Figure 3B ) . By contrast , E . coli cells that were intoxicated by SitA3-CTD retained DAPI staining , though their nucleoids became more compact ( Figure 3B ) . Together , these results suggest that SitA1-CTD has DNase activity . HMM-HMM comparison ( HHpred [Söding et al . , 2005] ) of C-terminal residues 699–783 of SitA2 revealed distant homology to another CdiA-CTD from Y . pseudotuberculosis YPIII ( locus tag , Ga0077885_11586 ) , which was previously characterized as a DNase ( Morse et al . , 2015 ) . To examine this possibility , we expressed each SitA-CTD in M . xanthus under the control of an IPTG-inducible promoter . Expression of SitA2-CTD in M . xanthus resulted in cell filamentation and reduced DAPI staining ( Figure 2—figure supplement 2B ) , suggesting that SitA2-CTD degrades DNA . Expression of SitA1-CTD and SitA3-CTD in the cytoplasm of M . xanthus cells yielded similar results as when they were expressed in E . coli ( Figure 2—figure supplement 2B ) , although the DAPI signal from SitA3-CTD expressing cells was brighter than the control and many cells contained two distinct nucleoids ( Figure 2—figure supplement 2B ) , suggesting a block in cell division . To determine the phylogenetic distribution of SitA toxins , we conducted a BLAST search using the N-terminal domains of SitA1/2 ( which are homologous ) and SitA3 as query sequences . This search recovered >100 sitA orthologs that are common in the Myxococcales ( Supplementary file 1 ) . More sensitive search algorithms such as HMMER ( Finn et al . , 2011 ) failed to identify significant homologs outside of the Myxococcales . Consistent with the finding that SitA is delivered through OME , all orthologs contain lipoboxes and are only found in species that contain traAB . Moreover , SitA C-terminal domains are variable and typically show homology to nuclease domains when subjected to HMM-HMM comparison using HHpred ( Supplementary file 3 ) . Interestingly , many sitA genes are not linked to upstream sitB orthologs , particularly in species that are distantly related to M . xanthus ( Supplementary file 1 ) . This suggests that SitB may not be required for SitA function , or perhaps that SitB proteins , encoded at unlinked loci , function promiscuously between multiple SitA proteins . Notably , some sitA loci are found outside of Mx-alpha-like elements , however , these genes are typically adjacent to other mobile genetic elements . To test cross-genotype compatibility of SitA orthologs , we cloned two sitA gene cassettes from Myxococcus fulvus HW1 for heterologous expression in M . xanthus . One of these operons ( sitAI3Mf1 ) does not contain a sitB gene . As predicted , M . xanthus cells that express heterologous sitAI3Mf1 or sitBAI1Mf1 outcompeted the parental strain in a traA-dependent manner ( Figure 2—figure supplement 3 ) . These results indicate that SitA toxin delivery is not limited by its specific species/strain of origin , and that the systems are functional after horizontal gene transfer ( HGT ) of a minimal set of components ( sitAI ) . Our results show that sitBAI-expressing cells inhibit OME-compatible strains that lack a cognate SitI immunity protein . We hypothesized that this antagonism should be sufficient to mediate territorial exclusion . Territorial exclusion promotes physical segregation of nonself organisms ( Gibbs and Greenberg , 2011 ) , which in turn drives both diversification and maintenance of cooperation ( Papke and Ward , 2004; Velicer and Vos , 2009 ) . Territorial exclusion between wild M . xanthus genotypes , including those that are closely related and that are isolated in close proximity to one another has been well studied , but the specific determinants underlying this behavior are unknown ( Vos and Velicer , 2009; Rendueles et al . , 2015; Vos and Velicer , 2006; Wielgoss et al . , 2016 ) . We tested whether SitA is sufficient for territorial exclusion by conducting colony merger assays with every combination of DK1622 strains expressing one of the five described sitAI alleles . In these assays , two liquid cultures are spotted next to one another on agar , and the colonies are allowed to swarm toward each other . If the converging swarms merge , then the strains are considered compatible . For each combination , the expression of different sitAI cassettes resulted in dramatic lines of demarcation between the two strains ( Figure 4A ) , and the formation of these demarcation zones was traA-dependent ( Figure 4B ) . These results show that otherwise isogenic strains are rendered socially incompatible and geographically isolated by the acquisition of a single sitAI cassette . 10 . 7554/eLife . 29397 . 010Figure 4 . sitAI alleles determine the social compatibility of M . xanthus swarms . ( A ) M . xanthus colonies expressing identical sitAI cassettes merge ( as illustrated by the green arrow ) when spotted adjacent to one another ( top of each column ) . Strains that express different sitAI cassettes form demarcation zones between colonies ( illustrated by the red arrow ) . Labels on the left indicate toxin expressed by left colony , while top labels indicate toxin expressed by colony on the right . Green borders indicate colony merging and red indicates demarcation . ( B ) Demarcation zone formation is traA-dependent . Bar , 5 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 010 We previously found that swarm inhibition occurs efficiently even when motile cells outnumber the antagonistic nonmotile strain 40 to 1 ( Dey and Wall , 2014 ) . This observation suggests that an individual SitA-expressing cell can inhibit many targets . To further explore this phenomenon , we quantified viable target cells in a series of competition co-cultures in which the ratio of SitA producing cells to targets was progressively decreased by factors of 10 . To facilitate CFU enumeration in these experiments , we used ∆pilA cells , which are unable form type IV pili-dependent biofilms . We found that target cell CFU were reduced approximately 106-fold when the strains were mixed at a 1:1 ratio ( Figure 5A ) . The higher degree of killing reported here ( compared to competitive index in Figure 2B ) provides a clearer understanding of the killing efficiency because the CFU assay measures a broader dynamic range of viable cell number . Remarkably , the SitA-producing strain still reduced target cell viability >104-fold in co-cultures seeded at a 1:1000 ratio of inhibitors to target cells ( Figure 5A ) . These observations imply that each inhibitor cell intoxicates several thousand target cells during co-culture . In one explanation we hypothesized that OME delivery allows a series of SitA transfer events from one target cell to other cells . We consider this serial transfer mechanism plausible if translocation of all toxin molecules from the target cell OM to the cytoplasm is not completed before subsequent OME events occur . This model predicts that SitA toxins could spread through the population like an infectious agent , intoxicating target cells that never made direct contact with the producer . To test this hypothesis , we conducted three-strain competitions with ( 1 ) a sitBAI1 inhibitor strain that contains M . fulvus traAB ( traABMf ) as its only traAB alleles , ( 2 ) a susceptible target strain that contains M . xanthus traABDK1622 alleles and thus is incompatible for OME with the inhibitor , and ( 3 ) a susceptible intermediary strain that carries both traABMf and traABDK1622 ( Figure 5B ) . If serial toxin transfer occurs , the traAB merodiploid strain should act as an intermediary carrier/conduit to deliver toxin to traABDK1622 targets ( Figure 5B ) . As a control , we first showed that traABMf inhibitors do not inhibit traABDK1622 targets ( Figure 5C ) , consistent with the incompatibility of their traAB alleles . Importantly , inclusion of intermediary cells , which are inhibited ( Figure 5D ) , also resulted in the inhibition of traABDK1622 target cells ( Figure 5E ) . As expected , the intermediary strain ( which lacks sitBAI1 ) did not inhibit traABDK1622 targets in co-cultures containing only those two strains ( Figure 5F ) . To exclude a SitA-independent mechanism of target cell inhibition , we conducted the same three-strain competition , but provided traABDK1622 targets with the sitI1 immunity gene . In this latter co-culture , the intermediary strain was inhibited , but traABDK1622 targets were not ( Figure 5G ) . Finally , we tested an intermediary strain that lacks traA and found that neither intermediary nor target cells were inhibited during co-culture ( Figure 5H ) . 10 . 7554/eLife . 29397 . 011Figure 5 . SitA toxins are serially transferred by OME . ( A ) Viable cells ( CFU ) of a target population as a function of inhibitor to target cell ratio quantifies the efficiency of SitA1 and OME delivery . Strains were co-cultured on agar for 48 hr at indicated ratios before determining CFU of the marked ( Kmr ) target strain . ( B ) Experimental design to test serial transmissibility of SitA1 . The grey cell produces the SitA1 toxin and contains traABMf alleles . The target cells ( green ) are susceptible , but carry incompatible traABDK1622 alleles that preclude OME with inhibitors . Intermediary cells ( red ) express both traAB alleles . ( C–F ) Competitive indices of intermediary ( red ) and target ( green ) strains from two- and three-strain co-cultures . ( G ) Three-strain competition when the target strain expresses SitI1 . ( H ) Three strain competition when the intermediary strain is ∆traA . Competition outcomes were determined at 24 hr by fluorescent microscopy . Competitive index was calculated relative to the inhibitor ( C–E , G–H ) or relative to intermediary strain ( F ) . Starting ratio was 1:5:5 inhibitor to intermediary to target . ( I ) Serial transfer of the SitA1-mCherry fusion . The left panel shows a 10:1:1 mixture of sitA1-mCherry cells to intermediary to target visualized at 0 and 6 hr . Red arrow indicates a representative example of an intermediary cell that expresses cytoplasmic tdTomato ( which does not transfer ) . Yellow arrows indicate GFP-labeled target cells that have acquired an OM-localized mCherry signal at 6 hr . Boxes represent the number of mCherry positive GFP cells out of 100 . Right panel: otherwise identical experiment omitting the intermediary strain . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 01110 . 7554/eLife . 29397 . 012Figure 5—figure supplement 1 . TraA is not transferred during OME . ( A ) TraA-mCherry fusion protein is functional . Cells expressing TraA-mCherry were mixed with either traA+ ( top ) or ∆traA ( bottom ) donors that express an OM lipoprotein reporter , SSOM-sfGFP . Arrows indicate a TraA-mCherry cell that has acquired SSOM-sfGFP after 6 hr on agar media . White insets indicate the number of recipient cells where transfer was detected from a total of 100 scored recipient cells . ( B ) TraA-mCherry is not transferred during OME . Top: positive control in which donors that express an OM lipoprotein reporter , SSOM-mCherry , were mixed with cytoplasmic GFP expressing cells for 24 hr on agar media . Arrows indicate a GFP cells that has acquired the mCherry signal . Bottom: TraA-mCherry expressing cells were mixed with cells expressing cytoplasmic GFP for 24 hr . No mCherry signal was visible in the GFP cells . Bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 012 To directly visualize serial transfer , we co-cultured traABMf cells that express the SitA1-mCherry fusion ( described in Figure 2 ) with traAB merodiploid intermediary cells that express cytoplasmic tdTomato , and GFP-labeled , traABDK1622 target cells . Microscopic examination of target cells at 6 hr revealed that all had acquired mCherry fluorescence , with the signal localized to the cell envelope ( Figure 5I , left panels ) . By contrast , no SitA1-mCherry transfer was detected when the intermediary was absent from the co-culture ( Figure 5I , right panels ) . Although these results are consistent with serial transfer of SitA , it is also possible that target cells acquire TraAMf and/or inhibitor cells acquire TraADK1622 by OME-dependent exchange of TraA with the intermediary strain , which would then allow direct transfer of SitA between inhibitor and target cells . To test if TraA is transferred during OME , we used a TraA-mCherry fusion to monitor transfer of TraA . The TraA fusion promoted efficient transfer of an OM sfGFP reporter ( Figure 5—figure supplement 1A ) , demonstrating that it is functional to catalyze OME . However , TraA-mCherry itself was not transferred ( Figure 5—figure supplement 1B ) . One explanation for why TraA does not transfer is that it interacts with TraB , which contains an OmpA domain known to bind the cell wall . For this reason , we suspect TraA is anchored to the cell envelope and is unable to transfer ( Cao and Wall , 2017 ) . Taken together , these results indicate that SitA1 can be transferred from the initial target to secondary recipients , supporting a model in which SitA acts like an infectious agent that disseminates through a population by OME . In M . xanthus and its close relatives , sitA is typically accompanied by an overlapping sitB cistron . In the case of DK101 the genes overlap by 11 bp in all three sitBAI cassettes . SitB shows no significant homology to other proteins or domains using HMMER or HHpred , though it does contain a type I signal sequence ( SignalP 4 . 1 [Petersen et al . , 2011] ) . I-TASSER ( Zhang , 2008 ) predicts that SitB adopts a transmembrane β-barrel structure characteristic of OM proteins . To examine the role of SitB1 , we tested the activity of inhibitor cells that express either sitBAI1 or sitAI1 ( cells lack sitB1 ) against a susceptible target strain . The inhibitors in these experiments also carried a ∆sitB3 mutation to eliminate the possibility of promiscuous interactions between SitB3 and SitA1 . At a 1:1 ( inhibitor to target ) ratio , sitAI inhibitors had less of an advantage against targets than sitBAI inhibitors , but still retained activity compared to the mock inhibitor control ( Figure 6A , left ) . The sitAI inhibitors were less effective at a 1:10 ratio , and at 1:100 were indistinguishable from mock inhibitors ( Figure 6A ) . In contrast , sitBAI1 inhibitors were equally effective at outcompeting the target strain at each of the three ratios ( Figure 6A ) . Thus , SitB1 contributes significantly to SitA1-mediated inhibition . 10 . 7554/eLife . 29397 . 013Figure 6 . SitB contributes to SitA function and serial transfer . ( A ) The indicated SitA1 inhibitor strains were co-cultured with target cells at three different inhibitor to target ratios . Competitive index was measured at 24 hr by counting the ratios of fluorescently marked cells . Asterisks indicate level of statistical significance , ns = not significant . P-values of indicated comparisons from left to right: 0 . 0002 , 0 . 006 , 0 . 0257 , 0 . 9359 . ( B ) Serial transfer was monitored as in Figure 5 using sitBAI or sitAI inhibitors that express traABMf . Co-cultures were seeded at a 10:1:1 ratio of inhibitor to intermediary to target strains . Significance indicators refer to comparisons between the inhibitor strains . P-values from left to right: 0 . 6341 , 0 . 0193 . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 013 Progressive loss of function at increasing target to inhibitor ratios with the sitAI inhibitors could indicate defects in serial toxin transfer compared to sitBAI inhibitors . Therefore , we tested serial transfer using three-strain co-cultures as described in Figure 5B . To improve the sensitivity of this assay , we increased the inhibitor to intermediary to target strain ratio to 10:1:1 to compensate for the inhibition defect of sitAI1 cells . As observed in Figure 5 , sitBAI1 cells outcompeted both the intermediary and target strains ( Figure 6B ) . Because of the high inhibitor cell ratio , the sitAI1 cells outcompeted the intermediary strain to a similar extent as sitBAI cells , but importantly , sitAI cells had little to no effect on target cells ( Figure 6B ) . We extended the experiment to 48 hr and also performed experiments with SitA-resistant intermediary cells , to increase the number of conduit cells , but again we did not observe SitAI-mediated antagonism of target cells . These results support the hypothesis that SitB promotes SitA transfer , including the serial transfer from primary to secondary target cells . Although SitB1 clearly contributes to SitA1-mediated inhibition , it is not strictly required , which may explain why many myxobacterial sitA genes are not linked to sitB . Finally , these results are congruent with our above conclusion that TraA is not transferred , because if it was , then direct transfer of SitA1 would occur between sitAI inhibitors and the target strain . However , this did not occur because the target strain was not inhibited . M . xanthus uses multiple inhibitory mechanisms to antagonize non-kin . However , because SitA toxins are serially transferred between cells , we hypothesized that they should be powerful determinants of competitive outcomes during inter-genotype conflict . To test this hypothesis , we quantified the contribution of TraA and SitA to competitive outcomes in co-cultures of DK1622 with wild M . xanthus soil isolates . Isolates A66 and A88 ( from Tübingen , Germany ) ( Vos and Velicer , 2006 ) and DK801 ( from California , USA ) ( Martin et al . , 1978 ) contain traA alleles in the same recognition group as DK1622 ( originally isolated from Iowa , USA ) ( Pathak et al . , 2013; Dey et al . , 2016 ) . We compared the fitness outcomes of WT , ∆traA , and ∆sitBAI3 genotypes when co-cultured with these environmental isolates by monitoring the ratio of fluorescently labeled DK1622-derived cells to isolate cells at 4 , 8 and 24 hr . As a second metric , we enumerated CFU of the DK1622-derived strains at the 24 hr time point . In every case , mutant strains that cannot deploy SitA3 had dramatically decreased competitive fitness outcomes and viability compared to WT ( Figure 7A ) . Remarkably , against all three isolates , the presence of traA was the determining factor in which strain prevailed , demonstrating up to a 106-fold swing in strain ratio ( vs . A66 ) and a near 107-fold difference in CFU ( vs . DK801 ) between WT and ∆traA strains ( Figure 7A ) . These results indicate that the ability to deliver SitA3 is a dominant determinant of competitive fitness under these conditions . Not surprisingly , the finding that environmental isolates outcompeted and killed ∆traA strains ( Figure 7A ) , confirms the existence of OME-independent killing mechanisms at play . Interestingly , these experiments revealed that ∆traA cells had less competitive fitness than ∆sitBAI3 , indicating a competitive fitness defect in ∆traA cells beyond just the inability to deploy SitA3 . As a control , we competed the DK1622 genotypes against isolates A23 and A47 ( from Tübingen [Vos and Velicer , 2006] ) , which are outside the DK1622 TraA recognition group ( Pathak et al . , 2013 ) . ∆traA and ∆sitBAI3 genotypes resulted in similar competitive outcomes as WT as would be expected when inter-strain OME does not occur and SitA3 cannot be deployed ( Figure 7B ) . These results show that SitA contributes significantly to fitness during competition with TraA-compatible non-kin genotypes , and likely plays a key role in competition and survival in nature . 10 . 7554/eLife . 29397 . 014Figure 7 . TraA and SitA are dominant determinants of competitive outcome within TraA recognition groups . ( A ) Line graphs represent strain ratio over time when the three indicated , DK1622-derived strains , which were fluorescently labelled , were competed with wild isolates ( A66 , A88 , DK801 ) . These isolates belong to the same TraA recognition group as DK1622 . Histograms indicate viable cells ( CFU ) of the DK1622-derived strains ( Tcr ) after the 24 hr competition . ( B ) Identical experiments as in A , except the lab strains were competed with wild isolates that belong to different TraA recognition groups . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 014 Here , we describe a novel family of proteins that carry polymorphic toxin domains and are delivered between myxobacteria by OME . SitA proteins are similar to other polymorphic toxins in that they carry diverse C-terminal domains , are neutralized by cognate immunity proteins , and are delivered in a cell contact-dependent manner . However , they are unique in their N-terminal domains , and in that they are lipoproteins transferred with other cargo during OME . To our knowledge , this is the first example of a polymorphic toxin system in which the toxin itself is a lipoprotein . Unlike CDI , T4SS , T6SS and T7SS toxins , there appears to be no requirement for a specialized apparatus to export the toxins . Instead , toxins are exchanged bi-directionally and simultaneously during OME . Therefore , SitA delivery likely only requires OM localization of the toxin and compatible TraA receptors . This discovery highlights the diversity of mechanisms used by bacteria to deliver polymorphic toxins . Importantly , the SitA toxin family constitutes a second identity constraint upon OME with partner cells . For two cells to engage in OME , not only must they present compatible TraA receptors , but they must also contain immunity proteins to each other’s toxins . In this ‘recognize and verify’ system ( Wall , 2016 ) , if the latter constraint is not met , then the recipient of the toxin is poisoned . TraA homotypic interactions alone are considered kind or greenbeard recognition , in which social interactions are based on a single gene locus . The finding that myxobacteria verify relatedness with sitAI confirms the notion that myxobacteria apply a bona fide kin discrimination mechanism during OME by requiring identity verification at multiple polymorphic loci . Interestingly , this system allows TraA interactions to promote contrasting behaviors – cooperative or antagonistic – depending on relatedness . Either outcome makes OME potentially beneficial regardless of the partner by conferring the ability to both share goods with clonemates and poison non-kin . SitA delivery range is restricted to within a single TraA recognition group . Considering traA allele diversity is high ( Pathak et al . , 2013 ) , this significantly limits the use of SitA to related but nonself individuals . This suggests that one of the primary functions of SitA is the discrimination of exchange partners , consistent with the notion that sharing large amounts of goods with non-kin is costly . Within TraA recognition groups , OM material is a shared good; a resource to be guarded from exploitation by OME compatible , yet nonself populations . Myxobacteria achieve this safeguard by inextricably linking the delivery of these goods with the delivery of SitA toxins . Another example is the Burkholderia thailandensis CDI system , which couples a communication signal with polymorphic toxin delivery during biofilm formation ( Anderson et al . , 2014 ) . Similarly , the CDI system of E . coli mediates both antagonism and cooperative intercellular adhesion to related cells ( Ruhe et al . , 2013 , 2015 ) . In Proteus mirabilis , IdsD/IdsE interactions communicate identity and may promote cooperative behaviors ( Cardarelli et al . , 2015 ) . This intercellular communication is also coupled to toxic T6SS effector delivery ( Wenren et al . , 2013 ) . In these examples , organisms link goods , signals , and/or cooperative behaviors to polymorphic toxin delivery , which ensures that potential cooperators are related . Differential acquisition of antagonistic systems can affect cooperation compatibility between originally identical genotypes . We demonstrated that when two otherwise isogenic colonies express different SitA toxins , they are no longer able to merge swarms . The acquisition of a single sitAI operon would therefore alter strain identity and population structure between previously clonal cells . However , the impact of SitA on strain identity and population structure in natural soil habitats depends on two factors: ( 1 ) strains that belong to the same traA recognition group must exist in proximity , and ( 2 ) sitA loci must be sufficiently diverse to ensure different toxin/immunity types are represented at fine geographic scales . Velicer and colleagues have examined the compatibility of natural M . xanthus isolates obtained from a centimeter-scale plot of soil ( Vos and Velicer , 2009; Wielgoss et al . , 2016 ) . Colony merger assays between these geographically proximal strains reveal compatibilities among the most closely related isolates , but also strong incompatibilities between strains that differ by only several dozens of mutations outside of the Mx-alpha region ( Wielgoss et al . , 2016 ) . Indeed , these incompatibilities correlate with gene variation at hyper-variable Mx-alpha loci , where sitBAI genes commonly reside . Our analysis of their published sequences reveals 69 total and 15 unique sitA alleles distributed over 22 isolates ( see Materials and methods for search criteria ) . Between these strains there are two traA alleles that are known to be incompatible ( Pathak et al . , 2013 ) . Within each TraA recognition group we have observed an apparently high degree of correlation between the published colony merger compatibility of the strains ( Wielgoss et al . , 2016 ) and the sitA genes they possess . This suggests that sitBAI polymorphisms contribute to swarm incompatibility and inter-strain competition among natural soil isolates . We are currently investigating this possibility . Moreover , because Mx-alpha produces defective phage particles that promote specialized transduction ( Starich and Zissler , 1989; Starich et al . , 1985 ) , these elements are apparent hotspots for HGT , perhaps explaining the high degree of Mx-alpha variation discovered between otherwise related isolates ( Wielgoss et al . , 2016 ) . These variations likely contribute to the emergence of new compatibility types which underlie complex population structures , and explain the observation of rapidly evolving social antagonisms in M . xanthus ( Velicer and Vos , 2009 ) . Thus , TraA and SitA may act as powerful evolutionary drivers of myxobacteria diversification . sitAI genes are often associated with prophage-like elements or other mobile elements and thus are likely acquired by HGT . By conferring a fitness advantage to their host , they may play an important role in transmission and retention of mobile DNA . For example , a HGT event into one cell in a population endowing it with a novel sitBAI operon allows that cell to infect its clonemates with toxins , thus ensuring the propagation of that element within the population . Similarly , the loss of the element would be lethal because the surrounding cells harbor this toxin-immunity pair would kill susceptible cells . Importantly , this model explains why many lab strains have stably maintained three large tandem repeats of Mx-alpha , which is expected to be genetically unstable ( Starich and Zissler , 1989; Roth et al . , 1996 ) . In cases in which strains have spontaneously lost Mx-alpha units ( Dey et al . , 2016 ) , those events likely occurred during propagation in liquid media , where OME cannot occur . Mx-alpha has the attributes of a selfish or addictive element that exploits the social nature of myxobacteria and OME . However , the origin of sitBAI loci is unclear because these genes also reside outside of selfish elements in myxobacterial genomes . SitA toxins are uniquely powerful determinants of identity , likely because they are transmitted as infectious agents between recipient cells . The infectious model is consistent with the observations that individual M . xanthus cells typically make contact with multiple cells simultaneously within a swarm , OME is constitutively active , and that prodigious amounts of material are transferred during OME ( Nudleman et al . , 2005 ) . Furthermore , SitA entry into the cytoplasm occurs by a secondary and uncoupled pathway to OME ( Dey et al . , 2016; Dey and Wall , 2014 ) . Thus , it is possible that SitA lingers in the OM of the primary target long enough to allow transfer to secondary target cells through subsequent OME events ( Figure 8A ) . Perhaps a cellular protein is required for SitA cell entry ( Dey et al . , 2016; Dey and Wall , 2014 ) , but this protein is outnumbered by SitA molecules in the OM , making cytoplasmic entry a rate-limiting step . Based on these inferences , we propose two non-exclusive models that explain serial transfer: ( 1 ) Following transfer of SitA to a primary infected cell , OME with a secondary cell occurs before the full complement of SitA enters the cytoplasm of the initial recipient ( Figure 8B ) ; or ( 2 ) that three or more cells are engaged in OME simultaneously ( Figure 8C ) . Our results suggest that SitB functions to promote serial transfer . Given that ∆sitB inhibitors have defects in direct transfer , serial transfer could be blocked simply by a decrease in number of SitA molecules delivered or a decrease in rate of delivery . Alternatively , SitB could stabilize SitA in the OM of the inhibitor and/or target cell , thereby increasing OM dwell times to promote serial transfer . The finding that SitB is an accessory protein is consistent with bioinformatics analysis in which many of the sitA genes that reside outside of the M . xanthus species are not linked with a sitB gene . In the case of M . xanthus isolates ( Wielgoss et al . , 2016 ) , we found only 3 of 69 sitA genes were not associated with sitB . The mechanism of serial transfer , the function of SitB , and the ability of SitA to traverse the cell envelope are topics for future study . With respect to the latter point , in prior work we found that a mutation that disrupts the inner membrane protein , OmrA , renders target cells resistant to SitA1 ( Dey et al . , 2016; Dey and Wall , 2014 ) . Thus , as proposed for CDI toxins , one possibility is that SitA toxins exploit inner-membrane proteins to gain access to the cytoplasm ( Willett et al . , 2015 ) . 10 . 7554/eLife . 29397 . 015Figure 8 . Model for serial transfer of SitA . ( A ) SitA is delivered cell-to-cell by OME . After OME , SitA may enter the cytoplasm or linger ( lag ) in the OM . ( B ) In the delayed entry model , the infected cell can undergo OME of SitA to another naïve cell before all SitA molecules enter the cytoplasm and before cell death . n = number of target cells poisoned by infected cell . ( C ) Alternative , but non-exclusive model in which OME and SitA transfer occurs between three or more cells simultaneously . Here SitA is delivered to a tertiary cell via an intermediary cell . DOI: http://dx . doi . org/10 . 7554/eLife . 29397 . 015 Remarkably , SitA appears to determine the competitive outcome of two-strain co-cultures between DK1622 and wild-isolates , despite the T6SS and a host of other antagonistic machinery at play ( Konovalova et al . , 2010; Smith and Dworkin , 1994 ) . We hypothesize that the rapid spread of SitA by serial transfer may disrupt the antagonistic capabilities of competitors . When deployed between converging swarms , serial transfer provides a mechanism to inhibit cells behind the front-lines . Thus , SitA mediated antagonism results in the formation of distinct territorial boundaries ( see Figure 4 ) , which in turn minimize OME , resource sharing , and social interactions between non-clonal swarms . The selective pressure from SitA antagonism within a TraA recognition group may help drive the generation and fixation of traA polymorphisms that determine recognition specificity . For instance , we recently showed that simply substituting a single amino acid residue in TraA can alter recognition specificity while retaining OME function ( Cao and Wall , 2017 ) . More broadly , it is a puzzle how organisms select and maintain genetic variation in social genes such as traA . Diversification of beneficial greenbeard genes is theorized to be selected against because social groups with more common alleles receive benefit more often than those with less common alleles . This pressure to possess the allele that gains the most benefit is thought to ultimately erode allele diversity that originally allowed discrimination . This problem is known as Crozier’s paradox ( Strassmann et al . , 2011; Crozier , 1986 ) . Our results provide one solution to this paradox in that variation at a second locus ( sitA , a ‘harming greenbeard' [Gardner and West , 2010] ) exerts selective pressure to diversify TraA , a helping greenbeard ( Wall , 2016 ) . For example , if one strain is killed by another via SitA , a TraA mutation that alters specificity within the losing population would be immune and retain OME function , and would thus be selected for . We suggest that other greenbeard systems could involve a similar balance between antagonism and cooperation that promotes maintenance of diversity for beneficial greenbeard genes . Our results provide a description of a novel polymorphic toxin system that helps direct cellular goods to clonemates to promote multicellular cooperation . Polymorphic toxin systems are widely prevalent in bacteria and their role in population structure , ecology and evolution of microbes is only beginning to be understood . Importantly , this study highlights the diversity of delivery mechanisms for these domains and how they have adapted to the lifestyle of their host genomes . All strains are listed in Supplementary file 2A . M . xanthus was routinely grown in CTT medium [1% casitone; 10 mM Tris⋅HCl ( pH 7 . 6 ) ; 8 mM MgSO4; 1 mM KH2PO4] in the dark at 33°C . E . coli TOP10 and MG1655 were grown in LB media at 37°C . As needed for selection or induction , media were supplemented with kanamycin ( 50 µg/mL ) , oxytetracycline ( 10 µg/mL ) , ampicillin ( 100 µg/mL ) , streptomycin ( 50 µg/mL ) , arabinose ( 0 . 2% ) , or IPTG ( 1 mM ) . TPM buffer ( CTT without casitone ) or PBS was used to wash cells . CTT or LB agar was used as a solid growth medium for routine strain maintenance . For all assays , strains were grown to logarithmic growth phase , washed , and re-suspended to the appropriate density . All plasmids and primers are listed in Supplementary file 2B , C . Plasmids were constructed and maintained in E . coli and subsequently electroporated into M . xanthus . In the case of cloning IPTG-dependent sitA-CTDs , plasmids were maintained in XL1-Blue , which overexpresses LacI and reduces clone toxicity . Insertion mutations were created by amplifying an approximately 500 bp fragment of the gene of interest by PCR and cloning the fragment into the pCR-TOPO XL or pCR-TOPO 2 . 1 vectors ( Invitrogen , Carlsbad , CA ) . For gene expression in M . xanthus , we cloned the appropriate gene ( s ) into pMR3487 using XbaI and NdeI restriction sites and T4 DNA ligase . If the gene ( s ) of interest contained these restriction sites , we used Gibson Assembly ( Gibson et al . , 2009 ) ( New England Biolabs , Ipswich , MA ) for plasmid construction . pMR3487 recombines at a specific site in the M . xanthus chromosome and expression is induced with IPTG ( Iniesta et al . , 2012 ) . Traditional restriction endonuclease cloning was used to create pBAD30 , pCH450 and pKSAT derived plasmids in which the sitA-CTD fragments had an ATG start codon engineered into the insert . Expression of sitI genes in pKSAT is constitutive , driven by the Kmr promoter . pCV10 for GFP expression was created by ligating tandem rRNA promoters , used for expressing lacI from the pMR3487 plasmid , with EGFP and the pSWU19 plasmid backbone . This plasmid recombines into the M . xanthus chromosome at the Mx8 phage attachment site . The deletion of sitBAI3 was constructed by cloning in-frame regions flanking and partially overlapping the start and stop codons of sitB3 and sitI3 , respectively , into pBJ114 using Gibson Assembly . After recombination in M . xanthus , mutants were grown in CTT for 24 hr and plated on CTT containing 2% galactose to select for spontaneous loss of the galK marker . Deletion mutants were distinguished from WT by PCR with primers that flanked the deletion site . Competition experiments , unless otherwise noted , were done using 1:1 strain mixtures of 3 × 108 cells per mL spotted ( 20 µL ) on agar plates containing 0 . 5× CTT with 2 mM CaCl2 and 1 mM IPTG ( competition media ) . Culture spots were harvested at 24 hr and observed on glass slides by microscopy to quantify strain ratios based on fluorescent labels . Typically , between 200 and 800 cells were counted . Competitive index in all assays was quantified by calculating the change in ratio of target to toxin-producing inhibitor cells over 24 hr . For example , if the 0 hr ratio was 1 to 1 and the 24 hr ratio was 1 to 100 , the competitive index was . 01 , indicating that the target strain was outcompeted . Swarm inhibition experiments were done identically but cells were not collected and instead were imaged after 72 hr . To determine the potency of killing , competition assays were conducted where the target cell volume and density were held constant ( 50 μL , 3 × 108 cells per mL ) while the number of toxin producing cells were titrated 1 to 10 for each sample . Cells were harvested at 48 hr , serially diluted and plated on CTT containing Km to enumerate viable target cells . For the SitA1 serial transfer assay , competition was done at 1:5 or 1:5:5 mixtures of inhibitors to target ( s ) using a culture density of 3 × 109 cells per mL . For competition with environmental isolates , liquid cultures of the DK1622-derived strain and the environmental isolate were adjusted to 3 × 109 cells per mL liquid culture , mixed ( 100 µL DK1622 to 50 µL isolate ) , and spread onto competition media to ensure there were enough non-lysed cells to count by fluorescent microscopy . At the indicated time points , 2 mL of TPM was added to the agar plate , agitated with a plate spreader , and collected by pipette . Cells were centrifuged at low speed to help prevent clumping and either cell ratio was quantified by microscopy ( as above ) or the cells were resuspended in 1 mL TPM for CFU determination . Clumping was not an issue due to a non-isogenic mix of strains and massive cell lysis interfering with cell-cell adhesion . CFU of DK1622-derived strains were enumerated by 1 to 10 serial dilution and plating on CTT with oxytetracycline . All figures that contain error bars indicate the experiments were done in triplicate on different days . All statistical tests comparing two results are unpaired , two-tailed t-tests . To test for growth inhibition by toxin expression , overnight cultures of each strain grown in LB with appropriate antibiotics were adjusted to OD600 = 1 . 0 and back diluted 1 to 10 into fresh media containing 0 . 2% arabinose ( to induce expression from pBAD ) , ampicillin 100 μg/mL and streptomycin 50 μg/mL . Cultures were then incubated in a shaker at 37°C for 5 hr . 1 to 10 serial dilutions of each culture were plated on LB with 0 . 2% arabinose and 100 μg/mL ampicillin and imaged after overnight growth at 37°C . For DAPI staining , cells were grown as described above . At 0 hr and 6 hr post arabinose induction cells were adjusted to OD600 = 0 . 4 and 1 mL was collected by centrifugation . Cells were fixed in freshly prepared 4% ( vol/vol ) formaldehyde in PBS for 15 min , rotating at room temperature ( RT ) . The reaction was quenched by addition of an equal volume of 250 mM glycine ( pH 7 . 5 ) in PBS . Cells were collected and washed 3x and resuspended to 100 μL . Cells were then spotted on poly-L-lysine coated slides and incubated for 10 min . Excess liquid was removed and cells were rinsed with water . 5 μL of fluorogel-II with DAPI ( Electron Microscopy Sciences , Hatfield , PA ) was placed on cells , a coverslip was applied and cells were imaged at 100× magnification with a Nikon E800 microscope coupled to a digital imaging system ( Wei et al . , 2011 ) . To assess tRNA processing , total RNA was isolated from E . coli clones using guanidine isothiocyanate-phenol extraction as described ( Garza-Sánchez et al . , 2006 ) . Here , cultures were grown to mid-log phase and then diluted to OD600 = 0 . 05 and grown for 30 min before 0 . 4% arabinose was added to induce toxin expression . tRNAs were analyzed by Northern blot hybridization using the following 5´-radiolabeled oligonucleotide probes: tRNAUGCAla ( 5´ - TCC TGC GTG CAA AGC AG ) , tRNAICGArg ( 5´ - CCT CCG ACC GCT CGG TTC G ) , tRNACGASer ( 5´ - GTA GAG TTG CCC CTA CTC CGG ) , and tRNAGCACys ( 5´ - GGA CTA GAC GGA TTT GCA A ) . These assays used a modified competition media ( replacing 1 . 5% agar with 1 . 0% agarose for imaging and addition of 5 μg/mL oxytetracycline ) . A multichannel pipette was used to simultaneously pipette competing strain suspensions ( 1 . 5 × 109 cells per mL ) until culture spots were nearly touching . Aliquots were air dried and plates were then incubated in a humid chamber at 33°C for 3 days . Spots were viewed on an Olympus SZX10 stereomicroscope coupled to a digital imaging system . To discover SitA homologs , we performed BLAST ( Altschul et al . , 1990 ) analysis against the IMG database ( Markowitz et al . , 2012 ) , using the conserved N-terminal sequence ( first 507 amino acids of SitA1 and the first 441 amino acids of SitA3 ) as queries . HMMER ( Finn et al . , 2011 ) , HHPred ( Söding et al . , 2005 ) and I-TASSER ( Zhang , 2008 ) analysis were performed as described above using default parameters . To detect any correlation between colony merger and sitA genes from the referenced study ( Wielgoss et al . , 2016 ) , we first identified SitA homologs by performing local BLAST analysis of the sequences published at http://www . odose . nl/u/michiel/h/22-myxo-genomes-w-annotation ( Wielgoss et al . , 2016 ) using SitB1 as the query . Any gene located downstream of a SitB homolog that contained the described features of SitA was considered a SitA homolog . We also used SitA homolog sequences as queries to find any sitA genes without an accompanying sitB . Redundant sequences were removed using NCBI FASTA Tools Unique Sequences webpage . The sequences were then clustered according to >96% pairwise amino acid identity and each cluster was considered a unique polymorphic toxin group . The two closest toxin groups were 90% identical . This analysis was done blind with respect to the published colony merger compatibility types ( Wielgoss et al . , 2016 ) . Log-phase reporter and target cell liquid cultures were adjusted to 1 . 5–3 × 109 cells per mL , mixed at the indicated ratios , and plated on ½ CTT agar with 2 mM CaCl2 , +/- IPTG , as needed . Strains were incubated at 33°C for the indicated time-period , collected , and visualized by fluorescent microscopy as described ( Wei et al . , 2011 ) . Log phase cultures were resuspended to 6 × 108 cells per mL in reduced osmolarity PBS solution ( “mPBS” = 7 . 2 mM NaCl , 5 . 4 mM KCl , 10 mM Na2HPO4 , 1 . 8 mM KH2PO4 ) . 100 μL of 10% paraformaldehyde solution ( in mPBS ) and 1 μL of 5% glutaraldehyde solution ( in H2O ) were added to 400 μL of cell suspension . Each mixture was spotted on a poly-lysine coated slide . Fixation proceeded for 30 min at RT . After rinsing with mPBS , cells were permeabilized with 0 . 025% Triton X-100 for 10 min and washed . Cells were blocked for 1 hr at RT with 4% BSA in mPBS , then probed with a 1:1000 final concentration of anti-FLAG antibody ( in 4% BSA , Sigma , St . Louis , MO ) for 1 hr at RT and subsequently washed 2× for 5 min and 1× for 10 min with mPBS . Cells were then probed with a 1:2000 final concentration of secondary antibody ( in 4% BSA , Alexa Fluor 488-conjugated donkey anti-rabbit IgG; Jackson ImmunoResearch , Westgrove , PA ) for 30 min at RT and subsequently washed as before . SlowFade Gold antifade reagent ( Invitrogen ) was added to the slide and the cells were visualized with a 100× objective lens . Western blot was performed according to standard protocols , using anti-FLAG antibody described above , and horseradish peroxidase-conjugated goat anti-rabbit secondary antibody ( Thermo Scientific , Waltham , MA ) . Each SitA-CTD was expressed from an IPTG-inducible promoter . For liquid growth-inhibition , cells were grown to log-phase and diluted to 5 × 107 cells per mL in two separate flasks containing fresh media ( CTT , 2 . 5 μg/mL oxytetracycline ) . To one of two flasks , IPTG was added to 1 . 0 mM . Cells were grown in the dark with shaking at room temperature . Culture growth was monitored at the indicated time points by measuring turbidity using a Klett meter . For DAPI staining , cells were grown as above for 30 hr , washed and resuspended in mPBS to a concentration of 1 . 5 × 109 cells per mL . 1 μL of a 50 μg/mL DAPI solution ( Life Technologies , Carlsbad , CA ) was added for each mL of culture . Cells were incubated for 20 min in the dark with rotation , washed , concentrated , and visualized by fluorescent microscopy .	Most people do not think of bacteria as having a social life . However , some groups , such as myxobacteria , are highly cooperative . Although these microbes exist as individual cells , they can also move and hunt in coordinated packs and when nutrients are low , about a million cells come together to build spore-filled structures . To do so , myxobacteria need to recognize their sibling cells among the vast number of different species of microbes found in soil . One way that the bacteria recognise their kin is by displaying a variable cell surface protein , called TraA , that identifies other individuals that display the same protein on their surface . Upon recognition , cells exchange resources by briefly fusing their outer membranes . This allows bacteria to help to rejuvenate damaged sibling cells by delivering healthy cell components to them . Now , using a genetic approach , Vassallo et al . present evidence that bacteria can also exchange toxins . The newly identified toxin-exchange system works alongside the TraA kin recognition system to allow myxobacteria to recognize and verify their true sibling cells in diverse environments . The cells involved in the exchange must contain matching immunity proteins to survive the interaction – thus the exchange does not harm sibling cells . Strikingly , once the toxic proteins are delivered , they can be passed on to other cells by a series of transfers , much like an infection spreads throughout a population . The study performed by Vassallo et al . provides a new framework for understanding how microbes recognize their kin to build a community . These insights will help investigators to explore other microbial ecosystems , including those found inside the human body . Additionally , the results also suggest ways in which cells can be engineered to specifically recognize other cells to transfer materials between them . This system could be adapted to program different cell types so that they interact with specific partners and perform complex tasks .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2017	Infectious polymorphic toxins delivered by outer membrane exchange discriminate kin in myxobacteria
Dinitrogen reduction in the biological nitrogen cycle is catalyzed by nitrogenase , a two-component metalloenzyme . Understanding of the transformation of the inert resting state of the active site FeMo-cofactor into an activated state capable of reducing dinitrogen remains elusive . Here we report the catalysis dependent , site-selective incorporation of selenium into the FeMo-cofactor from selenocyanate as a newly identified substrate and inhibitor . The 1 . 60 Å resolution structure reveals selenium occupying the S2B site of FeMo-cofactor in the Azotobacter vinelandii MoFe-protein , a position that was recently identified as the CO-binding site . The Se2B-labeled enzyme retains substrate reduction activity and marks the starting point for a crystallographic pulse-chase experiment of the active site during turnover . Through a series of crystal structures obtained at resolutions of 1 . 32–1 . 66 Å , including the CO-inhibited form of Av1-Se2B , the exchangeability of all three belt-sulfur sites is demonstrated , providing direct insights into unforeseen rearrangements of the metal center during catalysis . The reduction of substrates by nitrogenase entails multiple cycles of association and dissociation between two component proteins for sequential transfer of electrons ( Burgess and Lowe , 1996; Howard and Rees , 2006; Hoffman et al . , 2014; Hageman and Burris , 1978 ) . In the transient complex , electrons are transferred from the [4Fe:4S]-cluster of the homodimeric Fe-protein to the MoFe-protein in a reaction requiring adenosine triphosphate ( ATP ) hydrolysis ( Burgess and Lowe , 1996; Howard and Rees , 1994 ) . The MoFe-protein , an ( αβ ) 2 tetramer , contains two types of unique metal centers per catalytic αβ-unit: the P-cluster [8Fe:7S] and the FeMo-cofactor [7Fe:9S:C:Mo]-R-homocitrate ( Kim and Rees , 1992; Einsle et al . , 2002; Spatzal et al . , 2011 ) . The P-cluster is the initial electron acceptor with subsequent transfer to the FeMo-cofactor , one of the most elaborate metalloclusters found in nature and the catalytic center of biological nitrogen reduction . For substrates and inhibitors to bind , the FeMo-cofactor resting state must be reduced by two to four electrons provided by the Fe-protein ( Burgess and Lowe , 1996; Thorneley and Lowe , 1985 ) . The characterization of bound species including reaction intermediates has proven to be experimentally challenging due to the transitory nature of these states , inevitably leading to a recovery of the FeMo-cofactor resting state . While a number of studies have investigated substrate and inhibitor interactions ( Lee et al . , 1997; George et al . , 1997; Pickett et al . , 2004; Seefeldt et al . , 2004; Davis et al . , 1979 ) , only the recent crystal structure of Azotobacter vinelandii MoFe-protein ( Av1 ) with the inhibitor CO ( Av1-CO ) has provided high resolution details of a bound ligand ( Spatzal et al . , 2014 ) . In addition , the high symmetry and complex electronic structure of the FeMo-cofactor complicate spectroscopic studies ( Spatzal , 2015 ) . Thus , an atomically explicit description of the catalytic mechanism remains obscure . In this study , we have undertaken an alternative approach to follow events during catalysis by site-specifically introducing a reporter in the FeMo-cofactor ( Figure 1A , B , C ) . Because the S2B position of the active site can be reversibly replaced with CO ( Spatzal et al . , 2014 ) , it represents a potential site for other substitutions by substrates and inhibitors . Se , a structural surrogate for S in [Fe:S] clusters ( Meyer et al . , 1992; Zheng et al . , 2012 ) , has crystallographic and spectroscopic properties that make it an excellent probe , hence , potential Se containing compounds were investigated . Based upon the previous recognition , that thiocyanate ( SCN- ) is both a substrate and an inhibitor of nitrogenase ( Rasche and Seefeldt , 1997 ) , we evaluated the kinetic properties of selenocyanate ( SeCN- ) . We found SeCN- ( pKa < 1 [Boughton and Keller , 1966] ) to be a poor substrate as measured by methane production ( Figure 1—figure supplement 1 ) , a product also observed in thiocyanate ( Rasche and Seefeldt , 1997 ) and cyanide ( Li et al . , 1982 ) reduction . In addition , SeCN- is a potent , yet reversible inhibitor of acetylene reduction with an inhibition constant 30 times lower than observed for SCN- ( Ki ( SeCN- ) = 410 ± 30 μM versus Ki ( SCN- ) = 12 . 7 ± 1 . 2 mM ( Figure 1—figure supplement 2 ) . In contrast to inhibition of acetylene reduction , proton reduction activity is retained , although at a decreased level ( Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 11620 . 003Figure 1 . Selective Se-incorporation into the active site of the MoFe-protein . ( A ) Side view of FeMoSe-cofactor ( [7Fe:8S:1Se:Mo:C]-R-homocitrate ) in Av1-Se2B at a resolution of 1 . 60 Å , highlighting the S2B replacement by Se . ( B ) View along the Fe1-C-Mo axis . The electron density ( 2Fo-Fc ) map is contoured at 5 . 0 σ and represented as grey mesh . The 2Fo-Fc density at the Se2B site is significantly increased compared to the S5A and S3A sites . ( C ) Same orientation as B ) superimposed with the anomalous difference Fourier map calculated at 12 , 662 eV ( green ) at a resolution of 1 . 60 Å contoured at 5 . 0 σ showing the presence of anomalous electron density arising from Se . Fe atoms are shown in orange , S in yellow , Se in green , Mo in turquoise , C in grey , and O in red . ( D ) Acetylene reduction activity of Av1 ( black ) compared to Av-Se ( red ) . ( E ) Ammonia formation from reduction of the natural substrate , N2 , was determined for Av1 ( black ) and Av1-Se ( red ) . Error bars represent standard deviations from three measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00310 . 7554/eLife . 11620 . 004Figure 1—source data 1 . Numerical data for the graphs depicted in Figure 1D and 1E . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00410 . 7554/eLife . 11620 . 005Figure 1—source data 2 . Numerical data for the graphs depicted in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00510 . 7554/eLife . 11620 . 006Figure 1—source data 3 . Numerical data for the graphs depicted in Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00610 . 7554/eLife . 11620 . 007Figure 1—source data 4 . Numerical data for the graphs depicted in Figure 1—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00710 . 7554/eLife . 11620 . 008Figure 1—figure supplement 1 . CH4 production based on KSeCN and KSCN as substrates . Methane production was determined based on 0 . 05 , 0 . 1 , 0 . 2 , 0 . 5 , 1 , 2 , 5 mM KSCN ( red ) or KSeCN ( black ) as substrates . Maximum CH4 production from KSeCN was obtained at a concentration of 1 mM , whereas CH4 production from KSCN does not reach maximum within the tested range . Error bars represent standard deviations from three measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00810 . 7554/eLife . 11620 . 009Figure 1—figure supplement 2 . Inhibition of acetylene reduction by KSeCN and KSCN . Inhibitory properties of KSeCN and KSCN were determined using a modified acetylene reduction assay . Concentrations for substrate ( C2H2 ) were below saturation and concentrations for inhibitors ( KSCN , KSeCN ) were at low inhibition to allow for analysis . Dixon plots were prepared by plotting 1/v versus inhibitor concentration . Ki was determined from the intersection point derived from unrestrained linear fits of data points . ( A ) Dixon plot for KSCN , showing a Ki of 12 . 7 ± 1 . 2 mM KSCN . Concentrations of C2H2 were varied as follows: 20 ( grey ) , 30 ( red ) , 40 ( blue ) , 60 ( magenta ) , 100 ( green ) , 500 ( teal ) μL per 9 mL total headspace volume . Concentrations of KSCN were: 0 , 1 , 2 , 3 , 4 mM . ( B ) Dixon plot for KSeCN , showing a Ki of 410 ± 30 uM KSeCN . Concentrations of C2H2 were varied as follows: 20 ( grey ) , 30 ( red ) , 40 ( blue ) , 60 ( magenta ) , 100 ( green ) , 500 ( teal ) μL per 9 mL total headspace volume . Concentrations of KSeCN were: 0 , 50 , 100 , 200 , 500 μM . ( C ) Acetylene reduction activity in the presence of KSeCN ( black ) or KSCN ( red ) at varied concentrations: 50 μM , 75 μM , 100 μM , 500 μM , 1 mM , 5 mM , 10 mM , 15 mM and 20 mM . Error bars represent standard deviations from three measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 00910 . 7554/eLife . 11620 . 010Figure 1—figure supplement 3 . Influence of KSeCN and KSCN on proton reduction . Proton reduction activity as a function of KSeCN ( black ) or KSCN ( red ) concentrations ( 0 , 0 . 5 , 1 , 5 , 10 mM ) . H2 production in the presence of 10 mM KSeCN is approximately 65% when compared to 10 mM KSCN , and approximately 38% in comparison to the KSCN/KSeCN free reduction activity . Error bars represent standard deviations from three measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 010 By screening a range of experimental parameters , we found that Se from SeCN- could be incorporated into the FeMo-cofactor by incubating nitrogenase under assay ( turnover ) conditions with concentrations of SeCN- sufficient to inhibit acetylene reduction . The 1 . 60 Å resolution crystal structure of Av1 ( Av1-Se2B ) , isolated from assays containing SeCN- ( Figure 1A , B , C ) , revealed that Se quantitatively replaced belt-S position S2B in the FeMo-cofactor thereby generating the [7Fe:8S:1Se:Mo:C]-R-homocitrate cluster ( FeMoSe-cofactor ) form of the MoFe-protein . The essentially exclusive substitution of S2B by Se was validated by anomalous difference Fourier maps calculated from data measured at the Se-K edge f’’-peak position of 12 , 662 eV ( Figure 1C ) . At this resolution , no perturbation of the cofactor structure or its environment was detected , reflecting the small increase ( 3 . 8% ) in the ionic radius of Se relative to S ( Zheng et al . , 2012 ) . Except for a low occupancy site ( ranging from 0–20% under different conditions ) adjacent to the previously identified potential sulfur-binding site ( located ~22 Å away from the FeMo-cofactor at the interface of the α and β subunits [Spatzal et al . , 2014] ) , no other Se sites were identified in the anomalous difference Fourier map . Significantly , Av1-Se2B retains acetylene and dinitrogen reduction activity when compared to native Av1 ( Figure 1D , E ) , with the difference that the activity time course for Av1-Se2B with acetylene and dinitrogen exhibits a longer initial lag phase than for native Av1 ( Figure 1D inset , E ) . Consequently , the ability to prepare site-specifically labeled FeMoSe-cofactor provided the starting point for a structural investigation of the active site during substrate ( acetylene ) turnover . To trace the inserted Se-label , we developed a method to terminate the acetylene-turnover reaction at defined time points by freeze quenching ( fq ) followed by sample workup for crystallographic investigation . Crystal structures of the enzyme were determined at seven time points after initiation of substrate turnover corresponding to 2 to 5360 acetylene reduced per active site . These structures ( designated Av1-Se-fq-2 to Av1-Se-fq-5360 ) , at resolutions of 1 . 32–1 . 66 Å , represent the first example of time-dependent structural snapshots of the nitrogenase active site during turnover ( Figure 2A , B ) . ( Note: Av1-Se2B refers to the selectively labeled active site , while Av1-Se refers to Se substituted Av1 , where the site of Se in the FeMo-cofactor is not specified ) . 10 . 7554/eLife . 11620 . 011Figure 2 . Se-migration in the active site during substrate reduction . Se incorporation into all belt-S positions based on Av1-Se2B ( FeMoSe-cofactor ) . ( A ) Se-occupancy in the active site as a function of numbers of acetylene reduced per cofactor . Se-occupancy of site x2B-position is shown in dark-grey , x5A in blue , and x3A in red . Sum of ( x2B , x5A , x3A ) is shown in light grey . ( B ) Structural models of Se-incorporated FeMo-cofactor during turnover . 1 ) FeMoSe-cofactor resting state in Av1-Se2B . 2–8 ) Cofactor structures obtained at seven time points according to numbers of acetylene reduced per active site: 2 , 46 , 341 , 921 , 1785 , 2141 and 5361 . Crystal structure resolutions in the order 1–8: 1 . 60 , 1 . 50 , 1 . 45 , 1 . 32 , 1 . 64 , 1 . 66 , 1 . 65 and 1 . 48 Å , respectively . Anomalous difference Fourier maps ( calculated at 12 , 662 eV ) allowing for the quantification of Se are shown as green mesh , and are contoured at 5 . 0 σ . Color scheme is according to Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 011 The time course series of crystallographic data demonstrates a correlation between enzyme catalysis and the migration of Se from its initial Se2B site into the two remaining belt sulfur positions ( S5A and S3A ) . Consequently all belt-S atoms of FeMo-cofactor , separated by 5 . 7 Å in the resting state , can interchange during substrate ( acetylene ) reduction ( Figure 2 , Supplementary file 1A ) . The Se migration from S2B favors S5A over S3A at the earliest time points ( Figure 2 , Supplementary file 1A ) , but whether this reflects an ordered sequence or a more complex mechanism cannot be established from these data . Remarkably , with a sufficient number of catalytic turnovers ( several thousand , Figure 2A ) , the total Se is lost and apparently replaced by S , leading to a recovery of the original FeMo-cofactor ( Figure 2B ) . Neither the source of this S , nor the pathway ( s ) by which Se exits the MoFe-protein , have been identified . The observation that Se2B can migrate to S5A and S3A and is ultimately chased from the cofactor , clearly implies that all three belt-positions are labilized during the reduction of acetylene . In contrast , little or no migration of Se2B was observed during proton reduction . Our previous study of CO inhibited Av1 found replacement of S2B by CO ( Spatzal et al . , 2014 ) . The CO inhibition of Av1-Se2B during catalytic turnover has the potential to evaluate both the reactivity at the Se2B site and the location of the displaced chalcogen . The Av1-Se-CO structure at a resolution of 1 . 53 Å revealed that Se2B was ca . 90% replaced by a bridging CO with a geometry nearly identical to that observed for Av1-CO ( Figure 3 ) ( Spatzal et al . , 2014 ) . Unexpectedly , Se was not expelled from the FeMo-cofactor but migrated to the other two belt positions with ca . 88% overall retention ( Se occupancy: 10% ( x2B; where x reflects a mixture of S and Se ) , 35% ( x3A ) and 44% ( x5A ) ) . This result augments our findings from acetylene turnover , namely that the catalytic state of the cofactor capable of interacting with CO is similar to that which reacts with substrates . In contrast to the simple model of CO displacement and loss of S from S2B previously envisioned ( Spatzal et al . , 2014 ) , the path of net loss of sulfur from the cofactor is through either one or both of the other belt positions . 10 . 7554/eLife . 11620 . 012Figure 3 . Se-migration upon CO-binding to Av1-Se2B . Structure of Av1-Se-CO at a resolution of 1 . 53 Å , highlighting the Se2B replacement by CO and migration of Se to the remaining belt-S sites . ( A ) View along the Fe1-C-Mo axis of the metal center . The electron density ( 2Fo-Fc ) map is contoured at 5 . 0 σ and represented as grey mesh . The electron density at the CO site is significantly decreased compared to the x5A and x3A sites and in excellent agreement with CO when residual Se-density is subtracted . ( B ) Same orientation as A ) superimposed with the anomalous difference Fourier map calculated at 12 , 662 eV ( green ) at a resolution of 1 . 53 Å contoured at 5 . 0 σ showing the presence of anomalous electron density arising from Se . Numbers in parentheses indicate the fractional occupancies of the specified groups . DOI: http://dx . doi . org/10 . 7554/eLife . 11620 . 012 The incorporation and migration of the site-specific reporter , Se , demonstrates unanticipated dynamics of cofactor elements , specifically that the belt sulfur atoms , S2B , S5A , and S3A , can interchange and exchange with exogenous ligands under turnover conditions . The lability of the S2B position towards ligand exchange in non-resting states of the FeMo-cofactor highlights the likely role for the Fe2-Fe6 edge as a primary interaction site for substrates and inhibitors . Both CO ( Spatzal et al . , 2014 ) and Se from SeCN- , with different chemical properties , are incorporated at this site . Significantly , the side-chain residues flanking this position were previously identified to be catalytically important as demonstrated by mutagenesis studies ( Benton et al . , 2003 ) . The lack of substrate or inhibitor binding to the FeMo-cofactor resting state could be the result of the sulfurs serving as protecting groups that shield the iron core ( Howard and Rees , 2006 ) . The displacement of belt S in the catalytically active reduced states would effectively de-protect these Fe sites , thereby activating them for reaction with substrates . The interchange of all belt-S atoms is suggestive that substrates may also migrate to different sites of the trigonal six-iron prism during catalysis . While the underlying mechanism ( s ) are not known , one can speculate this may involve a twist-type mechanism interconverting trigonal prism and octahedral forms of the six-iron core , where swapping Fe-S bonding partners in the latter would result in interchange of the belt-S . Our results indicate that the resting state structure of FeMo-cofactor does not capture key features of the catalytic state , and a detailed understanding of how nitrogenase reduces dinitrogen must include the role of cofactor rearrangements during turnover . Furthermore , the incorporation of selenium into FeMo-cofactor opens a novel route to probe the substrate reduction mechanism of nitrogenase by using its unique crystallographic and spectroscopic properties . Cell growth and protein purification were carried out as previously described ( Spatzal et al . , 2011; 2014 ) . The specific activity of Av1 was 2350 ± 100 nmol min-1 mg-1and of Av2 was 2060 ± 55 nmol min-1 mg-1 when measured by acetylene reduction at saturation of each component . Protein concentrations were determined by absorbance at 410 nm using extinction coefficients of 76 mM-1 cm-1 for Av1 and 9 . 4 mM-1 cm-1 for Av2 . Nitrogenase activity was determined by monitoring acetylene and ethylene in the headspace ( 9 mL ) of reaction mixtures ( 1 mL ) that consisted of 20 mM creatine phosphate , 5 mM ATP , 5 mM MgCl2 , 25 units/mL phosphocreatine kinase , and 25 mM Na2S2O4 , in 50 mM Tris/Cl ( pH 7 . 5 ) buffer ( Yang et al . , 2014; Wolle et al . , 1992 ) . All reaction mixtures were made anaerobic ( Schlenk line technique ) and kept under an Ar-atmosphere . 1 mL of the headspace was replaced by 1 mL acetylene , followed by incubation for 5 min at 30°C . The reaction was initiated by addition of the nitrogenase component proteins ( Av2:Av1 = 4:1 , active site ratio = 2:1 , 0 . 125 mg Av1/0 . 135 mg Av2 per assay ) and terminated at specific time points by the addition of 1 mL 3 M citric acid . Ethylene and acetylene in the assay headspace were measured by gas chromatography ( activated alumina 60/80 mesh column , flame ionization detector ) . Calibration curves were constructed using defined amounts of acetylene in the headspace of protein-free assay mixtures . N2 reduction was monitored by determining ammonia formation , based on a modification of the previously described fluorescence method ( Barney et al . , 2005; Corbin , 1984 ) . The 1 . 0 mL assay contained 20 mM creatine phosphate , 5 mM ATP , 5 mM MgCl2 , 25 units/mL phosphocreatine kinase , and 25 mM Na2S2O4 , in MOPS buffer ( pH=7 . 5 ) . The headspace ( 9 mL ) of the assay vial was flushed with N2 before incubation for 5 min at 30°C . Reactions were initiated by addition of the nitrogenase component proteins ( Av2:Av1 = 4:1 , active site ratio = 2:1 ) , and terminated at specific time points by the addition of 300 μL 0 . 5 M EDTA ( pH=8 . 0 ) . The liquid chromatography step in the previously described method was replaced by filtering the assay mixture ( Amicon Ultra 3kDa centrifugal filter ) and collecting the filtrate , subsequently used for fluorescence measurement using a Flexstation 3 plate reader ( λexcitation = 410 nm , λemission = 472 nm ) . Dihydrogen from proton reduction was measured by gas chromatography ( molecular sieve 5A-80/100 column ) equipped with a thermal-conductivity detector . The assay was identical to the acetylene reduction assay with the acetylene omitted . Ar was used as the reference/carrier gas . Calibration curves were prepared using 10% H2 ( balance Ar ) as a standard . Methane , one product of the KSeCN and KSCN reduction , was determined by gas chromatography in the headspace of the assay as described above for the acetylene reduction assay . The assay mixture was identical to the acetylene reduction assay except that the acetylene was omitted and KSeCN or KSCN ( 0 . 05 , 0 . 1 , 0 . 2 , 0 . 5 , 1 , 2 , 5 mM ) were added . Calibration curves were determined with pure methane gas . Av1-Se2B was prepared based on the above described proton reduction activity assay protocol in the presence of 25 mM KSeCN , providing conditions commensurate with full inhibition of acetylene to ethylene reduction . Av1-Se2B was isolated from the assay mixtures by ultrafiltration using a 100 kDa cut-off membrane and washed two times ( dilution ratio of 1:100 each ) with 200 mM NaCl , 50 mM Tris/Cl buffer pH = 7 . 5 containing 5 mM Na2S2O4 to remove excess KSeCN . The protein was further purified by size-exclusion chromatography ( Superdex-200 , 450 mL , 50 mM Tris/Cl buffer pH = 7 . 5 containing 5 mM Na2S2O4 ) . The final protein concentration was adjusted to 30 mg/mL . Significantly , incorporation of Se required full nitrogenase turnover as SeCN- incubation with Av1 alone was ineffective; likewise , the non-substrate Na2Se failed to serve as a Se donor even when incubated under turnover conditions . Av1-Se freeze quenched samples were obtained by applying the above described acetylene reduction activity assays , with the replacement of wild-type Av1 with Av1-Se2B . Additionally , an alteration of the protein concentration per assay as well as a variation of the active site ratio of Av2:Av1 = 1:2 to 4:1 was required to either slow down acetylene reduction for the isolation of samples corresponding to low numbers of turnover , or to allow for the isolation of high turnover samples and to ensure non-limiting assay conditions . Termination of protein activity at distinct time points was achieved by rapid freezing the activity assay mixtures in liquid nitrogen , simultaneously measured by the formation of ethylene from acetylene . Time points and corresponding numbers of turnover per active site ( # ) at given Av2:Av1 active site ratios for the prepared samples were: #2 ( t = 0 . 05 min , Av2:Av1 = 1:2 ) , #46 ( t = 0 . 5 min , Av2:Av1 = 1:1 ) , #341 ( t = 2 min , Av2:Av1 = 1:1 ) , #921 ( t = 5 min , Av2:Av1 = 1:1 ) , #1785 ( t = 10 min , Av2:Av1 = 1:1 ) , #2141 ( t = 40 min , Av2:Av1 = 1:1 ) and #5361 ( t = 80 min , Av2:Av1 = 4:1 ) . The samples were subsequently processed at 3oC . Av1 was isolated by ultrafiltration using 100 kDa cut-off membranes and twice was washed with 200 mM NaCl , 50 mM Tris/Cl pH = 7 . 5 , 5 mM Na2S2O4 buffer to remove other assay components . The final protein concentration was adjusted to 30 mg/mL and 21°C for crystallization . For the preparation of Av1-Se-CO , CO-inhibited activity assays were prepared as described earlier ( Spatzal et al . , 2014 ) with the substitution of Av1-Se2B for Av1 in the absence of acetylene . The inhibited activity assays were concentrated using an Amicon ultrafiltration cell with a 100 kDa cut-off membrane under 15 psi CO overpressure . The protein was concentrated to 30 mg/mL and subsequently crystallized in solutions saturated with CO ( Spatzal et al . , 2014 ) . Av1-Se2B and all Av1-Se freeze-quenched samples were crystallized based on the sitting drop vapor diffusion method at 21°C in an anaerobic chamber containing a 95% Ar / 5% H2 atmosphere . The reservoir solution contained 24–28% PEG 8000 ( v/v ) , 0 . 75–0 . 85 M NaCl , 0 . 1 M imidazole/malate ( pH 7 . 5 ) , 1% glycerol ( v/v ) , 0 . 5% 2 , 2 , 2-trifluoroethanol ( v/v ) and 2 . 5 mM Na2S2O4 . Additionally , a seeding strategy was applied to accelerate the crystallization process and to optimize crystal shape . The cystals of the respective proteins ( Av1-Se2B , Av1-Se-fq-2 , Av1-Se-fq-46 , Av1-Se-fq-341 , Av1-Se-fq-921 , Av1-Se-fq-1785 , Av1-Se-fq-2141 , Av1-Se-fq-5361 and Av1-Se-CO ) were obtained between 6 and 24 hr after setting up the crystallization experiment . Cryo-protection was achieved by transferring crystals into a 5 uL drop of reservoir solution containing 8–12% MPD ( v/v ) . Diffraction data were collected at 12 , 662 eV ( 0 . 97918 Å , experimentally determined f’’ peak position of the Se-K edge ) at the Stanford Synchrotron Radiation Lightsource ( SSRL ) beamline 12–2 equipped with a Dectris Pilatus 6M detector . The data were indexed , integrated , and scaled using iMosflm , XDS and Scala ( Leslie , 2006; Kabsch , 2010; Winn et al . , 2011 ) . Phase information were obtained by molecular replacement using the 1 . 0 Å resolution structure ( PDB-ID: 3U7Q ) as a model . Structural refinement and rebuilding was carried out in REFMAC5 and COOT embedded in CCP4 ( Winn et al . , 2011 ) . All protein and active site structures were rendered in PYMOL . Anomalous electron density maps were calculated based on the data collected at 12 , 662 eV using a combination of CAD and FFT embedded in the CCP4 program suite ( Winn et al . , 2011 ) . Quantification of Se/Fe/S anomalous electron densities at 12 , 662 eV ( f’’ ( Se ) = 3 . 84 e; f’’ ( Fe ) = 1 . 50 e; f’’ ( S ) = 0 . 24 e ) based on the refined structural models was performed using a MAPMAN-dependent script , allowing a free choice of radius of integration and B-factor cut-off , as described previously ( Einsle et al . , 2002; Spatzal et al . , 2011; 2014 ) . The quantification of Se-occupancies was carried out by determining relative anomalous densities arising from Se within an integration sphere radius of 1 . 0 Å at the respective position based on the anomalous difference Fourier maps . This strategy was shown to be the most robust approach for density quantification within the FeMo-cofactor ( Spatzal et al . , 2011; Spatzal , 2015 ) . Comparison of the integrated Se-density values to the average anomalous density values for a full occupancy iron atom , taking into account the differences in f” for Fe and Se at 12 , 662 eV , yielded Se occupancy values . A total of 30 iron atoms ( 2x7 Fe from the two copies of FeMo-cofactor , and 2x8 Fe from the two copies of P-cluster per Av1 heterotetramer ) were used for averaging and yielded the internal reference for anomalous scattering at 12 , 662 eV . The standard deviation between anomalous density values for the Fe-atoms was ≤ 4% . No adjustments were made for variation in B factors since the Fe and S components of both metalloclusters have similar values ( average and standard deviation = 9 . 3 ± 1 Å2 for isotropically refined B-factors in the Av1-Se2B structure ) . For Fe and S in the FeMo-cofactor , the corresponding B values are 9 . 0 ± 0 . 4 Å2 , with the B for Se2B = 9 . 9 ± 0 . 5 Å2 and the average B for the entire protein structure = 13 . 8 Å2 . Quantification of Se-occupancies was further improved by including the contribution of the residual anomalous density of sulfur , with the anomalous scattering of S being 6 . 3% that of Se ( f’’ ( S ) /f’’ ( Se ) ) at an energy of 12 , 662 eV . The largest deviation between Se-anomalous densities observed for the two crystallographically independent copies of FeMo-cofactor was found to be ~5% , which provides an estimate of the uncertainty in the occupancy values . The root mean square value of the anomalous electron density map is ~2% of the Se peak value , which sets a lower threshold on the minimum occupancy that can be detected at this site .	The element nitrogen is required for all forms of life , and is an essential component of important biological molecules such as DNA and proteins . The most abundant form of nitrogen is dinitrogen , which comprises 78% of the Earth’s atmosphere . However , dinitrogen is highly unreactive , and so the nitrogen must be converted into a more reactive form before it can be used biologically . The only known enzyme capable of carrying out this reaction is called nitrogenase , but how this enzyme performs this difficult task is still not understood . Enzymes contain a region known as the active site , to which substrates – the molecules that the enzyme acts upon – bind . The active site of nitrogenase contains a region called the FeMo-cofactor , which must transform from an inactive to an active state to catalyze the conversion of dinitrogen to ammonia . Another substrate of the nitrogenase enzyme is a molecule called selenocyanate , which is made up of atoms of selenium , carbon and nitrogen . Spatzal , Perez et al . examined the structure of the active site of nitrogenase taken from the bacteria species Azotobacter vinelandii while the enzyme transformed selenocyanate . This revealed unexpected structural changes of the FeMo-cofactor that significantly challenge previous assumptions about how the active site works . For example , a single selenium atom from selenocyanate can be incorporated into a specific position of the FeMo-cofactor , which highlights the importance of this position for the enzyme’s initial interaction with substrates . Spatzal , Perez et al . then used the inserted selenium atom as a probe to investigate the changes in the active site structure that occur when either reacting with a substrate called acetylene or being inhibited by carbon monoxide . This revealed that selenium can migrate into the positions taken up by three of the FeMo-cofactor’s nine sulfur atoms ( the three “belt-sulfurs” ) during these interactions . The active site was not previously thought to be active in this way: this will need to be taken into account in all future models that describe how dinitrogen is converted into a biologically useful form . In the future , Spatzal , Perez et al . will investigate in detail how these “belt-sulfur” atoms exchange with atoms from the substrate , where the removed sulfur is stored , and the pathway by which it returns . Further experiments will also characterize the active site during the transformation of dinitrogen .	[ "Abstract", "Introduction", "Methods" ]	[ "short", "report", "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Catalysis-dependent selenium incorporation and migration in the nitrogenase active site iron-molybdenum cofactor
Chitin is a fungal microbe-associated molecular pattern recognized in Arabidopsis by a lysin motif receptor kinase ( LYK ) , AtCERK1 . Previous research suggested that AtCERK1 is the major chitin receptor and mediates chitin-induced signaling through homodimerization and phosphorylation . However , the reported chitin binding affinity of AtCERK1 is quite low , suggesting another receptor with high chitin binding affinity might be present . Here , we propose that AtLYK5 is the primary chitin receptor in Arabidopsis . Mutations in AtLYK5 resulted in a significant reduction in chitin response . However , AtLYK5 shares overlapping function with AtLYK4 and , therefore , Atlyk4/Atlyk5-2 double mutants show a complete loss of chitin response . AtLYK5 interacts with AtCERK1 in a chitin-dependent manner . Chitin binding to AtLYK5 is indispensable for chitin-induced AtCERK1 phosphorylation . AtLYK5 binds chitin at a much higher affinity than AtCERK1 . The data suggest that AtLYK5 is the primary receptor for chitin , forming a chitin inducible complex with AtCERK1 to induce plant immunity . As sessile organisms , plants have developed effective immune systems to defend against invading pathogens . Pathogen perception in plants can be divided into two different layers ( Jones and Dangl , 2006 ) . The initial response , mediated by perception of microbe-associated molecular patterns ( MAMPs ) , is termed MAMP-triggered immunity ( MTI ) ( Boller and Felix , 2009; Macho and Zipfel , 2014 ) . MTI is characterized by a wide range of physiological responses , including production of reactive oxygen species ( ROS ) , mitogen activated protein kinase ( MAPK ) phosphorylation , calcium influx , ion channel activation , callose deposition , growth inhibition , and expression of defense-related genes ( Macho and Zipfel , 2014 ) . However , adapted pathogens can inhibit MTI through the secretion of specific effector proteins or small RNAs into the cell ( Weiberg et al . , 2013 , 2014 ) . In response , plants have evolved polymorphic nucleotide-binding site leucine-rich repeat ( NBS-LRR ) proteins that either directly or indirectly recognize these effectors; thereby , restoring plant immunity . This form of immunity is termed effector-triggered immunity ( ETI ) ( Bent and Mackey , 2007; Boller and Felix , 2009 ) . Compared to MTI , ETI is much stronger and often associated with a hypersensitive response ( HR ) , which involves programmed cell death ( Jones and Dangl , 2006 ) . Plants use cell-surface localized pattern-recognition receptors ( PRRs ) to detect MAMPs activating MTI ( Boller and Felix , 2009; Macho and Zipfel , 2014 ) . In plants , several PRRs have been well characterized , including FLAGELLIN-SENSING 2 ( FLS2 ) and elongation factor-TU ( EF-Tu ) RECEPTOR ( EFR ) , which detect bacterial flagellin and EF-Tu , respectively ( Gomez-Gomez and Boller , 2000 , 2002; Zipfel et al . , 2006 ) . These two receptors belong to the leucine rich repeat receptor like protein kinase ( LRR-RLK ) family ( Shiu and Bleecker , 2001 ) . Activation of these receptors by ligand binding induces the association with BRASSINOSTEROID INSENSITIVE 1 ( BRI1 ) - Associated receptor Kinase 1 ( BAK1 ) and phosphorylation of Botrytis-Induced Kinase ( BIK1 ) ( Chinchilla et al . , 2007; Heese et al . , 2007; Lu et al . , 2010; Roux et al . , 2011; Zhang et al . , 2010 ) . Phosphorylated BIK1 dissociates from the receptor and subsequently phosphorylates the respiratory burst oxidase homolog D ( RBOHD ) protein , which controls ROS production in a calcium-independent manner ( Kadota et al . , 2014; Li et al . , 2014 ) . Other MAMPs , such as the oligosaccharides bacterial peptidoglycan ( PGN ) and fungal chitin [degree of polymerization ( dp ) ≥ 6] , are detected by lysin-motif ( LysM ) containing proteins . The chitin receptor was first reported in rice with the identification of the chitin-elicitor binding protein ( CEBiP ) ( Kaku et al . , 2006 ) , which contains an extracellular LysM motif and a transmembrane domain , but lacks an intracellular kinase domain . Data indicate that CEBiP forms a complex with the rice chitin-elicitor receptor kinase 1 ( OsCERK1 ) to mediate MTI in response to chitin ( Hayafune et al . , 2014; Shimizu et al . , 2010 ) . OsCERK1 has an active , intracellular kinase domain . The data suggest that OsCERK1 does not bind chitin but its intracellular kinase domain is activated by chitin binding to OsCEBiP ( Hayafune et al . , 2014 ) . In Arabidopsis thaliana , AtCERK1 was shown to be a key chitin receptor involved in chitin perception ( Miya et al . , 2007; Wan et al . , 2008 ) . For example , Atcerk1 mutant plants completely lose the ability to respond to chitin elicitation . There are three homologs of CEBiP in Arabidopsis , but a triple knock-out mutant , Atlym1/Atlym2/Atlym3 lacking these three proteins was fully competent to respond to chitin treatment ( Wan et al . , 2012 ) . However , two CEBiP like proteins appear to function in conjunction with CERK1 to mediate recognition of bacterial PGN ( OsLYP4 and OsLYP6 in rice , AtLYM1 and AtLYM3 in Arabidopsis ) ( Liu et al . , 2012a; Willmann et al . , 2011 ) . Another CEBiP-like protein , AtLYM2 , was demonstrated to act independently of AtCERK1 to mediate chitin-induced suppression of intracellular flux through plasmodesmata ( Faulkner et al . , 2013 ) . In Arabidopsis , there are five members of the lysin-motif receptor like kinase family ( LYKs ) , that is , AtCERK1/LysM RLK1/AtLYK1 , and AtLYK2-5 ( Wan et al . , 2012 ) . AtCERK1 was reported as the primary chitin receptor based on the mutant phenotype ( Miya et al . , 2007; Wan et al . , 2008 ) but also the fact that the protein can be precipitated by binding to chitin beads ( Iizasa et al . , 2010; Petutschnig et al . , 2010 ) . The X-ray crystal structure of the ectodomain of AtCERK1 was solved by Liu et al . ( 2012b ) . The structure predicted interaction of chitin oligomers with the second LysM motif in the extracellular domain . These authors suggested a model by which long chain chitin oligomers ( dp ≥ 6 ) bound to the LysM domains on two monomers , resulting in homodimerization of AtCERK1 . This dimerization was shown to activate the intracellular kinase domain ( Liu et al . , 2012b; Petutschnig et al . , 2010 ) . However , there remains the possibility that , similar to the situation in rice , the active chitin receptor is composed of more than one protein . For example , mutations in AtLYK4 were shown to significantly reduce the plant response to chitin ( Wan et al . , 2012 ) , although the phenotype was not as pronounced as that of Atcerk1 mutant plants . While the X-ray crystal structure of the ectodomain of AtCERK1 provided evidence that it is indeed a chitin binding protein , a puzzling aspect of this work is the low binding affinity ( chitooctaose , Kd = 45 µM ) reported , based on calorimetry ( Liu et al . , 2012b ) . Another puzzling aspect is that mutations in AtCERK1 , predicted to block chitin binding ( AtCERK1A138H ) , did not block chitin-induced AtCERK1A138H phosphorylation ( Liu et al . , 2012b ) . These data led us to consider the possibility that a second protein may be involved that mediates high affinity chitin binding and works with AtCERK1 to activate MTI . In this study , we show that mutations in AtLYK5 result in a significant reduction in the plant chitin response . AtLYK5 is required for chitin-induced AtCERK1 homodimerization and phosphorylation . AtLYK5 binds to chitin with a much higher affinity than AtCERK1 . The data suggest that AtLYK5 is the primary receptor for chitin , forming a chitin-inducible complex with AtCERK1 to induce plant innate immunity . Arabidopsis has five LysM receptor kinases ( LYKs ) ( Figure 1—figure supplement 1 ) . Therefore , plants mutated in each of these genes were tested for their ability to induce reactive oxygen species ( ROS ) in response to chitin elicitation . As expected from previous publications , mutations in AtCERK1 showed strongly reduced ROS production ( Miya et al . , 2007; Wan et al . , 2008 ) , while mutations in AtLYK4 also showed a slight reduction in ROS production upon chitin elicitation ( Wan et al . , 2012 ) . In previous publications , which involved screening Atlyk1-5 mutants , we reported that a transposon insertion in AtLYK5 did not affect chitin-induced MTI ( Wan et al , 2008 , 2012 ) . This conclusion was based on measuring AtWRKY53 expression upon chitin addition . At the time of these studies , the only Atlyk5 mutant available was in the Lansberg ( Ler ) background ( Atlyk5-1 ) . As part of a new round of screening , we again examined the chitin response of Atlyk5-1 mutant plants . qRT-PCR analysis showed that chitin treatment induced similar expression of AtWRKY53 in both Ler wild-type and Atlyk5-1 mutant plants ( Figure 1—figure supplement 2 ) ; data consistent with the previously published results ( Wan , et al , 2008 , 2012 ) . However , in contrast to these results , the expression of AtWRKY33 15 min after chitin treatment was significantly reduced in Atlyk5-1 mutant plants relative to Ler wild-type plants ( Figure 1—figure supplement 2 ) . Chitin-triggered MAP kinase ( MPK ) phosphorylation was also significantly reduced in Atlyk5-1 mutant plants compared with Ler wild-type plants ( Figure 1—figure supplement 2 ) . The phosphorylated AtCERK1 triggered by chitin elicitation can be detected as a band shift based on immunoblots using anti-AtCERK1 antibody ( Figure 1—figure supplement 2 ) ( Liu et al . , 2012b; Petutschnig et al . , 2010 ) . Chitin-triggered AtCERK1 phosphorylation was detected in Ler wild-type plants but was reduced in Atlyk5-1 mutant plants ( Figure 1—figure supplement 2 ) . In general , based on chitin-triggered ROS production , Ler wild-type plants showed a lower response to chitin than Col-0 plants , while Atlyk5-1 mutant plants showed similar ROS production to the wild-type when treated with chitin ( Figure 1—figure supplement 2 ) . Taken together , these experiments suggested that our original conclusion concerning AtLYK5 may not be correct; that is , this protein may be involved in chitin response . What is clear is that the Atlyk5-1 mutant , with a transposon insertion in the 3′ region of the gene , does not exhibit a strong phenotype under all conditions . The analysis of the chitin response in the Ler ecotype is further complicated by the generally weak response to chitin elicitation . Given these concerns , we identified and characterized a Col-0 Atlyk5 mutant ( Atlyk5-2 ) from the SALK population ( Figure 1—figure supplement 3 ) . This line has a T-DNA in the extracellular domain of AtLYK5 ( Figure 1—figure supplement 3 ) . It should be noted that in our original publications ( Wan et al . , 2008 , 2012 ) , all of the Atlyk mutants , with the exception of Atlyk5-1 , were derived from the Col-0 ecotype . As shown in Figure 1A , chitin-induced ROS production was significantly lower in the Atlyk5-2 mutant plants compared to Col-O wild-type plants ( Figure 1A ) . Calcium influx is activated by exposure of wild-type plants to chitin . Similar treatment of Atlyk5-2 mutant plants showed a 90% reduction and a significant delay in the calcium response , while Atcerk1 mutant plants showed essentially no calcium response to chitin ( Figure 1B ) . Flagellin-triggered calcium influx in both Atcerk1 and Atlyk5-2 mutant plants were similar to Col-0 wild-type plants ( Figure 1—figure supplement 4 ) , indicating that these defects were specific to chitin and not to a general effect on MTI . MPK3 and MPK6 are specifically phosphorylated upon chitin elicitation in Col-0 wild-type but not in Atcerk1 mutant plants . Significant reduction in MPK phosphorylation was detected in Atlyk5-2 mutant plants after chitin treatment ( Figure 1C ) . Consistent with these findings , the Atlyk5-2 mutant plants showed an intermediate response with regard to chitin-induced AtWRKY29 , AtWRKY30 , AtWRKY33 , and AtWRKY53 expression compared to the Col-0 wild-type and the Atcerk1 mutant plants ( Figure 1D , E , Figure 1—figure supplement 4 ) . Both Atcerk1 and Atlyk5-2 mutant plants showed increased susceptibility to the fungal pathogen Alternaria brassicicola compared with Col-0 wild-type plants . Pretreatment with chitooctaose enhanced resistance to A . brassicicola in wild-type plants but not in Atcerk1 or Atlyk5-2 mutant plants ( Figure 1F ) . Untreated Atcerk1 and Atlyk5-2 mutant plants showed wild-type levels of resistance to the bacterial pathogen Pseudomonas syringae pv . tomato DC3000 . However , only the wild-type showed increased bacterial resistance when plants were pretreated with chitooctaose to induce MTI ( Figure 1G ) . In order to confirm that the loss of chitin response was due to the AtLYK5 mutation , transgenic plants expressing the full-length AtLYK5 gene were generated under control of its native promoter in the Atlyk5-2 mutant genetic background ( Figure 2—figure supplement 1 ) . Expression of AtLYK5 in Atlyk5-2 mutant plants complemented all of the chitin-induced responses , including ROS production and MAPK phosphorylation ( Figure 2—figure supplement 1 ) . These data indicate that AtLYK5 is essential for a strong response to chitin elicitation . The response of all five Atlyk mutant plants was tested based on chitin-induced ROS prodction ( Figure 1—figure supplement 5 ) , confirming that AtCERK1 , AtLYK4 and AtLYK5 , but not AtLYK2 or AtLYK3 are involved in chitin signaling . 10 . 7554/eLife . 03766 . 003Figure 1 . Atlyk5 mutant plants are defective in chitin-triggered immune responses . ( A ) ROS production was measured from Col-0 wild-type and Atlyk5-2 mutant plants for 30 min after treatment with different chitin oligomers . 5mer: chitopentaose , 6mer: chitohexaose , 7mer: chitoheptaose , and 8mer: chitooctaose . Data are mean ± SE ( n = 8 ) . Asterisks indicate significant difference relative to H2O treated Col-0 wild-type plants . ( p < 0 . 01 , Student's t test ) . ( B ) Calcium influx in the wild-type , Atcerk1 and Atlyk5-2 mutant plants expressing aequorin was recorded for 30 min after chitooctaose treatment . ( C ) MAP kinase phosphorylation after chitooctaose treatment was detected by immunoblot using anti-P44/P42 antibody . ( D ) AtWRKY29 ( At4g23550 ) and ( E ) AtWRKY30 ( At5g24110 ) gene expression was analyzed using qRT-PCR in the wild-type , Atcerk1 and Atlyk5-2 mutant plants with or without treatment with chitooctaose , 8mer . UBQ10 ( At4g05320 ) was used a control . Data are mean ± SE ( n = 3 ) . Asterisks indicate significant difference relative to H2O treated Col-0 wild-type plants . ( p < 0 . 01 , Student's t test ) . ( F ) 4-week-old leaves from Col-0 wild-type , Atcerk1 , Atlyk5-2 , and Atlyk4/lyk5-2 mutant plants were inoculated with Alternaria brassicicola 24 hr after hand-infiltration with H2O or 1 µM chitooctaose . The diameter of the lesion area was measured 4 days after inoculation . Data are mean ± SE ( n = 12 ) . Asterisks indicate significant difference relative to H2O treated Col-0 wild-type plants . ( p < 0 . 05 , Student's t test ) . ( G ) Leaf populations of Psuedomonas syringae pv . tomato DC3000 3 days after inoculation . 4-week-old plants were either pretreated with H2O or 1 µM chitooctaose 24 hr before inoculation with P . syringae . Data are mean ± SE ( n = 9 ) . Asterisk indicates T-test significant difference compared with H2O-treated Col-0 plants at p < 0 . 05 , Student's t test . ( H ) AtCERK1 , AtLYK4 and AtLYK5 gene expression in different plant ages and plant tissue . RNA from whole seedling of 5 day , 10 day , 20 day old plants and leaf and root tissues from 20 day old plants were used for reverse transcript and qRT-PCR was performed using specific primers . Data are mean ± SE ( n = 3 ) . Asterisks indicate significant difference relative to chitiooctaose treated Col-0 wild-type plants ( p < 0 . 01 , Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00310 . 7554/eLife . 03766 . 004Figure 1—figure supplement 1 . Arabidopsis LYK gene family . A complete alignment based on the full-length sequences of each protein was used to draw the phylogenetic tree using Clustal X software . AtCERK1: At3g21630 , AtLYK2: At3g01840 , AtLYK3: At1g51940 , AtLYK4: At2g23770 , AtLYK5: At2g33580 . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00410 . 7554/eLife . 03766 . 005Figure 1—figure supplement 2 . Chitin response in Ler lyk5-1 mutant plants . ( A–B ) WRKY53 gene ( A ) and WRKY33 gene ( B ) expression was analyzed using qRT-PCR in Ler wild-type and Atlyk5-1 mutant plants with or without treatment with 1 µM chitooctaose . 8mer: chitooctaose . Data are ± SE ( n = 3 ) , *p < 0 . 05 . ( C–D ) MPK phosphorylation in Ler wild-type and Atlyk5-1 mutant plants revealed by immunoblot . Leaf discs from 5-week-old plants ( C ) or 2-week-old seedlings ( D ) were subjected to the treatment with 1 µM chitooctaose for the time point shown in figures . Lower panel shows similar loading of total protein . ( E ) Chitin induces AtCERK1 phosphorylation . Mature leaves from Col-0 wild type plants were hand-infiltrated with 1 µM chitooctaose or H2O as control for 15 min . All samples were incubated with antarctic phosphatase or H2O as control at 37°C for 15 min . Anti-AtCERK1 antibody was used to detect AtCERK1 protein . ( F ) AtCERK1 phosphorylation in Ler wild-type and Atlyk5-1 mutant plants revealed by immunoblot using anti-AtCERK1 antibody after hand-infiltration with 1 µM chitooctaose for the time points shown in figures . ( G ) ROS production was measured from the Ler wild-type plants , Atlyk5-1 mutant plants , Col-0 wild-type plants , Atlyk5-2 mutant plants for 30 min after treatment with 0 . 5 µM chitooctaose . Data are mean ± SE ( n = 6 ) . Asterisk indicates significant difference . ( p < 0 . 01 , Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00510 . 7554/eLife . 03766 . 007Figure 1—figure supplement 3 . Characterization of Atlyk5 mutant plants . ( A ) Genomic structure of AtLYK5 and two insertion sites of two mutants . ( B ) Identification of T-DNA insertion by PCR using genomic DNA from Col-0 wild-type and Atlyk5-2 mutant plants . Location of primers used is shown in ( A ) . Primer sequences are listed in Supplemental file 1 . ( C ) RT-PCR was used to identify transcriptional expression of AtLYK5 in Col-0 WT and Atlyk5-2 mutant plants . Upper panel shows expression level of AtLYK5 in Col-0 and Atlyk5-2 mutant plants , lower panel shows expression of AtCERK1 as a control . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00710 . 7554/eLife . 03766 . 008Figure 1—figure supplement 4 . WRKY33 and WRKY53 gene expression in Atlyk5-2 mutant plants . ( A ) Calcium influx in the wild-type , Atcerk1 and Atlyk5-2 mutant plants expressing aequorin was recorded for 30 min after 100 nM flg22 treatment . ( B ) WRKY33 ( At2g38470 ) and ( C ) WRKY53 ( At4g23810 ) gene expression was analyzed using qRT-PCR in the wild-type , Atcerk1 and Atlyk5-2 mutant plants with or without treatment with chitooctaose , 8mer . UBQ10 ( At4g05320 ) was used a control . Data are mean ± SE ( n = 3 ) . Asterisks indicate significant difference relative to chitiooctaose treated Col-0 wild-type plants ( p < 0 . 01 , Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00810 . 7554/eLife . 03766 . 006Figure 1—figure supplement 5 . Chitin-induced ROS production in five lyk mutant plants . ROS production was measured from Col-0 wild-type , five Atlyk mutant , and Atlyk4/Atlyk5-2 double mutant plants for 30 min after treatment with 1 µM chitooctaose . Data are mean ± SE ( n = 8 ) . Asterisks indicate significant difference relative to chitooctaose treated Col-0 wild-type plants . ( p < 0 . 01 , Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 006 Analysis of the Atlyk5 mutant plants showed some residual response to chitin . Previously , we reported that AtLYK4 is also required for chitin elicitation ( Wan et al . , 2012 ) . Phylogenetic analysis shows that AtLYK5 and AtLYK4 are in the same branch ( Figure 1—figure supplement 1 ) . Therefore , we examined the possibility that AtLYK4 may provide some functional redundancy for the loss of AtLYK5; thus , explaining the low level response of the Atlyk5 mutants to chitin . To address this hypothesis , mutant plants defective in both AtLYK4 and AtLYK5 were generated through crossing ( Figure 2—figure supplement 1 ) and tested for their response to chitin . Similar to the Atcerk1 mutants , plants mutated in AtLYK4 and AtLYK5 lost all tested responses to chitin , including ROS production and MAPK phosphorylation , as well as resistance to the fungal pathogen A . brassicicola ( Figure 1F and Figure 2 ) . These genetic studies suggest that AtLYK5 and AtCERK1 are essential for a strong plant chitin response , while AtLYK4 can partially compensate for the loss of AtLYK5 . 10 . 7554/eLife . 03766 . 009Figure 2 . AtLYK5 has overlapping function with AtLYK4 . ( A ) ROS production was measured from Col-0 wild-type , Atyk4 , Atlyk5-2 , and Atlyk4/lyk5-2 mutant plants for 30 min after treatment with H2O ( as control ) or 1 µM chiooctaose . 8mer: chitooctaose . Data are mean ± SE ( n = 8 ) . ( B ) and ( C ) Western blot of total protein extracts from plants treated with 1 µM chitooctaose . Protein was separated by SDS-PAGE gel and visualized using anti-P44/P42 antibody . Upper panel in each figure shows phosphorylated MPK3 and MPK6 , lower panel shows similar loading of each lane stained with ponceau S solution . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 00910 . 7554/eLife . 03766 . 010Figure 2—figure supplement 1 . Complementation of Atlyk5-2 mutant plants . ( A ) HA-tagged AtLYK5 protein levels in different transgenic Atlyk5-2 mutant plants expressing full length AtLYK5 under the control of its native promoter . ( B ) ROS production was measured from the wild type plants , and Atlyk5-2 mutant plants , and the Atlyk5-2 mutant plants expressing AtLYK5 under the control of its native promoter for 30 min after chitooctaose . AtLYK5 indicates complementation lines ( #1 and #3 were used ) . Data are mean ± SE ( n = 8 ) . Asterisks indicate significant difference relative to H2O treated Col-0 wild type plants . ( p < 0 . 01 , Student's t test ) . ( C ) MPK phosphorylation in different transgenic plants shown above prior ( − ) or 15 min ( + ) after 1 µM chitooctaose treatment detected using anti-P44/P42 antibody . Lower panel shows similar loading of each lane stained with ponceau S solution . #1 and #3 indicate two different transgenic plants expressing AtLYK5 in Atlyk5-2 mutant plants . ( D ) Identification of Atlyk4/Atlyk5-2 double knockout mutant plants . Whole genomic DNA was extracted from Atlyk4+/−/Atlyk5-2+/− and Atlyk4−/−/Atlyk5-2−/− mutant plants . PCR was used to characterize homozygous Atlyk4/Atlyk5-2 double mutant plants . Primer pairs used as following: 1 and 3: AtLYK4 LP + AtLYK4 RP , 2 and 4: AtLYK4 LB + AtLYK5 RP , 5 and 7: AtLYK5 LP + AtLYK5 RP , 6 and 8: AtLYK5 LB + AtLYK5 RP . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 01010 . 7554/eLife . 03766 . 011Figure 2—figure supplement 2 . Tissue-specific expression of AtCERK1 , AtLYK4 , and AtLYK5 . ( A ) AtCERK1 ( At3g21630 ) , AtLYK4 ( At2g23770 ) and AtLYK5 ( At2g33580 ) gene expression in different developmental stages predicted by AtGenExpress Visualization Tool ( AVT ) . Results can also be seen at http://jsp . weigelworld . org/expviz/expviz . jsp . ( B ) Subcellular localization of AtLYK5 in Nicotiana bethamiana . Leaf expressing AtLYK5-GFP was stained with FM4-64 before monitoring epifluorescence signal using confocal microscopy ( Left panel ) . Right panel showed expression of AtLYK5-GFP detected by immunoblot using anti-GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 011 Because AtCERK1 , AtLYK5 , and AtLYK4 are all involved in the chitin response , we measured their transcriptional level in different plant tissues at different plant ages using quantitative PCR ( qRT-PCR ) . At different growth stages of 5 , 10 , and 20 days after germination , as well as leaf and root tissues from 20-day-old plants , the transcript levels of three genes were similar . However , in the root tissue tested , AtLYK5 expression was higher than AtCERK1 and AtLYK4 , while both showed similar expression levels in roots ( Figure 1H ) . These data are consistent with the results predicted by the AtGenExpress Visualization Tool ( AVT ) showing that AtLYK5 is co-expressed with AtCERK1 , with some variation in the root ( Figure 2—figure supplement 2 or online at http://jsp . weigelworld . org/expviz/expviz . jsp ) . Previous reports showed that both AtCERK1 and AtLYK4 are localized on the plasma membrane ( Miya et al . , 2007; Wan et al . , 2012 ) . The AtLYK5 was fused with c-terminal GFP and transiently expressed under the CaMV 35S promoter in Nicotiana benthamiana . Confocal microscopy showed that AtLYK5-GFP co-localized with the plasma membrane dye , FM4-64 , and western blots showed the correct size of the AtLYK5-GFP protein ( Figure 2—figure supplement 2 ) . These results indicate that together with AtCERK1 and AtLYK4 , AtLYK5 is a membrane-localized protein . Given the low reported affinity of AtCERK1 for chitooctaose ( Liu et al . , 2012b ) , we tested the ability of AtLYK5 to bind to chitin . HA-tagged versions of each of the five AtLYK proteins were expressed in Arabidopsis protoplasts and chitin-magnetic beads were used to pull down any chitin binding proteins . As shown in Figure 3—figure supplement 1 , besides AtCERK1 , only AtLYK4 and AtLYK5 were also pulled down by chitin beads . The binding with AtLYK5 was strongly inhibited by chitoheptaose and chitooctaose , whereas the binding with AtCERK1 or AtLYK4 was only slightly reduced by the same competitors , consistent with AtLYK5 having a higher binding affinity for chitooctaose relative to AtCERK1 . In order to further investigate this possibility , the chitin binding ability of the extracellular domains of AtCERK1 and AtLYK5 were measured using isothermal titration calorimetry ( ITC ) . As shown in Figure 3 , the binding affinity of AtLYK5 for chitooctaose was measured ( Kd = 1 . 72 µM ) , which is roughly 200-fold higher than measured for AtCERK1 under the same conditions ( Kd = 455 µM ) ( Figure 3A , B ) . In control experiments , no chitin binding affinity was detected using buffer titrated with chitooctaose ( Figure 3—figure supplement 1 ) . As expected , AtLYK5 showed no binding affinity for elicitor-inactive chitotetraose ( Figure 3—figure supplement 1 ) . As a positive control , wheat germ agglutinin ( WGA ) showed strong chitin binding affinity ( Kd = 1 . 6 µM; Figure 3—Figure supplement 1 ) . Therefore , the data indicate that AtCERK1 shows very low affinity for chitooctaose , while AtLYK5 shows an affinity very close to the well-characterized chitin binding protein WGA . 10 . 7554/eLife . 03766 . 012Figure 3 . AtLYK5 shows stronger chitin binding affinity than AtCERK1 . The binding affinity of AtLYK5 ( A ) and AtCERK1 ( B ) to chitooctaose ( GlcNAc ) 8 was measured using isothermal titration calorimetry ( ITC ) . Proteins were purified from E . coli . Upper panels and lower panels indicate raw data and integrated heat values , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 01210 . 7554/eLife . 03766 . 013Figure 3—figure supplement 1 . AtLYK5 has chitooctaose binding affinity . ( A ) Chitin binding of AtLYK proteins detected using chitin-magnetic beads . HA-tagged AtCERK1 and AtLYK2-5 were expressed from the 35S promoter in protoplasts made from Col-0 wild type plants . Protoplasts were treated either with 10 µM of the specific chitin oligomer noted above for 15 min ( + ) or with H2O ( − ) before harvest . Proteins were pulled-down using chitin magnetic beads ( New England Biolabs , Ipswich , MA ) . Upper panel shows input of each protein , lower panel shows proteins after chitin binding . IB , immnunoblot with anti-HA antibody . ( B–C ) Wheat germ agglutinin binds to chitin with high affinity . Binding of wheat germ agglutin in the ( B ) presence of chitin ( GlcNAc ) 8 or ( C ) buffer . Binding was measured using isothermal titration calorimetry ( ITC ) . Upper panels and ( lower panels ) indicate raw data and integrated heat values , respectively . ( D ) Binding affinity of AtLYK5 in presence of chitin ( GlcNAc ) 4 . Binding was measured using isothermal titration calorimetry ( ITC ) . Upper panels and ( lower panels ) indicate raw data and integrated heat values , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 013 The AtCERK1 crystal structure predicted chitooctaose binding to the second LysM motif of the extracellular domain leading to homodimerization and kinase activation ( Liu et al . , 2012b ) . However , AtCERK1 appears to be a very weak chitin binding protein raising questions as to the biological relevance of the AtCERK1 homodimer model . Therefore , in order to predict the chitin binding site ( s ) within the AtLYK5 extracellular domain , a computational model of the AtLYK5 ectodomain was built by homology modeling against the known crystal structure of the fungal ECP6 ( Sanchez-Vallet et al . , 2013 ) , a LysM effector protein , which binds chitin with very high affinity ( binding at pM levels; Figure 4A , B and Figure 4—figure supplement 1 ) . Based on the docking model of AtLYK5 with chitooctaose , the binding affinity was calculated at −8 . 9 kcal mol−1 ( Figure 4A , B ) , a value comparable to the computational binding affinity of ECP6 ( −9 . 0 kcal mol−1 ) ( Figure 4—figure supplement 1 ) . Four residues , that is , Thr-72 , Tyr-128 , Ser-206 , and Ser-216 , were predicted to form hydrogen bonds and hydrophobic interactions with chitooctaose based on docking model ( Figure 4—figure supplement 1 ) . Point mutations were introduced at each of these residues and transgenically expressed in Atlyk5 mutant plants from the native promoter . As shown in Figure 4C , AtLYK5S206P and AtLYK5Y128G transgenic plants could not rescue the Atlyk5-2 mutant phenotype as measured by chitin-triggered ROS production . In contrast , expression of AtLYK5T72G and AtLYK5S216P mutant proteins in the Atlyk5-2 mutant plants did restore the chitin response . Consistent with these results , AtLYK5S206P mutant proteins did not bind to chitin beads , while AtLYK5Y128G mutant proteins showed a strong reduction in chitin binding using this same assay ( Figure 4D ) . Binding of the AtLYK5T72G and AtLYK5S216P mutant proteins to the chitin beads was similar to wild-type AtLYK5 ( Figure 4D ) . These data indicate that residues Tyr-128 and Ser-206 of AtLYK5 are important for chitin binding and that chitin binding is essential for biological activity . 10 . 7554/eLife . 03766 . 014Figure 4 . Tyr-128 and Ser-206 are important for AtLYK5-mediated chitin response . ( A ) A computational ribbon structure of the AtLYK5 ectodomain was built based on crystal structure of fungal ECP6 . The model shows the three AtLYK5 LysM domains , i . e . LysM1-3 . Each LysM domain contains two beta strands and two helixes interconnected via loops . ( B ) The binding affinity was calculated at −8 . 9 kcal mol−1 . The binding site was formed by 3 LysM motifs . Green lines depict hydrogen bonds formed between ligand atoms and their corresponding residues atoms . ( C ) Reactive oxygen species ( ROS ) was measured within 30 min after chitin treatment . The AtLYK5 wild-type gene or versions with specific point mutations were transformed into Atlyk5-2 mutant plants . Eight individual transgenic plants were used for this measurement . Data are mean ± SE . Asterisks indicate significant difference relative to H2O treated Col-0 wild-type plants . ( p < 0 . 01 , Student's t test ) . ( D ) Chitin binding affinity of AtLYK5 and AtLYK5 mutant proteins as labeled in ( C ) detected by anti-HA antibody . Upper panel shows input of each transgenic plant , lower panel shows western blot after pull down with chitin-magnetic beads . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 01410 . 7554/eLife . 03766 . 015Figure 4—figure supplement 1 . Computational model of the extracellular domain of AtLYK5 . ( A–C ) The docking model of the ectodomain with chitooctaose shown in surface ( A ) and ribbon form ( B ) and a close-up surface ( C ) . The binding affinity was calculated at −8 . 9 kcal mol−1 . The model shows the three AtLYK5 LysM domains , that is , LysM1-3 . Each LysM domain contains two beta strands and two helixes interconnected via loops . ( D–E ) Docking of chitooctaose to the ECP6 . ( D ) A ribbon structure represents the docking model of ECP6 ( gray color ) and chitooctaose ( blue , red and yellow sticks ) . The binding affinity was calculated at −9 . 0 kcal mol−1 . ( E ) A molecular surface of ECP6 with chitooctaose binding site formed by 3 LysM motifs . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 015 Given that both AtCERK1 and AtLYK5 are required for a strong chitin response , we hypothesized that AtLYK5 might interact with AtCERK1 . Indeed , co-immunoprecipitation ( Co-IP ) assays showed that AtCERK1-HA strongly interacts with AtLYK5-Myc in the presence of chitin ( Figure 5A ) . Chitin-induced association between AtLYK5 and AtCERK1 was much stronger than homodimerization of AtCERK1 ( Figure 5A ) . We also tested the chemical specificity of this response using chitin oligomers of increasing length ( Figure 5B ) . Chitopentaose which can not trigger immune responses in plants did not induce AtLYK5-CERK1 association , while chitin oligomers ( dp ≥ 6 ) , which are strong elicitors , induced the interaction between AtLYK5 and AtCERK1 ( Figure 5B ) . We also tested the association between AtLYK4 and AtCERK1 before and after chitin treatment . As shown in Figure 5—figure supplement 1 , AtCERK1-HA could be co-immunoprecipitated with AtLYK4-Myc; however , this interaction was independent of the presence of chitin . 10 . 7554/eLife . 03766 . 016Figure 5 . AtLYK5 regulates chitin-induced phosphorylation and homodimerization of AtCERK1 . ( A ) AtLYK5 associates with AtCERK1 after chitin treatments . HA-tagged AtCERK1 and Myc-tagged AtLYK5 or AtCERK1 were co-expressed in protoplasts made from Col-0 wild-type plants . Protoplasts were harvested with or without the treatment with 1 µM chitooctaose as labeled above . Co-immunoprecipitation was made using anti-Myc antibody . Left panel and right panel are cropped from the same gel . ( B ) The association between AtCERK1 and AtLYK5 is induced by different chitin oligomers . Protoplasts were treated with different chitin oligomers ( 1 µM ) as shown above for 15 min . ( C ) AtLYK5 regulates chitin-induced AtCERK1-AtCERK1 association . HA-tagged AtCERK1 and Myc-tagged AtCERK1 were copexpressed in protoplasts made from Col-0 wild-type or Atlyk5-2 mutant plants . Protoplasts were harvested with or without the treatment with 1 µM chitooctaose . Co-immunoprecipitation was made using anti-Myc antibody . ( D ) AtLYK5 controls chitin-induced phosphorylation of AtCERK1 . Plant leaves from wild-type and the Atlyk5-2 mutant plants were treated with 1 µM chitooctaose for the time shown above . Anti-AtCERK1 antibody was used to detect the phosphorylation status of AtCERK1 shown as a shift in protein migration . Lower panel shows a non-specific band used to assess similar loading of each lane . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 01610 . 7554/eLife . 03766 . 017Figure 5—figure supplement 1 . AtLYK4 associates with AtCERK1 before and after chitin treatment . ( A ) Interaction between AtCERK1 and AtLYK4 . HA-tagged AtCERK1 and Myc-tagged AtLYK4 were co-expressed in protoplasts from Col-0 wild type plants . Protoplasts either prior ( − ) or after treatment with 1 µM chitooctaose were harvested . Co-immunoprecipitation used anti-Myc antibody . IB , immunoblot detected with either anti-HA or anti-Myc antibodies . ( B ) Chitin-induced phopshorylation of AtCERK1 . Mature leaves from Col-0 wild-type , Atlyk4 , and Atlyk5-2 mutant plants were treated with 1 µM chitin oligomers for 15 min . Total protein was separated on 7% SDS-PAGE and immunoblots were detected with anti-AtCERK1 antibody . Protein band shift indicates phosphorylation . 6mer: chitohexaose , 7mer: chitoheptaose , and 8mer: chitooctaose . Upper panel shows immunoblot probed with anti-AtCERK1 antibody , lower panel shows a non-specific band detected by anti-AtCERK1 antibody as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 017 Previously , it was postulated that chitin-induced AtCERK1 homodimerization leads to AtCERK1 phosphorylation ( Liu et al . , 2012b; Petutschnig et al . , 2010 ) , required for downstream activation of plant innate immunity . Consistent with this model , co-immunoprecipitation demonstrated AtCERK1 dimerization upon chitin addition in wild-type plants ( Figure 5C ) . However , this association disappeared in the Atlyk5-2 mutant plants ( Figure 5C ) . Chitin-induced phosphorylation of AtCERK1 was detected as a protein mobility-shift on SDS-PAGE gels . In contrast , no phosphorylation of AtCERK1 was detected after chitin elicitation of Atlyk5-2 mutant plants ( Figure 5D and Figure 5—figure supplement 1 ) . These data clearly demonstrate that AtLYK5 is essential for both AtCERK1 dimerization and subsequent activation of protein phosphorylation . A comparison was made using the amino acid sequences of the intracellular kinase domain of AtLYK5 and other known receptor kinases , including AtCERK1 ( Miya et al . , 2007; Wan , et al . , 2008 ) , AtBRI1 ( Li and Chory , 1997 ) , AtBAK1 ( Li et al . , 2002; Nam and Li , 2002 ) , AtLYK4 ( Wan et al . , 2012 ) , and Does not Respond to Nucleotides 1 ( AtDORN1 ) ( Choi et al . , 2014 ) . The analysis showed that several residues generally considered essential for kinase activity are missing from the AtLYK5 sequence ( Figure 6—figure supplement 1 ) , including those in the P-loop sub-domain I , RD domain in sub-domain VIa , and DFG domain in sub-domain VII ( Figure 6—figure supplement 1 ) , suggesting that AtLYK5 lacks kinase activity ( Hanks et al . , 1988 ) . In order to test this directly , the AtLYK5 kinase domain was expressed and purified from Escherichia coli and used in an in vitro kinase assay . As shown in Figure 6A , as a positive control , the AtCERK1 kinase domain showed strong kinase activity as previously reported ( Miya et al . , 2007 ) , while no kinase activity was detected using the AtLYK5 kinase domain . In order to investigate the role of AtLYK5 kinase activity in vivo , wild-type AtLYK5 , a mutant AtLYK5K395E ( mutation of Lys to Glu predicted to disrupt ATP binding ability ) , and mutant AtLYK5ΔKD ( deletion of kinase domain ) proteins were transgenically expressed in Atlyk5-2 mutant plants from the native promoter ( Figure 6—figure supplement 1 ) . The data show that expression of AtLYK5 or AtLYK5K395E could complement the Atlyk5-2 mutant phenotype , as measured by chitin-triggered ROS production , MAPK phosphorylation , and AtCERK1 phosphorylation , whereas transgenic expression of AtLYK5ΔKD did not complement the Atlyk5-2 mutant ( Figure 6B–E ) . These results indicate that although AtLYK5 kinase activity is not required for chitin signaling , the intracellular kinase does have a function , which may include mediating protein–protein interactions . Indeed , the AtLYK5ΔKD mutant protein could not be co-immunoprecipitated with AtCERK1 after chitin treatment , while AtCERK1 interacted with AtLYK5K395E and wild-type AtLYK5 normally ( Figure 6F ) . These data strongly demonstrate that the kinase domain of AtLYK5 is necessary for the association of AtLYK5 and AtCERK1 . 10 . 7554/eLife . 03766 . 018Figure 6 . The kinase domain of AtLYK5 is critical for chitin signaling . ( A ) In vitro kinase activities of AtCERK1 ( 255–617 aa ) and AtLYK5 ( 309–664 aa ) were measured by incorporation of γ-[32P]-ATP . Left panel shows autoradiography , and right panel shows gel stained with coomassie brilliant blue . ( B–F ) Plant tissues were harvested before ( − ) or 15 min after ( + ) treatment or at the time point shown in each figure of treatment with 1 µM chitooctaose . ( B–E ) AtLYK5K395E but not AtLYK5ΔKD ( 1–320 aa ) complemented the Atlyk5-2 mutant as determined by chitin-triggered ROS production . Asterisks indicate significant difference relative to H2O treated Col-0 wild-type plants . ( Data are mean ± SE ( n = 8 ) , p < 0 . 01 , Student's t test ) , MPK phosphorylation , and chitin-induced AtCERK1 phosphorylation . Upper panel of each figure show immunoblot data , lower panel shows either rubisco band stained with ponceau S solution ( C and D ) or a non-specific band ( E ) to show similar loading of each lane . Plant tissues were harvested before ( − ) or after ( + ) 15 min treatment with 1 µM chitooctaose . ( F ) AtLYK5K395E but not AtLYK5ΔKD coimmunoprecipitates with AtCERK1 after chitin elicitation . Co-IP was made using anti-AtCERK1 antibody with proteins from transgenic Arabidopsis Atlyk5-2 mutant plants expressing either AtLYK5 , or AtLYK5K395E or AtLYK5ΔKD . 8mer: chitooctaose , AtLYK5 , K395E , and ΔKD indicate transgenic Arabidopsis expressing AtLYK5 , AtLYK5K395E , and AtLYK5ΔKD , respectively . Different number indicates different transgenic lines used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 01810 . 7554/eLife . 03766 . 019Figure 6—figure supplement 1 . AtLYK5 is a kinase inactive protein . ( A ) AtLYK5 lacks conserved residues required for kinase activity . Sequence alignment was done using the entire intracellular domain of AtLYK5 and other , selected receptor kinases . Each kinase subdomain is labeled by as Roman numeral and the conserved residues of each subdomain are shown below these numbers . Amino acids sequences of each kinase used for alignment are shown as following: AtBAK1 ( At4g33430 ) : 297–662 aa , AtBRI1 ( At4g39400 ) : 815–1196 aa , AtDORN1 ( At5g60300 ) : 310–718 aa , AtLYK5 ( At2g33580 ) : 301–664 aa , AtLYK4 ( At2g23770 ) : 295–612 aa , AtCERK1 ( At3g21630 ) : 255–671 aa . ( B–C ) Expression of AtLYK5K395E ( B ) and AtLYK5ΔKD ( C ) with a C-terminal HA tag from the native promoter in Atlyk5-2 mutant plants . Upper panel shows protein level detected using anti-HA antibody , lower panel shows similar loading of each lane stained with ponceau S solution . AtLYK5ΔKD: 1–320 aa . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 019 Receptor homodimerization or oligomerization is a common mechanism for ligand-mediate receptor activation ( Jiang and Hunter , 1999; Mellado et al . , 2001; Pang and Zhou , 2013; Stock , 1996 ) . Therefore , we tested whether AtLYK5 could form homodimers and whether this required AtCERK1 and/or chitin treatment . As shown in Figure 7 , AtLYK5 homodimers were detected even in the absence of chitin and this association was independent of the presence of AtCERK1 ( Figure 7A ) . AtLYK5 homodimers were also detected in vivo ( Figure 7B ) . In the presence of dithiothreitol , a reducing reagent that disrupts disulfide bonds , AtLYK5 proteins became monomers ( Figure 7B ) . In addition , we did not observe any molecules larger than AtLYK5 dimers , suggesting no oligomerization of AtLYK5 ( Figure 7B ) . 10 . 7554/eLife . 03766 . 020Figure 7 . AtLYK5 forms a homodimer . ( A ) Homodimeriztion of AtLYK5 is independent on the presence of CERK1 or chitin elicitation . AtLYK5-HA and AtLYK5-Myc , or AtCERK1-HA and AtCERK1-Myc were co-expressed in protoplasts made from Col-0 and Atcerk1 mutant plants . Protoplasts were harvested before ( − ) or 15 min after ( + ) treatment with 1 µM different chitin oligomers . Co-immunoprecipitation was made using anti-Myc antibody . ( B ) Dithiothreitol ( DTT ) treatment converts AtLYK5 dimer to monomer . Crude protein was extracted from transgenic Arabidopsis expressing AtLYK5-HA in Atlyk5-2 mutant plants . Plant tissues were harvested before ( − ) and 15 min after ( + ) treatment with 1 µM chitooctaose . Crude proteins from these tissues were boiled for 5 min before ( − ) or after ( + ) adding 50 mM DTT . Left panel and right panel of the immunoblot detected with anti-HA antibody are from the same gel . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 02010 . 7554/eLife . 03766 . 021Figure 7—figure supplement 1 . A possible working model of chitin receptor in Arabidopsis . Before chitin treatment , AtLYK5 is present as a homodimer . After chitin elicitation , AtCERK1 associates with AtLYK5 to form a possible tetremer to mediate chitin signaling , and AtCERK1 will be phosphorylated at the same time . In this model , AtLYK5 ( or AtLYK4 ) serves as a chitin perception , while AtCERK1 is responsible for chitin signaling transduction due to lack of kinase activity of AtLYK5 ( or AtLYK4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03766 . 021 The current model for AtCERK1 function indicates that this protein directly binds long chain chitooligosaccharides ( dp ≥ 6 ) , leading to homodimerization , kinase activation and downstream induction of MTI ( Bohm et al . , 2014; Kadota et al . , 2014; Liu et al . , 2012b ) . The problem with this model is that it does not account for the relatively low binding affinity of AtCERK1 for chitin and the fact that mutations ( AtCERK1A138H ) predicted to disrupt chitin binding did not block AtCERK1 autophosphorylation ( Liu et al . , 2012b ) . These discrepancies can now be explained by the interaction of AtCERK1 with AtLYK5 , which binds chitin with a significantly higher affinity than AtCERK1 . Indeed , AtLYK5 is required for AtCERK1 dimerization and also kinase activation . The data suggest that AtLYK5 exists in the cell , in the absence of chitin , as a homodimer . Chitin binds to AtLYK5 leading to its association with AtCERK1 , dimerization and kinase activation . In this way , chitin signaling is transduced downstream through AtCERK1 kinase activity . Our model , together with the studies from rice , suggests that AtCERK1 is not the primary chitin receptor . In rice , one long-chain chitooligosaccharide is sandwiched between two OsCEBiP monomers . Chitin treatment induces the association of OsCEBiP with OsCERK1 , which activates the OsCERK1 intracellular kinase domain ( Hayafune et al . , 2014; Shimizu et al . , 2010 ) . However , in Arabidopsis , the three CEBiP-like proteins are not required for chitin-triggered innate immunity ( Shinya et al . , 2012; Wan et al . , 2012 ) . Instead , AtLYK5 appears to play a similar role as CEBiP as a chitin-binding receptor to mediate AtCERK1 activation through ligand-induced association . These two systems are very similar since both OsCEBiP and AtLYK5 lack the intracellular kinase function . OsCEBiP has no intracellular kinase domain , while AtLYK5 contains an intracellular kinase domain that lacks activity . However , the AtLYK5 intracellular kinase domain does appear to have biological function , likely by mediating protein–protein interactions . It is likely that the chitin receptor complex functions as a heterotetramer in both rice ( Hayafune et al . , 2014 ) and Arabidopsis ( Figure 7—figure supplement 1 ) , although this needs to be confirmed . Such an oligomer complex could explain why in vitro receptor binding affinities ( ∼1 µM ) do not correlate well with the measured responses of plants to chitin elicitation at nM concentrations ( Zhang et al . , 2002 ) . An oligomeric AtCERK1 receptor complex is also consistent with the finding that AtLYM1 and AtLYM3 ( Willmann et al . , 2011 ) are necessary for PGN-induced MTI . All these studies suggest that the primary function of AtCERK1/OsCERK1 is not chitin binding per se , but to serve as a receptor scaffold for interaction with chitin binding co-receptors and , most importantly , to provide intracellular kinase activity . Based on ITC data , AtLYK5 has similar chitooctaose binding affinity to that of wheat germ agglutinin ( WGA ) while , under the same conditions , AtCERK1 showed extremely low chitooctaose binding activity . Previously published data , also using ITC , measured the chitopentaose binding affinity of WGA in the low µM range , while the binding affinity ( Kd ) to shorter chitin oligomers ( GlcNAc ) 2–4 was significantly lower ( i . e . , mM level; Asensio , et al . , 2000 ) . ITC is one of the best methods to study protein–protein interactions or protein-ligand association . Since ITC detects protein–protein or protein-ligand interaction directly which is label-free , the binding affinity value might be slightly different from other methods . For example , in the case of WGA , ITC values ( Kd ) for binding to chitobiose and chitotriose were 1 . 6 mM and 117 µM , respectively , ( Asensio , et al . , 2000 ) , while similar measurements using surface plasmon resonance ( SPR ) gave Kd values of 165 µM and 45 µM , respectively ( Lienenmann et al . , 2009 ) . Published reports for the chitin binding affinity for AtCERK1 also vary widely . For example , ( Iizasa et al . , 2010 ) measured the affinity of AtCERK1 for chitin on a solid surface ( commercial chitin magnetic beads ) and reported a value ( Kd ) of 82 nM . It is difficult to reconcile this number with other values obtained by ITC . One possible explanation is that the use of a solid surface promoted oligomerization of AtCERK1 resulting in an enhancement of binding affinity . In contrast , using ITC , the binding affinity ( Kd ) of AtCERK1 for chitotetrose , chitopentaose , and chitooctaose was measured as 159 µM , 66 µM , and 45 µM , respectively ( Liu et al . , 2012b ) . However , these studies were performed with significantly higher protein and ligand concentration ( i . e . , 0 . 1 mM AtCERK1 protein vs 4 mM chitotetraose or chitopentaose , and 2 . 4 mM chitooctaose ) than used in our study . Again , such high protein concentrations could have promoted oligomerization of the AtCERK1 protein , which would affect the binding affinities . Arabidopsis has five LYK genes . We can now assign biological function to four of these proteins . AtLYK3 appears to be involved in recognizing short chain ( dp = 4–5 ) lipo-chitooligosaccarides and chitooligosaccarides ( Liang et al . , 2013 ) . However , this recognition leads to a suppression of MTI . The role of AtLYK2 remains unknown but it should be noted that this gene shows very low expression in all tissues examined ( Liang et al , 2014 ) . AtCERK1/AtLYK1 , AtLYK4 and AtLYK5 all appear to mediate the plant response to chitin elicitation leading to MTI . Previously , we reported that the Ler ecotype Atlyk5-1 mutant was not defective in chitin-induced MTI . However , our current data show that this mutant likely represents a weak mutant allele , perhaps due to the 3′ position of the transposon insertion . The Ler ecotype is also not ideal for studying chitin signaling since it shows a generally weaker response compared to Col-0 . In contrast , the Col-0 ecotype Atlyk5-2 mutant showed a reduced response to chitin treatment in all assays tested , including pathogen response . The phenotype of the Atlyk5-2 mutant was reversed by genetic complementation with the wild-type AtLYK5 gene . Chitin-induced MTI requires long chain chitin oligomers ( dp ≥ 6 ) ( Miya et al . , 2007; Wan et al . , 2008 , 2012 ) . The functions of AtLYK4 and AtLYK5 are partially redundant in that single mutants retain residual responses to chitin , while Atlyk4/Atlyk5-2 double mutant plants show a complete lack of response to chitin , similar to the Atcerk1 mutants . However , AtLYK5 appears to have the predominant role in chitin elicitation , as judged by the relative strength of the Atlyk4 and Atlyk5-2 mutant phenotypes . There are biochemical differences in the function of AtLYK4 and AtLYK5; for example , AtLYK4 interacts with AtCERK1 independently of the presence of chitin , while AtLYK5-AtCERK1 interaction is chitin dependent . The biological significance of these differences and the functional redundancy of AtLYK4 and AtLYK5 are currently unclear . One possibility is that AtLYK4 also functions as a chitin binding receptor to mediate the plant chitin repsonse . Receptor homodimerization or oligomerization and subsequent phosphorylation are common mechanisms for ligand-mediate receptor activation . Arabidopsis examples include AtBRI1 homodimerization and heterodimerization with AtBAK1 ( Li et al . , 2002; Nam and Li , 2002; Santiago , et al . , 2013; Sun , et al . , 2013a; Wang et al . , 2005 ) . AtFLS2 appears to exist in the cell as a homodimer before and after ligand perception and then associates with AtBAK1 upon flagellin treatment ( Albert , et al . , 2013; Chinchilla et al . , 2007; Heese et al . , 2007; Sun et al . , 2012; Sun , et al . , 2013b ) . The suggested mechanism is that ligand perception triggers autophosphorylation between the homodimer or trans-autophosphorylation between/among hetero-interacting proteins to initiate cellular signaling . It appears that all LysM receptors may function as a protein complex . For example , in leguminous plants , the LysM domain , Nod factor receptors ( e . g . , Lotus japonicas Nod factor receptor 1 and 5; LjNFR1 and LjNFR5 ) function as a heterodimer to mediate high affinity binding to the lipo-chitooligosaccharide Nod factor ( Broghammer et al . , 2012; Gust et al . , 2012; Madsen et al . , 2011; Radutoiu et al . , 2003 ) . It should be emphasized that this model for Nod factor binding is very similar to that of the Arabidopsis chitin receptor . In both cases , one LysM RLK ( CERK1/NFR1 ) has an active intracellular kinase domain but interacts with a second LysM RLK ( LYK5/NFR5 ) that lacks intracellular kinase activity . This similarity underlines the now well recognized evolutionary link between chitin , Nod factor and mycorrhizal ( Myc ) factor recognition ( Liang et al . , 2014 ) . Arabidopsis mutant plants Atcerk1 ( GABI-KAT 096F09 ) , Atlyk2 ( SAIL_318C08 ) , Atlyk3 ( SALK_140374 ) , Atlyk4 ( CS850683 ) , Atlyk5-2 ( SALK_131911C ) , and wild-type Col-0 plants were used in this study . Typically , 4-week-old Arabidopsis plants grown in a condition of 16 hr light/8 hr dark cycle at 22–23°C were used for diverse treatments . For chitin treatment , usually 1 µM chitooctaose ( Sigma , St Louis , MO ) , unless otherwise mentioned , was used to treat plant tissues . Homozygous Atlyk5-2 mutant and Atlyk4/Atlyk5-2 double mutants plants were genotyped using primers listed in Supplemental file 1 . All primers used for gene cloning are listed in Supplemental file 1 . The full-length cDNAs of AtLYK2 , AtLYK3 , AtLYK4 , and AtLYK5 were amplified using the template made by Wan et al . ( 2012 ) and cloned into pDONR-Zeo plasmid by BP cloning . The resultant plasmid was then used for LR cloning with destination plasmids pUC-GW14 and pUC-GW17 ( Cao et al . , 2013 ) . The resultant plasmids were used for protein expression in Arabidopsis protoplasts . Two genomic DNA fragments , containing the 1 . 5 kb promoter of AtLYK4 and AtLYK4 coding region up to the stop codon and the 1 . 8 kb promoter of AtLYK5 and AtLYK5 coding region up to the stop codon , and cDNA of AtLYK5 , were individually amplified with pfu Ultra II HF ( Agilent Technologes , Santa Clara , CA ) and cloned into the pDONR-Zeo vector using BP clonase ( Invitrogen , Carlsbad , CA ) . The resulting plasmids were then recombined into the destination binary vector pGWB13 or pGWB5 ( Nakagawa et al . , 2007 ) for Arabidopsis transformation or transient expression in N . benthamiana . Kinase domains of AtLYK5 ( 309–664 aa ) and AtCERK1 ( 255–617 aa ) were amplified and inserted into pGEX 5X-1 between EcoR I and Xho I and between EcoR I and Sal I , separately , and transformed into BL21 ( DE3 ) for recombinant protein expression . Arabidopsis protoplasts were prepared from 4-week-old plants according to the protocol described by Yoo et al . ( 2007 ) . For protein expression , protoplasts ( 200 µl , about 2 × 105 cells ) were transfected with 20 µg plasmids . For co-immunoprecipitation assays , protoplasts ( 1 ml; ∼106 cells ) were transfected with 100 µg plasmid . After incubation in a growth chamber at 23°C overnight ( 14–16 hr ) , the transfected protoplasts were treated with chitooctaose for the times shown in the figure legends and frozen in liquid nitrogen and stored in −80°C for further use . Samples from either protoplasts or plant tissues were lysed in a buffer containing 50 mM Tris ( PH 7 . 6 ) , 150 mM NaCl , 0 . 5% Triton X-100 and 1 × protease inhibitor ( Sigma , MO ) . The resulting extract was centrifuged at 14 , 000 rpm for 15 min at 4°C . Either anti-Myc ( Covance , Princeton , New Jersey ) or anti-AtCERK1 antibody was used for CoIP experiments according the method described by Cao et al . ( 2013 ) . Anti-AtCERK1 antibody was made based on the peptide N′-CNFQNEDLVSLMSGR-C′ located at c-terminal of AtCERK1 by GenScript Company ( Piscataway , NJ ) . 1 µM chitooctaose or H2O was hand-infiltrated into leaves of different plants . At the time points shown in the figures , leaves were harvested and frozen in liquid nitrogen and ground using a Bead Ruptor Homogenizer ( Omni , Kennesaw , GA ) . Samples were placed on ice for 30 min during the lysis . After centrifuging at 13 , 000 rpm at 4°C for 15 min , the supernatants were boiled for 5 min in 1 × SDS loading buffer . The phosphorylated AtCERK1 was separated on 7% SDS-PAGE gel at low voltage ( 60–80 V ) for 4 hr or until the protein ladder with 70 kDa reached the bottom of the gel . For dephophorylation assay , antarctic phosphatase ( New England Biolabs , Ipswich , MA ) was used for treatment for 15 min at 37°C . AtCERK1 was detected with anti-AtCERK1 antibody . ROS production and MAPK phosphorylation assays were performed as described by Liang et al . ( 2013 ) . The disease assay with A . brassicicola was conducted as described by Wan et al . ( 2008 ) . 1 µM chitooctaose or H2O ( control ) was infiltrated into leaves from 4-week-old plants 24 hr before bacterial infiltration . The bacterial pathogen assay was carried out according to the method described by Cao et al . ( 2013 ) . Aequorin transgenic seeds were kindly provided by Dr Marc R Knight ( University of Oxford ) . The Atlyk5-2 mutant was crossed with Col-0 aequorin to make an Atlyk5-2 aequorin line . Calcium influx assays were done using the same method described by Liang et al . ( 2013 ) . Purification of AtCERK1 and AtLYK5 intracellular domains fused with N-terminal GST tag and in vitro kinase assay were performed as described by Cao et al . ( 2013 ) . Briefly , cultures of E . coli strains harboring plasmid were supplemented with 0 . 1 mM isopropyl b-D-thiogalactopyranoside ( IPTG ) at OD600 0 . 8 at 18°C for 12 hr . The cells were lysed in a buffer containing 1 × PBS ( MP Biomedicals , France ) supplemented with 1 mM EDTA , 0 . 1% Triton X-100 , 1 mg/ml lysozyme and placed on ice for 30 min with slow shaking before sonication . After centrifuge at 4°C for 30 min , the supernatant was applied to a column containing glutathione Sepharose 4B ( GE Healthcare , Milwaukee , WI ) for protein purification . The column was washed five-times with 1 × PBS buffer . Recombinant proteins were eluted with 10 mM reduced glutathione . Proteins were dialyzed with a buffer containing 50 mM Tris ( PH 7 . 6 ) , 50 mM KCl , and 10% glycerol . For in vitro kinase assay , proteins were incubated in the buffer [50 mM Tris ( PH 7 . 6 ) , 50 mM KCl , 10 mM MnCl2 , 10 mM MgCl2 , 10 mM ATP , and 10 mCi γ-[32P]-ATP at room temperature for 30 min . Autoradiography was performed using a phosphor screen and a phosphorimager . Total RNA was extracted from 10-day-old seedlings using an RNase easy kit ( Invitrogen , Grand Island , NY ) according to the manufacturer's instructions . qRT-PCR was carried out as previously described by Tanaka et al . ( 2011 ) . The primers used are listed in Supplemental file 1 . All the binary vectors were electroporated into Agrobacterium tumefaciens GV3101 ( pMP90 ) and transformed by the floral dip method ( Clough and Bent , 1998 ) into Atlyk5-2 mutant plants . Transgenic plants were selected on half strength MS with 25 mg/l hygromycin after seed surface sterilization described by Clough and Bent ( 1998 ) . AtLYK5 cDNA in pDONR-Zeo vector was recombined into binary vector pGWB5 using LR reaction . The resultant plasmid was electroporated into agrobacterial strain GV3101 for transiently expression in N . benthamiana according the method described by Wan et al . ( 2012 ) . 2 days after infiltration , the infiltrated leaf was used to monitor fluorescence signal using confocal microscope . The ectodomain of AtLYK5 ( 27–278 aa ) was amplified and inserted into pMCSG73 using the ligation-independent procedure ( PMID: 18988021 ) . A thrombin cleavage site ( LVPRGS ) was inserted right before the His tag using site-directed mutagenesis PCR . Cultures of E . coli BL21 , expressing the appropriate plasmid , were supplemented with 0 . 1 mM IPTG at OD600 0 . 8 at 18°C for 24 hr . Recombinant protein was purified with Talon metal affinity resin and eluted with 300 mM imidazole . Eluted protein was then incubated with thrombin at 4°C overnight and subsequently dialyzed with ( 50 mM Tris ( PH 8 . 0 ) 150 mM NaCl ) . The resultant protein was purified over Strep-Tactin resin . The ectodomain of AtCERK1 ( 24–231 aa ) was amplified and inserted into pMAL-c2G vector at the BamH I and Sal I sites . After expression as described above , the recombinant protein was purified over amylose affinity resin according to the manufacturer's protocol ( New England Biolabs , Ipswich , MA ) . All protein samples including the ectodomains of AtLYK5 and AtCERK1 , and wheat germ agglutinin ( Sigma , St Louis , MO ) were dialyzed against a buffer containing 10 mM HEPES , pH 7 . 5 , 100 mM NaCl . Chitooctaose used for ITC was purchased from Sigma . 20 µM AtLYK5 or AtCERK1 purified protein or 10 µM wheat germ agglutinin ( Sigma , St Louis , MO ) was titrated separately against either 400 µM chitooctaose or 200 µM chitooctaose at 25°C . As a control , 10 µM AtLYK5 was titrated against 200 µM chitotetraose . Buffer lacking protein was used as control and titrated against 200 µM chitooctaose . The heat from each experiment was measured by MicroCal VP-ITC . Homology modeling of the LysM domain of AtLYK5 was performed in the Modeller 9 . 12 ( Marti-Renom et al . , 2000 ) using the crystallized structure of ECP6 was used as the template ( PDB code 4B8V ) . The query sequence and template structure alignment was first performed using the Modeller Align module , and then manually inspected to ensure the best alignment for generating a pool of 2000 models . The best model was selected and analyzed based on Modeller's probability density function ( Discrete Optimized Protein Energy score ) and the Ramachandran plot ( Laskowski et al . , 1993 ) , and subsequently refined using the Loop refine module of Modeller . In order to perform docking and calculate a binding affinity , the LysM model and the chitooctaose ligand ( NCBI identifier 24978517 ) were prepared using the MGLTools-1 . 5 . 6 software ( Morris et al . , 2009 ) to satisfy docking requirements such as addition of hydrogen atoms , calculation of partial charge using the AMBER force field ( Cornell et al . , 1995 ) , selection of flexible bonds for the ligand and residues , and adjustment of docking position and grid space ( parameters not shown ) . The putative binding site of the LysM domain model of AtLYK5 was inferred from the template structure . The docking experiment was then performed with the AutoDockVina software ( Trott and Olson , 2010 ) , and the docking model with the lowest binding energy ( expressed in kcal per mol ) was selected and visualized in the Chimera software ( Pettersen et al . , 2004 ) . A detailed interaction map between the ligand and surrounding residues was generated by the LigPlus software ( Wallace et al . , 1995 ) . To calculate the binding affinity of ECP6 to chitooctaose as a positive control , only the crystal model of ECP6 was retrieved for the docking task . The docking experiment with the chitooctaose ligand was performed with the AutoDock Vina software as mentioned above . The docking model with the lowest binding energy was selected and visualized .	Invading fungi are responsible for many of the plant diseases that affect global crop production . Plants have to be able to identify these fungi , and activate the right defense strategies if they are to protect themselves . Chitin is a polymer that is found in the cell walls of all fungi , but not in plants , so if the plant detects chitin , it knows that a potentially harmful fungus may be nearby . The detection of chitin , and the resulting activation of a plant's defenses , requires a receptor protein called CERK1 . In rice , CERK1 needs to interact with another receptor protein called CEBiP , which binds to chitin . However , in Arabidopsis thaliana—which is widely studied in plant research—CERK1 can bind to chitin on its own , although this interaction is very weak , so it has been suggested that a second protein may be involved . Cao et al . have now found that a receptor protein called LYK5 , which is very similar to CERK1 , is much better at attaching to chitin in A . thaliana . It can also bind to CERK1 , but only when chitin is present , and is required for activation of basic plant defenses . The experiments suggest that LYK5 detects chitin on behalf of CERK1 , in a similar way to how CEBiP works in rice . The next step in this research is to work out how CERK1 and LYK5 are able to activate plant defenses .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology" ]	2014	The kinase LYK5 is a major chitin receptor in Arabidopsis and forms a chitin-induced complex with related kinase CERK1
Multivalent presentation of viral glycoproteins can substantially increase the elicitation of antigen-specific antibodies . To enable a new generation of anti-viral vaccines , we designed self-assembling protein nanoparticles with geometries tailored to present the ectodomains of influenza , HIV , and RSV viral glycoprotein trimers . We first de novo designed trimers tailored for antigen fusion , featuring N-terminal helices positioned to match the C termini of the viral glycoproteins . Trimers that experimentally adopted their designed configurations were incorporated as components of tetrahedral , octahedral , and icosahedral nanoparticles , which were characterized by cryo-electron microscopy and assessed for their ability to present viral glycoproteins . Electron microscopy and antibody binding experiments demonstrated that the designed nanoparticles presented antigenically intact prefusion HIV-1 Env , influenza hemagglutinin , and RSV F trimers in the predicted geometries . This work demonstrates that antigen-displaying protein nanoparticles can be designed from scratch , and provides a systematic way to investigate the influence of antigen presentation geometry on the immune response to vaccination . Multivalent antigen presentation , in which antigens are presented to the immune system in a repetitive array , has been demonstrated to increase the potency of humoral immune responses ( Bennett et al . , 2015; Snapper , 2018 ) . This has been attributed to increased cross-linking of antigen-specific B cell receptors at the cell surface and modulation of immunogen trafficking to and within lymph nodes ( Irvine et al . , 2013; Tokatlian et al . , 2019 ) . An ongoing challenge has been to develop multimerization scaffolds capable of presenting complex oligomeric or engineered antigens ( Sanders and Moore , 2017; Jardine et al . , 2013; McLellan et al . , 2013a ) , as these can be difficult to stably incorporate into non-protein-based nanomaterials ( e . g . liposomes , polymers , transition metals and their oxides ) . Epitope accessibility , proper folding of the antigen , and stability are also important considerations in any strategy for antigen presentation . Several reports have utilized non-viral , naturally occurring protein scaffolds , such as self-assembling ferritin ( Kanekiyo et al . , 2013; Sliepen et al . , 2015; Darricarrère et al . , 2018 ) , lumazine synthase ( Sanders and Moore , 2017; Abbott et al . , 2018 ) , or encapsulin ( Kanekiyo et al . , 2015 ) nanoparticles , to present a variety of complex oligomeric or engineered antigens . These studies have illustrated the advantages of using self-assembling proteins as scaffolds for antigen presentation ( López-Sagaseta et al . , 2016; Kanekiyo et al . , 2019 ) , including enhanced immunogenicity and seamless integration of antigen and scaffold through genetic fusion . More recently , computationally designed one- and two-component protein nanoparticles ( Hsia et al . , 2016; King et al . , 2014; Bale et al . , 2016 ) have been used to present complex oligomeric antigens , conferring additional advantages such as high stability , robust assembly , ease of production and purification , and increased potency upon immunization ( Marcandalli et al . , 2019; Brouwer et al . , 2019 ) . The ability to predictively explore new structural space makes designed proteins ( Parmeggiani et al . , 2015; Brunette et al . , 2015 ) attractive scaffolds for multivalent antigen presentation . In our previous work with computationally designed nanoparticle immunogens ( Marcandalli et al . , 2019; Brouwer et al . , 2019 ) , the nanoparticles were generated from naturally occurring oligomeric proteins without initial consideration of geometric compatibility for antigen presentation . A more comprehensive solution would be to de novo design nanoparticles which present complex antigens of interest . For homo-oligomeric class I viral fusion proteins , a large group that includes many important vaccine antigens ( Harrison , 2015 ) , a close geometric match between the C termini of the antigen and the N termini of a designed nanoparticle component would enable multivalent presentation without structural distortion near the glycoprotein base , and potentially allow for better retention of antigenic epitopes relevant to protection . More generally , precise control of antigen presentation geometry through de novo nanoparticle design would enable systematic investigation of the structural determinants of immunogenicity . We sought to develop a general computational method for de novo designing protein nanoparticles with geometries tailored to present antigens of interest , focusing specifically on the prefusion conformations of the trimeric viral glycoproteins HIV-1 Env ( BG505 SOSIP ) ( Wang et al . , 2017; Sanders et al . , 2013 ) , influenza hemagglutinin ( H1 HA ) ( Kadam et al . , 2017 ) , and respiratory syncytial virus ( RSV ) F ( DS-Cav1 ) ( McLellan et al . , 2013a ) . To make the antigen-tailored nanoparticle design problem computationally tractable , we employed a two-step design approach ( Figure 1 ) . In the first step , we de novo designed antigen-tailored trimers , featuring N termini geometrically matched to the C termini of the viral glycoproteins . In the second step , we generated tetrahedral , octahedral , and icosahedral two-component nanoparticles by designing secondary interfaces between a designed trimer ( fusion component ) and a de novo homo-oligomer ( assembly component ) ( Fallas et al . , 2017 ) . This design approach yielded nanoparticles tailored to present 4 , 8 , or 20 copies of the viral glycoproteins in defined geometries ( Figure 1d ) . Sequences for all designed trimers and homo-oligomers , two-component nanoparticles , and antigen-fused components in this study can be found in Supplementary file 1A , B , and C , respectively . Details on each step of the design approach are described in the following sections . We chose to design our antigen-tailored trimers from monomeric repeat proteins composed of rigidly packed 20- to 50-residue tandem repeat units ( Parmeggiani et al . , 2015; Brunette et al . , 2015; Urvoas et al . , 2010; Kajander et al . , 2007; Main et al . , 2003 ) , as their high stability and tunable length ( through variation of repeat number ) are desirable properties for the design of protein-based nanomaterials . These structurally diverse alpha-helical repeat proteins featured three to six repeat modules and total lengths between 119 and 279 residues . They were docked into C3-symmetric trimers using our RPX docking method , which identifies configurations likely to accommodate favorable side chain packing at the de novo designed interface ( Fallas et al . , 2017 ) . To identify trimeric configurations with N termini compatible for fusion to the C termini of the three viral glycoproteins , docks with an RPX score above 5 . 0 were screened using the sic_axle protocol ( Marcandalli et al . , 2019 ) . Geometrically compatible docks ( non-clashing termini separation distances of 15 Å or less ) were subjected to full Rosetta C3-symmetric interface design and filtering ( see Materials and Methods ) , and twenty-three designs were selected for experimental characterization ( Figure 1—figure supplement 1 ) . Synthetic genes encoding each of the designed trimers were expressed in E . coli and purified from lysates by Ni2+ immobilized metal affinity chromatography ( Ni2+ IMAC ) followed by size-exclusion chromatography ( SEC ) . Twenty-two designs were found to express in the soluble fraction , and nine formed the intended trimeric oligomerization state as assessed by SEC in tandem with multi-angle light scattering ( SEC-MALS; examples in Figure 2 top panel , second row; SEC-MALS chromatograms for the remaining designs are in Figure 2—figure supplement 2 and data in Figure 2—figure supplement 1—source data 1; SEC chromatograms for remaining designs with off-target retention volumes are in Figure 2—figure supplement 2 ) . Four of the designs that were trimeric and expressed in high yield , 1na0C3_2 , 3ltjC3_1v2 , 3ltjC3_11 , and HR04C3_5v2 , were selected for solution small angle X-ray scattering ( SAXS ) experiments . The proteins exhibited scattering profiles very similar to those computed from the corresponding design models , suggesting similar supramolecular configuration ( Figure 2 top panel , third row; metrics in Table 1 and Figure 2—source data 1 ) . These four trimers were derived from three distinct designed helical repeat proteins from TPR , HEAT , or de novo topological families ( 1na0 , 3ltj , and HR04 , see Materials and Methods ) ( Brunette et al . , 2015; Urvoas et al . , 2010; Main et al . , 2003 ) . Crystals were obtained for the two designs 1na0C3_2 and 3ltjC3_1v2 . Structures were determined at resolutions of 2 . 6 and 2 . 3 Å , revealing a backbone root mean square deviation ( r . m . s . d . ) between the design model and structure of 1 . 4 and 0 . 8 Å , respectively ( Figure 2—figure supplement 3 , and Figure 2—figure supplement 3—source data 1 , crystallization conditions , structure metrics , and structure-to-model comparisons are described in Materials and Methods ) . The structures confirmed in both cases that the designed proteins adopt the intended trimeric configurations , and that most of the atomic details at the de novo designed interfaces are recapitulated . As secondary assembly components were required to design our antigen-tailored nanoparticles , validated trimers were docked pairwise with de novo designed symmetric homo-oligomers ( Fallas et al . , 2017 ) to generate tetrahedral , octahedral , and icosahedral nanoparticle configurations using the TCdock program ( King et al . , 2014; Bale et al . , 2016 ) . To increase the probability of generating icosahedra which confer the highest valency among the targeted symmetries , three naturally occurring homopentamers were also included in the docking calculations ( PDB IDs 2JFB , 2OBX , and 2B98 ) . Analogously to the designed trimers , nanoparticle docks were scored and ranked using the RPX method ( Fallas et al . , 2017 ) to identify configurations likely to accommodate favorable side chain packing at a secondary de novo designed interface . High-ranking and non-redundant nanoparticle configurations featuring outward-facing N termini for antigen presentation were selected for Rosetta interface design ( King et al . , 2014; Bale et al . , 2016 ) . Fifty-three nanoparticle designs across all three targeted symmetries that exhibited the best interface metrics were selected for experimental characterization ( see Materials and Methods ) . The nomenclature for the eleven tetrahedra , twenty-one octahedra , and twenty-one icosahedra indicate the symmetry of the nanoparticle ( T , O , or I ) , the oligomeric state of the first component ( A ) and second component ( B ) used in each design , the letters “dn” reflecting the de novo nature of the input oligomers , and the rank by RPX score from the docking stage ( e . g . , “I53_dn5” indicates an icosahedral nanoparticle constructed from a pentameric and trimeric component , ranked 5th in RPX-scoring for the two input oligomers ) . Synthetic genes encoding each of the two-component nanoparticles were obtained with one of the components fused to a His6-tag , and the designs were purified using Ni2+ IMAC ( see Materials and methods ) . Pairs of proteins at the expected molecular weights were found to co-elute by SDS-PAGE for twenty-four of the designs , consistent with spontaneous assembly of the nanoparticles followed by pulldown His6-tagged component ( featured co-eluting designs are presented in Figure 2—figure supplement 4 ) . SEC chromatograms revealed that nineteen designs did not form assemblies of the expected size or that the resulting assemblies were heterogeneous ( Figure 2—figure supplement 5 ) . Five designs comprising a panel of unique geometric configurations , T33_dn2 , T33_dn5 , T33_dn10 , O43_dn18 , and I53_dn5 , ran as monodisperse particles of the predicted molecular mass by SEC-MALS and were further investigated by SAXS . The experimental solution scattering curves closely matched the scattering curves computed from the design models ( Schneidman-Duhovny et al . , 2010 ) for all five designs ( Figure 2 , bottom panel and Figure 2—figure supplement 6; metrics in Table 1 and Figure 2—source data 1 , bottom five designs ) . Due to its high valency and production yield , we selected the I53_dn5 nanoparticle to investigate the capacity of its two components to be separately produced and assembled in vitro . The two components of I53_dn5 were re-cloned , expressed , and separately purified ( pentameric "I53_dn5A" with His6-tag and trimeric “I53_dn5B” ) . Nanoparticle assembly appeared to be complete within minutes after equimolar mixing ( Figure 2—figure supplement 7 ) . This capability is noteworthy as it enables production of each component independently , even from different host systems , which provides more flexibility in nanoparticle manufacturing . In vitro assembly also confers more control over nanoparticle assembly and composition , for example by assembling with a mixture of components fused to different antigens ( Boyoglu-Barnum et al . , 2020 ) . The five SAXS-validated nanoparticles were structurally characterized using negative stain electron microscopy ( NS-EM ) ( Lee and Gui , 2016; Ozorowski et al . , 2018 ) . 2 , 000–5000 particles were manually picked from the electron micrographs acquired for each designed nanoparticle and classified in 2D using the Iterative MSA/MRA algorithm ( see Materials and Methods ) . 3D classification and refinement steps were performed in Relion/3 . 0 ( Zivanov et al . , 2018 ) . Analysis of the NS-EM data confirmed high sample homogeneity for all five nanoparticle designs as evident from the micrographs and 2D class-averages ( Figure 3 ) . While some free nanoparticle components were detected in the T33_dn5 sample , suggesting a certain propensity towards disassembly , analysis of the reconstructed 3D maps revealed that all five nanoparticles assemble as predicted by the design models , at least to the resolution limits of NS-EM . In order to obtain higher-resolution information , three designs , T33_dn10 , O43_dn18 , and I53_dn5 , representing one nanoparticle from each targeted symmetry ( T , O , I ) , were subjected to cryo-electron microscopy ( cryo-EM ) . Cryo-EM data acquisition was performed as described in the Materials and Methods section and data acquisition statistics are displayed in Figure 4—source data 1 . The data processing workflow is presented in Figure 4—figure supplement 1 . Appropriate symmetry ( T , O , and I for T33_dn10 , O43_dn18 , and I53_dn5 , respectively ) was applied during 3D classification and refinement and maps were post-processed in Relion/3 . 0 ( Zivanov et al . , 2018 ) . The final resolutions of the reconstructed maps for the T33_dn10 , O43_dn18 , and I53_dn5 nanoparticles were 3 . 9 , 4 . 5 , and 5 . 3 Å , respectively . Some structural heterogeneity was observed in the cryo-EM data , particularly in the case of I53_dn5 . In 2D classification results we generated particle projection averages that range from spherical to ellipsoid shape ( Figure 4—figure supplement 1c ) , indicating some degree of flexibility . There is less evidence of flexibility in T33_dn10 and O43_dn18 , in agreement with the higher final map resolution for these nanoparticles . Nanoparticle design models were relaxed into the corresponding EM maps by applying multiple rounds of Rosetta relaxed refinement ( Wang et al . , 2016 ) and manual refinement in Coot ( Emsley and Crispin , 2018 ) to generate the final structures . Refined model statistics are shown in Figure 4—source data 2 . Reconstructed cryo-EM maps for T33_dn10 , O43_dn18 , and I53_dn5 and refined models are superimposed in Figure 4 . Overall , the refined structures show excellent agreement with the corresponding Rosetta design models . Backbone r . m . s . d . values estimated for the asymmetric unit ( consisting of a single subunit of component A and component B ) were 0 . 65 , 0 . 98 , and 1 . 3 Å for T33_dn10 , O43_dn18 , and I53_dn5 , respectively ( Table 1 ) . To explore the capability of the designed nanoparticles to present viral glycoproteins , we produced their trimeric fusion components genetically linked to a stabilized version of the BG505 SOSIP trimer . Synthetic genes for BG505 SOSIP fused to the N termini of T33_dn2A , T33_dn10A , and I53_dn5B ( BG505 SOSIP–T33_dn2A , BG505 SOSIP–T33_dn10A , and BG505 SOSIP–I53_dn5B ) were transfected into HEK293F cells . The secreted fusion proteins were then purified using a combination of immuno-affinity chromatography and SEC . The corresponding assembly component for each nanoparticle was produced recombinantly in E . coli , and in vitro assembly reactions were performed as equimolar mixtures of the two components overnight . Assembled nanoparticles were purified by SEC and analyzed by NS-EM to assess particle assembly and homogeneity . ~ 1000 particles were manually picked and used to perform 2D classification and 3D classification/refinement in Relion ( Zivanov et al . , 2018 ) . Models for the BG505 SOSIP-displaying nanoparticles fit into their reconstructed 3D maps are displayed in Figure 5 ( left ) . BG505 SOSIP trimers are clearly discernible in 2D class-averages and reconstructed 3D maps . However , the trimers appear less well-resolved than the corresponding nanoparticle core in the three reconstructions , likely due to the short flexible linkers between the BG505 SOSIP trimer and the fusion component . The self-assembling cores of the antigen-fused T33_dn2 , T33_dn10 , and I53_dn5 nanoparticles were very similar to the NS-EM maps of the unmodified nanoparticles ( at least to the resolution limits of NS-EM ) , demonstrating that fusion of the BG505 SOSIP trimer did not induce any major structural changes to the underlying nanoparticle scaffolds . Free components were detected in raw EM micrographs of BG505 SOSIP–I53_dn5 , indicating some degree of disassembly . This finding is supported by stability data reported in a parallel study , where BG505 SOSIP–I53_dn5 demonstrated sensitivity to various physical and chemical stressors ( Antanasijevic et al . , 2020 ) . To further characterize the capability of the designed nanoparticles to present viral glycoproteins , we characterized the structures and antigenic profiles of I53_dn5 fused to the prefusion influenza HA and RSV F glycoproteins ( HA–I53_dn5 and DS-Cav1–I53_dn5 ) . Constructs were generated with each glycoprotein genetically linked to the N terminus of the I53_dn5B trimeric fusion component , and the proteins were secreted from HEK293F cells and purified by Ni2+ IMAC . The fusion proteins were mixed with equimolar pentameric I53_dn5A for HA–I53_dn5 or I53_dn5A . 1 ( a stabilized and redox-insensitive variant of I53_dn5A lacking cysteines , see Materials and Methods ) for DS-Cav1–I53_dn5 , and the assembly reactions purified by SEC . For both assemblies , the majority of the material migrated in the peak expected for assembled nanoparticles , and NS-EM analysis showed formation of I53_dn5 nanoparticles with spikes emanating from the surface ( Figure 5—figure supplement 1 and 2 ) . In both cases , there was considerable variation in the spike geometry , again suggesting some flexibility between the glycoproteins and the underlying scaffold . The GG linker connecting DS-Cav1 to I53_dn5 likely accounts for the observed flexibility and suboptimal definition of the glycoprotein trimer in two-dimensional class averages ( Figure 5—figure supplement 1 , bottom right ) . There was no engineered linker between the glycoprotein and fusion component in the case of HA–I53_dn5 , and more clearly defined spike density was observed in the class averages ( Figure 5—figure supplement 2 , bottom right ) . To determine if the presented glycoproteins were properly folded , we examined their reactivity with conformation-specific monoclonal antibodies ( mAbs ) . The DS-Cav1–I53_dn5 nanoparticle was found by an enzyme-linked immunosorbent assay ( ELISA ) to bind the RSV F-specific mAbs D25 ( McLellan et al . , 2013b ) , Motavizumab ( Cingoz , 2009 ) , and AM14 ( Gilman et al . , 2015 ) similarly to soluble DS-Cav1 trimer with foldon ( McLellan et al . , 2013a ) , indicating that the F protein is presented in the desired prefusion conformation on the nanoparticle ( Figure 4—figure supplement 1 , top ) . Biolayer interferometry binding experiments with anti-HA head - and stem-specific mAbs ( Krause et al . , 2011; Ekiert et al . , 2009 ) analogously showed that both the HA–I53_dn5 nanoparticle and the HA–I53_dn5B trimer presented the head and stem regions with wild-type-like antigenicity ( Figure 5—figure supplement 2 , top ) . Previous work involving icosahedral nanoparticle scaffolds presenting HIV-1 Env trimers has shown that antigen crowding can modulate the accessibility of specific epitopes and thereby influence the humoral immune response ( Brouwer et al . , 2019; Brinkkemper and Sliepen , 2019 ) . The nanoparticle scaffolds developed in this work were specifically designed to exhibit varying geometries and valencies , providing a unique way to manipulate and understand epitope accessibility in the context of nanoparticle vaccines . We selected the BG505 SOSIP–T33_dn2 tetrahedral nanoparticle ( assembled in vitro using BG505 SOSIP–T33_dn2A and T33_dn2B ) to compare against a previously published SOSIP-displaying icosahedral nanoparticle ( BG505 SOSIP–I53-50 ) ( Brouwer et al . , 2019 ) through surface plasmon resonance ( SPR ) experiments . BG505 SOSIP–T33_dn2 presents four copies of the BG505 SOSIP trimer with much greater spacing than BG505 SOSIP–I53-50 with twenty copies . This difference derives from the angles between neighboring three-fold rotational symmetry axes—where the displayed BG505 trimers are located on the nanoparticle surfaces—in icosahedral and tetrahedral point group symmetries ( 41 . 8° and 109 . 5° , respectively ) . To first validate mAb binding to BG505 SOSIP–T33_dn2A , NS-EM class averages and a 3D reconstruction were generated in complex with the VRC01 Fab ( Walker et al . , 2011 ) , confirming the expected binding mode ( Figure 6a ) . Next , in part to simulate surface B cell receptors , a panel of anti-Env mAbs targeting epitopes ranging from the apex to the base of the BG505 SOSIP trimer were immobilized on SPR sensor chips ( Figure 6b ) . BG505 SOSIP–T33_dn2A trimer or BG505 SOSIP–T33-dn2 nanoparticle was flowed over the mAbs and the ratio of macromolecule bound was calculated from the binding signal as previously described ( Brouwer et al . , 2019 ) . For mAbs that target apical , V3-base , and CD4-binding site epitopes ( PGT145 , PGT122 , 2G12 , and VRC01 ) ( Gilman et al . , 2015; Walker et al . , 2011; Trkola et al . , 1996; Wu et al . , 2010 ) , the number of molecules of trimer or nanoparticle bound was relatively similar ( ratio ~ 1 ) . However , for mAbs that target more base-proximate epitopes in the gp120-gp41 interface ( ACS202 , VRC34 , and PGT151 ) ( van Gils et al . , 2017; Kong et al . , 2016; Falkowska et al . , 2014 ) , an inter-protomeric gp41 epitope ( 3BC315 ) ( Klein et al . , 2012 ) , and the main autologous neutralizing antibody epitope in the glycan hole centered on residues 241 and 289 ( 11B ) ( McCoy et al . , 2016 ) , the accessibility was reduced in the nanoparticle format . Binding to a mAb directed to the trimer base ( 12N ) ( McCoy et al . , 2016 ) was not observed for nanoparticle BG505 SOSIP–T33_dn2 ( Figure 6b ) . We compared epitope accessibility of BG505 SOSIP–T33_dn2 to that of BG505 SOSIP–I53-50 for six different mAbs ( Brouwer et al . , 2019 ) . As for BG505 SOSIP–T33_dn2 , mAbs to the apex , V3-base , and CD4-binding site ( PGT145 , PGT122 , and VRC01 ) gave molar ratios of ~ 1 compared to BG505 SOSIP–I53-50 . However , for mAbs that target the more base-proximate epitopes in the gp120-gp41 interface ( VRC34 and PGT151 ) , there was nearly three-fold higher epitope accessibility on T33_dn2 compared to I53-50 ( Figure 6c ) . Further down the trimer , no accessibility difference was again observed for a mAb that targets the gp41 inter-protomeric epitope ( 3BC315 ) , which was relatively inaccessible on both nanoparticles , likely due to steric hindrance by neighboring trimers . These findings demonstrate that antigen epitope accessibility can be finely tuned through presentation geometry , which could be used as a strategy to target the immune response against specific epitopes of interest . Strong BCR signaling is required for eliciting robust humoral immune responses , but the molecular mechanisms by which this can be accomplished are not fully understood . Historically , live-attenuated or inactivated viruses and engineered virus-like particles ( VLPs ) have been able to confer protective immunity without pathogenicity , but the empirical discovery and compositional complexity of such vaccines has hampered understanding of possible mechanisms for obtaining sufficient levels of protective antibodies . De novo designed protein nanoparticles provide a modular way to present antigens to the immune system in defined geometries and of known composition . Multivalent antigen presentation can enhance antigen-specific antibody titers by orders of magnitude ( Bennett et al . , 2015; Snapper , 2018 ) , but presentation of complex antigens is challenging due to the required geometric compatibility between antigen and scaffold . The design approach described here , in which nanoparticles incorporate de novo designed homo-oligomers tailored to present antigens of interest , is a general solution to this problem . More broadly , the ability to build protein-based nanomaterials with geometric specifications from scratch represents an important step forward in computational protein design , and provides a systematic way to investigate the influence of antigen presentation geometry on immune response . The ability to fully tailor structures of nanoparticle scaffolds could be particularly useful for HIV-1 Env-based immunogens . While previous studies of HIV-1 Env trimers presented on nanoparticle scaffolds have demonstrated enhanced immunogenicity ( Escolano et al . , 2019 ) , the effects are often modest compared to those observed for other antigens ( Bennett et al . , 2015; Snapper , 2018; Marcandalli et al . , 2019; Brinkkemper and Sliepen , 2019 ) . While there may exist intrinsic peculiarities to HIV-1 Env that limit increases in antibody responses upon multivalent presentation ( Klasse et al . , 2012; Ringe et al . , 2019 ) , limitations associated with epitope inaccessibility caused by crowding of the trimers on nanoparticle surfaces have also been identified ( Sanders and Moore , 2017; Brouwer et al . , 2019 ) . This observation strongly motivates the exploration of antigen presentation across a range of scaffolds to identify geometries that most effectively elicit the desired immune response , particularly when it is of interest to direct the humoral immune response to specific epitopes . Indeed , the SPR experiments presented here demonstrate that epitopes proximate to the BG505 SOSIP base were markedly more accessible to immobilized mAbs on BG505 SOSIP–T33_dn2 than BG505 SOSIP–I53-50 , directly implicating steric crowding on the nanoparticle surface as a determinant of antigenicity . Furthermore , the availability of multiple antigen-displaying nanoparticles makes possible the usage of different scaffolds during prime and boost immunizations , which could limit immune responses directed toward the scaffolds while boosting antigen-specific antibody responses . Finally , the ability to tune antigen presentation geometry through de novo nanoparticle design provides a route to investigate the effects of this parameter on B cell activation , as well as the potency and breadth of the ensuing humoral response . This design approach could help overcome the intrinsically low affinity of germline BCRs for viral glycoproteins , and enable development of broadly neutralizing antibodies . Listed below are the RCSB Protein Data Bank entries for monomeric repeat proteins used for trimer docking and design in this study . An additional set of monomeric repeat proteins is provided in which experimental SAXS data agreed with the computational model ( Fallas et al . , 2017 ) . X-ray Structures ( PDB ID ) SAXS Validated Models1na0 ( 1NA0 ) tpr13ltj ( 3LTJ ) HR002fo7 ( 2FO7 ) HR04 ( 5CWB ) HR07 ( 5CWD ) HR10 ( 5CWG ) The monomeric repeat proteins were used as input to C3-symmetric trimer docking and design against the three viral antigens of interest: HIV-1 BG505 SOSIP , influenza H1 HA , and RSV F protein ( PDB IDs 5VJ6 res . 518–664 , 5KAQ res . 11–501 , 5TPN res . 27–509 ) ( Wang et al . , 2017; Mousa et al . , 2017; Joyce et al . , 2016 ) . A C3-symmetric docking search was first performed , and output was assessed by the previously described RPX scoring method which discerns docks with more potential favorable pair-wise interactions than others ( Fallas et al . , 2017 ) . Up to the top-scoring 100 docks for each repeat protein monomer were aligned against each of the three antigens along the shared C3 axis of symmetry and sampled translationally along and rotationally about the axis in 1 Å and 1° increments , respectively . For each sample , the distance between the C-terminal residue of the antigen and N-terminal residue of the docked trimer was measured until a minimum , non-clashing distance was determined ( Figure 1—figure supplement 1 ) . Solutions for docks that were less than or equal to 15 Å for one or more of the three antigens were selected for full Rosetta symmetric interface design as described in previously published methods ( Fallas et al . , 2017 ) . Individual design trajectories were filtered by the following criteria: difference between Rosetta energy of bound ( oligomeric ) and unbound ( monomeric ) states less than −30 . 0 Rosetta energy units , interface surface area greater than 700 Å2 , Rosetta shape complementarity ( sc ) greater than 0 . 65 , and less than 50 mutations made from the respective native monomer . Designs that passed these criteria were manually inspected and refined by single point reversions , and one design per unique docked configuration was added to the set of trimers selected for experimental validation . Two-component nanoparticle docks were scored and ranked using the RPX method ( Fallas et al . , 2017 ) as opposed to prior methods involving only interface residue contact count ( King et al . , 2014; Bale et al . , 2016 ) . High-scoring and non-redundant nanoparticle configurations were selected for Rosetta interface design with an added caveat that they include trimers with outward-facing N termini for antigen fusion . The design protocol took a single-chain input . pdb of each component , and a symmetry definition file ( DiMaio et al . , 2011 ) containing information for a specified cubic point group symmetry . The oligomers were then aligned to the corresponding axes of the symmetry using the Rosetta SymDofMover , taking into account the rigid body translations and rotations retrieved from the . dok file output from TCdock ( King et al . , 2014; Bale et al . , 2016 ) . A symmetric interface design protocol was employed which included pair-wise interaction motifs found from the RPX method ( Fallas et al . , 2017 ) within each Rosetta symmetric interface design trajectory ( King et al . , 2014; Bale et al . , 2016 ) . Individual design trajectories were filtered by the following criteria: difference between Rosetta energy of bound and unbound states less than −30 . 0 Rosetta energy units , interface surface area greater than 700 Å2 , sc greater than 0 . 6 , and less than 50 mutations made from each native oligomer . Designs that passed these criteria were manually inspected and refined by single point reversions for mutations that did not appear to contribute to stabilizing the bound state of the interface . The sequence with the best overall metrics for each unique docked configuration was selected for experimental characterization . Synthetic genes for designed proteins were optimized for E . coli expression and assembled from genes ( purchased through Genscript or Gen9 ) ligated into the pET21b ( + ) ( designed trimers ) or pET28b ( + ) ( designed nanoparticles ) vector at restriction sites NdeI and XhoI or NcoI and XhoI , respectively . A second ribosome-binding site was inserted between the open-reading frames of individual components of nanoparticle designs ( ‘AGAAGGAGATATCAT’ ) , such that the two proteins would be co-expressed and screened for co-elution by SDS-PAGE . Plasmids were cloned into BL21 or Lemo21 ( DE3 ) E . coli competent cells ( New England Biolabs ) . Transformants were inoculated and grown in 5 mL of either LB or TB medium with either 100 mg/L carbenicillin or 100 µg/L kanamycin at 37°C overnight . Subsequently , liquid cultures were inoculated 1:100 ( v:v ) and grown at 37°C until an OD600 of 0 . 5–0 . 8 . Isopropyl-thio-β-D-galactopyranoside ( IPTG ) was then added at a concentration of 0 . 5–1 mM and growth temperature was reduced to 18°C to induce protein expression , or cultures were left at 37°C and auto-induced by media-included galactose according to the Studier protocols ( Studier , 2005 ) . Expression proceeded for 20 hr until the cell cultures were harvested by centrifugation . Cell pellets were resuspended in 25 mM Tris , 150 mM NaCl , 5 mM imidazole , DNase , EDTA-free protease inhibitors ( Pierce ) , pH 8 . 0 . and lysed by sonication or microfluidization . Each protein was then purified from lysate by Ni2+ IMAC with Ni-NTA Superflow resin ( Qiagen or GE ) . Resin with bound cell lysate was washed with 15 column volumes of 25 mM Tris , 150 mM NaCl , 40 mM imidazole , pH 8 . 0 . Proteins were eluted with five column volumes of 25 mM Tris , 150 mM NaCl , 400 mM imidazole , pH 8 . 0 for further purification by SEC . Elution samples for each designed protein were concentrated down using a 10 , 000 MWCO protein concentrator ( Novagen ) and fractionated by size on an AKTA pure chromatography system using a Superdex 200 ( for designed trimers ) or Superose 6 10/300 GL column ( for designed nanoparticles ) in 25 mM Tris , 150 mM NaCl , pH 8 . 0 ( TBS ) . Sizing profiles were noted based on absorption at 220 nm and 280 nm wavelength light for each fraction . Molecular weights for predominant species in each protein trace were estimated by comparison to the corresponding monomeric profile . Fractions containing single predominant species from an initial SEC purification were concentrated down with 10 , 000 MWCO protein concentrators ( Novagen ) to a concentration of 1 . 0–2 . 0 mg/mL and run through a high-performance liquid chromatography system ( Agilent ) using a Superdex 200 or Superose 6 10/300 GL column ( GE Life Sciences ) in TBS buffer . These fractionation runs were coupled to a multi-angle light scattering detector ( Wyatt ) to determine the absolute molecular weights for each designed protein complex . Designed proteins that predominantly formed the target oligomeric species were re-expressed and purified for low-resolution solution structure determination by SAXS at the SIBYLS High Throughput SAXS Advanced Light Source in Berkeley , California ( Dyer et al . , 2014 ) . A beam exposure time of between 0 . 3 and 10 s was used to obtain averaged diffraction data ( SAXS FrameSlice Application ) , which are represented in plots of log intensity ( I ) vs . q . A 11kEV/1 . 125A X-ray beam was used with a 2 m beamstop . Design 1na0C3_2 was found to crystallize in 1 M LiCl , 100 mM citrate , 20% w/v PEG 6000 , pH 4 , and was frozen using 25% glycerol as cryoprotectant . Design 3ltjC3_1 crystallized in 1 mM DL-glutamic acid monohydrate , 100 mM DL-alanine , 100 mM glycine , 100 mM DL-lysine monohydrochloride , 100 mM DL-serine; 100 mM Tris , 100 mM BICINE , 20% v/v ethylene glycol , 10% w/v PEG 8000 , pH 8 . 5 . Diffraction data for each of these designs were collected at the Advanced Light Source ( Beamline 8 . 2 . 1 ) at Lawrence Berkeley National Laboratory in Berkeley , California . Both designed trimers contained uncleaved C-terminal His6-tags in crystallized conditions . Diffraction data for 3ltjC3_1 was collected on beamline 5 . 0 . 1 at the Advanced Light Source ( Berkeley , CA ) and 1na0C3_2 on beamline 8 . 2 . 1 , both using an ADSC Q315R CCD area detector . Both datasets were scaled and merged in HKL2000 ( Otwinowski and Minor , 1997 ) . The structures were phased by molecular replacement , with the computational design as the search model , using the program PHASER ( McCoy et al . , 2007 ) in the PHENIX software suite ( Adams , 2012 ) . Iterative rounds of manual model building and refinement were conducted in Coot ( Emsley and Crispin , 2018 ) and Phenix . refine ( Afonine et al . , 2012 ) , respectively for both structures . Hydrogens were added for all refinement runs . The geometric quality of the final model was assessed using the Molprobity server ( Chen et al . , 2010 ) . Resolution cutoff was determined by monitoring the refinement statistics in the context of the reflection data completeness and the CC1/2 and I/σI values ( Karplus and Diederichs , 2012 ) . Genes for the individual nanoparticle components were cloned into expression vectors and expressed independently in E . coli . The His6-tagged proteins were purified following the purification protocol described for the designed trimers . Initial SEC chromatograms for the components were obtained on a Superdex 200 10/300 GL column , and predominant peak species were stoichiometrically mixed in TBS buffer for 20 min at 25°C . A secondary SEC step was performed on a Superose 6 10/300 GL column to assess assembly of the intended particle based on expected retention volume . Multiple rounds of designs were performed to remove native unpaired cysteines from I53_dn5A . In the first round of design , cysteines were mutated to alanines ( C94A , C119A ) , which caused the protein to bind and retain through purification a bright yellow metabolite . Further mutations were introduced to knock out metabolite binding in the native enzymatic active site ( W18G ) , which led to protein precipitation during purification . Additional mutations were made ( K84R , M88P , E91D , L117I , L120D ) to re-stabilize the protein , named I53_dn5A . 1 . Synthetic genes were optimized for mammalian expression and subcloned into the pPPI4 vector . BamHI and NheI restriction sites were used for insertion of different nanoparticle components to the C terminus of BG505 SOSIP . Quick Ligation kit , BamHI-HF , and NheI-HF restriction enzymes were purchased from New England Biolabs . BG505 SOSIP variant used for all early optimizations steps was engineered with a combination of v5 . 2 ( Torrents de la Peña et al . , 2017 ) ( mutations: E64K , A73C , A316W , A561C ) and MD3D ( Steichen et al . , 2016 ) ( mutations: M271I , A318Y , R585H , L568D , V570H , R304V , F519S ) stabilizing mutations , and had glycosylation sites introduced at positions 241 and 289 ( mutations: P240T , S241N , F288L , T290E , P291S ) . This construct was termed BG505 SOSIP . v5 . 2 ( 7S ) . For epitope-accessibility experiments ( by surface plasmon resonance ) , a version of this construct was designed without the 241 and 289 glycans . HEK 293F ( RRID:CVCL_6642 ) cells were grown in suspension using FreeStyle 293 Expression Medium ( Thermo Fisher Scientific ) at 135 RPM , 8% CO2 , 80% humidity , 37°C . At confluency of ~ 1 × 106 cells/ml , the cells were co-transfected with pPPI4 DNA vectors encoding the appropriate fusion component ( 250 µg per 1 L of cells ) and furin protease ( 80 µg per 1 L of cells ) . Polyethylenimine ( Polysciences Inc ) was used as a transfection reagent ( 1 mg per 1 L of cells ) . Cells were incubated for 6 days , after which they were spun down by centrifugation ( 7 , 000 RPM , 1 hr , 4°C ) and the protein-containing supernatant was further clarified by vacuum-filtration ( 0 . 45 µm , Millipore Sigma ) . For immuno-affinity chromatography steps , Sepharose 4B columns with immobilized PGT145 IgG were used ( RRID:AB_2491054 ) . Fusion components were eluted with 3 M magnesium chloride , 250 mM L-Arginine buffer , pH 7 . 2 into an equal volume of SEC buffer ( 25 mM Tris , 250 mM L-Arginine , 500 mM NaCl , 5% glycerol , pH 7 . 4 ) . Eluates were concentrated and buffer exchanged into SEC buffer . A Sephacryl S200 16/600 column was used for subsequent SEC purification . Synthetic genes were optimized for mammalian expression and subcloned into the CMV/R vector ( VRC 8400 ) ( Barouch et al . , 2005 ) . XbaI and AvrII restriction sites were used for insertion of I53_dn5B component to the C terminus of the H1 HA ectodomain ( residues 1–676 from A/Michigan/45/2015 ) , which also contained a Y98F mutation to prevent sialic-acid binding and self-aggregation during expression ( Whittle et al . , 2014 ) . Gene synthesis and cloning was performed by Genscript . HEK 293 F cells were grown in suspension using Expi293 Expression Medium ( Thermo Fisher Scientific ) at 150 RPM , 8% CO2 , 70% humidity , 37°C . At confluency of ~ 2 . 5 × 106 cells/mL , the cells were co-transfected with the vector encoding HA–I53_dn5B ( 1000 µg per 1 L of cells ) . Expifectamine was used as a transfection reagent according to the manufacturer’s protocol . Cells were incubated for 96 hr , after which they were spun down by centrifugation ( 4 , 000 RPM , 20 min , 4°C ) and the protein-containing supernatant was further clarified by vacuum-filtration ( 0 . 45 µm , Millipore Sigma ) . For nickel-affinity chromatography steps , a background of 50 mM Tris , 350 mM NaCl , pH 8 . 0 was added to clarified supernatant . For each liter of supernatant , 4 mL of Ni Sepharose excel resin ( GE ) was rinsed into phosphate-buffered saline ( PBS ) using a gravity column and then added to the supernatant , followed by overnight shaking at 4°C . After 16–24 hr , resin was collected and separated from the mixture and washed twice with 50 mM Tris , 500 mM NaCl , 30 mM imidazole , pH 8 . 0 prior to elution of desired protein with 50 mM Tris , 500 mM NaCl , 300 mM imidazole , pH 8 . 0 . Eluates were concentrated and applied to a HiLoad 16/600 Superdex 200 pg column pre-equilibrated with PBS for purification by SEC . Gene synthesis and cloning was performed by Genscript . HEK 293 F cells ( RRID:CVCL_6642 ) were grown in suspension using Expi293 expression medium ( Thermo Fisher Scientific ) at 150 RPM , 8% CO2 , 70% humidity , 37°C . At confluency of ~ 2 . 5 to 3 × 106 cells/ml , the cells were transiently transfected with the vector encoding DS-Cav1–I53_dn5B ( 1 mg per 1 L of cells ) . Expifectamine was used as a transfection reagent according to the manufacturer’s protocol . Cells were incubated for 96 hr and spun down by centrifugation ( 4 , 000 RPM for 20 min at 4°C ) . Supernatant was vacuum-filtered ( 0 . 45 µm , Millipore Sigma ) and 50 mM Tris , 350 mM NaCl , pH 8 . 0 was added for nickel-affinity chromatography . Ni Sepharose resin ( GE ) was washed three times with PBS by centrifugation ( 2 , 000 RPM for 5 min at 4°C ) and added to the supernatant . Nickel-supernatant was incubated either overnight at 4°C or for 2 hr at room temperature . Resin was collected and separated from the mixture and washed twice with 50 mM Tris , 500 mM NaCl , 30 mM imidazole , pH 8 . 0 prior to elution of desired protein with 50 mM Tris , 500 mM NaCl , 300 mM imidazole , pH 8 . 0 . Eluates were concentrated and applied to a HiLoad 10/300 Superdex 200 Increase GL column pre-equilibrated with PBS for purification by SEC . Several reactions containing 5–10 µg of the purified fusion component and an equimolar amount of the corresponding assembly component were prepared and incubated under different conditions ( varying temperature and assembly buffer ) for 24 hr . Native PAGE Bis-Tris gels ( Thermo Fisher Scientific ) and NS-EM was used for detection of assembly . Following the identification of optimal assembly conditions , milligram quantities of particles were assembled and purified by SEC ( Superose six or Sepharose 500 column ) with TBS as the running buffer . Fractions corresponding to the fusion component were pooled and concentrated ( Amicon Ultra Centrifugal Filter Units , Millipore Sigma ) . NS-EM experiments were performed as described previously ( Lee and Gui , 2016; Ozorowski et al . , 2018 ) . Fusion components and assembled nanoparticle samples ( with and without antigen ) were diluted to 20–50 µg/ml and loaded onto the carbon-coated 400-mesh Cu grid that had previously been glow- discharged at 15 mA for 25 s . VRC01 ( RRID:AB_2491019 ) complexes with BG505 SOSIP–T33_dn2A were formed by combining the trimer with a six-fold molar excess of the VRC01 Fab and subsequent incubation for 1 hr at room temperature . Complex sample was diluted to 20 µg/ml and loaded onto the glow discharged Cu grids . Grids were negatively stained with 2% ( w/v ) uranyl-formate for 60 s . Data collection was performed on a Tecnai Spirit electron microscope operating at 120 keV . The magnification was 52 , 000 × with a pixel size of 2 . 05 Å at the specimen plane . The electron dose was set to 25 e-/Å ( Snapper , 2018 ) . All imaging was performed with a defocus value of −1 . 50 µm . The micrographs were recorded on a Tietz 4k × 4 k TemCam-F416 CMOS camera using Leginon automated imaging interface . Data processing was performed in Appion data processing suite . For BG505 SOSIP-fused nanoparticle samples ( v5 . 2 ( 7S ) ) , approximately 500–1000 particles were manually picked from the micrographs and 2D-classified using the Iterative MSA/MRA algorithm . For non-fused nanoparticle samples , 2 , 000–5000 particles were manually picked and processed . For BG505 SOSIP-fused trimer samples and Fab complexes , 10 , 000–40 , 000 particles were auto-picked and 2D-classified using the iterative MSA/MRA algorithm . 3D classification and refinement steps were done in Relion/2 . 1 ( RRID:SCR_015701 ) ( Kimanius et al . , 2016 ) . The resulting EM maps have been deposited to EMDB with IDs: 21162 ( T33_dn2 ) , 21163 ( T33_dn5 ) , 21164 ( T33_dn10 ) , 21165 ( O43_dn18 ) , 21166 ( I53_dn5 ) , 21167 ( BG505 SOSIP–T33_dn2A ) , 21168 ( BG505 SOSIP–T33_dn2A + VRC01 Fab ) , 21169 ( BG505 SOSIP–T33_dn2 nanoparticle ) , 21170 ( BG505 SOSIP–T33_dn10 nanoparticle ) , 21171 ( BG505 SOSIP–I53_dn5 nanoparticle ) . The HA–I53_dn5 complex was adsorbed onto glow-discharged carbon-coated copper mesh grids for 60 s , stained with 2% uranyl formate for 30 s , and allowed to air dry . Grids were imaged using the FEI Tecnai Spirit 120 kV electron microscope equipped with a Gatan Ultrascan 4000 CCD Camera . The pixel size at the specimen level was 1 . 60 Å . Data collection was performed using Leginon ( Suloway et al . , 2005 ) with the majority of the data processing carried out in Appion ( Lander et al . , 2009 ) . The parameters of the contrast transfer function ( CTF ) were estimated using CTFFIND4 ( Mindell and Grigorieff , 2003 ) . All particles were picked in a reference-free manner using DoG Picker ( Voss et al . , 2009 ) . The HA–I53_dn5 particle stack from the micrographs collected was pre-processed in Relion ( RRID:SCR_015701 ) . Reference-free two-dimensional ( 2D ) classification with cryoSPARC was used to select a subset of particles , which were used to generate an initial model using the Ab-Initio reconstruction function in CryoSPARC . The particles from the best class were used for non-uniform refinement in CryoSPARC to obtain the final 3D reconstruction . The sample was diluted with a buffer containing 10 mM HEPES pH 7 . 0 and 150 mM NaCl to a concentration of 0 . 025 mg/ml and adsorbed for 15 s to a glow-discharged carbon-coated copper grid . The grid was washed with the same buffer and stained with 0 . 7% uranyl formate . Images were collected at a nominal magnification of 57 , 000 × using EPU software on a ThermoFisher Talos F200C electron microscope equipped with a 4k × 4 k Ceta camera and operated at 200 kV . The pixel size was 0 . 253 nm . Particles were picked automatically using in-house written software ( unpublished ) and extracted into 200 × 200 pixel boxes . Reference-free 2D classifications were performed using Relion 1 . 4 ( Zivanov et al . , 2018 ) and SPIDER ( Frank et al . , 1996 ) . Grids were prepared on Vitrobot mark IV ( Thermo Fisher Scientific ) . Temperature was set to 10°C , humidity at 100% , wait time at 10 s , while the blotting time was varied in the 4–7 s range with the blotting force at 0 . The concentrations of T33_dn10 , O43_dn18 , and I53_dn5 nanoparticle samples were 4 . 2 , 3 . 0 , and 1 . 9 mg/ml , respectively . n-Dodecyl β-D-maltoside ( DDM ) at a final concentration of 0 . 06 mM was used for sample freezing . Quantifoil R 2/1 holey carbon copper grids ( Cu 400 mesh ) were pre-treated with Ar/O2 plasma ( Solarus plasma cleaner , Gatan ) for 10 s prior to sample application . Concentrated nanoparticle samples were mixed with appropriate volumes of stock DDM solution and 3 µl applied onto the grid . Excess sample and buffer was blotted off and the grids were plunge-frozen into nitrogen-cooled liquid ethane . Cryo-grids were loaded into a Titan Krios ( FEI ) operating at 300 kV , equipped with the K2 direct electron detector camera ( Gatan ) . Exposure magnification of 29 , 000 was set with the resulting pixel size of 1 . 03 Å at the specimen plane . Total dose was set to ~ 50 e-/Å ( Snapper , 2018 ) with 250 ms frames . Nominal defocus range was set to −0 . 6 to −1 . 6 µm for all three nanoparticle samples . Automated data collection was performed using Leginon software ( Suloway et al . , 2005 ) . Data collection information for acquired datasets is shown in Figure 4—source data 1 . MotionCor2 ( Zheng et al . , 2017 ) was run to align and dose-weight the movie micrographs . GCTF v1 . 06 was applied to estimate the CTF parameters . Initial processing was performed in cryoSPARC 2 . 9 . 0 . Template-picked particles were extracted and subjected to 2D classification . Multiple rounds of heterogeneous refinement were performed to further clean-up particle stacks in three acquired datasets . Selected subsets of particles were then transferred to Relion 3 . 0 ( RRID:SCR_015701 ) ( Zivanov et al . , 2018 ) for further processing . Reference models were generated using Ab-Initio reconstruction in cryoSPARC v2 . 9 . 0 ( Punjani et al . , 2017 ) with the application of appropriate symmetry ( tetrahedral , octahedral , and icosahedral for T33_dn10 , O43_dn18 , and I53_dn5 , respectively ) . Several rounds of 3D classification and refinement were used to sort out a subpopulation of particles that went into the final 3D reconstructions . Tetrahedral , octahedral , and icosahedral symmetry restraints were applied for all 3D refinement and classification steps during the processing of T33_dn10 , O43_dn18 , and I53_dn5 datasets , respectively . Soft solvent mask around the nanoparticle core was introduced during the final 3D classification , refinement , and post-processing . The resolutions of the final reconstructed maps were 3 . 86 Å for T33_dn10 , 4 . 54 Å for O43_dn18 , and 5 . 35 Å for I53_dn5 . The resulting EM maps have been deposited to EMDB with IDs: 21172 ( T33_dn10 ) , 21173 ( O43_dn18 ) and 21174 ( I53_dn5 ) . A graphical summary of the data processing approach and relevant information for each dataset are displayed in Figure 4—figure supplement 1 . The antigenicity of BG505 SOSIP–T33_dn2A trimer and BG505 SOSIP–T33_dn2 nanoparticle was analyzed on a BIAcore 3000 instrument at 25 °C and with HBS-EP ( GE healthcare Life sciences ) as running buffer , as described ( Brouwer et al . , 2019 ) . Affinity-purified goat anti-human IgG Fc ( Bethyl Laboratories , Inc ) and goat anti-rabbit IgG Fc ( Abcam ) were amine-coupled to CM3 chips and the anti-HIV-1 Env human and rabbit mAbs were captured to an average density of 320 ± 1 . 5 RU ( s . e . m ) . BG505 SOSIP–T33_dn2A or BG505 SOSIP–T33_dn2 ( both v5 . 2 ( 7S ) without N241/N289 ) ( Torrents de la Peña et al . , 2017; Steichen et al . , 2016 ) was allowed to associate for 300 s and then dissociate for 600 s at a concentration of 10 nM assembled macromolecule ( trimer or nanoparticle ) . The low background binding in parallel flow cells with only anti-Fc was subtracted . The lack of binding of nanoparticles lacking Env was ascertained for each mAb . To illustrate how epitope accessibility affects the relative binding of the trimers and nanoparticles , we converted the signals , which are proportional to mass bound , to moles bound and calculated the ratio for nanoparticles/trimers . For this comparison historic data on icosahedral nanoparticles were included ( Brouwer et al . , 2019 ) . The number of moles binding to the immobilized IgG at the end of the association phase was calculated: n = R ⋅ m ⋅ AM where n is the number of moles of macromolecules , R the response at 300 s ( RU ) , m the mass bound per area and RU ( g/ ( mm [Snapper , 2018] RU ) ) , A the interactive area of the chip ( mm [Snapper , 2018] ) , and M the molar mass of the macromolecule ( g/mol ) . This analysis corrects for the greater mass ( and thereby greater signal ) for each bound nanoparticle such that the number of binding events by differing macromolecules can be directly compared . To produce biotin-labeled antibodies specific to the H1 HA head , mAb 5J8 ( Krause et al . , 2011 ) in PBS was mixed with a 20 × molar excess ( relative to complete antibodies ) of EZ-Link NHS-LC-Biotin ( Thermo Fisher Scientific ) and allowed to sit for 2 hr at 4°C , followed by two rounds of overnight dialysis against PBS at 4°C to remove excess biotinylation reagent . All biosensors were hydrated in assay buffer ( 25 mM Tris , 150 mM NaCl , 0 . 5% bovine serum albumin , 0 . 01% TWEEN 20 , pH 8 . 0 ) before use . Biotinylated 5J8 ( 20 μg/mL in assay buffer ) was immobilized on SA biosensors , then briefly dipped in assay buffer prior to exposure to designed H1 HA fusions ( 500 nM per asymmetric unit , in assay buffer ) . The biosensor was again dipped in assay buffer and then exposed to the stem-specific mAb CR6261 ( 20 μg/mL in assay buffer ) ( Throsby et al . , 2008 ) . Purified HA-displaying nanoparticles or trimers were applied to a Sephacryl S-500 column pre-equilibrated with 25 mM Tris , 2 M NaCl , 5% glycerol , pH 8 . 0 . Sizing profiles were recorded based on absorption at 280 nm wavelength light . ELISA was used to measure binding kinetics of DS-Cav1–I53_dn5 to RSV F-specific mAbs D25 , Motavizumab , and AM14 . D25 is a prefusion specific mAb that binds site Ø ( McLellan et al . , 2013b ) . Motavizumab binds site II of the pre and post-fusion conformations ( Cingoz , 2009 ) . AM14 is trimer-specific binding across protomers of the prefusion conformation ( Gilman et al . , 2015 ) . 96-well ELISA plates were coated with 2 μg/mL DS-Cav1 nanoparticles . Plates were incubated at 4°C overnight and blocked with PBS containing 5% skim milk at 37°C for 30 min . mAbs listed were serially diluted in fourfold steps , and then added to the plates and incubated at 37°C for 45 min . Horseradish peroxidase ( HRP ) -conjugated anti-human IgG ( Southern Biotech . , Birmingham , AL ) was added and incubated at 37°C for 30 min , followed by 3 , 3′ , 5′ , 5- Tetramethylbenzidine ( TMB; Sigma-Aldrich , St . Louis , MO ) HRP substrate , and yellow color that developed after the addition of 1 M H2SO4 was measured by absorbance at 450 nm .	Vaccines train the immune system to recognize a specific virus or bacterium so that the body can be better prepared against these harmful agents . To do so , many vaccines contain viral molecules called glycoproteins , which are specific to each type of virus . Glycoproteins that sit at the surface of the virus can act as ‘keys’ that recognize and unlock the cells of certain organisms , leading to viral infection . To ensure a stronger immune response , glycoproteins in vaccines are often arranged on a protein scaffold which can mimic the shape of the virus of interest and trigger a strong immune response . Many scaffolds , however , are currently made from natural proteins which cannot always display viral glycoproteins . Here , Ueda , Antanasijevic et al . developed a method that allows for the design of artificial proteins which can serve as scaffold for viral glycoproteins . This approach was tested using three viruses: influenza , HIV , and RSV – a virus responsible for bronchiolitis . The experiments showed that in each case , the relevant viral glycoproteins could attach themselves to the scaffold . These structures could then assemble themselves into vaccine particles with predicted geometrical shapes , which mimicked the virus and maximized the response from the immune system . Designing artificial scaffold for viral glycoproteins gives greater control over vaccine design , allowing scientists to manipulate the shape of vaccine particles and test the impact on the immune response . Ultimately , the approach developed by Ueda , Antanasijevic et al . could lead to vaccines that are more efficient and protective , including against viruses for which there is currently no suitable scaffold .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "immunology", "and", "inflammation" ]	2020	Tailored design of protein nanoparticle scaffolds for multivalent presentation of viral glycoprotein antigens
Decisions are accompanied by a degree of confidence that a selected option is correct . A sequential sampling framework explains the speed and accuracy of decisions and extends naturally to the confidence that the decision rendered is likely to be correct . However , discrepancies between confidence and accuracy suggest that confidence might be supported by mechanisms dissociated from the decision process . Here we show that this discrepancy can arise naturally because of simple processing delays . When participants were asked to report choice and confidence simultaneously , their confidence , reaction time and a perceptual decision about motion were explained by bounded evidence accumulation . However , we also observed revisions of the initial choice and/or confidence . These changes of mind were explained by a continuation of the mechanism that led to the initial choice . Our findings extend the sequential sampling framework to vacillation about confidence and invites caution in interpreting dissociations between confidence and accuracy . Many decisions benefit from the acquisition of multiple samples of evidence acquired sequentially in time . In that case , a decision maker must decide not only about the proposition in question but also about when to terminate deliberation . The ensuing tradeoff between speed and accuracy is explained by sequential sampling with optional stopping models in which evidence is accumulated to some stopping criterion or bound ( Link , 1975; Ratcliff and Rouder , 1998 ) . The mechanism receives experimental support from human psychophysics and neural recordings in monkeys and rats ( Gold and Shadlen , 2007; Brunton et al . , 2013; Shadlen and Kiani , 2013; Hanks et al . , 2015 ) . The same framework also explains the confidence that a decision is correct ( Kiani and Shadlen , 2009; Kiani et al . , 2014a ) . This is because the quantity that is accumulated , termed a decision variable ( DV ) , when combined with elapsed decision time , maps to the probability that a decision rendered on its value will be correct ( Kiani and Shadlen , 2009; Drugowitsch et al . , 2014 ) . The attribution of confidence is important for guiding subsequent decisions , learning from mistakes and exploring alternatives . Thus , when the decision maker terminates deliberation , the choice is accompanied by a degree of certainty ( i . e . , confidence ) , based on the same stream of evidence that supported that decision ( Fetsch et al . , 2014 ) . This last point remains controversial , however , for there are many instances when the confidence in a decision and the decision itself are dissociable . For example , human decision makers tend to overestimate their certainty about choices based on truly ambiguous evidence ( Fischoff et al . , 1982; Baranski and Petrusic , 1994; Erev et al . , 1994; Drugowitsch et al . , 2014; Kiani et al . , 2014a ) , and they can perform above chance level yet report they are guessing ( Kunimoto et al . , 2001 ) . These and other observations have led psychologists to suggest that confidence and choice may be guided by different sources of evidence ( Pleskac and Busemeyer , 2010; Zylberberg et al . , 2012; Moran et al . , 2015 ) , or that the evaluation of the same evidence differs fundamentally in the way that it affects choice and confidence ( Fleming and Dolan , 2012; Maniscalco and Lau , 2012; De Martino et al . , 2013; Ratcliff and Starns , 2013 ) . The latter distinction is captured by the notion of a 1st-order confidence that is based rationally on the evidence in support of the decision and a 2nd-order confidence that can depart from this evidence . As this distinction rests on a proper understanding of the mechanism that supports choice and confidence , it is possible that some of the observations taken as support for higher order explanations of confidence are simply accounted for by deficiencies of the theory of 1st order choices . Naturally , if a decision maker acquires additional information after committing to a choice , she might wish to revise a decision , or the confidence in that decision , or both . Such changes of mind occur occasionally , even when there appears to be no additional information available after the initial decision has been rendered . The sequential sampling framework ( e . g . , bounded evidence accumulation ) offers a natural account of this phenomenon , because the mechanism incorporates processing delays , which leave open the possibility that the brain might have access to additional evidence that did not influence the initial decision and which might instead influence a revision . Evidence for such a process was adduced to explain reversals of perceptual decisions in humans and monkeys ( Rabbitt , 1966; Rabbitt and Vyas , 1981; McPeek et al . , 2000; Caspi and Beutter , 2004; Van Zandt and Maldonado-Molina , 2004; Resulaj et al . , 2009; Burk et al . , 2014; Kiani et al . , 2014b; Moher and Song , 2014 ) . Here , we address the possibility that this same mechanism can account for a revision in the confidence a decision maker assigns to his or her choice . We hypothesized that such revisions might account for an apparent dissociation between degree of confidence and choice . We asked humans to decide about the net direction of motion in a dynamic random dot display , using a variety of difficulty levels . They simultaneously indicated both their choice and the confidence in that choice by moving a handle with their arm . We show that choice , confidence , and reaction time are explained by a sequential sampling mechanism operating on a common evidence stream . On a small fraction of trials , subjects changed their initial decision about the direction of motion and , more frequently , about their confidence . We show that confidence and choice are informed by the same evidence , both at the initial choice and on any subsequent revision . However , changes of confidence arising through post-decision processing can contribute to an apparent dissociation between confidence and decision . We wish to account for the effect of stimulus difficulty on the initial direction choice , the confidence level associated with the choice and the time taken to make the decision . Not surprisingly , stronger motion supported more accurate ( Figure 2a ) and faster decisions ( Figure 2b ) , that tended to be assigned the higher confidence ( Figure 2c ) . The interplay between confidence , accuracy and reaction time can be appreciated from the breakdown of the data within the panels of Figure 2 ( indicated by color ) . When subjects were more confident , they tended to be more accurate ( Figure 2a; P<0 . 0001 for all subjects ) and faster ( Figure 2b; P<0 . 0001 for all subjects ) . In addition , subjects reported high confidence more often on correct choices than on errors ( Figure 2c ) , and they tended to be more confident on those errors associated with stronger motion ( P<1e-5 for 3 subjects and P = 0 . 50 for S3 ) . 10 . 7554/eLife . 12192 . 004Figure 2 . Interplay between initial confidence , accuracy , and reaction time . ( a ) Proportion correct responses as a function of motion coherence split by high ( blue ) and low ( red ) confidence decisions . ( b ) Mean reaction time as a function of motion strength as in ( a ) . ( c ) Probability of a high confidence initial choice as a function of motion coherence , split by correct ( green ) and error ( magenta ) trials . Data are means and s . e . m . ; curves are model fits . Only data with 10 or more trials are plotted . For clarity some of the points have been jittered horizontally . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 00410 . 7554/eLife . 12192 . 005Figure 2—figure supplement 1 . Initial choice behavior for Subject 5 . Data and fits in the same format as Figure 2 . This subject responded with high confidence on over 95% of trials . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 00510 . 7554/eLife . 12192 . 006Figure 2—figure supplement 2 . Non-stationary behavior of Subject 4 . ( a ) Cumulative of number of trials with high confidence ( back line ) as a function of trial number shows a change in slope around trial 5000 , corresponding to an increase in the rate of responding with high confidence . Red line shows linear regression fit to the first 5000 trials . ( b ) Data and fits for trials of this subject’s data in the same format as Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 006 These regularities are consistent with a common mechanism of bounded evidence accumulation to support choice , reaction time and confidence ( Kiani et al . , 2014a ) . The process involves a race between a mechanism that accumulates evidence for right , and against left , and a process that integrates evidence for left , and against right ( Usher and McClelland , 2001; Mazurek et al . , 2003 ) . Since these processes obtain evidence originating in the same stimulus , they will tend to accumulate noisy evidence of opposite sign ( i . e . , anti-correlated ) . However , the neural noise associated with the accumulations need not be perfectly anticorrelated , so they are depicted separately ( Figure 3a ) . The first process to reach an upper bound determines the choice and decision time . 10 . 7554/eLife . 12192 . 007Figure 3 . Information flow diagram showing visual stimulus and neural events leading to an initial decision , response time , and a possible change of mind . ( a ) Competing accumulator model of the initial decision . Noisy sensory evidence from the random dot motion supports a race between a mechanism that accumulates evidence for right ( and against left ) and a mechanism that integrates evidence for left ( and against right ) . Samples of momentary evidence are drawn from two ( anti-correlated ) Gaussian distributions with opposite means , which depend on the direction and motion strength of the stimulus . In this case the motion is rightward; therefore , the momentary evidence for right has a positive mean , and the rightward accumulator has positive drift . The first accumulation to reach an upper decision bound determines the choice and decision time . In this case the decision is for a rightward response . ( b ) Evolution of the decision variables ( top ) and log-odds ( bottom ) in the task . For simplicity we plot both accumulations on the same graph . At the initial decision time , the state of both the winning and losing processes as well as decision time confer the log-odds that a decision rendered on the evidence is likely to be correct , what we term confidence or belief . The bottom plot shows the log-odds of a rightward choice being correct , calculated from the decision variable and time . We assume that subjects adopt a consistent criterion θ on “degree of belief” to decide in favor of high or low confidence . Note that the decision is terminated by the decision variable ( top ) , not the log-odds ( bottom ) . Although the motion stimulus is displayed up to the reaction time , the decision does not benefit from all of the information , owing to sensory and motor delays ( ts and tm , respectively ) . In the post-decision period , the accumulation therefore continues . Changes of confidence and/or decision are determined by which region the log-odds is in after processing for an additional time , tpip ≤ ts + tm . To incorporate energetics costs of changing a decision and having to reach mid-movement to a new target we allow the initial bounds to move from their initial levels ( δθ1-δθ3 ) . This example uses the parameter fits for Subject 1 and a 3 . 2% coherence with an initial high confidence correct rightward decision followed by a change of confidence to a rightward , low-confidence decision ( for an initial low-confidence example see Figure 3-figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 00710 . 7554/eLife . 12192 . 008Figure 3—figure supplement 1 . A second example of the evolution of decision variables ( top ) and log-odds ( bottom ) in the task . The example is in the same format as Figure 3b but for an initial low confidence , correct decision with a change to high confidence . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 008 The state of both the winning and losing processes as well as decision time ( Figure 3b shows the races on the same plot ) confer an expectation that a decision rendered on the evidence is likely to be correct , what we term confidence or belief . Since the winning process is at a fixed bound , the confidence is adequately summarized by the height of this bound , decision time , and the state of the losing accumulator . Thus the degree of belief is based on the balance of evidence ( i . e . , difference ) between winning and losing DVs , as well as the decision time . This model has been tested in neurophysiology and a previous report in humans ( Kiani et al . , 2014a ) which extends ideas from signal detection theory , race , and Bayesian models ( Vickers , 1979; Kepecs et al . , 2008; Maniscalco and Lau , 2012; Drugowitsch et al . , 2014 ) . Its main distinction is the recognition that the time to decision influences confidence ( Fetsch et al . , 2014; Kiani et al . , 2014a ) . For example , this explains why confidence associated with errors increases with stronger coherences: stronger motion induces faster decision times for correct and errors alike . For simplicity , we assumed that subjects adopt a consistent criterion on “degree of belief” ( log odds , Figure 3b ) to decide in favor of high or low confidence . The smooth curves in Figure 2 are fits of the model ( 4 parameters; see Materials and methods ) , which capture the main features of the subjects’ choices . Naturally , subjects differed in their sensitivity , speed/accuracy and confidence which was accounted for by the parameters of the fits ( Table 1 ) . They also differed in their criteria for categorizing degree of belief into high and low , which can be appreciated by the separation of choice functions by confidence report . To maximize the number of points , given the reward structure ( see Materials and methods ) , subjects should choose high confidence if they believe the probability of a correct choice is greater than two-thirds , which corresponds to a log-odds of 0 . 69 . Interestingly the inferred criteria for all the subjects were close to this optimal criterion , although all subjects were somewhat risk-averse ( probability thresholds of 0 . 71 , 0 . 78 , 0 . 71 and 0 . 75; thresholds in log-odd units shown in Table 1 ) . There are some noteworthy discrepancies between model and data ( e . g . , proportion of high confidence choices at 0% coherence for 3 of 4 subjects ) , but the model captures the trend in the confidence ratings on error trials , mentioned above ( 3 of 4 subjects; Figure 2c ) , even though the fits themselves are dominated by the more numerous correct choices . The model is vastly superior to two alternative formulations , which would explain confidence on balance of evidence or deliberation time but not both ( compared to the original model , log likelihood is reduced by 63–1180 and 2301–3749 across participants for these models , respectively; see Materials and methods ) . 10 . 7554/eLife . 12192 . 009Table 1 . Parameter fits for the initial decision parameters and post-initiation processing parameters . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 009Subject 1Subject 2Subject 3Subject 4Initial decision parametersB0 . 740 . 731 . 070 . 74κ13 . 6412 . 868 . 6919 . 50μtnd ( s ) 0 . 4610 . 4210 . 4090 . 427θ0 . 891 . 260 . 871 . 12Post-initiation processing parameterstpip ( s ) 0 . 3950 . 2350 . 3900 . 285δθ10 . 771 . 321 . 160 . 79δθ20 . 240 . 320 . 440 . 06δθ3-0 . 36-0 . 57-0 . 26-0 . 69 Importantly , the fits of the initial choices allow us to characterize the state of the DV—including decision time—leading to the subjects’ choice and the mapping of this DV to confidence . This is the starting point to address the possibility of revising the initial choice and/or confidence rating afterwards . The critical assumptions are ( i ) the same information was used to reach an initial choice about direction and confidence , and ( ii ) there is additional information available to the decision maker after committing to a choice and confidence category , despite the disappearance of the random dot motion upon response initiation . Both assumptions are supported by the model-free analyses depicted in Figure 4 . We calculated psychophysical kernels using the information from the stochastic motion displays to determine the time frame over which information in the stimulus affected both components of the choice . To do this we used only weak motion strengths and examined the residual motion energy after removing the means associated with motion direction and strength ( see Materials and methods ) . The traces in Figure 4a show the averages of these residuals grouped by choice ( Figure 4a , top ) and grouped by confidence rating ( Figure 4a , bottom ) . The similar time course of the confidence and choice kernels implies that the initial direction and confidence choices were supported by a common evidence stream . Moreover , by discounting the lag and smoothing introduced by the motion filter ( inset ) , it is apparent that subjects relied on information from the beginning of the display until ~400 ms before movement onset ( arrows; see Materials and methods and Figure 4—figure supplement 1 ) to guide both their direction choices and their confidence . These values are consistent with the estimates of non-decision time ( tnd ) obtained from the model fits to the initial choices and RTs ( Table 1 ) . We next evaluate our hypothesis that some of the additional ~400 ms of information , which did not inform the initial direction and confidence choices , affects changes of mind about direction , confidence or both . 10 . 7554/eLife . 12192 . 010Figure 4 . Influence of motion information on choice and confidence . ( a ) Stimulus information supporting initial choice and confidence coincide . Motion-energy residuals were obtained by applying a motion energy filter to the sequence of random dots presented on each trial , and subtracting the mean of all trials having the same coherence and direction of motion . Positive ( negative ) residuals indicate an excess of rightward ( leftward ) motion . In each panel , data are aligned to stimulus onset ( left ) and movement initiation ( right ) . Only motion coherences ≤6 . 4% are included in the analysis . Inset shows the impulse response of the motion filter to a two-stroke rightward motion “impulse” at t = 0 . The upper panel shows the average of the motion energy residuals for rightward ( blue ) and leftward ( red ) choices , irrespective of confidence level . Arrows indicate , for each subject , the time prior to the movement initiation when the motion energy fluctuations cease to affect choice . The estimates correct for the delays of the filter ( see Figure 4—figure supplement 1 ) . The lower panel shows the difference in motion energy residuals between high and low confidence , for each direction choice . Shading indicates s . e . m . ( b ) Influence of motion energy residuals on changes of mind about direction and confidence . When subjects changed their initial decision about direction ( top panel ) , motion information changed sign just before movement initiation . When confidence changed from high to low ( middle panel ) , residuals were positive or negative for the two direction choices , respectively , and attenuated or reversed sign just before movement initiation . In contrast , late information provided additional support for the initial choice when confidence changed from low to high ( bottom panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 01010 . 7554/eLife . 12192 . 011Figure 4—figure supplement 1 . Estimation of the non-decision times from the psychophysical kernels . Although some previous studies measured the non-decision time as the point in time at which motion fluctuations no longer exert a significant influence on the initial choice , this method is biased because estimates of latency become shorter if non-decision times are more variable across trials , or if more trials are included . We used an alternative approach , which involves fitting a function , f ( t ) , to the psychophysical kernels . ( a ) The shape of f ( t ) was derived assuming that the slow decay of the psychophysical kernels when aligned on movement onset is due to: ( i ) trial-to-trial variability in the non-decision time , and ( ii ) the smoothing introduced by the impulse response of the motion energy ( inset; same as in Figure 4a ) . Without these influences , the influence of motion fluctuations on choice would step to zero at a fixed latency ( μtnd ) before movement ( black solid line ) . Inter-trial variability in the non-decision time reduces the number of trials that contribute to the psychophysical kernel for times closer to movement onset . If this variability is assumed Gaussian , the step function is smoothed into a cumulative Gaussian ( g[t μtnd , σtnd]; dashed line ) . To fit the psychophysical kernels , we also need to consider the additional smoothing introduced by the motion filter , which we do by convolving g ( t ) with the impulse response of the motion filter , IR ( t ) , such thatf ( t ) =αg ( t\|μtnd , σtnd ) IR ( t ) , where is α is an arbitrary scaling parameter . The final fit , that is f ( t ) , is shown by the black line . To increase the statistical power , we combined the motion energy residuals from rightward and leftward choices , such that positive residuals indicate an excess of motion in the direction of the initial choice . The green shaded area represents s . e . m . for the average of the motion energy residuals , including trials from all subjects . We fit μtnd , σtnd and α to minimize the deviance between f ( t ) and the average of the motion energy residuals . The best-fitting parameters μtnd and σtnd are indicated in the panel . ( b ) Same analysis as in ( a ) , but conducted separately for each subject . Latencies are similar to those obtained by fitting a bounded accumulation model ( Table 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 011 After indicating their initial choices , subjects occasionally changed the direction of their hand movement to indicate a different direction or level of confidence or both ( Table 2 and Figure 5 ) . The frequency of these changes of mind varied between subjects , from 2 . 0 to 8 . 8% of trials . Changes of decision were more likely if the initial decision was an error than if it was correct ( Figure 5a ) . These corrections were more likely if the motion information was stronger . Of course , errors were less frequent with stronger motion and less frequent than correct responses ( Figure 2a ) . For three of the four subjects , changes of decision corrected an initial error more often than they spoiled an initially correct choice ( P<0 . 001 , P<0 . 001 , P = 0 . 084 , P<0 . 001 ) , consistent with previous reports ( Resulaj et al . , 2009; Burk et al . , 2014 ) . 10 . 7554/eLife . 12192 . 012Table 2 . Pattern of changes of mind for each subject . Total trials performed with percentage of trials for different types of changes of mind . The average additional points earned is the difference in the points earned on change of minds trials compared to those that would have been earned had the subject not changed their mind , divided by the total number of change of mind trials . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 012SubjectTotal trials%TrialsAverage additional points earned per trial with a changeΔconfidence onlyΔdecision onlyΔconfidence & ΔdecisionAll changes190224 . 972 . 571 . 308 . 831 . 12290232 . 781 . 310 . 404 . 490 . 90390181 . 380 . 490 . 172 . 030 . 72450004 . 262 . 401 . 628 . 280 . 9010 . 7554/eLife . 12192 . 013Figure 5 . Changes of confidence and decision . ( a ) Probability of changes of decision when the initial decision was an error ( dark red ) or correct ( light red ) as a function of motion strength . Circles show subject data ( mean ± s . e . m ) and curves are model fits . ( b ) Probability of changes of confidence when the initial decision was low confidence ( dark blue ) or high confidence ( light blue ) as a function of motion strength . ( c ) Proportion of trials with changes of confidence ( blue ) and changes of decision ( red ) as a function of motion strength . These predictions ( curves ) are evaluated only at the motion strengths that were presented to the subjects , because they were obtained by using the fits from ( a and b ) together with the proportion of actual initial choices ( error/correct and high/low confidence ) for each motion strength . Figure 5—figure supplement 1 shows the empirical and model fit proportion of trials corresponding to panels and ( b ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 01310 . 7554/eLife . 12192 . 014Figure 5—figure supplement 1 . Proportion of trials with a change of decision or confidence . In the main figure , we show the conditional probability that a subject would change her decision about direction or confidence , given the initial decision . Here , we show the empirical proportions . ( a ) ( a ) Proportion of trials with a change of decision from error to correct ( dark red ) or correct to error ( light red ) as a function of motion strength . ( b ) Proportion of of trials with a change of confidence from low to high ( dark blue ) and high to low ( light blue ) as a function of motion strength . Circles show subject data ( mean ± s . e . m ) . Model fits ( curves ) are evaluated only at the motion strengths that were presented , because they were obtained by using the fits from Figure 5a and b together with the observed proportion of initial choices ( error/correct and low/high confidence ) for each participant . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 014 The novel insights from the present study derive from the changes in the confidence ratings ( Figure 5b ) . For all subjects , changes of confidence were more frequent than changes of decision ( Table 2 and Figure 5c; p<1e-4 for all participants ) . Further , changes of confidence were more likely if the initial decision was low confidence than if it was high confidence ( Figure 5b ) . These changes to high confidence were more likely when the initial low confidence accompanied a stronger motion . Note that Figure 5a shows the conditional probability of changing the decision about direction , given the initial choice was correct or incorrect ( the actual proportions are shown in Figure 5—figure supplement 1a; and change of confidence shown in Figure 5—figure supplement 1b ) . Changes of mind were beneficial to the participants in that on changes of mind trials ( either decision , confidence or both ) the gain in points over not changing one’s mind ranged from 0 . 72–1 . 12 points across the subjects ( Table 2 ) . Changes of both decision and confidence were less common ( 0 . 4–1 . 6% ) , which may be partly due to an energy cost associated with crossing the workspace ( Burk et al . , 2014; Moher and Song , 2014 ) . Those double changes that did occur tended to move from high to low confidence ( 1 . 03 , 0 . 35 , 0 . 12 , 1 . 36% ) compared to trials from low to high ( 0 . 26 , 0 . 04 , 0 . 04 , 0 . 26%; p<0 . 0001 for three subjects and p<0 . 05 for S3 ) . We hypothesized that a change of confidence , like a change of decision , can be explained by continuation of the processing of visual information that arrived after the subject had committed to her initial choice ( Figure 3b ) . We first evaluated this hypothesis by elaborating the model-free analysis of motion energy ( Figure 4 ) , described above . This analysis implies that ~400 ms of stimulus information , which did not affect the initial decision , might be available to revise the initial decision about confidence and direction . Indeed , when subjects changed their confidence rating from low to high , the motion information in the post-decision period supported the initial choice ( Figure 4b , bottom ) , whereas changes from high to low confidence were associated with motion information in support of the direction opposite to the one chosen ( Figure 4b , middle ) , and this trend was amplified for changes of mind about direction ( Figure 4b , top ) . A simple extension of the bounded accumulation model explains the frequency of these changes as well as their dependency on features of the stimulus and the subject’s initial report . The model scheme is illustrated in Figure 3b . We assumed that evidence was accumulated past the point of the initial decision for a fixed amount of time ( a free parameter , tpip ) that is less than the non-decision time . The state of the belief at tpip determines the final confidence and choice ( see Materials and methods ) . As shown in Figure 3b , we allowed for the possibility that subjects might not apply the identical criteria for confidence and choice in the pre- and post-initiation epoch ( the δθ parameters; see Materials and methods ) . The settings of tpip and the δθ parameters capture aspects of the energetic costs , by suppressing changes that would occur with small fluctuations in the evidence or very late in the movement . The trace in Figure 3b illustrates an initial high-confidence , rightward choice . Evidence that arrived in the post-initiation period ( i . e . , too late to affect the initial choice ) tended to favor leftward , detracting enough from the belief to support a change of confidence but not enough to support a change of decision in favor of leftward . The example shows a resistance to change because the total evidence actually favors leftward , but the model asserts a change of decision bound that is somewhat below the neutral evidence level . The curves in Figure 5 are fits of the model , which capture the main features of the subjects’ revisions of both choice and confidence . Note that all parameters of these fits are fixed from the fits to the initial choice and confidence , except for the three δθ parameters and tpip ( fit values in Table 1 ) . The variations in the participants’ behavior are accounted for by the different settings of the model parameters , which are illustrated graphically in Figure 6 for an initial low and initial high confidence rightward decision . Although the relation between the precise values of the δθ parameters and behavior are hard to intuit , the trends are consistent . After an initial low confidence decision ( Figure 6 , top row ) , all subjects required more belief to change to a high confidence decision than would have been required for an initial high confidence decision ( δθ1>0 ) . Similar hysteresis is seen for initial high confidence decisions ( Figure 6 , bottom row ) . Moreover , all subjects required more evidence to change their initial decision about direction than a simple reversal in sign of the evidence ( δθ2>0 ) . Finally , all participants relaxed their belief criterion when they changed their direction decision while either maintaining or changing to high confidence ( δθ3<0 ) . Because most initial choices were high confidence , this strategy would reduce motor effort by avoiding changes of both direction and confidence , as this requires crossing the workspace with a complete reversal of the initial movement . 10 . 7554/eLife . 12192 . 015Figure 6 . Log-odds thresholds for initial decisions and changes of mind for each participant . For initial decisions , the confidence threshold ( ± θ for t<0 , marked by the boundaries between light and dark colors ) determines whether the decision has high or low confidence . The initial direction decision is consistent with the sign of the log-odds ( rightward for green and leftward for red ) . Rows show the thresholds for an initial low- ( top ) and high- ( bottom ) confidence rightward decision ( i . e . , initial decisions that end in the region indicated by the vertical black bar in the plot for Subject 1 ) . Thresholds in the post-initiation stage ( right of the initial decision line ) tend to “move away” from the initial log-odds threshold , consistent with resistance to change . For all participants , the opposite-choice , high-confidence threshold “moves towards” the initial decision , thereby reducing the width of the pink zone . Because initial confidence was more often high ( lower row ) , the narrow pink zone can be interpreted as resistance to a double change of mind about both direction and confidence . Note that the model parameterization requires the pink and red regions to be same for both initial high and low confidence decisions . The non-decision time ( tnd ) and time of post-initiation processing ( tpip ) are also shown for each participant ( labels in left top graph ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 015 We wish to emphasize that confidence and direction reports remain coupled via a common mechanism at the time of both the initial and final decision . The observation supports a parsimonious account of decision confidence which is somewhat at odds with recent literature on meta-cognition ( see Discussion ) . Evidence for a meta-cognitive process is adduced from a dissociation between the signal-to-noise properties of the evidence that would support a level of choice accuracy versus the level associated with a degree of confidence ( e . g . , meta-d′; see Fleming and Lau , 2014; Maniscalco and Lau , 2012 ) . The meta-d′ statistic is not well behaved in the sequential sampling framework , but the basic logic remains applicable ( see Materials and methods ) . In Figure 7 , we compare two odds ratios ( OR ) . Both express the relative probability of a high-confidence rating , given a correct or error choice . Along the abscissa , we base the OR on the initial confidence ratings , whereas on the ordinate , we use the final confidence ratings , thus accommodating the change of confidence . In essence , we are pretending that the subject indicated her choice and subsequently told us her confidence . Clearly , there is a systematic discrepancy between the OR pairs ( P<1e-4 , sign test ) . All but one of the points are above the main diagonal , and the difference in OR is statistically reliable for 13 of the 19 individual points ( 4 subjects × 5 non-zero motion strengths with at least 1 error; P<0 . 05 , bootstrap; see Materials and methods ) . This discrepancy might be regarded as a sign of 2nd-order confidence , whereas it is simply the result of a continuation of the 1st-order process that couples choice and confidence , via bounded evidence accumulation . 10 . 7554/eLife . 12192 . 016Figure 7 . The possibility of change of confidence introduces an apparent dissociation between accuracy and confidence . The odds ratio ( OR ) statistic captures the relative tendency to report high confidence on correct versus error trials . The scatter plot compares ORs calculated from the participants’ data using the initial and final confidence report . Both ORs use the correct/error designation for the initial direction decision . The ORs calculated from the initial confidence report establish a baseline: the breakdown of confidence associated with the information that explains the accuracy at each motion strength . The ORs calculated from the final confidence report are larger and might thus be mistakenly interpreted as support for a dissociation between determinants of choice and confidence . Symbol shapes correspond to the four subjects; symbol shading denotes motion strength . An odds ratio can only be calculated if there are error trials ( or the ratio is infinite ) . This necessitated exclusion of one of the points at the highest coherence level . DOI: http://dx . doi . org/10 . 7554/eLife . 12192 . 016 Amongst the three manifestations of choice behavior , confidence is perhaps the most important and the least understood—compared to speed and accuracy—because in the world outside the laboratory , we do not always receive immediate feedback about our choices . Often , all we know about the accuracy of a choice is a degree of confidence . The attribution of confidence is important for guiding subsequent decisions , learning from mistakes and exploring alternatives . The present study establishes that this assessment evolves in time and , like the choice itself , can undergo revision with additional evidence . The study of perceptual decision-making offers insight into the process because the stream of evidence can be controlled experimentally . This is especially so in the present study using random dot kinematograms because the temporal stream of evidence is effectively a sequence of independent , statistically stationary samples . In other types of decisions—and most perceptual ones as well—the samples of evidence are derived from internal evaluations and memory processes , which are less well characterized , but which also evolve as a function of time . When the timing of a choice is under the control of the decision maker , a common strategy is to terminate deliberation upon a sufficient level of evidence . The class of bounded sequential sampling models , including many variants ( Wald , 1947; Stone , 1960; Laming , 1968; Link , 1975; Good , 1979; Luce , 1986; Ratcliff and Rouder , 1998 ) explains the tradeoff between speed and accuracy , and many of the essential steps have correlates in neurophysiology ( Romo et al . , 2002; Heekeren et al . , 2004; Ploran et al . , 2007; Heitz and Schall , 2012; White et al . , 2012; Kelly and O'Connell , 2013; Hanks et al . , 2014; Hanks et al . , 2015 ) . The mechanism ( and models ) also expose a discrepancy between the information supplied to the decision maker and the information that is actually used to make the decision . Specifically , there are processing delays between the arrival of information from the environment and updating its representation in working memory as a decision variable bearing on a proposition , and there are delays between commitment to a decision and communicating this decision through the motor system . Even in relatively fast perceptual decisions , this so-called non-decision time typically exceeds 300 ms . In previous studies , we showed that the motor system receives a continuous updating of the state of the evidence leading to a choice ( Gold and Shadlen , 2000; Selen et al . , 2012 ) and can utilize the late arriving information to revise an initial choice by continuing the process of deliberation ( Resulaj et al . , 2009; Burk et al . , 2014 ) . The present study extends these observations to the metacognitive operation of assigning a degree of belief that the decision is correct . The finding is important for several reasons . First , it establishes that deliberation in the post-initiation epoch is as rich a process as the deliberation preceding initiation . This implies that termination has a kind of specificity; it marks the end of deliberation for purposes of response initiation , but it does not actually terminate all deliberation . We suspect that this is just one aspect of a more general property . For example , the same information can bear on a variety of potential actions and cognitive operations , which need not share identical speed-accuracy tradeoffs and thus deliberate for different amounts of time . Second , the finding explains a potential dissociation between the information that decision makers might use to guide confidence and choice . We explain the initial choice and confidence using the identical stream of information , as in a previous study ( Kiani et al . , 2014a ) . The model is by no means perfect ( e . g . , it overestimates confidence at the lowest motion strength ) , but it demonstrates , nonetheless , that on many trials , the final confidence judgments are clearly based on additional information that did not affect the initial choice . This assertion is supported by the analysis of motion energy leading to the confidence choices ( Figure 4 ) . Just a few hundred milliseconds of additional evidence led to changes in the confidence associated with the initial choice , ranging from 1 . 5–6 . 3% of trials across our subjects , and this was enough to induce a clear departure from the level of confidence that ought to be associated with the evidence available at the time of the initial choice . In other words , had we asked our subjects only for their confidence rating after they indicated their initial choice—without an opportunity to revise those choices—we would have detected a different relationship between confidence and accuracy ( Figure 7 ) . When confidence is assessed after the initial choice , as it is in most experiments , then it ought to come as no surprise that choice and confidence are explained by only partially overlapping sources of information . Importantly , however , this does not imply a dissociation between confidence and the decision ( Pleskac and Busemeyer , 2010 ) , as the latter can also be revised based on the same additional information . Of course , in many settings , it is possible that a decision maker might acquire new information after an initial choice , either from the environment or from memory ( e . g . , reconsidering the evidence , weights and costs ) , and this could lead to a divergence of choice and confidence . However , our result invites caution when interpreting such divergence as indicators of metacognitive processes . Does the divergence necessarily implicate a process of a different nature , as implied by the distinction between 1st and 2nd order processes ( Fleming and Dolan , 2012 ) ? Or , is it possible that the mechanism responsible for assigning confidence based on the additional evidence is compatible with the one that supports a choice—perhaps a different choice were the decision maker given an opportunity to revise ( consistent with the original notion of type-1 and type-2 decisions; Clarke et al . , 1959; Galvin et al . , 2003 ) ? The evaluation of additional evidence may explain why confidence ratings are more strongly correlated with accuracy when they are reported with less time pressure ( Yu et al . , 2015 ) , and why some individuals appear to be better than others at discriminating correct from incorrect decisions ( e . g . Ais et al . , 2016 ) . The central question is whether this additional information affects confidence in a manner that is fundamentally different from the process that would tie these attributes together . In our study , the answer seems to be negative , and we speculate that the interesting aspects of more real-world distinctions involving re-evaluation of evidence using memory , for example , will require better understanding of the way that memory and decision processes interact but not a fundamentally different mechanism for associating confidence with the evidence arising from that process . We conclude that confidence , choice and reaction time can be understood in a common framework of bounded evidence accumulation . By definition , confidence may be regarded as metacognitive , simply because it is a report about the decision process itself . Yet the operations leading to confidence seem neither mysterious nor dissociable from the decision process ( cf . De Martino et al . , 2013 ) . That said , there are many unknown features of the underlying mechanism . We know little about the establishment of the mapping between belief and the representation of evidence and time . Nor do we know how a criterion is applied to this mapping to render the categorical choices in our study , or in post-decision wagering decisions ( Kiani and Shadlen , 2009 ) , or in confidence ratings ( Maniscalco and Lau , 2012; Fleming and Lau , 2014; Kiani et al . , 2014a ) . An appealing idea is that this also looks like a threshold crossing in the brain with dynamic costs associated with time ( e . g . , urgency Thura et al . , 2012; Drugowitsch et al . , 2012 ) and dynamic biases ( Hanks et al . , 2011 ) . Presumably , downstream structures that represent confidence-related values such as reward prediction error , must approximate the mapping between decision variable—represented in LIP and other brain areas—and elapsed time ( or number of samples ) to achieve this . Of course , downstream structures that control the arm must have access to the decision about both direction and confidence to control the initial reach and possibly revise the movement . To explain our findings , we assumed that choice and confidence are related but processed as if separate attributes , via a correspondence between decision variable and belief . It is intriguing to think that the two dimensions , which are bound together into a single action by the motor reach system in our task , could be dissociated in cognition and memory . Six naïve right-handed subjects , between the ages of 21 and 34 , participated in this study . The Cambridge Psychology Research Ethics Committee approved the experimental protocol , and subjects gave informed consent . Two of the subjects were excluded from the analyses based on poor task performance ( see below ) . Subjects were seated and used their right hand to hold the handle of a vBOT manipulandum that was free to move in the horizontal plane and allowed the recording of the position of the handle at 1000 Hz ( Howard et al . , 2009 ) . Subjects were prevented from seeing their arm by a horizontal mirror that was used to overlay virtual images of a downward facing CRT video display , mounted above the mirror , into the plane of the movement . A headrest ensured a viewing distance of around 40 cm . Subjects discriminated the direction of motion in a dynamic random-dot motion stimulus ( Roitman and Shadlen , 2002 ) presented within an aperture subtending 5 degrees of visual angle . The dots were displayed for one frame ( 13 . 3 ms , 75 Hz refresh ) and then three frames later a subset of these dots was displaced in the direction of motion while the rest of the dots were displaced randomly . Thus the positions of the dots in frame four , say , could only be correlated with dots in frames one and/or seven but not with dots in frames two , three , five and six . The dot density was 12 . 5 dots deg-2s-1 and displacements were consistent with a motion speed of 5 deg/s . The difficulty of the task was manipulated through the coherence of the stimulus , defined as the probability that each dot would be displaced as opposed to randomly replaced . Figure 1a show a schematic of the experimental setup . A trial began when the subject’s hand ( displayed to the subject as a 0 . 5 cm radius red circle ) was inside the home region , i . e . , within 1 cm of a grey cross approximately 30 cm in front of their body . After a random delay , sampled from a truncated exponential distribution ( range , 0 . 5–2 . 0 s; mean , 0 . 82 s ) , a dynamic random-dot stimulus appeared at the home position . On each trial , stimulus coherence was selected randomly from the set ± 0 , ± 3 . 2 , ± 6 . 4 , ± 12 . 8 , ± 25 . 6 , and ± 51 . 2% , where negative coherences correspond to leftward motion and positive coherences to rightward motion . The sign on the 0% coherence is arbitrary but determined which direction would be rewarded ( see below ) . Four circular choice targets with a radius of 1 . 5 cm were displayed at the corners of a 17 x 17 cm square centered on the home position . The two choice targets on the left corresponded to a leftward motion decision and the two on the right to a rightward motion decision . To encourage participants to also report the confidence in their decision , the two choice targets for each motion direction decision had different payoffs for correct and incorrect choices . One target was low-risk with a reward of 1 point for a correct choice and a loss of 1 point for an incorrect choice . The other target was high-risk with 2 points for a correct choice and a loss of 3 points for an incorrect choice . The designation correct/incorrect was assigned randomly on 0% coherence trials . Subjects judged the direction of the moving random dots and reached to a choice target when ready . We encouraged them to make quick decisions without sacrificing accuracy . They were free to interpret this instruction as they wished; they received no verbal instruction to aim for any particular speed/accuracy regime . Critically , when the movement was initiated—that is , the hand was more than 1 cm from the central cross—the random-dot stimulus was extinguished . The trial ended when the subject reached one of the four choice targets . The time course of a trial is shown in Figure 1b . If the movement had not been initiated within 3 s after stimulus onset , an error message appeared ( “Too Slow” ) and the trial would be repeated later in the session . After each trial , auditory feedback was given with a pleasant chime or a low-pitched tone corresponding to a correct and incorrect choice , respectively , and the number of points earned or lost was displayed on the screen . Subjects were instructed to maximize the number of points per trial . To encourage this , a running sum of the points was displayed at the top of the display in a bar graph . Each experimental session consisted of four blocks of 180 or 192 trials each ( 15 or 16 trials of the 12 coherences ) . We generated 48 stimuli with a rightward motion direction ( 8 for each of the 6 different coherence values ) . The leftward stimuli were generated by using the same dot locations but horizontally mirrored about the center of the aperture . This ensured that across the stimuli there was no left-right bias due to the motion energy of the stimuli . In most experiments , we used this “double-pass” procedure so that these 96 stimuli were displayed twice , in a random order . In a given session , the vertical orientation of the targets changed from block to block; in half of the blocks ( A ) , the two high-risk targets were at the top of the display , while in the other half ( B ) , the high-risk targets were at the bottom ( Figure 1a ) . A session consisted of four blocks ( 768 trials ) ordered ABBA or BAAB , and this alternated from session to session . Each subject took part in 12 experimental sessions ( 9024–9216 trials ) . All subjects received extensive training on the motion task , beginning with variable duration viewing , controlled by the computer and choice-reaction-time testing without confidence categories . Subjects passed to the main experiment when choice and reaction time functions were stable . We required subjects to have sufficient perceptual skills and motivation to perform the task . One subject was excluded based on poor discrimination performance: at the end of the first training session , this subject still performed at chance level on all coherences . A second subject was excluded because s/he responded with high confidence on 95% of trials ( and with 90% high confidence even at 0% coherence ) . After 5 sessions ( 3330 trials ) we decided to replace this subject . We found that despite the idiosyncrasy in this subject’s data , our model can fit their initial choices ( Figure 2—figure supplement 1 ) . Subject 4 showed a qualitative change in behavior on the second half of their data ( after ~5000 trials; Figure 2—figure supplement 2a ) , reporting high confidence on nearly all trials . As no stationary model can account for such nonstationary behavior we included only the first 5000 trials in our analysis . However , our model could still fit the initial choices for the omitted second half of this participant's data ( Figure 2—figure supplement 2b ) . We excluded from analysis any trials with a reaction time less than 150 ms ( 9 trials ) . For each trial , the final decision was determined by the choice target reached . To determine whether a change of decision had taken place we calculated the area between the hand’s path over the first 1 cm of movement and the vertical line through the hand’s initial starting location ( i . e . the line separating the left and right choice targets ) . A change of decision was reflected in the area on the side opposite to the final choice being greater than 0 . 1 cm2 ( Resulaj et al . , 2009 ) . The same procedure was applied to determine changes of confidence with the area now calculated relative to the horizontal line separating the high and low confidence choice targets . On each trial , we thus obtain the initial and final decision: choice ( left/right ) , confidence ( low/high ) , and reaction time ( time between stimulus onset and movement initiation ) . We show combined data for the two target configurations ( high/low confidence targets at top/bottom ) , having reassured ourselves that the arrangement had no detectable effects on choice accuracy ( p≥0 . 4; Fisher’s exact test ) and only small effects on reaction time ( RT ) for 2 subjects ( magnitudes<3% , p<0 . 01; t-test ) . There was a subtle bias for the bottom targets ( 50 . 6 , 51 . 9 , 52 . 5 and 51 . 4% , respectively ) , possibly due to kinematic factors , but recall that the orientation was balanced across the experiment . To examine effects of motion strength on confidence , we fit a logistic model to the probability of reporting high confidence , as a function of absolute value of motion coherence ( C ) Phigh=[1+exp ( −b0−b1\|C\| ) ]−1 where and bi are fitted coefficients . To examine whether confidence judgments were associated with more accurate choices we fit a logistic model to the direction choice data for each subject where the probability of choosing right is given by Pright = 1+exp ( -b0-b1C-b2I-b3IC-1 where C is the signed motion strength , I is an indicator variable ( zero for a low confidence choice and one for a high confidence choice ) . To test for improved sensitivity ( accuracy ) with high confidence , we evaluated the null hypothesis ( H0: b3 ≤ 0 ) . To examine whether confidence judgments were associated with different reaction times we analyzed each subject's reaction time as an ANOVA with categorical factors of unsigned coherence and confidence . All comparisons of event frequency ( e . g . changes of mind ) were performed with the Fisher exact test . For the model-free analyses of the time course of motion information on choice and confidence ( Figure 4 ) , we derived choice and/or confidence conditioned averages of stimulus motion energy ( psychophysical kernels ) . Due to the stochastic nature of the motion stimuli , the strength of motion will vary from trial to trial , and even within a trial . To quantify the fluctuations of motion along the horizontal axis , we convolved the sequence of random dots shown on each trial with a pair of spatiotemporal oriented filters , selective for rightward and leftward motion . The filters were matched to the speed and displacement of the coherently moving dots ( see details in Adelson and Bergen , 1985; Kiani et al . , 2008 ) . The results of the convolution were summed across space to yield the motion energy for each direction and as a function of time . The net motion energy was obtained by subtracting leftward from rightward motion energy . To average data across trials , we removed the average motion energy associated with each trial’s coherence and direction of motion . Because fluctuations have a stronger impact when motion is weak , only the lowest motion strengths ( ≤6 . 4% coh ) were included in these analyses . The influence of motion fluctuations on choice and confidence becomes negligible a few hundred milliseconds before movement onset ( Figure 4a ) . To obtain empirical estimates of non-decision time , we fit a function to the psychophysical kernel . The shape of the function f ( t ) was derived assuming that the psychophysical kernels decay slowly when aligned on movement onset ( Figure 4a , right ) because of: ( i ) the trial-to-trial variability in the non-decision time ( assumed Gaussian ) , which gradually reduces the number of trials that contribute to the psychophysical kernel , and ( ii ) the additional smoothing introduced by the impulse response of the motion energy ( inset of Figure 4a ) . With these assumptions , f ( t ) becomes: f ( t ) = α ( 1-Φ ( t , μtnd , σtnd ) ) IR ( t ) where α is a scaling parameter , Φ represents a cumulative Gaussian distribution with parameters μtnd and σtnd , IR ( t ) is the impulse response of the motion filter , and * indicates convolution . The curve-fitting procedure entails fitting μtnd , σtnd and α to match f ( t ) to the psychophysical kernels ( least-square fit ) . Figure 4—figure supplement 1 illustrates the fitting procedure and shows best fitting parameters for each subject . Although we do not make use of meta-d′ in this paper , we refer to the concept and therefore provide a brief definition . In signal detection theory ( SDT ) , d′ refers to the difference between the means of two standard Normal distributions , X1~N ( μ1 , 1 ) and X2~N ( μ2 , 1 ) , where Xn is a random variable and N ( μ1 , σ ) is a Normal distribution with mean μ1 and standard deviation σ . The two distributions might represent signal-plus-noise and noise-alone or the distributions of firing rates of neurons tuned to opposite directions of motion ( e . g . , Britten et al . , 1992 ) . For binary classification we can conceive of a single distribution of the difference between samples of X1 and X2 , such that X∆~N ( d′ , √2 ) , assuming independence . X∆ thus represents a DV whose sign identifies the more likely alternative . There is a one-to-one correspondence between proportion correct and d′ . Applied to binary classification tasks , meta-d′ is the value of d′ that would support the proportions of high-confidence ratings on error and correct choices , respectively , based on the assumption that the confidence designation is based on comparison of X∆ to some arbitrary but fixed criterion . Thus , within the SDT framework , if meta-d′≠d′ then it is not possible to account for the confidence ratings using the same signal:noise relationship that supports the choice accuracy , and this discrepancy has been interpreted as a sign of meta-cognitive confidence ( Maniscalco and Lau , 2012; Fleming and Lau , 2014 ) . However , the SDT framework must be extended to account for RT and choice . Under sequential sampling ( e . g . , bounded drift diffusion ) , meta-d′ ≠ d′ , even when the choice and confidence are explained by the same decision variable . Hence we pursued a more empirical approach . To examine the extent to which initial and final confidence are related to the correctness of the initial choice ( Figure 7 ) , we first calculated the odds of a high confidence rating , for correct and incorrect initial decision . For example , odds ( high confidence \| correct ) = P ( high confidence \| correct ) /P ( low confidence \| correct ) . From this we calculated the odds ratio OR = odds ( high confidence \| correct ) /odds ( high confidence \| error ) which indicates whether an event is more likely to occur in the first condition ( i . e . more often for correct choices , hence OR >1 ) or second condition ( i . e . more often for errors , OR <1 ) . We compared the ORs for the initial and final confidence , both with regard to the initial choice . The ORs calculated from the initial confidence report establish a baseline: the breakdown of confidence associated with the information that explains the accuracy at each motion strength , analogous to meta-d′ equal to d′ in the SDT framework described above . A larger OR for the final vs . initial confidence would indicate that a final high confidence response became more probable for correct choices compared to incorrect initial choices . We used a bootstrap ( N = 1000 ) to evaluate the reliability of the inequalities in ORs , depicted in Figure 7 . We fit the initial decision data with a model in which the decision process is a race between two accumulators ( Vickers , 1979; Usher and McClelland , 2001; Gold and Shadlen , 2007; Churchland et al . , 2008 ) , one that accrues momentary evidence for right ( and against left; R-L ) and another that accrues evidence for left ( and against right; L-R ) . Momentary evidence was modeled as draws from a bivariate Gaussian with a mean that depends on the coherence ( C ) such that drift rate ( /s ) is κC for the rightward and -κC for the leftward accumulator , giving a mean of ( κC , -κC ) . The bivariate Gaussian has a negative covariance ρ which determines the extent to which the two accumulators share noise ( for example arising from fluctuations in the stimulus ) . A decision is made when one of the two races crosses a decision bound B . The reaction time is determined by the time to reach the decision bound and an additional non-decision time ( e . g . , due to sensory and motor latencies ) which was modeled as a normal distribution with mean μtnd and standard deviation σtnd . The state of both the winning and losing race together with decision time map directly to the log-odds of being correct ( see Kiani et al . , 2014a ) . To model confidence , we included a log-odds threshold θ which separated high from low confidence judgments . For simplicity , we chose a time-invariant threshold . We have also fit data using time-dependent confidence bounds , which improve the fits for all subjects but does not affect the figures visibly and does not affect our conclusions . To fit the accuracy , confidence , and reaction time of the initial choices , we determined the number of trials for the 4 possible initial choices ( correct/high-confidence , correct/low-confidence , error/high-confidence & error/low-confidence ) for each ( unsigned ) coherence , as well as the corresponding mean reaction times . For any setting of the parameters , the model predicts the probability of each of the 4 possible initial choices and the mean reaction times for each coherence level . To fit model parameters we minimized the negative log likelihood ( i . e . , cost ) , using a multinomial distribution for the 4 choice types and Gaussian distribution for the sample mean RTs . For analytic simplicity ( see below ) , we used a flat bound ( i . e . , stationary rather than collapsing ) , which does not capture the shape of the RT distributions and the mean RT on error trials ( Drugowitsch et al . , 2012 ) . Therefore we used only sample mean RT and its associated s . e . m . for correct trials ( and all 0% coherence trials ) in the cost function . To further reduce the degrees of freedom of the model to four we fixed both the covariance ρ to -0 . 5 ≈ -0 . 71 and non-decision time standard deviation σtnd to 60 ms ( within the normal range from a previous study ( Burk et al . , 2014 ) and supported by an analysis of motion energy; see Figure 1—figure supplement 1 ) . This choice of covariance assumes that about half the noise is shared between the two accumulators ( i . e . arises from the stimulus ) and is within the normal range of fitted covariances ( Kiani et al . , 2014a ) . Importantly , this choice of covariance also allowed us to use an analytic solution to the race model when fitting . That is , we generalized the method of images used in ( Moreno-Bote , 2010 ) which provides analytic solution for covariances of 0 and -0 . 5 ( requiring 3 and 5 images respectively ) . Increasing the number of images leads to a variety of possible covariances ( but not to arbitrary covariances ) . We chose a covariance of -0 . 5 which requires the use of only 7 images thereby allowing efficient fitting of the subjects’ data . Using this analytic method precluded the use of collapsing decision bounds ( and no lower reflecting bounds; cf . , Kiani et al . , 2014a ) . The fitting was performed for each participant using multiple ( 30 ) runs of Matlab’s fminsearchbnd with a wide range of different initial parameters settings . The variability across runs was minimal , suggesting that the optimization procedure converged to global maximum . To model changes of mind , we assumed that once the initial decision had been made , the accumulators continued to integrate information that was not accessible at the time of the initial decision ( due to latencies in the sensory and motor system ) for a further post-initiation period tpip ( constrained to be less than tnd ) . A change in confidence or choice would occur if at the end of this period the log odds crossed a confidence or choice threshold , respectively ( Figure 3b ) . Since there are motor costs involved in changes of mind , we included additional parameters that could move the confidence and choice thresholds away from their initial values in the post-initiation stage by δθ1 , δθ2 and δθ3 ( Figure 3b ) . The three thresholds split the final decision into four zones and the specification of these thresholds depended on whether the initial choice was low or high confidence . For an initial high-confidence choice , the three thresholds were θ–δθ1 , -δθ2 , and –θ–δθ3 , respectively ( see Figure 3b ) . For an initial low-confidence decision , the first threshold was instead θ+δθ1 . To fit the model to each subject’s data , for each unsigned coherence we calculated the number of trials corresponding to the 16 possible events that could occur over the trial ( 4 possible initial choices and 4 possible final choices ) and again used maximum likelihood to fit the four free parameters carrying over κ and θ from the initial decision fits . Note that basing the change of decision about direction on log odds represents a simplification of a process similar to the one for the initial decision , for example , one that would operate on the evolving DVs represented by the competing accumulators . In principle , such a mechanism could terminate post-initiation processing more flexibly , but the frequency of change of mind about choice and confidence is too small to evaluate this possibility . We also evaluated two alternative models for initial choices only , which differ in how they assign high vs . low confidence . Model 1 exploits a classic idea from signal-detection-theory , and assigns confidence based on a threshold on the balance of evidence ( ignoring deliberation time ) ( Vickers , 1979; Kepecs et al . , 2008; Kepecs and Mainen , 2012; Wei and Wang , 2015 ) . Model 2 , assigns confidence based on threshold on decision time only , thus ignoring the balance of evidence . We fit both models to our participants’ data . They have the same number of parameters ( 4 ) as the original model . A 4-choice model for the initial choice-confidence decision ( based on dynamic programming ) is worthy of consideration but we have yet to find a satisfactory account of our data with this approach .	To understand how the brain makes decisions is to understand how we think – how we deal with information , interpret it and agree with a particular interpretation of the information . Neuroscience has begun to uncover the mechanisms that underlie these processes by linking the activity of nerve cells in the brain to different aspects of making decisions . These include how long it takes to reach a decision , why we make errors and how confident we feel about a decision . Sometimes when we make a decision and have committed to an answer , we then change our minds . Now , van den Berg et al . have asked whether the brain mechanisms that support a change of mind also support a change in confidence . To investigate this problem , human volunteers were asked to perform a difficult task where they had to decide whether a field of randomly moving dots had a tendency to drift to the left or to the right . During the experiment , van den Berg et al . recorded how long the volunteers took to make their decision , how confident the volunteers felt about their choice , and whether they were correct . Analyzing this data revealed that all of these measures could be explained by a mechanism where the brain accumulates evidence only until there appears to be enough evidence to favor one choice over the other . This process specifies how confident an individual should be based on the quality of the sensory evidence and how long it takes to make a decision . In addition , van den Berg et al . found that occasionally a volunteer changed their mind about how confident they were about a decision after they’d made it , as if they had continued to think about it . This was despite the volunteers receiving no more information about the task or how well they had done once they had made their decision . Therefore , it appears that the brain processed additional information that had already been detected but did not have time to affect the initial choice . The activity of the nerve cells in the brain was not recorded as the volunteers made their decisions . Future experiments that incorporate these measurements could help reveal how the brain performs the necessary computations and account for the time delay seen in processing some of the data . Where is this delayed information processed in the brain , and how does it lead to a change of mind ?	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	A common mechanism underlies changes of mind about decisions and confidence
Thousands of long noncoding RNAs ( lncRNAs ) have been discovered , yet the function of the vast majority remains unclear . Here , we show that a p53-regulated lncRNA which we named PINCR ( p53-induced noncoding RNA ) , is induced ~100-fold after DNA damage and exerts a prosurvival function in human colorectal cancer cells ( CRC ) in vitro and tumor growth in vivo . Targeted deletion of PINCR in CRC cells significantly impaired G1 arrest and induced hypersensitivity to chemotherapeutic drugs . PINCR regulates the induction of a subset of p53 targets involved in G1 arrest and apoptosis , including BTG2 , RRM2B and GPX1 . Using a novel RNA pulldown approach that utilized endogenous S1-tagged PINCR , we show that PINCR associates with the enhancer region of these genes by binding to RNA-binding protein Matrin 3 that , in turn , associates with p53 . Our findings uncover a critical prosurvival function of a p53/PINCR/Matrin 3 axis in response to DNA damage in CRC cells . The tumor suppressor p53 functions as a sequence-specific master regulatory transcription factor that controls the expression of hundreds of genes ( Riley et al . , 2008; Vogelstein et al . , 2000 ) and is mutated at a high frequency in human cancer types ( Oren , 1992; Vogelstein et al . , 2000; Vousden and Lane , 2007 ) . Although p53 exerts its tumor suppressor effects by regulating a wide variety of cellular processes , it has context-dependent functions ( Aylon and Oren , 2016; Vousden , 2000; Zilfou and Lowe , 2009 ) that are determined by various factors including cell-type , genetic background of the cell , extracellular environment , and the nature and duration of stress . Depending on the cellular context , p53 can have opposite effects on cell survival , cell migration , differentiation and metabolism ( Aylon and Oren , 2016; Kruiswijk et al . , 2015; Zilfou and Lowe , 2009 ) . Consistent with these pleiotropic effects of p53 , the expression of genes that have opposing effects on the above-mentioned processes are regulated by p53 ( Aylon and Oren , 2016; Riley et al . , 2008 ) . For example , in the context of DNA damage , p53 induces the expression of prosurvival genes such as CDKN1A ( p21 ) , 14-3-3σ and BTG2 ( Chan et al . , 1999; Polyak et al . , 1996; Rouault et al . , 1996 ) that cause cell cycle arrest , as well as proapoptotic genes such as PUMA , BAX and NOXA ( Riley et al . , 2008 ) that cause cell death . Interestingly , these prosurvival and proapoptotic genes are all upregulated by p53 in a cell regardless of the effect of p53 on cellular outcome . Therefore , it is important to investigate the function of a p53 target gene in the appropriate cellular context . While the protein-coding genes regulated by p53 have been extensively studied and we and others have identified critical roles of microRNAs ( miRNAs ) in the p53 pathway ( Chang et al . , 2007; Hermeking , 2012; Lal et al . , 2011; Raver-Shapira et al . , 2007 ) , the function of the newly discovered long noncoding RNAs ( lncRNAs ) in p53 signaling remains largely unknown . LncRNAs are transcripts > 200 nucleotides ( nt ) long that lack a functional open reading frame . Growing evidence suggests critical roles of lncRNAs in multiple cellular processes including differentiation , dosage compensation , genomic stability , metabolism , metastasis and DNA repair ( Arun et al . , 2016; Dey et al . , 2014; Fatica and Bozzoni , 2014; Lee , 2012; Lee et al . , 2016; Ling et al . , 2013; Mueller et al . , 2015; Redis et al . , 2016; Sharma et al . , 2015; Tripathi et al . , 2013 ) . Some p53-regulated lncRNAs including lincRNA-p21 , PANDA , PINT , LED , NEAT1 and DINO have been shown to function as downstream effectors of p53 ( Adriaens et al . , 2016; Blume et al . , 2015; Dimitrova et al . , 2014; Huarte et al . , 2010; Hung et al . , 2011; Léveillé et al . , 2015; Marín-Béjar et al . , 2013; Schmitt et al . , 2016 ) . However , the function and mode of action of most p53-regulated lncRNAs has yet to be elucidated . In this study , we focused on a previously uncharacterized lncRNA that we named PINCR ( p53-induced noncoding RNA ) . We show that during DNA damage , PINCR has a context-dependent function . RNA pulldowns from cells expressing endogenous PINCR fused to an S1-RNA aptamer show that PINCR binds to the RNA-binding protein Matrin 3 to regulate the induction of a subset of prosurvival p53 targets by associating with the enhancers of these genes via a Matrin 3-p53 complex . Our results identify PINCR as a lncRNA that functions as a context-dependent prosurvival gene in the p53 pathway . To identify lncRNAs regulated by p53 in multiple cell lines , we performed microarray analysis ( Affymetrix HT2 . 0 ) from three colorectal cancer ( CRC ) cell lines ( HCT116 , RKO and SW48 ) following activation of p53 with Nutlin-3 ( Figure 1—figure supplement 1A and Figure 1—figure supplement 1—source data 1 ) , a pharmacological inhibitor of MDM2 . Using a cut-off of 1 . 50-fold change , 66 transcripts were upregulated in all three lines ( Figure 1—figure supplement 1B , C and Supplementary file 1 ) . Forty-eight of the 66 transcripts were also identified in a recent p53 GRO-seq study in HCT116 cells ( Allen et al . , 2014 ) indicating that they may be direct p53 targets . The 66 transcripts included several known p53 targets including BTG2 , BAX , CDKN1A ( p21 ) , GADD45A , MDM2 and RRM2B . Four out of 66 transcripts were annotated lncRNAs ( Supplementary file 2 ) . Among the four lncRNAs , RP3-326I13 . 1 , a ~2 . 2 kb long spliced intergenic lncRNA with unknown function , transcribed from the X-chromosome , was strongly induced upon p53 activation ( Supplementary file 2 ) . We validated this result by quantitative reverse transcription PCR ( qRT–PCR ) after Nutlin-3 treatment ( Figure 1A ) . Due to this strong induction upon p53 activation , we named this lncRNA PINCR . Notably , although this lncRNA was also strongly and directly upregulated by p53 upon ectopic overexpression of p53 in a mutant p53-expressing CRC line ( Hünten et al . , 2015 ) , its function has not been elucidated . Therefore , we decided to investigate the role of PINCR in the p53 network . 10 . 7554/eLife . 23244 . 003Figure 1 . PINCR is a nuclear lncRNA directly induced by p53 after DNA damage . ( A ) qRT-PCR analysis from HCT116 , SW48 and RKO cells untreated or treated with Nutlin-3 for 8 hr . Error bars represent SD from two independent experiments . ( B ) qRT-PCR analysis for PINCR and the known p53 target PUMA from isogenic p53-WT and p53-KO HCT116 cells untreated or treated with DOXO for the indicated times . ( C ) Snapshot of p53 ChIP-seq data of the PINCR promoter from MCF7 and U2OS cells untreated or treated with Nutlin or 5-FU or RITA . ( D ) HCT116 cells were untreated or treated with DOXO for 16 hr and qPCR using primers spanning the p53RE of p21 and PINCR was performed from Input and p53-ChIP . ( E ) HCT116 cells were co-transfected for 48 hr with pGL3 or pGL3 containing the PINCR promoter , and pCB6 or pCB6-p53 . Luciferase assays were performed using pRL-TK as internal control . ( F ) Luciferase assays were performed from untreated ( CTL ) or DOXO-treated HCT116 cells co-transfected for 48 hr with the internal control pRL-TK and pGL3 containing the PINCR wild-type ( WT ) promoter or pGL3 containing the PINCR promoter in which the p53RE was deleted ( △p53RE ) . ( G ) Maximum CSF scores of PINCR as well as other coding and noncoding RNAs determined by analysis with PhyloCSF . ( H ) qRT-PCR analysis from nuclear and cytoplasmic fractions of DOXO-treated HCT116 cells; the cytoplasmic GAPDH mRNA and the nuclear lncRNA MALAT1 were used as controls . Error bars in B , D-F represent SD from three independent experiments . #p<0 . 01; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00310 . 7554/eLife . 23244 . 004Figure 1—figure supplement 1 . Identification of p53-regulated lncRNAs . ( A ) Isogenic p53-WT and p53-KO HCT116 cells were untreated or treated with Nutlin-3 for 8 hr and immunoblotting was performed for p53 and the loading control GAPDH . ( B ) Heat map is shown for the differentially expressed mRNAs and lncRNAs identified by microarrays performed in duplicate from HCT116 , SW48 and RKO cells untreated or treated with Nutlin-3 for 8 hr . Upregulated genes are shown in red and downregulated genes in green . PINCR ( RP3-326I13 . 1 ) is shown in the red box . ( C ) Venn diagram showing the overlap between the transcriptomes upregulated ≥1 . 5-fold after Nutlin-3 treatment of HCT116 , SW48 and RKO cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00410 . 7554/eLife . 23244 . 005Figure 1—figure supplement 1—source data 1 . p53 immunoblot for Figure 1—figure supplement 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00510 . 7554/eLife . 23244 . 006Figure 1—figure supplement 2 . PINCR is highly induced after DNA damage in SW48 cells . ( A ) qRT-PCR analysis of PINCR and the known p53 target PUMA from isogenic p53-WT and p53-KO SW48 cells untreated or treated with DOXO for the indicated times . ( B ) Pictorial representation of PINCR locus and the p53 response element ( p53RE ) located 118 bp upstream of the first exon . Error bars represent SD from three biological replicates . p<0 . 05; #p<0 . 01; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00610 . 7554/eLife . 23244 . 007Figure 1—figure supplement 3 . RNA-seq was performed in duplicate from HCT116 cells untreated ( CTL ) or treated with DOXO ( 300 nM ) for 16 hr ( Li et al . , unpublished ) . Snapshot of the RNA-seq data for the PINCR locus ( A ) , 5’end of PINCR ( B ) and 3’end of PINCR ( C ) is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00710 . 7554/eLife . 23244 . 008Figure 1—figure supplement 4 . RT-PCR analysis of full-length PINCR . RT-PCR for PINCR was performed using cDNA prepared from HCT116 DOXO-treated total RNA or pCB6-PINCR . The forward primer starts at the 5’end of the annotated PINCR RNA and the reverse primer is located near the 3’end of PINCR RNA . Location of the primers and conditions for PCR are shown in ( A ) . Picture of the agarose gel after electrophoresis is shown in ( B ) . pCB6-PINCR was used as positive control . RT minus refers to a control reaction in which the total RNA was used as template for PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00810 . 7554/eLife . 23244 . 009Figure 1—figure supplement 5 . PINCR molecules per HCT116 cell . The number of molecules of PINCR RNA per HCT116 cell was determined by two approaches . First ( A–C ) , using RNA-seq from HCT116 cells ( Li et al . , unpublished ) : The FPKM of PINCR from CTL and DOXO-treated HCT116 cells was compared to NORAD , a lncRNA known to be expressed at 500–1000 molecules/HCT116 cell ( Lee et al . , 2016 ) . Second ( D ) , by qRT-PCR from DOXO-treated HCT116 cells using in vitro transcribed PINCR RNA as standard . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 00910 . 7554/eLife . 23244 . 010Figure 1—figure supplement 6 . Conservation of PINCR . Schematic diagrams of PINCR ( RP3-326I12 . 1 ) genomic DNA alignments include multiple alignment of some mammalian species from the ‘Multiz Alignments of 46 Vertebrates’ track ( shown in green ) , measurements of evolutionary conservation by the PhastCons method ( Mammal Cons , Vertebrate Cons ) , and primate genomes ( net track , grey ) alignments . The net track shows the best human/other chain for PINCR genomic DNA . In the graphical display of primate genomes alignment , the boxes represent ungapped alignments and the line represent gaps . The diagram was downloaded and adapted from the UCSC Genome Browser ( https://genome . ucsc . edu ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01010 . 7554/eLife . 23244 . 011Figure 1—figure supplement 6—source data 1 . Multiple sequence alignment of mature PINCR transcript for Figure 1—figure supplement 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01110 . 7554/eLife . 23244 . 012Figure 1—figure supplement 6—source data 2 . Multiple sequence alignment of PINCR promoter for Figure 1—figure supplement 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 012 Given the well-established role of p53 after DNA damage , we next assessed changes in PINCR expression during DNA damage induced by Doxorubicin ( DOXO ) in isogenic p53 wild-type ( p53-WT ) and p53 knockout ( p53-KO ) HCT116 and SW48 cells . The final concentration of DOXO in this and all subsequent experiments was 300 nM , unless stated otherwise . The known p53 target PUMA ( Nakano and Vousden , 2001 ) was used as a positive control . Although PINCR was almost undetectable at the basal level , after DNA damage it was significantly induced as early as 8 hr after DOXO treatment and was induced >100 fold after 24 hr , in a p53-dependent manner in both lines ( Figure 1B and Figure 1—figure supplement 2A ) . To determine if PINCR is a direct target of endogenous p53 , we first utilized publicly available p53 ChIP-seq ( Chromatin immunoprecipitation sequencing ) data ( Menendez et al . , 2013; Nikulenkov et al . , 2012 ) . Upon p53 activation , we observed a single p53 ChIP-seq peak in a region ~118 bp upstream of the first exon of PINCR in MCF7 ( breast cancer ) and U2OS ( osteosarcoma ) cells ( Figure 1C and Figure 1—figure supplement 2B ) . We validated this result in HCT116 cells by ChIP-qPCR ( Figure 1D ) . We next inserted a ~2 kb region of the PINCR promoter into a promoterless luciferase reporter vector ( pGL3 ) and co-transfected this construct in HCT116 cells along with a mammalian expression vector ( pCB6 ) or pCB6 overexpressing p53 ( pCB6-p53 ) . We found that the PINCR promoter drives luciferase expression upon p53 overexpression ( Figure 1E ) . Deletion of the p53-response element ( p53RE ) in the PINCR promoter resulted in significant decrease in luciferase activity ( Figure 1F ) . These results suggest that PINCR is a direct target of p53 . A detailed subsequent analysis of PINCR revealed many features of this lncRNA: ( 1 ) PINCR is a noncoding RNA because its coding potential was comparable to the noncoding RNA NEAT1 ( Figure 1G ) ; ( 2 ) PINCR is highly enriched in the nucleus ( Figure 1H ) , similar to the nuclear-retained lncRNA MALAT1 ( Hutchinson et al . , 2007 ) ; ( 3 ) the 5’and 3’ends of PINCR matched the annotated transcript based on analysis of our RNA-seq data from HCT116 cells ( Li et al . , unpublished ) ( Figure 1—figure supplement 3 ) ; ( 4 ) analysis of the length of the PINCR transcript by RT-PCR revealed two closely migrating bands ( Figure 1—figure supplement 4 ) that matched the expected size of the amplicon ( ~1 . 8 kb ) ; ( 5 ) PINCR is expressed at ~13–26 molecules per HCT116 cell after DNA damage and less than one molecule per cell without DNA damage ( Figure 1—figure supplement 5A–C ) based on comparison of the FPKM of PINCR with the lncRNA NORAD , known to be expressed at 500–1000 molecules per HCT116 cell ( Lee et al . , 2016 ) . As an alternative approach , qRT-PCR using in vitro transcribed PINCR RNA showed that PINCR is expressed at ~27 molecules per HCT116 cell after DNA damage ( Figure 1—figure supplement 5D ) ; ( 6 ) PINCR promoter including the p53RE , mature PINCR transcript and the transcription start site are quite conserved among primates but poorly conserved between human and mouse ( Figure 1—figure supplement 6 , Figure 1—figure supplement 6—source data 1 and Figure 1—figure supplement 6—source data 2 ) . The strong p53-dependent induction of PINCR after DNA damage led us to hypothesize that PINCR mediates the effect of p53 by regulating G1 and/or G2/M arrest after DNA damage . To begin to test this hypothesis , we used the CRISPR/Cas9 technology to delete the PINCR genomic locus in HCT116 cells ( Figure 2—figure supplement 1A and B ) . Targeted deletion of PINCR in 2 PINCR-KO clones ( KO#1 and KO#2 ) was confirmed by Sanger sequencing ( Figure 2—figure supplement 1C and D ) and loss of PINCR expression was validated by qRT-PCR ( Figure 2A ) . As negative controls , we selected two clones that were WT for PINCR ( WT#1 and WT#2 ) . The p53RE in the PINCR promoter was partially deleted in PINCR-KO#1 but fully intact in PINCR-KO#2 ( Figure 2—figure supplement 1C and D ) and as expected , we observed significantly impaired p53 binding in PINCR-KO#1 but not in PINCR-KO#2 ( Figure 2—figure supplement 2 ) . 10 . 7554/eLife . 23244 . 013Figure 2 . Loss of PINCR impairs G1 arrest and results in increased cell death after DNA damage . ( A ) qRT-PCR analysis from PINCR-WT ( WT#1 and WT#2 ) and PINCR-KO clones ( KO#1 and KO#2 ) untreated or treated with DOXO for 16 hr . ( B , C ) PINCR-WT and PINCR-KO clones were untreated or treated with DOXO for the indicated time points and cell cycle analysis was performed using Propidium iodide ( PI ) staining followed by flow cytometry analysis ( FACS ) . ( D ) PINCR-KO cells were stably transfected with pCB6 or pCB6-PINCR and qRT-PCR was performed . ( E ) PINCR-KO cells stably expressing PINCR were untreated or treated with DOXO in biological duplicates at the indicated times and cell death ( sub-G1 cells ) was assessed by PI staining followed by FACS . ( F ) Immunostaining for Nucleoporin and cleaved caspase-3 from PINCR-WT and PINCR-KO clones with or without DOXO treatment for 72 hr . DNA was counterstained with DAPI . ( G ) PINCR-WT and PINCR-KO clones were untreated or treated with the indicated DOXO concentrations for 4 hr and colony formation assays were performed after 10 days . Error bars in A and D represent SD from three independent experiments . #p<0 . 01; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01310 . 7554/eLife . 23244 . 014Figure 2—figure supplement 1 . CRISPR deletion of PINCR locus . ( A ) Pictorial representation of PINCR genomic locus . Yellow boxes represent annotated exons and blue boxes represent regions targeted by gRNAs ( E5 . 1 and E3 . 1 ) . HEK293T cells were transfected with Cas9 alone ( Lanes 1 and 2 ) or Cas9 along with gRNAs E5 . 1 and E3 . 1 ( Lanes 3 and 4 ) . Deletion was confirmed by genomic PCR using PINCR-CRISPR analysis primers . ( B ) Schematic of generation of PINCR-KO HCT116 cells . ( C , D ) Partial genomic sequence of PINCR from Sanger sequencing and snap-shot of PINCR locus ( using BLAT ) for the PINCR-KO#1 and PINCR-KO#2 HCT116 clones is shown . The p53RE is underlined and in green font . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01410 . 7554/eLife . 23244 . 015Figure 2—figure supplement 2 . p53 binding to the p53RE of PINCR in PINCR-WT and PINCR-KO cells was assessed by ChIP-qPCR from HCT116 cells ( PINCR-WT or PINCR-KO ) treated with 5-FU for 24 hr . Fold enrichment relative to IgG is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01510 . 7554/eLife . 23244 . 016Figure 2—figure supplement 3 . Cell cycle profiles for Figure 2 . ( A ) Raw cell cycle profiles corresponding to the data in Figure 2B and C are shown . ( B ) Raw cell cycle profiles corresponding to a representative experiment for Figure 2E are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01610 . 7554/eLife . 23244 . 017Figure 2—figure supplement 4 . Quantitation for immunostaining and colony formation . ( A ) Quantitation for Figure 2F . Percentage cleaved Caspase-3-positive cells in PINCR-WT and PINCR-KO clones after 72 hr DOXO treatment is shown . ( B ) Quantitation for Figure 2G . Number of colonies in PINCR-WT and PINCR-KO clones untreated or treated for 4 hr with the indicated DOXO concentrations . Error bars represent SD from three experiments . p<0 . 05; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01710 . 7554/eLife . 23244 . 018Figure 2—figure supplement 5 . Loss of PINCR results in reduced G1 arrest after Nutlin-3 treatment . PINCR-WT and PINCR-KO clones were untreated or treated with Nutlin-3 for 24 hr and the cell cycle profiles were examined by PI-staining and FACS analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 018 We next treated the PINCR-WT and PINCR-KO cells with DOXO for 24 , 48 and 72 hr and examined the effect on cell cycle arrest . In PINCR-KO cells , G1 arrest was substantially impaired as early as 24 hr after DNA damage ( Figure 2B ) and these cells displayed increased apoptosis as measured by the elevated sub-G1 population after 48 and 72 hr of DOXO treatment but not at 24 hr ( Figure 2C and Figure 2—figure supplement 3A ) . Notably , loss of PINCR did not alter the cell cycle profile in the absence of DNA damage ( Figure 2—figure supplement 3A ) . To make sure that the observed phenotypes were not DNA-dependent but due to loss of PINCR RNA , we performed a rescue experiment . We inserted the full-length PINCR RNA into pCB6 and reintroduced PINCR in the PINCR-KO cells by stable transfection . The extent of PINCR overexpression was not supraphysiological; we observed ~40 fold increase in PINCR expression ( Figure 2D ) which is less than the ~100 fold induction that we had observed for endogenous PINCR . Although , reintroduction of PINCR in the PINCR-KO cells significantly rescued apoptosis at both 48 and 72 hr after DNA damage ( Figure 2E ) , we did not observe a rescue of G1 arrest ( Figure 2—figure supplement 3B ) . The incomplete rescue may be because unlike endogenous PINCR that is induced ~100 fold after DOXO-treatment , the extent of exogenous PINCR overexpression was ~40 fold . Another possibility is that in the rescue experiments , we overexpressed the annotated isoform , whereas we had found that HCT116 cells express at least two isoforms of PINCR . In response to DOXO treatment , HCT116 cells arrest in G1 but the majority arrest in G2 . p53 has been shown to play a critical role in the G1 arrest and in keeping the cells in G2 ( Bunz et al . , 1998; Kuerbitz et al . , 1992; Levine , 1997 ) . To determine if in addition to its role in G1 arrest , PINCR also regulates G2 arrest , we examined the integrity of the nuclear envelope by performing immunostaining for Nucleoporin after treating PINCR-WT and PINCR-KO cells with DOXO for 72 hr . We found that the nuclear membrane was intact in both PINCR-WT and PINCR-KO cells suggesting that loss of PINCR does not result in aberrant entry into mitosis ( Figure 2F ) . Immunostaining for cleaved caspase-3 , a marker of apoptosis , further confirmed increased apoptosis after DNA damage upon loss of PINCR ( Figure 2F and Figure 2—figure supplement 4A ) . This hypersensitivity to DNA damage was persistent and also observed in colony formation assays ( Figure 2G and Figure 2—figure supplement 4B ) . In this experiment , we did not observe a difference in clonogenicity upon loss of PINCR in untreated cells , which is consistent with the unaltered cell cycle profile upon loss of PINCR in untreated cells . To confirm that PINCR is involved in p53-dependent G1 arrest , we performed cell cycle analysis from PINCR-WT and PINCR-KO cells after Nutlin-3 treatment . As expected , in both PINCR-WT and PINCR-KO cells , Nutlin-3 treatment resulted in dramatic reduction in the population of cells in S-phase ( Figure 2—figure supplement 5 ) . As compared to PINCR-WT cells , we observed reduced G1 population and increased G2/M population in both PINCR-KO clones after Nutlin-3 treatment . These data indicate that PINCR plays a role in p53-dependent G1 arrest and it has a prosurvival function in response to DNA damage . If the major function of PINCR after DNA damage is to arrest cells in G1 , the effect of PINCR loss should be more pronounced if the DNA damaging agent mainly causes G1 arrest . We therefore examined the effect on G1 arrest and apoptosis 48 hr after treatment of PINCR-WT and PINCR-KO cells with three chemotherapeutic drugs: DOXO ( 300 nM ) , the radiomimetic NCS ( Neocarzinostatin , 400 ng/ml ) and 5-Fluorouracil ( 5-FU , 100 µM ) . After confirming the induction of PINCR in response to NCS and 5-FU treatment ( Figure 3—figure supplement 1A and B ) , we performed cell cycle analysis . In PINCR-WT cells , the percentage of cells arrested in G1 was smallest ( 11% ) for NCS and largest ( 63% ) for 5-FU ( Figure 3A ) . Loss of PINCR resulted in decreased G1 arrest for NCS , DOXO and 5-FU . However , in PINCR-KO cells , the sub-G1 population was highest ( 36% ) after 5-FU treatment indicating that hypersensitivity of PINCR-KO cells to chemotherapeutic drugs is dependent on G1 arrest . Importantly , this impaired G1 arrest and increased apoptosis after 5-FU treatment upon loss of PINCR was observed in both PINCR-KO clones ( Figure 3B and C ) and was further confirmed by immunoblotting for the apoptosis marker cleaved-PARP ( Figure 3D and Figure 3—source data 1 ) . Furthermore , loss of PINCR significantly impaired clonogenicity after 5-FU treatment ( Figure 3E and F ) . 10 . 7554/eLife . 23244 . 019Figure 3 . PINCR knockout cells are hypersensitive to 5-FU in vitro and poorly tumorigenic in vivo . ( A ) PINCR-WT and PINCR-KO clones were untreated or treated with NCS , DOXO , 5-FU for 48 hr and PI staining followed by FACS analysis was performed . ( B , C ) PINCR-WT#1 and PINCR-KO ( KO#1 and KO#2 ) clones were untreated or treated with 5-FU in biological duplicates for 48 hr and the effect on G1 arrest and cell death ( sub-G1 ) was assessed by PI staining followed by FACS . ( D ) PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU for 48 hr and immunoblotting for cleaved PARP and loading control GAPDH was performed . ( E , F ) PINCR-WT and PINCR-KO cells were untreated or treated with indicated 5-FU concentrations for 4 hr , and colony formation assays were performed after 10 days . ( G , H ) Untreated PINCR-WT#1 and PINCR-KO ( KO#1 and KO#2 ) cells were injected subcutaneously into the flanks of athymic nude mice ( five mice for each group , two tumors per mice ) . Average tumor volume ( G ) and tumor mass ( H ) are shown . Error bars in F represent SD from three experiments . Tumor mass data in H is shown as median +/- interquartile range and p-values were calculated using the Krusal-Wallis test . p<0 . 05; #p<0 . 01; p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 01910 . 7554/eLife . 23244 . 020Figure 3—source data 1 . Cleaved PARP immunoblot for Figure 3D . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02010 . 7554/eLife . 23244 . 021Figure 3—figure supplement 1 . PINCR is induced by 5-FU or NCS . ( A ) qRT-PCR analysis from PINCR-WT and PINCR-KO cells untreated or treated with 5-FU for 24 hr . ( B ) qRT-PCR for PINCR from HCT116 cells untreated or treated with NCS for 24 hr . Error bars represent SD from three biological replicates . p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02110 . 7554/eLife . 23244 . 022Figure 3—figure supplement 2 . CRISPR knockout of PINCR in SW48 cells . ( A ) Partial genomic sequence of PINCR from Sanger sequencing and snap-shot of PINCR locus ( using BLAT ) for the PINCR-KO SW48 clone is shown . The p53RE is underlined and in green font . ( B ) qRT-PCR analysis from PINCR-WT and PINCR-KO SW48 cells untreated or treated with 5-FU for 24 hr . Error bars represent SD from three experiments . ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02210 . 7554/eLife . 23244 . 023Figure 3—figure supplement 3 . Cell cycle profiles for PINCR-WT and PINCR-KO SW48 cells . PINCR-WT and PINCR-KO SW48 cells were untreated or treated with 5-FU for 72 hr or 96 hr and the effect on cell cycle and apoptosis was determined by PI-staining and FACS analysis . Raw cell cycle profiles from a representative experiment are shown in ( A ) and the results from three independent experiments are shown in ( B , C ) . Error bars represent SD from three experiments . p<0 . 05; p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02310 . 7554/eLife . 23244 . 024Figure 3—figure supplement 4 . Long term proliferation for PINCR-WT and PINCR-KO SW48 cells . PINCR-WT and PINCR-KO SW48 cells were untreated or treated with 5-FU . After 7 days , the effect on long-term of these cells was assessed by staining the cells with crystal violet . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02410 . 7554/eLife . 23244 . 025Figure 3—figure supplement 5 . Effect of different doses of 5-FU on PINCR levels and cell survival . ( A ) qRT-PCR analysis from HCT116 cells untreated or treated for 24 hr with different doses of 5-FU . ( B ) HCT116 PINCR-WT and PINCR-KO cells were untreated or treated for 48 hr with different doses of 5-FU and cell cycle profiles were examined by PI-staining and FACS analysis . Error bars represent SD from three experiments . p<0 . 05; #p<0 . 01; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02510 . 7554/eLife . 23244 . 026Figure 3—figure supplement 6 . Over-expression of PINCR has no significant effect on cell cycle . HCT116 cells were transfected with pCB6 or pCB6-PINCR for 24 hr and then left untreated cells or treated with 5-FU for 48 hr . PINCR levels were measured by qRT-PCR ( A ) and the effect on cell cycle was determined by PI-staining and FACS analysis ( B , C ) . Raw cell cycle profiles from a representative experiment are shown in ( B ) . Cell cycle profiles from three independent experiments are shown in ( C ) . Error bars represent SD from three experiments . p<0 . 05; *p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02610 . 7554/eLife . 23244 . 027Figure 3—figure supplement 7 . Immunohistochemical analysis from PINCR-WT and PINCR-KO cells . Immunohistochemical staining of the PINCR-WT and PINCR-KO tumors for the proliferation marker Ki67 ( A ) , apoptosis marker Cleaved Caspase-3 ( B ) and H and E staining ( C ) . Arrows indicate Ki67 ( A ) or Cleaved Caspase-3 ( B ) positive cells . Error bars represent SD from four biological replicates . ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02710 . 7554/eLife . 23244 . 028Figure 3—figure supplement 7—source data 1 . Ki67 staining images of PINCR-WT and PINCR-KO tumors for Figure 3—figure supplement 7A . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 028 Next , to determine if the observed phenotypes are not restricted to HCT116 , we knocked out PINCR in SW48 cells ( Figure 3—figure supplement 2A and B ) . In response to DNA damage induced by 5-FU , we observed reduced G1 arrest and increased apoptosis in the PINCR-KO clone as compared to PINCR-WT clones ( Figure 3—figure supplement 3A–C ) . Moreover , following extended treatment with 5-FU , the PINCR-KO clone was markedly more sensitive than PINCR-WT clones ( Figure 3—figure supplement 4 ) . These data confirm that the phenotypic effects observed upon loss of PINCR are not unique to HCT116 . We next employed several different concentrations of 5-FU ( 0 to 375 µM ) and measured the extent of induction of PINCR and PUMA , and examined the sub-G1 population . Although we found an increase in sub-G1 population with increasing dose of 5-FU , the extent of induction of PINCR or PUMA did not change significantly ( Figure 3—figure supplement 5A ) . At all doses of 5-FU , PINCR-KO cells were more sensitive than PINCR-WT cells ( Figure 3—figure supplement 5B ) . These data indicate that the extent of PINCR induction may not be determinant of the likelihood of the cells to die rather than undergo G1 arrest . In addition , we found that over-expression of PINCR did not significantly affect the cell cycle in untreated cells or in response to DNA damage induced by 5-FU ( Figure 3—figure supplement 6 ) . To determine the function of PINCR in an in vivo setting , we subcutaneously injected NOD-SCID mice with HCT116-PINCR-WT or PINCR-KO cells , untreated or treated with 5-FU for 4 hr followed by a 4-hr recovery . Although mice injected with 5-FU-treated PINCR-WT or PINCR-KO cells did not form tumors , in untreated condition the rate of tumor growth was substantially reduced ( 7–10-fold ) upon loss of PINCR ( Figure 3G and H ) . All mice injected with untreated PINCR-WT cells developed detectable tumors , whereas the untreated PINCR-KO cells displayed significantly reduced tumor growth as early as day 12 post-injection ( p<0 . 05 ) ( Figure 3G ) . Immunohistochemical staining of the tumors for the proliferation marker Ki67 and the apoptosis marker cleaved caspase-3 revealed that PINCR-WT and PINCR-KO tumors had a high proportion of Ki67-positive cells ( >50% ) and a very low proportion of cleaved caspase-3-positive cells ( <1% ) ( Figure 3—figure supplement 7 ) . As compared to PINCR-WT tumors , the PINCR-KO tumors had significantly decreased Ki67-positive cells ( Figure 3—figure supplement 7A and Figure 3—figure supplement 7—source data 1 ) , suggesting that the observed reduced tumor volume is due to inhibition of cell proliferation . To determine if PINCR mediates its effect by regulating gene expression , we performed mRNA microarrays from three biological replicates of PINCR-WT and PINCR-KO cells , untreated or treated with 5-FU for 24 hr ( Supplementary file 3 ) . Gene set enrichment analysis ( GSEA ) for the upregulated genes identified the p53 pathway as the top upregulated pathway after DNA damage in both PINCR-WT and PINCR-KO cells ( Figure 4—figure supplement 1A and Figure 4—figure supplement 1—source data 1 ) suggesting that loss of PINCR does not alter global p53 signaling . Consistent with this , we observed comparable p53 induction in both PINCR-WT and PINCR-KO cells after 5-FU treatment ( Figure 4—figure supplement 1B ) and the majority of known p53 targets including the G1 regulator p21 were induced to similar levels . Interestingly , the normalized enrichment score ( NES ) for the p53 pathway in PINCR-KO cells was significantly lower ( NES = 2 . 673 ) than that in PINCR-WT cells ( NES = 3 . 045 ) indicating that the induction of a subset of p53 targets may be abrogated in PINCR-KO cells ( Figure 4—figure supplement 1A ) . Thus , the induction of a subset of p53 targets appeared to be PINCR-dependent ( Figure 4A ) . Further analysis indicated that the induction of 11 direct p53 targets that were also identified in the p53 GRO-seq study ( Allen et al . , 2014 ) including BTG2 , GPX1 , RRM2B was less pronounced in PINCR-KO cells ( Supplementary file 3 ) . 10 . 7554/eLife . 23244 . 029Figure 4 . PINCR regulates the induction of select p53 target genes important for G1 arrest after DNA damage . ( A ) Schematic representation of a subset of p53 target genes upregulated after 5-FU treatment in a PINCR-dependent or PINCR-independent manner . ( B ) qRT-PCR analysis from PINCR-WT and PINCR-KO cells untreated or treated with 5-FU for 24 hr . ( C ) PINCR-WT and PINCR-KO cells were treated with 5-FU for 24 hr and qPCR for the p53RE of BTG2 , RRM2B and GPX1 was performed from IgG-ChIP and p53-ChIP . ( D–F ) HCT116 cells were reverse transfected with CTL-ASO or PINCR-ASO for 48 hr . The cells were then left untreated or treated with 5-FU for 24 hr ( D ) or 48 hr ( E , F ) following which qRT-PCR analysis ( D ) , PI staining and FACS analysis was performed ( E , F ) . Error bars in B-F represent SD from three independent experiments . p<0 . 05; #p<0 . 01; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 02910 . 7554/eLife . 23244 . 030Figure 4—figure supplement 1 . Loss of PINCR results in impaired induction of a subset of p53 targets without altering induction of p53 levels . ( A ) Gene set enrichment analysis ( GSEA ) for the genes upregulated in the microarrays performed in biological triplicates from untreated or 5-FU-treated PINCR-WT and PINCR-KO cells . ( B ) PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU for 24 hr and immunoblotting for p53 and loading control GAPDH was performed . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03010 . 7554/eLife . 23244 . 031Figure 4—figure supplement 1—source data 1 . p53 immunoblot for Figure 4—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03110 . 7554/eLife . 23244 . 032Figure 4—figure supplement 2 . Loss of PINCR does not markedly alter total p21 protein levels , Rb phosphorylation or subcellular localization of p21 . Immunoblotting was performed for p21 ( A ) and phospho-Rb ( pRb ) ( C ) from whole cell extracts prepared from PINCR-WT and PINCR-KO HCT116 untreated or treated with 5-FU for 24 hr . GAPDH was used as loading control . ( B ) The effect of loss of PINCR on subcellular localization of p21 was determined by immunoblotting from nuclear and cytoplasmic extracts prepared from PINCR-WT and PINCR-KO HCT116 untreated or treated with 5-FU for 24 hr . Tubulin was used as cytoplasmic marker and Histone H3 was used as nuclear marker . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03210 . 7554/eLife . 23244 . 033Figure 4—figure supplement 2—source data 1 . p21 and phospho Rb immunoblots for Figure 4—figure supplement 2A , B and C . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03310 . 7554/eLife . 23244 . 034Figure 4—figure supplement 3 . Knockdown of the PINCR targets BTG2 , GPX1 or RRM2B phenocopies the effect of PINCR loss . ( A ) qRT-PCR analysis from PINCR-WT HCT116 cells transfected for 48 hr with control siRNA ( siCTL ) or two independent siRNAs ( I and II ) against BTG2 , GPX1 or RRM2B . PI-staining and FACS analysis was performed from PINCR-WT HC116 cells ( treated with 5-FU for 48 hr ) after knockdown of BTG2 , GPX1 or RRM2B . Error bars represent SD from three biological replicates . p<0 . 05; #p<0 . 01; *p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03410 . 7554/eLife . 23244 . 035Figure 4—figure supplement 4 . Raw cell cycle profiles showing that knockdown of the PINCR targets BTG2 , GPX1 or RRM2B phenocopies the effect of PINCR loss . Raw cell cycle profiles from the PI-staining and FACS analysis of Figure 4—figure supplement 4 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03510 . 7554/eLife . 23244 . 036Figure 4—figure supplement 5 . Knockdown of PINCR results in decreased G1 arrest and increased apoptosis . ( A ) HCT116 cells were reverse transfected with CTL-ASO or PINCR-ASO for 48 hr and then treated with DOXO for 48 hr . The effect on cell cycle was examined by PI-staining and FACS analysis . Results from three independent experiments are shown in ( A ) . Raw cell cycle profiles from a representative experiment from untreated , DOXO-treated or 5-FU-treated cells ( for A and Figure 4E–F ) are shown . Error bars represent SD from three experiments . p<0 . 05 , ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03610 . 7554/eLife . 23244 . 037Figure 4—figure supplement 6 . Knockdown of PINCR results in reduced colony formation . Colony formation assays were performed from HCT116 cells that were treated with 5-FU for 4 hr ( A ) or untreated ( B ) after knockdown of PINCR with PINCR-ASO . CTL refers to control ( CTL ) ASO . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 037 Among the 11 PINCR-dependent p53 targets , we selected BTG2 , RRM2B and GPX1 for further analysis due to evidence in the literature supporting their roles in induction of G1 arrest and inhibition of apoptosis after DNA damage . BTG2 encodes an antiproliferative protein critical in regulation of the G1/S transition ( Guardavaccaro et al . , 2000; Rouault et al . , 1996; Tirone , 2001 ) . Silencing RRM2B in p53-proficient cells reduces ribonucleotide reductase activity , DNA repair , and cell survival after exposure to various genotoxins ( Tanaka et al . , 2000; Xue et al . , 2007; Yanamoto et al . , 2005 ) . GPX1 attenuates DOXO-induced cell cycle arrest and apoptosis ( Gao et al . , 2008 ) . In subsequent experiments , we sought to use p21 as a negative control because p21 mRNA was induced to similar levels in both PINCR-WT and PINCR-KO cells ( Figure 4B ) . However , given the well-established role of p21 in controlling G1 arrest after DNA damage , it was important to make sure that loss of PINCR did not alter p21 protein levels , p21 subcellular localization and/or Rb-phosphorylation . Indeed , we found similar levels of total , nuclear or cytoplasmic p21 in PINCR-WT and PINCR-KO cells under untreated condition and after 5-FU treatment ( Figure 4—figure supplement 2A and B and Figure 4—figure supplement 2—source data 1 ) . The decrease in Rb phosphorylation in response to 5-FU treatment was comparable in PINCR-WT and PINCR-KO cells ( Figure 4—figure supplement 2C and Figure 4—figure supplement 2—source data 1 ) . These results indicate that p21 expression is not altered upon loss of PINCR and it can therefore be used as a negative control . We next asked the question if depletion of the PINCR targets BTG2 , GPX1 and RRM2B recapitulated the effects of PINCR depletion . We validated significant knockdown of these genes by qRT-PCR ( Figure 4—figure supplement 3A ) and found significantly increased apoptosis ( sub-G1 cells ) upon knockdown of each of these genes followed by 5-FU treatment ( Figure 4—figure supplement 3B–3D and Figure 4—figure supplement 4 ) . Significant reduction in G1 arrest after 5-FU treatment was observed after knockdown of GPX1 but not BTG2 or RRM2B . These data indicate that depletion of BTG2 , GPX1 or RRM2B recapitulates the effects of PINCR depletion in response to 5-FU treatment . Consistent with our microarray data , we observed impaired induction of BTG2 , GPX1 and RRM2B mRNAs upon loss of PINCR ( Figure 4B ) . Moreover , by p53 ChIP-qPCR , we observed substantial decrease in the binding of p53 to the p53RE of BTG2 , GPX1 and RRM2B upon loss of PINCR ( Figure 4C ) . There is evidence in the literature that p53 can directly bind to RNA including a recent report showing direct binding of p53 to the p53-regulated lncRNA DINO ( Riley and Maher , 2007; Schmitt et al . , 2016 ) . However , we found that PINCR does not directly bind to p53 ( data not shown ) . To make sure that the altered induction of BTG2 , GPX1 and RRM2B reflect a function of the PINCR transcript itself , we measured the induction of these genes after PINCR knockdown using antisense oligonucleotides ( ASOs ) . We tested 5 ASOs that potentially target PINCR RNA ( data not shown ) . Robust knockdown of PINCR in HCT116 cells was observed with one ASO that we designated as PINCR-ASO ( Figure 4D ) . Importantly , as observed with PINCR-KO cells , knockdown of PINCR followed by 5-FU treatment resulted in decreased induction of BTG2 , GPX1 and RRM2B but not p21 ( Figure 4D ) and caused decreased G1 arrest ( Figure 4E ) and increased apoptosis ( Figure 4F ) after 5-FU or DOXO treatment ( Figure 4—figure supplement 5 ) . In clonogenic survival assays , knockdown of PINCR resulted in reduced colony formation after 5-FU treatment ( Figure 4—figure supplement 6A ) . Surprisingly , unlike PINCR-KO cells that did not show significant difference in proliferation from PINCR-WT cells in untreated condition , decreased colony formation in untreated condition was observed after PINCR knockdown ( Figure 4—figure supplement 6B ) . Although this result indicates that basal PINCR levels can regulate proliferation despite low expression , this growth defect may be restored long-term during genetic deletion of PINCR using CRISPR/Cas9 . Taken together , the results from PINCR knockdown experiments corroborates our findings from the PINCR-KO clones . Because we found that PINCR does not bind to p53 , we hypothesized that PINCR binds to an RNA-binding protein that serves as an adaptor protein and mediates this effect of PINCR . To identify this adaptor protein , we incubated in vitro-transcribed biotinylated ( Bi ) -PINCR ( Bi-PINCR ) or Bi-Luciferase ( Bi-LUC ) RNA with untreated or DOXO-treated nuclear extracts and performed streptavidin pulldowns followed by mass spectrometry . Eleven proteins were enriched at least twofold in the Bi-PINCR pulldowns ( Supplementary file 4 ) as compared to Bi-LUC pulldowns in untreated condition as well as after DOXO treatment . Of these 11 proteins , the RNA- and DNA-binding nuclear matrix protein Matrin 3 showed the strongest enrichment ( eightfold in untreated condition; 16-fold after DOXO treatment ) ( Figure 5A and Supplementary file 4 ) . In a recent iCLIP ( Individual-nucleotide resolution UV crosslinking and immunoprecipitation ) study ( Coelho et al . , 2015 ) , the consensus RNA motif recognized by Matrin 3 was identified . We found that PINCR has six Matrin 3 binding motifs , and this motif was significantly enriched in the PINCR RNA as compared to the transcriptome ( Figure 5—figure supplement 1A and B ) . We next validated the specific PINCR-Matrin 3 interaction by performing streptavidin pulldowns followed by immunoblotting after incubating Bi-PINCR or Bi-LUC with HCT116 nuclear extracts ( Figure 5B and Figure 5—source data 1 ) or recombinant Matrin 3 ( rMatrin 3 ) ( Figure 5C and Figure 5—source data 1 ) . Moreover , we observed ~300-fold enrichment of PINCR in the Matrin 3 IPs from formaldehyde crosslinked HCT116 cells treated with DOXO ( Figure 5D ) ; p21 mRNA was not enriched ( Figure 5D and Figure 5—source data 1 ) , demonstrating the specificity of the PINCR-Matrin 3 interaction . 10 . 7554/eLife . 23244 . 038Figure 5 . Matrin 3 binds to PINCR and functions as a downstream effector of PINCR . ( A ) Peptide spectrum matches ( PSMs ) corresponding to Matrin 3 in the Bi-LUC and Bi-PINCR pulldowns from mass spectrometry analysis . ( B , C ) Streptavidin pulldowns followed by immunoblotting was performed following incubation of Bi-LUC and Bi-PINCR RNA with DOXO-treated HCT116 nuclear extracts ( B ) or recombinant Matrin 3 ( rMatrin 3 ) ( C ) . ( D ) Specific enrichment of PINCR in the Matrin 3 IPs was assessed by qRT-PCR from 24 hr 5-FU-treated formaldehyde cross-linked HCT116 cells . p21 mRNA was used as negative control . ( E ) PINCR-WT cells were transfected with CTL or two independent Matrin 3 siRNAs ( I and II ) for 48 hr and Matrin 3 knockdown was measured by immunoblotting . ( F ) PINCR-WT and PINCR-KO cells were transfected with CTL or Matrin 3 siRNAs and after 48 hr the cells were untreated or treated with 5-FU for 48 hr . The effect on the sub-G1 population was assessed by PI staining followed by FACS . ( G ) PINCR-WT and PINCR-KO cells were transfected with CTL or Matrin 3 siRNAs for 48 hr; transfected cells were left untreated or treated with 5-FU for 24 hr and qRT-PCR was performed . Error bars in D , F and G represent SD from three independent experiments . p<0 . 05; #p<0 . 01; p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03810 . 7554/eLife . 23244 . 039Figure 5—source data 1 . Matrin 3 immunoblot for Figure 5B , C and E . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 03910 . 7554/eLife . 23244 . 040Figure 5—figure supplement 1 . Matrin 3 motifs in PINCR RNA . ( A , B ) Putative Matrin 3 binding motif in PINCR RNA . ( A ) ‘N’ represents the number of times the motif appears in the PINCR RNA . ‘ES’ represents the enrichment score calculated as shown in ‘B’ . ( C ) HCT116 cells were reverse transfected with a control siRNA ( siCTL ) or siRNAs targeting Matrin 3 ( siMatrin 3-I and siMatrin 3-II ) for 48 hr and the extent of Matrin 3 knockdown was measured by qRT-PCR for Matrin 3 normalized to GADPH . Error bars represent SD from three independent experiments . ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04010 . 7554/eLife . 23244 . 041Figure 5—figure supplement 2 . Cell cycle analysis after Matrin 3 knockdown . PINCR-WT and PINCR-KO HCT116 cells were reverse transfected with a control siRNA ( siCTL ) or siRNAs targeting Matrin 3 ( siMatrin 3-I and siMatrin 3-II ) for 48 hr . Transfected cells were left untreated or treated with DOXO or 5-FU for 48 hr . The effect on G1 arrest and apoptosis was examined by PI-staining and FACS analysis . Shown are the results from three independent experiments after 5-FU treatment ( A ) or DOXO treatment ( B , C ) . Error bars represent SD from three independent experiments . #p<0 . 01 , ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04110 . 7554/eLife . 23244 . 042Figure 5—figure supplement 3 . Raw cell cycle profiles from a representative experiment for Figure 5F and Figure 5—figure supplement 3 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04210 . 7554/eLife . 23244 . 043Figure 5—figure supplement 4 . Matrin 3 regulates the induction of PINCR targets upon p53 activation by Nutlin-3 . PINCR-WT HCT116 cells were reverse transfected with CTL siRNA or Matrin-3 siRNAs for 48 hr and then treated with Nutlin-3 for 24 hr . The expression of PINCR targets and the negative control p21 , was assessed by qRT-PCR . Error bars represent SD from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04310 . 7554/eLife . 23244 . 044Figure 5—figure supplement 5 . Matrin 3 protein level and subcellular localization is not altered after DNA damage . HCT116 cells were untreated or treated with 5-FU for 24 hr and immunoblotting for Matrin 3 was performed from whole cell extracts ( A ) or nuclear and cytoplasmic extracts ( B ) . Tubulin was used as loading control for ( A ) . For ( B ) Histone H3 was used as nuclear marker and Tubulin was used as cytoplasmic marker . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04410 . 7554/eLife . 23244 . 045Figure 5—figure supplement 5—source data 1 . Matrin 3 immunoblot for Figure 5—figure supplement 5A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 045 Next , we knocked down Matrin 3 with two independent siRNAs ( Figure 5E and Figure 5—figure supplement 1C ) and determined the effect on G1 arrest and apoptosis of PINCR-WT and PINCR-KO . After 5-FU or DOXO treatment , there was more apoptosis upon Matrin 3 knockdown in PINCR-WT but this increase was not observed in the PINCR-KO ( Figure 5F and Figure 5—figure supplements 2 and 3 ) indicating that Matrin 3 is a downstream effector of PINCR . We did not observe a significant difference in G1 arrest , suggesting that Matrin 3 does not mediate the G1 arrest regulated by PINCR . These data indicate that there is an epistatic interaction between PINCR and Matrin 3 as virtually all the apoptotic effects of Matrin 3 after DNA damage are dependent on PINCR . Furthermore , in PINCR-WT cells silencing Matrin 3 resulted in less or no induction of the PINCR targets BTG2 , GPX1 and RRM2B but not p21 mRNA after 5-FU treatment ( Figure 5G ) . A role of Matrin 3 in regulating the induction of these p53 targets was also observed in response to Nutlin-3 treatment ( Figure 5—figure supplement 4 ) . Immunoblotting from untreated or 5-FU-treated HCT116 whole cell lysates and nuclear and cytoplasmic lysates indicated no change in Matrin 3 levels or subcellular localization ( Figure 5—figure supplement 5 and Figure 5—figure supplement 5—source data 1 ) . Collectively , these data suggest that the induction of the PINCR targets BTG2 , GPX1 and RRM2B is largely mediated by Matrin 3 and reveal an epistatic interaction between PINCR and Matrin 3 . To determine if Matrin 3 mediates the effect of PINCR by functioning as an adaptor protein , we performed co-IP experiments to determine if p53 and Matrin 3 form a complex . We found that p53 interacts with Matrin 3 in both untreated and 5-FU treated cells ( Figure 6A and B , Figure 6—figure supplement 1A , Figure 6—source data 1 and Figure 6—figure supplement 1—source data 1 ) . This interaction was not altered in the presence of RNase A or DNase , suggesting that p53 and Matrin 3 form a protein-protein complex ( Figure 6B ) . This result prompted us to examine the association of Matrin 3 on the p53RE of BTG2 , GPX1 and RRM2B . Matrin 3 ChIP-qPCR revealed that in untreated PINCR-WT cells , Matrin 3 binds to the p53RE of BTG2 and RRM2B but not GPX1 ( Figure 6C ) . For all three genes , in PINCR-WT cells , there was increased Matrin 3 binding to their p53RE after 5-FU treatment . Loss of PINCR impaired this binding of Matrin 3 to the p53RE of these genes ( Figure 6C ) but not the p53RE of p21 ( Figure 6—figure supplement 1B ) . Interestingly , after knockdown of Matrin 3 and 5-FU treatment in PINCR-WT cells , we did not observe a significant difference in the binding of p53 to the p53RE of these genes ( Figure 6D and Figure 6—figure supplement 2 ) suggesting that Matrin 3 does not control p53 occupancy at the p53RE in the promoters of these genes . 10 . 7554/eLife . 23244 . 046Figure 6 . Matrin 3 forms a complex with p53 complex and associates with the p53RE of select PINCR targets . ( A ) HCT116 cells were treated with 5-FU for 24 hr and immunoblotting for p53 was performed from input , no cell lysate control and IgG or Matrin 3 IP from whole cell extracts . ( B ) HCT116 cells were untreated or treated with 5-FU for 24 hr and the interaction between p53 and Matrin 3 was assessed by IPs from Mock ( no extract ) , RNase-treated or DNase-treated whole cell lysates . ( C ) PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU for 24 hr and qPCR with primers spanning the p53RE of BTG2 , RRM2B and GPX1 was performed from IgG-ChIP and Matrin 3-ChIP . ( D ) PINCR-WT HCT116 cells were reverse transfected with CTL or Matrin 3 siRNAs for 48 hr and then treated with 5-FU for 24 hr . The enrichment of p53 at the p53RE of PINCR targets was determined by ChIP-qPCR . Errors bars in C and D represent SD from three independent experiments . p<0 . 05; #p<0 . 01; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04610 . 7554/eLife . 23244 . 047Figure 6—source data 1 . p53 immunoblot for Figure 6A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04710 . 7554/eLife . 23244 . 048Figure 6—figure supplement 1 . Matrin 3 interacts with p53 . ( A ) Co-IP was performed from HCT116 cells in which nuclear extracts were incubated with IgG or p53 antibody and immunoblotting for Matrin 3 was performed . ( B ) Matrin 3 ChIP-qPCR was performed from PINCR-WT and PINCR-KO cells treated with 5-FU and the association with the p53RE in the p21 promoter was determined . Error bars represent SD from three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04810 . 7554/eLife . 23244 . 049Figure 6—figure supplement 1—source data 1 . Matrin 3 immunoblot for Figure 6—figure supplement 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 04910 . 7554/eLife . 23244 . 050Figure 6—figure supplement 2 . p53 binding to p21 promoter upon Matrin 3 knockdown . PINCR-WT HCT116 cells were reverse transfected with a control siRNA ( siCTL ) or siRNAs targeting Matrin 3 ( siMatrin 3-I and siMatrin 3-II ) for 48 hr . Transfected cells were treated with 5-FU for 24 hr and the binding of p53 to the p53RE of the p21 promoter was determined by ChIP-qPCR . Error bars represent SD from three independent experiments . ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 050 We next examined how Matrin 3 regulates the induction of the PINCR targets BTG2 , GPX1 and RRM2B after DNA damage , without altering p53 binding to their promoters . Given the evidence that Matrin 3 associates with enhancer regions ( Skowronska-Krawczyk et al . , 2005 ) , we reasoned that Matrin 3 modulates the induction of these genes by binding to their enhancer regions . More recently , it has been shown that proper enhancer-gene pairing is enabled by insulated neighborhoods formed by CTCF anchoring at domain boundaries and cohesion looping ( Hnisz et al . , 2016 ) and are mostly conserved across cell types . To test this possibility , we identified insulated neighborhoods around PINCR targets , ChIP-seq data tracks ( ENCODE Project Consortium , 2012 ) in HCT116 cells ( Figure 7A , Figure 7—figure supplement 1 and Figure 7—figure supplement 2 ) for ( 1 ) CTCF , a protein known to bind to chromatin domain boundaries , ( 2 ) the chromatin loop-enabling cohesion component RAD21 , ( 3 ) promoter associated histone mark H3K4me3 and ( 4 ) active promoter/enhancer associated mark H3K27ac . To determine the overlap of these peaks with p53 , we utilized p53 ChIP-seq data from MCF7 and U2OS cells ( Figure 7A , Figure 7—figure supplement 1 and Figure 7—figure supplement 2 ) . In addition , to determine potential chromatin looping near these PINCR targets , we utilized Hi-C data ( Figure 7A , Figure 7—figure supplement 1 and Figure 7—figure supplement 2 ) . For each of the three PINCR targets , the Hi-C data indicated chromatin looping with appropriate CTCF and cohesion signal consistent with insulated domain structure . Within the loop , we observed the following: ( 1 ) a strong p53 ChIP-seq peak corresponding to the p53RE in the promoter of these genes that was also marked by strong signal for H3K4me3 and H3K27ac; ( 2 ) a weak p53 ChIP-seq peak that was also marked by strong signal for H3K27ac and weak signal H3K4me3 . Promoters are marked by high H3K4me3 and high H3K27ac whereas enhancers typically have low H3K4me3 and high H3K27ac ( Ernst et al . , 2011 ) . Thus , our Hi-C and ChIP-seq data analysis indicates potential chromatin looping between a weak p53 binding region in the enhancer and strong p53 binding region in the promoter of the PINCR targets BTG2 , GPX1 and RRM2B . Notably , whereas the strong ChIP-seq peak at the promoters of these three genes had a canonical p53RE , the weak p53 binding region in their enhancers did not have a canonical p53RE indicating indirect association of p53 at these enhancers . 10 . 7554/eLife . 23244 . 051Figure 7 . PINCR modulates the association of Matrin 3 with enhancers of PINCR targets within insulated neighborhoods . ( A ) Topological domain and looping structure indicated by 3D contact domain profile ( top ) surrounding BTG2 gene and Hi-C data from ( Rao et al . , 2014 ) . ChIP-seq data tracks ( middle ) in HCT116 showing CTCF anchors and loop-enabling cohesion ( RAD21 ) , as well as promoter associated histone mark H3K4me3 and active promoter/enhancer associated mark H3K27ac . p53 ChIP-seq in untreated or treated MCF7 is also shown . Candidate regulatory enhancers ( used for ChIP-qPCR of Matrin3 ) are highlighted with an orange box at the tail of an arrow pointing toward the putative target gene . The p53 response element is shown , found at the promoter , where stronger ChIP-seq signal was present upon treatments . The one-dimensional genomic distance between the enhancer and the promoter is indicated between the zoomed in boxes ( bottom ) . ( B , C ) PINCR-WT and PINCR-KO HCT116 cells were treated with 5-FU for 24 hr and the association of Matrin 3 ( B ) or p53 ( C ) with the enhancer of PINCR targets was assessed by ChIP-qPCR . ( D ) PINCR-WT HCT116 cells were reverse transfected with CTL or Matrin 3 siRNAs for 48 hr and then treated with 5-FU for 24 hr . The enrichment of p53 at the enhancer regions of PINCR targets was determined by ChIP-qPCR . ( E ) A cartoon showing the chromatin looping between the enhancer and promoter region of BTG2 . Errors bars in B , C and D represent SD from three independent experiments . p<0 . 05; #p<0 . 01; p<0 . 005; ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05110 . 7554/eLife . 23244 . 052Figure 7—figure supplement 1 . Topological domain and looping structure indicated by 3D contact domain profile ( top ) surrounding GPX1 gene ( chr3:49 , 200 , 000–49 , 500 , 000 ) with Hi-C data from ( Rao et al . , 2014 ) . ChIP-seq data tracks ( middle ) in HCT116 showing CTCF anchors ( motif direction indicated by arrow ) and loop-enabling cohesion ( RAD21 ) , as well as promoter-associated histone mark H3K4me3 and active promoter/enhancer-associated mark H3K27ac . ChIP-seq of p53 in treated cells ( MCF7 and U2OS ) is also shown . Candidate regulatory enhancers ( used for ChIP-qPCR of Matrin3 ) are highlighted with an orange box at the tail of an arrow pointing toward the putative target gene . The p53RE is shown , found at the promoter . The one-dimensional distance between the enhancer and the promoter is indicated between the zoomed in boxes ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05210 . 7554/eLife . 23244 . 053Figure 7—figure supplement 2 . Topological domain and looping structure indicated by 3D contact domain profile ( top ) surrounding RRM2B gene ( chr8:103 , 250 , 000–103 , 800 , 000 ) with Hi-C data from ( Rao et al . , 2014 ) . ChIP-seq data tracks ( middle ) in HCT116 showing CTCF ( motif direction indicated by arrow ) , RAD21 , H3K4me3 and H3K27ac . ChIP-seq of p53 in treated cells ( MCF7 and U2OS ) is also shown . Candidate regulatory enhancers ( used for ChIP-qPCR of Matrin3 ) are highlighted with an orange box at the tail of an arrow pointing towards the putative target gene . The p53RE is shown , found at the promoter , where stronger ChIP-seq signal was present upon treatments . The one-dimensional genomic enhancer/promoter distance of 569 KB is indicated ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 053 Next , we sought to determine if Matrin 3 associates with the enhancer of BTG2 , GPX1 and RRM2B and if this association is dependent on PINCR . To test this , we performed ChIP-qPCR for Matrin 3 from PINCR-WT and PINCR-KO cells after 5-FU treatment . In PINCR-WT cells , there was strong enrichment of Matrin 3 at the enhancers of each of these genes ( Figure 7B ) . Loss of PINCR resulted in significant reduction in Matrin 3 occupancy on each of these enhancer regions ( Figure 7B ) . Moreover , after 5-FU treatment , we found significantly reduced p53 binding to these enhancer regions upon loss of PINCR ( Figure 7C ) or upon knockdown of Matrin 3 ( Figure 7D ) . These results indicate a role of Matrin 3 and PINCR in facilitating the association of p53 with the enhancers of specific p53 targets BTG2 , GPX1 and RRM2B and provide evidence of chromatin looping between the enhancers and promoters of these genes ( Figure 7E ) . We next explored the possibility that PINCR is also a part of the p53-Matrin 3 complex on the p53RE of BTG2 , GPX1 and RRM2B . To test this , we used a novel approach in which we tagged endogenous PINCR with an S1-tag and utilized the S1-tag to pulldown PINCR and then performed qPCR for the p53RE of BTG2 , GPX1 and RRM2B . The S1-tag is a 44 nt RNA aptamer that binds to streptavidin with high affinity and has been used in vitro to identify proteins that bind to S1-tagged RNAs ( Butter et al . , 2009; Iioka et al . , 2011; Srisawat and Engelke , 2001 , 2002 ) . To tag endogenous PINCR , we used CRISPR/Cas9 and knocked-in a single S1-tag at the 3’end of PINCR in HCT116 cells ( Figure 8A and Figure 8—figure supplements 1 and 2 ) . Importantly , the PINCR-S1 RNA was strongly upregulated ( >20 fold ) after DOXO treatment ( Figure 8—figure supplement 3A ) . Like endogenous untagged PINCR , the PINCR-S1 RNA was predominantly nuclear ( Figure 8—figure supplement 3B ) and expressed at levels comparable to PINCR ( Figure 8—figure supplement 3C ) . PINCR overexpression did not alter the expression of PINCR-S1 or BTG2 , GPX1 and RRM2B , suggesting that PINCR does not regulate its own expression and that PINCR over-expression is not sufficient to alter the expression of PINCR targets ( Figure 8—figure supplement 4A and B ) . 10 . 7554/eLife . 23244 . 054Figure 8 . PINCR associates with the enhancer regions of select p53 targets in a Matrin-3-dependent manner . Schematic showing knock-in of S1-tag at the 3’end of PINCR . ( B ) The enrichment of PINCR in the streptavidin pulldowns from PINCR and PINCR-S1 cells treated with 5-FU for 24 hr was assessed by qRT-PCR . ( C ) PINCR and PINCR-S1 cells were treated with 5-FU for 24 hr and followed by streptavidin pulldown . Interaction between PINCR and Matrin 3 was confirmed by immunoblotting for Matrin 3 or the control Tubulin . ( D ) PINCR and PINCR-S1 cells were treated with 5-FU for 24 hr and qPCR with primers spanning the p53RE of BTG2 , RRM2B and GPX1 was performed following streptavidin pulldown . ( E ) PINCR-S1 cells were reverse transfected with CTL siRNA , p53 siRNAs or two independent Matrin 3 siRNAs and then treated with 5-FU for 24 hr . The enrichment of PINCR-S1 at the p53RE and enhancer regions of PINCR targets was determined by ChIP-qPCR from the streptavidin pulldown material . Error bars represent SD from three independent experiments . p<0 . 05; *p<0 . 005 , ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05410 . 7554/eLife . 23244 . 055Figure 8—source data 1 . Matrin 3 immunoblot for Figure 8C . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05510 . 7554/eLife . 23244 . 056Figure 8—figure supplement 1 . Sequence alignment of gDNA from PINCR ( Seq_1 ) and PINCR-S1 ( Seq_2 ) clones . Yellow is the 44 nucleotide S1-tag inserted in PINCR-S1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05610 . 7554/eLife . 23244 . 057Figure 8—figure supplement 2 . Full sequence of the S1 targeting vector . Text highlighted in green corresponds to sgRNA target sequence , red is the S1-tag and blue is the 3’ end of PINCR . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05710 . 7554/eLife . 23244 . 058Figure 8—figure supplement 3 . PINCR-S1 is strongly induced after DNA damage and is a predominantly nuclear lncRNA . ( A ) PINCR-S1 cells were left untreated or treated with DOXO for 16 hr and the extent of induction of PINCR-S1 RNA was assessed by qRT-PCR . ( B ) qRT-PCR for PINCR , the cytoplasmic GAPDH and nuclear MALAT1 was performed from nuclear and cytoplasmic fractions of PINCR-S1 cells treated with DOXO for 16 hr . ( C ) The expression of endogenous PINCR and endogenous PINCR-S1 relative to GAPDH was measured by qRT-PCR from HCT116 cells or PINCR-S1 HCT116 cells untreated or treated with 5-FU for 16 hr . Error bars represent SD from three independent experiments . p<0 . 05 , ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05810 . 7554/eLife . 23244 . 059Figure 8—figure supplement 4 . Over-expression of PINCR in PINCR-S1 cells does not alter PINCR-S1 expression or the induction of PINCR targets . ( A , B ) PINCR-S1 HCT116 cells were transfected with pCB6 or pCB6-PINCR for 48 hr and then left untreated or treated with 5-FU for 16 hr . The expression of PINCR , PINCR-S1 , the PINCR targets BTG2 , GPX1 and RRM2B and the negative control p21 , was measured by qRT-PCR normalized to GAPDH . Error bars represent SD from three independent experiments . p<0 . 05 , *p<0 . 005 , ##p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 05910 . 7554/eLife . 23244 . 060Figure 8—figure supplement 5 . p53 knockdown in PINCR-S1 cells . PINCR-S1 HCT116 cells were reverse transfected with CTL or p53 siRNAs for 48 hr and immunoblotting was performed from whole cell lysates after treating the transfected cells with 5-FU for 16 hr . GAPDH was used as loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 06010 . 7554/eLife . 23244 . 061Figure 8—figure supplement 5—source data 1 . p53 immunoblot for Figure 8—figure supplement 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 23244 . 061 Streptavidin pulldowns from the PINCR-S1 expressing cells treated with 5-FU revealed >10-fold enrichment of the PINCR-S1 RNA ( Figure 8B ) and specific enrichment of Matrin 3 protein ( Figure 8C and Figure 8—source data 1 ) . To determine if PINCR-S1 associates with the p53RE of BTG2 , GPX1 and RRM2B , we performed streptavidin pulldowns from formaldehyde-crosslinked parental HCT116 cells ( negative control ) and the PINCR-S1 cells after 5-FU treatment . We found that the p53RE of each of these three genes but not the p21 p53RE , was specifically enriched in the PINCR-S1 pulldowns ( Figure 8D ) . Finally , we examined the association of PINCR-S1 with the p53RE and enhancer regions of PINCR targets in the presence and absence of p53 or Matrin 3 . To do this , we knocked down p53 ( Figure 8—figure supplement 5 and Figure 8—figure supplement 5—source data 1 ) or Matrin 3 with siRNAs and performed streptavidin pulldowns . As compared to the p53RE in the PINCR promoters , in CTL siRNA transfected cells treated with 5-FU , we found stronger association of PINCR-S1 with the enhancers of the PINCR targets ( Figure 8E ) . Silencing Matrin 3 or p53 resulted in dramatic reduction of association of PINCR-S1 with these enhancers and p53REs ( Figure 8E ) . Because p53 is important for PINCR expression , it is likely that the reduced association of PINCR-S1 to these regions after p53 knockdown is due to lack of expression . On the other hand , the observed loss in association of PINCR-S1 to the enhancers and promoters upon knockdown of Matrin 3 indicates that Matrin 3 recruits PINCR-S1 to these regions . Taken together , these results suggest that a p53-Matrin 3-PINCR complex associates with the p53RE and enhancers of BTG2 , GPX1 and RRM2B and plays a critical role in modulating the induction of these genes after DNA damage . In this study , we report the first functional characterization of PINCR , an intergenic nuclear lncRNA , strongly induced by p53 after DNA damage . Several p53-regulated lncRNAs have been recently identified and shown to play important roles in the p53 network . However , PINCR is unique from these recently characterized p53-regulated lncRNAs . Firstly , following p53 activation , the p53-regulated lncRNAs LED ( Léveillé et al . , 2015 ) and Linc-475 ( Melo et al . , 2016 ) regulate G1 arrest and prevent entry of cells into mitosis . However , PINCR-KO cells show a defect in G1 arrest but the cells arrest in the G2 phase after DNA damage . Secondly , the p53-regulated lncRNAs lincRNA-p21 ( Dimitrova et al . , 2014 ) , LED ( Léveillé et al . , 2015 ) and Linc-475 ( Melo et al . , 2016 ) regulate the levels of p21 . In addition , in a recent study , the p53-induced lncRNA DINO , was shown to directly bind to and regulate p53 levels ( Schmitt et al . , 2016 ) . However , PINCR does not alter p53 or p21 levels but instead regulates the expression of the p53 targets BTG2 , GPX1 and RRM2B that also regulate G1 arrest after DNA damage . Our study together with other recent studies shows that specific RNA-binding proteins and transcription factors play an important role in mediating the effects of a lncRNA . For example , in the context of p53 activation , lincRNA-p21 interacts with hnRNP-K and functions as a coactivator for p53-dependent p21 transcription ( Dimitrova et al . , 2014; Huarte et al . , 2010 ) . PANDA , another p53-regulated lncRNA upstream of p21 , associates with the transcription factor NF-YA to regulate the expression of pro-apoptotic genes during genotoxic stress ( Hung et al . , 2011 ) . The data presented here indicates that PINCR and Matrin 3 act as coactivators of p53 on a subset of p53 targets . It is known that Matrin 3 interacts with enhancer regions ( Romig et al . , 1992; Skowronska-Krawczyk et al . , 2014 ) , and our data shows that the induction of these genes may be mediated by chromatin looping between Matrin 3 bound to the enhancer regions of these genes and p53 bound to the p53RE in their promoters . PINCR recruits Matrin 3 to enhancers of PINCR-dependent p53 target genes . Future studies on the identification of genome-wide-binding sites of Matrin 3 and p53 and epigenetic marks in PINCR-WT and PINCR-KO cells in the absence or presence of DNA damage will be important . Interestingly , similar intrachromosomal interactions containing enhancer activity have been reported recently and shown to express enhancer RNAs ( eRNAs ) that are required for efficient transcriptional enhancement of interacting target genes and induction of a p53-dependent cell-cycle arrest ( Léveillé et al . , 2015; Melo et al . , 2013 ) . The development of new approaches to identify targets of endogenous lncRNAs is an active area of investigation and remains a major challenge in the lncRNA field . We developed a new approach in which we knocked-in an S1 tag at the 3'end of PINCR using CRISPR/Cas9 and determined the association of PINCR-S1 with the p53RE of specific p53 targets by qPCR following streptavidin pulldowns from crosslinked cells . Our results show a Matrin-3-dependent association of PINCR-S1 with the enhancer region of BTG2 , RRM2B and GPX1 and indicate a direct and specific role of PINCR in regulating these genes in response to DNA damage . Given the strong interaction between the S1 tag and streptavidin and studies utilizing transfected S1-tagged RNAs ( Vasudevan and Steitz , 2007 ) or in vitro transcribed S1-tagged RNAs ( Butter et al . , 2009; Iioka et al . , 2011; Srisawat and Engelke , 2001 , 2002 ) for the identification of interacting RNA-binding proteins or miRNAs , this method has the potential to identify the genome-wide targets of PINCR and other lncRNAs . In summary , our study suggests that PINCR is an important modulator of gene expression in the p53 pathway that regulates the induction of a subset of p53 targets and this effect is mediated in part via its interaction with Matrin 3 . Future investigations on PINCR in normal cells and in an expanded panel of cell lines will enhance our understanding of its role in tumorigenesis and tumor progression . The colorectal cancer cell lines HCT116 ( ATCC Number: CCL-247 ) , SW48 ( ATCC Number: CCL-231 ) and RKO ( ATCC Number: CRL-2577 ) and HEK293T ( ATCC Number: CRL-11268 ) cells were purchased from ATCC . The isogenic p53-WT and p53-KO HCT116 , RKO and SW48 were previously generated by Bert Vogelstein’s lab ( Johns Hopkins University ) . All cell lines were maintained in Dulbecco's modified Eagle's medium ( DMEM ) ( Thermo Fisher Scientific ) supplemented with 10% fetal bovine serum ( Thermo Fisher Scientific ) and 1% penicillin-streptomycin at 37°C , 5% CO2 . All cell lines were routinely checked for mycoplasma using the Venor Gem Mycoplasma detection kit ( Sigma-Aldrich , Catalog # MP0025-1KT ) . Cells were treated with 10 µM Nutlin-3 ( Skelleckchem , Catalog # S1061 ) , 300 nM Doxorubicin ( DOXO , Catalog # D1515 ) , 100 µM 5-Fluorouracil ( 5-FU; Calbiochem , Catalog # 343922 ) or 400 ng/ml NCS ( Sigma-Aldrich , Catlog#N9162 ) for the indicated time . The Allstars Negative ( CTL ) siRNAs were purchased from Qiagen and siRNAs for p53 ( SMARTpool siRNAs , Catalog # L-003329–00 ) , BTG2 ( I-Catalog # J-012308–06 and II-Catalog # J-012308–07 ) , GPX1 ( I-Catalog # J-008982–05 and II-Catalog # J-008982–07 ) , RRM2B ( I-Catalog # J-010575–05 and II-Catalog # J-010575–06 ) and Matrin 3 ( Catalog # J-017382–05 and J-017382–07 ) were purchased from Dharmacon . CTL-ASO and PINCR-ASO were designed and provided by Ionis Pharmaceuticals ( Supplementary file 5 ) . All siRNA and ASO transfections were performed by reverse transfection at a final concentration of 20 nM and 50 nM , respectively , using Lipofectamine RNAiMAX ( Life technologies ) as directed by the manufacturer . For gene expression analysis after PINCR or Matrin 3 knockdown , all the reverse transfections were performed for 48 hr followed by 24 hr DOXO or 5-FU or Nutlin-3 treatment . Total RNA from cell lines was isolated using RNeasy mini kit ( Qiagen ) . For qRT-PCR analysis , 500 ng total RNA was reverse-transcribed using iScript Reverse Transcription kit ( Bio-Rad ) , and qPCR was performed using Fast SYBR Green Master Mix ( Life technologies ) per the manufacturer’s instructions . Primer sequences are detailed in Supplementary file 5 . Nuclear and cytoplasmic extracts were prepared from HCT116 cells expressing PINCR or PINCR-S1 , PINCR-WT and PINCR-KO , untreated or treated with DOXO or 5-FU for 16 hr or as indicated in figure legend , using Digitonin as previously described ( Lal et al . , 2004 ) . RNA was isolated from cytoplasmic and nuclear fractions using Trizol reagent ( Invitrogen ) following the manufacturer’s protocol . Strand-specific genomic coordinates for all exons of human PINCR and NEAT1 , GAPDH , SDHA and UBC genes were downloaded from the UCSC Genome Browser ( GRCh37/hg19 ) in BED format . A Multiz alignment of 46 vertebrates aligned to GRCh37/hg19 ( http://hgdownload . soe . ucsc . edu/goldenPath/hg19/multiz46way/maf/ , in MAF format ) was downloaded separately for each gene based on the extracted coordinates for mature transcript accordingly to the UCSC annotation and uploaded to Galaxy ( https://usegalaxy . org/ ) . FASTA alignments were generated for each mature transcript separately using «reformat» and «concatenate» options in Galaxy for overlapping list of 29 mammals specified by the PhyloCSF phylogeny ( http://mlin . github . io/PhyloCSF/29mammals . nh . png ) . PhyloCSF was applied to generated FASTA alignments for assessing the coding potential ( the Codon Substitution Frequencies score - CSF ) of mature transcripts and individual exons of analyzed genes . The CSF score assigns a metric to each codon substitution observed in the input alignment based on relative frequency of that substitution in known coding and non-coding regions . The following parameters were used for analysis: PhyloCSF 29mammals input fasta_file --orf=ATGStop --frames=6 removeRefGaps --aa –allScores . Comparative analysis of all possible reading frames and estimation of the potential to encode any recognizable protein domains was created by BLASTX . Multiple alignments for complete PINCR mature transcripts and promoter regions were built using the Muscle program with default parameters ( Edgar , 2004 ) . Genome rearrangements were analyzed using the Owen program for pair-wise alignments ( Ogurtsov et al . , 2002 ) . CRISPR/Cas9-mediated deletion of PINCR gRNAs targeting the 5’ and 3’ ends of PINCR were designed using Zifit software ( http://zifit . partners . org/ZiFiT/ ) and were cloned into U6-gRNA vector ( Moriarity et al . , 2014 ) having BsmB1 restriction enzyme site . gRNAs sequence information is provided in Supplementary file 6 . gRNA oligos were ligated and phosphorylated using T4 ligation buffer ( NEB ) and T4 Polynucleotide Kinase ( NEB ) using a thermocycler with following parameter: 37°C for 30 min , 95°C for 5 min and then ramp down to 25°C at 5 °C/min . Annealed oligos were ligated with BsmB1 digested U6-gRNA vector ( 2 . 9 kb fragment ) using quick DNA ligase ( NEB ) . Ligation mix was transformed into E . coli DH5-alpha chemical competent cells and transformants were sequenced to confirm the presence of gRNAs . The efficiency of gRNAs was tested in HEK293T cells by transfecting Cas9 with the gRNAs . CRISPR-mediated PINCR-KO HCT116 or SW48 cells were then generated using piggyBac co-transposition method as previously described ( Moriarity et al . , 2014 ) . Cells were cotransfected with 2 μg each of hpT3 . 5Cagg5-FLAG-hCas9 and the 5′ and 3′ PINCR gRNAs cloned in U6-gRNA vector in addition to the 500 ng each of pcDNA-pPB7 transposase and pPBSB-CG-LUC-GFP ( Puro ) ( +CRE ) transposon vector using Lipofectamine 2000 . After 48 hr , transfected cells were treated with puromycin and incubated at 37°C for 1 week . Cells were then seeded at one cell per well in 96-well plates with puromycin containing DMEM media . Wells that produced single colonies were expanded and DNA was extracted . Clones were then genotyped for deletion of PINCR using standard PCR genotyping ( PINCR deletion analysis primer in Supplementary file 5 ) . Identified wild-type ( WT ) clones were used as controls . PCR products were sequenced to confirm the deletion of PINCR genomic locus . Also , total RNA was extracted from individual clones , with and without treatment with DOXO or 5-FU , and expression of PINCR was analyzed using qRT-PCR . Animal protocols were approved by the National Cancer Institute Animal Care and Use Committee following AALAAC guidelines and policies . PINCR-WT#1 , PINCR-KO#1 and PINCR-KO#2 cells were untreated or treated with 100 µM 5-FU for 4 hr , following which the drug was washed-off and fresh medium was added . After a 4 hr recovery , live cells were counted with trypan blue exclusion assays and equal numbers of live cells were injected for each sample . Cells ( 1 × 106 ) were mixed with 30% matrigel in PBS on ice , and the mixture was injected into the flanks of 6- to 8 week-old female athymic nude mice ( Animal Production Program , Frederick , MD ) ( each group N = 10 ) . Tumor volume was measured twice a week after 1 week of injection . To evaluate the effect of proliferation and/or apoptosis in the tumors , the xenograft tumors were collected from four PINCR-WT and PINCR-KO tumors and fixed in 10% neutral buffered formalin ( Sigma , St . Louis , MO ) . Paraffin sectioning , hematoxylin and eosin staining ( H and E ) , Ki67 staining and Cleaved Caspase-3 staining were performed by Histoserv , Inc ( Gaithersburg , MD ) . The following antibodies were used for immunohistochemistry staining: anti-Ki67 ( Abcam , Catalog # Ab16667 ) and anti-Cleaved Caspase-3 ( Cell Signaling , Catalog # 9661 ) . The images were acquired at 40× magnification . Full-length fragment of PINCR was PCR amplified from a pCB6 vector expressing full length PINCR using a forward primer containing the T7-promoter sequence at its 5’end and a gene-specific reverse primer ( Supplementary file 5 ) . The control luciferase cDNA was generated from vector pRL-TK ( Addgene ) linearized by BamH1 digestion . We then performed in vitro transcription to generate biotinylated PINCR ( Bi-PINCR ) and the control luciferase ( Bi-LUC ) RNAs using MEGAscript in vitro transcription kit ( Ambion ) and biotin RNA labeling mix ( Roche ) . The in vitro transcribed RNA was purified with RNeasy mini kit ( Qiagen ) . The biotinylated RNA was run on the bioanalyzer to check the quality . Nuclear extracts were prepared from HCT116 cells untreated or treated with DOXO for 24 hr as described above . The nuclear lysate was resuspended in RIP buffer ( 150 mM KCl , 25 mM Tris pH 7 . 4 , 0 . 5 mM DTT , 0 . 5% NP40 , 1 mM PMSF and protease Inhibitor ) and sonicated three times for 5 s and centrifuged at 14 , 000 x g at 4°C for 30 min . The nuclear lysate was precleared by incubation with Dynabeads M-280 Streptavidin ( Thermo Fisher Scientific ) for 4 hr at 4°C . In parallel , 40 µl Dynabeads were blocked with 1 mg BSA ( company ) and 50 µg tRNA ( company ) for 4 hr at 4°C . Twenty-five pmole Bi-PINCR or Bi-LUC RNA was incubated with 1 mg precleared nuclear lysate prepared above for 4 hr at 4°C . The biotinylated RNA-protein complexes were pulled down by incubation with preblocked Dynabeads for overnight at 4°C . Interacting proteins were fractionated by SDS-PAGE and each lane cut into 10 slices . The protein bands were then in-gel digested with trypsin ( Thermo ) overnight at 37°C . The peptides were extracted following cleavage and lyophilized . The dried peptides were solubilized in 2% acetonitrile , 0 . 5% acetic acid , 97 . 5% water for mass spectrometry analysis . They were trapped on a trapping column and separated on a 75 µm x 15 cm , 2 µm Acclaim PepMap reverse phase column ( Thermo Scientific ) using an UltiMate 3000 RSLCnano HPLC ( Thermo Scientific ) . Peptides were separated at a flow rate of 300 nl/min followed by online analysis by tandem mass spectrometry using a Thermo Orbitrap Fusion mass spectrometer . Peptides were eluted into the mass spectrometer using a linear gradient from 96% mobile phase A ( 0 . 1% formic acid in water ) to 55% mobile phase B ( 0 . 1% formic acid in acetonitrile ) over 30 min . Parent full-scan mass spectra were collected in the Orbitrap mass analyzer set to acquire data at 120 , 000 FWHM resolution; ions were then isolated in the quadrupole mass filter , fragmented within the HCD cell ( HCD normalized energy 32% , stepped ±3% ) , and the product ions analyzed in the ion trap . Proteome Discoverer 2 . 0 ( Thermo ) was used to search the data against human proteins from the UniProt database using SequestHT . The search was limited to tryptic peptides , with maximally two missed cleavages allowed . Cysteine carbamidomethylation was set as a fixed modification , and methionine oxidation set as a variable modification . The precursor mass tolerance was 10 ppm , and the fragment mass tolerance was 0 . 6 Da . The Percolator node was used to score and rank peptide matches using a 1% false discovery rate . For PINCR tagging , we used integration by Non-homologous end joining , which was accomplished by introducing a simultaneous double-strand break in genomic DNA and in the targeting vector at the 5’ of the S1-tag ( Brown et al . , 2016 ) . Plasmids encoding spCas9 and sgRNAs were obtained from Addgene ( Plasmids #41815 and #47108 ) . Oligonucleotides for construction of sgRNAs were obtained from Integrated DNA Technologies , hybridized , phosphorylated and cloned into the sgRNA plasmid or targeting vector using BbsI sites ( Brown et al . , 2017 ) . Target sequences for sgRNAs are provided in Supplementary file 6 . We prepared the targeting vector by first synthesizing two complementary oligonucleotides ( IDT ) with the sequence of the S1 tag followed by the sequence of RP3-326I13 . 1 located at the 3’ of the sgRNA-binding site , which potentially contains the native elements for termination of transcription . The oligonucleotides were dimerized , phosphorylated and cloned into the targeting vector using T4 ligase . Subsequently , we introduced at the 5’ of the S1 tag the sequence targeted by the RP3-326I13 . 1 sgRNA . The targeting vector also contained an independent PuroR expression cassette driven by a PGK promoter for facile isolation of clonal populations of cells that integrate the plasmid within the genome . The sequence of the targeting vector is provided in Figure 8—figure supplement 2 . HCT116 cells were transfected with 300 ng Cas9 , 300 ng of sgRNA and 300 ng of targeting vector using Lipofectamine 2000 ( Invitrogen ) according to the manufacturer’s instructions in 24-well plates . Three days after transfection , the cells were selected with 0 . 5 µg/ml Puromycin to generate clonal populations . Genomic DNA from each clone was isolated using DNEasy Blood and Tissue Kit ( Qiagen ) . PCRs to detect integration of the targeting vector at the target site were performed using KAPA2G Robust PCR kits ( Kapa Biosystems ) according to the manufacturer’s instructions . A typical reaction contained 20–100 ng of genomic DNA in Buffer A ( 5 µl ) , Enhancer ( 5 µl ) , dNTPs ( 0 . 5 µl ) , primers forward ( PINCR Det FP , 1 . 25 µl ) and reverse ( Targeting vector Det RP , 1 . 25 µl ) and KAPA2G Robust DNA Polymerase ( 0 . 5 U ) . The DNA sequences of the primers for each target are provided in Supplementary file 5 . PCR products were visualized in 2% agarose gels and images were captured using a ChemiDoc-It2 ( UVP ) . The PCR products were cloned into TOPO-TA cloning ( ThermoFisher ) and sequenced . For immunoprecipitation experiments , HCT116 clonal cell lines expressing PINCR or PINCR-S1 RNA were treated with 5-FU for 24 hr to induce PINCR expression . 2 × 107 cells were lysed in lysis buffer ( 150 mM NaCl , 50 mM Tris-HCl pH 7 . 5 , 0 . 5% Triton X-100 , 1 mM PMSF , Protease inhibitor cocktail and RNase inhibitor ) . Lysates were sonicated three times for 5 s and centrifuged at 14 , 000 x g at 4°C for 30 min . For IP , 500 µg of cellular extract was incubated overnight at 4°C with 25 µl Dynabeads M-280 Streptavidin ( Thermo Fisher Scientific ) . Beads were washed twice with high salt buffer ( 0 . 1% SDS , 1% Triton-X-100 , 2 mM EDTA , 20 mM Tris-HCl pH 8 and 500 mM NaCl ) followed by low salt buffer ( 0 . 1% SDS , 1% Triton-X-100 , 2 mM EDTA , 20 mM Tris-HCl pH 8 and 150 mM NaCl ) and TE buffer ( 10 mM Tris-HCl pH 8 and 2 mM EDTA ) . Bound proteins were eluted by boiling the samples for 5 min in SDS-PAGE sample buffer . Eluted proteins were subjected to SDS-PAGE and immunoblotting with Matrin 3 ( Bethyl labs ) or β-Tubulin ( Cell Signaling , Catalog # 2146S ) . Enrichment of PINCR RNA levels in the pulldown material was evaluated by directly adding Trizol to the beads , followed by RNA extraction and qRT-PCR . To test the binding of PINCR to the chromatin , PINCR and PINCR-S1 cells were treated with 5-FU for 24 hr to induce PINCR expression . Chromatin was cross-linked with 1% formaldehyde , and cells were lysed and sonicated in Buffer B ( 1% SDS , 10 mM EDTA , 50 mM Tris-HCl pH 8 , Protease inhibitor and RNase inhibitor ) . RNA-DNA-protein complexes were immunoprecipitated with Dynabeads M-280 Streptavidin , overnight using IP buffer ( 0 . 01% SDS , 0 . 5% Triton-X-100 , 1 . 2 mM EDTA , 16 . 7 mM Tris-HCl pH 8 and 167 mM NaCl ) . Beads were washed twice with high-salt buffer followed by TE buffer . Bound RNA-DNA-protein complexes were eluted from the beads using elution buffer ( 100 mM NaCl , 50 mM Hepes pH 7 . 4 , 0 . 5% NP40 , 10 mM MgCl2 and 5 mM Biotin ) , at room temperature for 20 min . Eluted material was incubated at 65°C for 2 hr ( 200 mM NaCl ) to reverse crosslink the bound proteins . The samples were treated with Proteinase K and eluted DNA was column purified ( Qiagen ) and analyzed by qPCR using primers flanking the p53-binding sites of different genes ( Supplementary file 5 ) . To test if PINCR binding to the chromatin is p53 and/or Matrin-3-dependent , PINCR-S1 cells were reverse transfected with CTL siRNAs , p53 siRNAs or two independent Matrin 3 siRNAs . After 48 hr , the cells were treated with 5-FU for 24 hr . The enrichment of PINCR-S1 at the promoter and enhancer regions of PINCR targets was determined by ChIP-qPCR followed by streptavidin pulldown as described above . For lncRNA profiling HCT116 , SW48 and RKO cells were untreated or treated with Nutlin-3 in duplicate for 8 hr . Total RNA was isolated using the RNeasy Mini kit ( Qiagen ) and hybridized to Affymetrix HT2 . 0 arrays that contain probes for ~11 , 000 lncRNAs . To identify the PINCR-regulated transcriptome PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU ( 100 uM ) for 24 hr . RNA samples were prepared as described above in triplicates and labeled using the IlluminaTotalPrep RNA amplification kit ( Ambion ) and microarrays were performed with the HumanHT-12 v4 Expression BeadChip kit ( Illumina ) . After hybridization , raw data were extracted with Illumina GenomeStudio software . Raw probe intensities were converted to expression values using the lumi package in Bioconductor with background correction , variance stabilization and quantile normalization . Differential expression between different conditions was computed by an empirical Bayes analysis of a linear model using the limma package in Bioconductor . Adjusted p-values were calculated with the Benjamini and Hochberg method , and differentially expressed genes were selected with adjusted p-value≤0 . 05 and a fold change ≥1 . 50 . All the microarray data for this study has been deposited in GEO . The Accession number is GSE90086 and the URL is https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE90086 . The unpublished RNA-seq data ( Li et al . , unpublished ) used in this study has been deposited in GEO . The Accession number is GSE79249 and the URL is https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE79249 . A 2 kb region upstream of the first exon of PINCR was PCR amplified ( primer sequences in Supplementary file 5 ) using 100 ng genomic DNA from HCT116 cells and inserted into upstream of Firefly luciferase of pGL3 luciferase vector ( Promega ) . To measure PINCR promoter activity , HCT116 cells were co-transfected with 100 ng of pGL3-empty vector or pGL3 expressing the PINCR promoter , along with pCB6-empty vector or pCB6 expressing p53 , and 10 ng pRL-TK expressing Renilla luciferase . After 48 hr , luciferase activity was measured using the dual-luciferase reporter system ( Promega ) . A 2 kb PINCR promoter region ( chrX: 43 , 034 , 255–43 , 036 , 255 ) with ( WT-p53RE ) and without ( Δp53RE ) the 20 bp p53RE ( GCCCTTGTCTGGACATGCCC ) was synthesized in pGL3 luciferase vector by GenScript . HCT116 cells were co-transfected with 100 ng of pGL3 expressing the WT or Δp53RE PINCR promoter and 10 ng pRL-TK expressing Renilla luciferase . After 48 hr , cells were left untreated or treated with DOXO for 24 hr and luciferase activity was measured using the dual-luciferase reporter system ( Promega ) . For cell cycle analysis , 3 . 0 × 105 PINCR-WT and PINCR-KO cells were seeded per well of a 6-well plate . After 24 hr , cells were untreated or treated with 300 nM DOXO or 100 µM 5-FU or 400 ng/ml NCS or 10 µM Nutlin-3 and the samples were collected at the indicated times . Cells were fixed with ice-cold ethanol for 2 hr and stained with propidium iodide ( Sigma ) in the presence of RNase A ( Qiagen ) . Cell cycle profiles were captured using FACS Calibur flow cytometer ( BD Biosciences ) , and the data were analyzed using FlowJo software ( FloJo , LLC ) . To perform cell cycle analysis after Matrin 3 knockdown , PINCR-WT and PINCR-KO cells were reverse transfected with siCTL and two independent Matrin 3 siRNAs using RNAiMAX at a final siRNA concentration of 20 nM . After 48 hr , cells were untreated or treated with 300 nM DOXO or 100 µM 5-FU and cell cycle profiles were captured as described above . For cell cycle analysis after BTG2/GPX1/RRM2B knockdown , PINCR-WT HCT116 cells were reverse transfected with siCTL and two individual siBTG2 , siGPX1 and siRRM2B using RNAiMAX at a final siRNA concentration of 20 nM . After 48 hr , cells were untreated or treated with 100 µM 5-FU and cell cycle profiles were captured as described above . Cell cycle analysis after PINCR knockdown using ASOs , HCT116 cells were reverse transfected with 50 nM CTL-ASO or PINCR-ASO . After 48 hr , cells were untreated or treated with 300 nM DOXO or 100 µM 5-FU and cell cycle profiles were captured as described above . For caspase-3 immunostaining , 3 . 0 × 105 PINCR-WT and PINCR-KO cells were seeded per well of a six-well plate . After 24 hr , cells were untreated or treated with DOXO for 72 hr and fixed with 4% paraformaldehyde for 10 min and permeabilized by 0 . 5% Triton X-100 for 10 min . Fixed cells were stained for 1 hr with primary antibodies anti-Mab414 ( Covance , Catalog # MMS120P ) for nuclear envelope and active caspase-3 ( Cell Signaling , Catalog # 9661S ) for apoptotic cells , followed by further staining with DAPI ( blue ) and secondary antibodies , anti-mouse 586 ( orange; Alexa Fluor 586 goat anti-mouse IgG , Life Technology , Catalog # A11031 ) and anti-rabbit 488 ( green; Alexa Fluor 488 donkey anti-rabbit IgG , life Technology , Catalog # A21206 ) for 1 hr . Images were taken by Ziess immunofluorescence microscope with x63 lens . For colony formation on plastic , 3 × 105 PINCR-WT and PINCR-KO HCT116 cells were seeded per well in six-well plates . After 24 hr , cells were untreated or treated with 100 nM or 300 nM DOXO or 10 µM , 50 µM or 100 µM 5-FU for 4 hr , following which the drug was washed-off and fresh medium was added . After a 4 hr recovery , cells were seeded in a 12-well plate at a density of 500 cells per well . After 2 weeks , colonies were fixed with ice-cold 100% methanol for 5 min , stained with crystal violet , and colonies were counted . For colony formation after ASO transfections , HCT116 cells were transfected with CTL-ASO and PINCR-ASO . After 48 hr , cells were untreated or treated with 100 µM 5-FU for 4 hr , following which the drug was washed-off and fresh medium was added . After a 4 hr recovery , cells were seeded in a 12-well plate at a density of 500 cells per well . After 2 weeks , colonies were fixed as described above . For long-term cell proliferation assays on plastic , 3 × 105 PINCR-WT and PINCR-KO SW48 cells were seeded in 12-well plate . After 24 hr , cells were untreated or treated with 100 µM 5-FU for 7 days . After 7 days , cells were fixed with ice-cold 100% methanol for 5 min and stained with crystal violet . The 2 . 2 kb transcript corresponding to PINCR RNA ( NR_110387 . 1 ) was cloned into pCB6 vector using EcoR1/Xba1 restriction enzyme cloning sites . PINCR-KO cells were transfected with pCB6 empty vector ( EV ) or pCB6 vector expressing PINCR . After 48 hr , the cells were treated with neomycin and incubated at 37°C for 4–5 days , for selection of stably transfected cells . Total RNA was extracted from the pool of cells and expression of PINCR was analyzed using qRT-PCR . For cell cycle analysis , 3 . 0 × 105 PINCR-KO cells expressing pCB6-EV or pCB6-PINCR were seeded per well in six-well plates . After 24 hr , cells were untreated or treated with 300 nM DOXO and the samples were collected at indicated times . FACS analysis was performed as described above . To determine the effect of PINCR overexpression on cell cycle and gene regulation , HCT116 cells expressing PINCR or PINCR-S1 were transfected with pCB6 or pCB6-PINCR expressing PINCR . After 48 hr , cells were treated with DOXO or 5-FU for indicated times . Expression of PINCR and other target genes was analyzed using qRT-PCR and FACS was performed as described above . To measure apoptosis after 5-FU treatment PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU ( 100 µM ) for 48 hr . Similarly , to determine the levels of p53 and/or p21 or phospho-Rb in PINCR-WT and PINCR-KO cells , the cells were untreated or treated with 5-FU for 24 hr . Whole-cell lysates were prepared using radioimmunoprecipitation ( RIPA ) buffer containing protease inhibitor cocktail ( Roche ) . Proteins were quantified using the bicinchoninic acid protein quantitation ( BCA ) kit ( Thermo Scientific ) . For immunoblotting , 20 µg whole cell lysate per lane was loaded onto a 12% SDS-PAGE gel , transferred to nitrocellulose membrane and immunoblotted with anti-Cleaved PARP ( Cell Signaling , Catalog # 2541 ) , anti-p53 ( DO-1 ) ( Santa Cruz , Catalog # sc-126 ) , anti-p21 ( Santa Cruz , Catalog # sc-397 ) , anti-Histone H3 ( Cell Signaling , Catalog # 4620 ) , anti-phospho-Rb ( Cell Signaling , Catalog # 9307P , 9208P , 9301P ) and anti-GAPDH ( Cell Signaling , Catalog # 14C10 ) . To determine the p53 and Matrin 3 knockdown efficiency PINCR-WT cells were reverse transfected with CTL siRNAs or p53 siRNAs or Matrin 3 siRNAs ( 20 nM ) respectively , and 48 hr after transfection cell lysates were prepared using RIPA buffer as described above , followed by immunoblotting with anti-p53 , Matrin 3 ( Bethyl labs Catalog # A300-591A ) or GAPDH antibodies . For co-immunoprecipitation experiments HCT116 cells were untreated or treated with 5-FU ( 100 µM ) for 24 hr and whole cell lysates were prepared in RIPA buffer and centrifuged at 14 , 000 x g at 4°C for 30 min . For IP , 25 µl Pierce protein A/G magnetic beads ( Thermo Scientific , Catalog # 88802 ) were incubated with 2 µg Matrin 3 antibody or IgG control ( Santa Cruz , Catalog # sc-2027 ) , for 4 hr at 4°C . Following this , 500 µg cellular extract was incubated for 4 hr at 4°C with A/G magnetic beads pre-coated with IgG or Matrin 3 . Beads were washed 5 times at 4°C with RIPA buffer and samples were untreated or treated with RNase A and DNase for 30 min at 37°C . Bound proteins were eluted by boiling the samples for 5 min in SDS-PAGE sample buffer . Eluted proteins were subjected to SDS-PAGE and immunoblotting using anti-p53 DO-1 antibody . To perform reciprocal IP nuclear lysates were prepared from HCT116 cells as described before and Immunoprecipitation was done as discussed above . IgG control ( Santa Cruz , Catalog # sc-2025 ) and DO-1 p53 antibodies were used for IP and anti-Matrin 3 antibody for immunoblotting . To determine the association of PINCR to Matrin 3 in intact cells , 2 × 107 HCT116 cells were treated with 5-FU ( 100 µM ) for 24 hr and then cross-linked with 1% formaldehyde . Crosslinked cells were resuspended in Buffer B ( 1% SDS , 10 mM EDTA , 50 mM Tris-HCl pH 8 , Protease inhibitor cocktail and RNAse inhibitor ) , followed by sonication . An aliquot of the sonicated cell lysates was subjected to IP using 2 µg IgG or Matrin 3 antibodies for 4 hr at 4°C on protein A/G magnetic beads , using IP buffer ( 0 . 01% SDS , 1 . 1% Triton X-100 , 1 . 2 mM EDTA , 16 . 7 mM Tris-HCl pH 8 , 167 mM NaCl ) . The IP material was washed twice with high salt buffer ( 0 . 1% SDS , 1% Triton-X-100 , 2 mM EDTA , 20 mM Tris-HCl pH 8 and 500 mM NaCl ) followed by TE buffer ( 10 mM Tris-HCl pH 8 and 2 mM EDTA ) . Bound RNA-protein complexes were eluted from the beads using elution buffer ( 0 . 1% SDS , 0 . 1M NaHCO3 , RNase inhibitor ) , at 37°C for 15 min followed by reverse cross-linking at 65°C for 2 hr by with 200 mM NaCl . Matrin 3 bound RNAs were isolated by phenol-chloroform extraction ( Ambion ) followed by ethanol precipitation and qRT-PCR was used to determine the enrichment of p21 ( negative control ) and PINCR in the Matrin 3 IPs . To determine the direct binding of PINCR to recombinant Matrin 3 ( rMatrin 3 ) , 200 ng of in vitro transcribed Bi-PINCR or Bi-LUC RNA was incubated with 500 ng recombinant Matrin 3 protein ( Creative BioMart , Catalog # MATR3-15H ) in 1X EMSA buffer ( 25 mM Tris-HCl pH 7 . 5 , 150 mM KCl , 0 . 1% Triton-X-100 , 100 µg/ml BSA , 2 mM DTT and 5% glycerol ) at room temperature for 2 hr . RNA–protein complex was immunoprecipitated at room temperature for 2 hr , by using Dynabeads M-280 Streptavidin . Beads were washed five to six times with 1X EMSA buffer without glycerol and bound material was subjected to SDS-PAGE and immunoblotting for Matrin3 . The following antibodies were used: anti-p53 ( DO-1 ) 1: 1000 dilution from Santa Cruz Biotechnology; anti-Matrin 3 1:1000 dilution from Bethyl Laboratories; anti-Cleaved PARP and anti-GAPDH at 1:1000 dilution from Cell Signaling . Chromatin IP was performed with the Active Motif ChIP kit ( Active Motif , Carlsbad , CA , USA ) as directed by the manufacturer . Briefly , 5 × 107 HCT116 cells grown in 15-cm plates were untreated or treated with DOXO ( 300 nM ) for 16 hr . Similarly , PINCR-WT and PINCR-KO cells were untreated or treated with 5-FU ( 100 µM ) for 24 hr . Chromatin was cross-linked with 1% formaldehyde , and cells were lysed and sonicated . Protein–DNA complexes were immunoprecipitated with control IgG or anti-p53 ( DO1 ) ( Santa Cruz ) or anti-Matrin 3 ( Bethyl labs ) antibody . The IP material was washed and heated at 65°C overnight to reverse crosslinks . ChIP DNA was column purified ( Qiagen ) and analyzed by qPCR . Primers flanking the p53 binding sites or the enhancer regions of different genes are listed in Supplementary file 5 . To test if association of p53 to the promoter and enhancer regions is Matrin-3-dependent 5 × 107 PINCR-WT cells were reverse transfected with CTL siRNAs and two independent Matrin 3 siRNAs . After 48 hr , cells were treated with 5-FU for 24 hr and enrichment of p53 at the promoter and enhancer regions of PINCR targets was determined by ChIP-qPCR as described above . Integrative Genome Browser ( IGV , software . broadinstitute . org/software/igv/ ) was used to download and visualize relevant ChIP-seq datasets from the ENCODE consortium data repository . Motif identification at CTCF peaks surrounding putative enhancer-gene pairs within insulated neighborhoods was gathered from the HOMER software package ( http://homer . ucsd . edu/homer/ngs/ ) . Genomic locations of p53 response elements were determined similarly from HOMER p53 motif datasets . Chromatin folding was inferred from 3D contact matrices calculated from in situ HiC data ( Rao et al . , 2014 ) and visualized using the Juicebox desktop application ( Durand et al . , 2016 ) .	Though DNA contains the information needed to build the proteins that keep cells alive , only 2% of the DNA in a human cell codes for proteins . The remaining 98% is referred to as non-coding DNA . The information in some of these non-coding regions can still be copied into molecules of RNA , including long molecules called lncRNAs . Little is known about what lncRNAs actually do , but growing evidence suggests that these molecules are important for a number of vital processes including cell growth and survival . When the DNA in an animal cell gets damaged , the cell needs to decide whether to pause growth and repair the damage , or to kill itself if the harm is too great . One of the best-studied proteins guiding this decision is the p53 protein , which increases the number of protein-coding genes needed to carry out either option in this decision . That is to say that , p53 regulates the genes needed to kill the cell and the genes needed to temporarily pause its growth and repair the damage , which instead keeps the cell alive . So , how does the p53 protein guide the decision , and are lncRNA molecules involved ? Using human colon cancer cells , Chaudhary et al . now report that when DNA is damaged , the levels of a specific lncRNA increase 100-fold . Further experiments showed that this lncRNA – named PINCR , which refers to p53-induced noncoding RNA – promotes the survival of cells . Chaudhary et al . showed that PINCR molecules do this by recruiting a protein called Matrin 3 to a certain region in the DNA called an enhancer and then links it to promoter region in the DNA of specific genes that temporarily pause cell growth but keep the cell alive . This in turn activates these ‘pro-survival genes’ . In further experiments , when the PINCR molecules were essentially deleted , p53 was not able to fully activate these genes and as a result more of the cells died . Together these findings increase our knowledge of how lncRNAs can work , especially in the context of DNA damage in cancer cells . A next important step will be to uncover other roles for the PINCR molecule in both cancer and healthy cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cancer", "biology" ]	2017	Prosurvival long noncoding RNA PINCR regulates a subset of p53 targets in human colorectal cancer cells by binding to Matrin 3
Biological homeostasis invokes modulatory responses aimed at stabilizing internal conditions . Using tunable photo- and mechano-stimulation , we identified two distinct categories of homeostatic responses during the sleep-like state of Caenorhabditis elegans ( lethargus ) . In the presence of weak or no stimuli , extended motion caused a subsequent extension of quiescence . The neuropeptide Y receptor homolog , NPR-1 , and an inhibitory neuropeptide known to activate it , FLP-18 , were required for this process . In the presence of strong stimuli , the correlations between motion and quiescence were disrupted for several minutes but homeostasis manifested as an overall elevation of the time spent in quiescence . This response to strong stimuli required the function of the DAF-16/FOXO transcription factor in neurons , but not that of NPR-1 . Conversely , response to weak stimuli did not require the function of DAF-16/FOXO . These findings suggest that routine homeostatic stabilization of sleep may be distinct from homeostatic compensation following a strong disturbance . Sleep architecture—the duration , timing , and order of individual stages of sleep—is derived from a combination of internal timekeeping pathways , a drive towards an appropriate baseline ( sleep pressure ) , and external constraints . Collectively , the use of both mammalian and non-mammalian models has suggested that sleep is phylogenetically ancient and evolutionarily conserved ( Campbell and Tobler , 1984; Sehgal and Mignot , 2011; Nelson and Raizen , 2013 ) . The key behavioral hallmarks of sleep are episodic reduced motion , reversibility , typical postures , sensory gating , and homeostasis ( Campbell and Tobler , 1984 ) . Generally , the homeostatic drive underlies correlations between the strength and duration of a disruption and the subsequent duration and quality of sleep . Behavioral signatures of homeostasis include faster time-courses of wake-to-sleep transitions , prolonged periods of sleep , and increased arousal thresholds following a period of deprivation that increases sleep pressure ( Moses et al . , 1975; Tobler , 1983; Hendricks et al . , 2000; Allada and Siegel , 2008; Raizen et al . , 2008 ) . The nematode Caenorhabditis elegans is the simplest model organism that has been shown to exhibit a sleep-like state to date ( Raizen et al . , 2008; Nelson and Raizen , 2013; Cho and Sternberg , 2014 ) . The 2–3 hr period of lethargus , a developmental stage that precedes the termination of each larval stage , is characterized by behavioral quiescence , a cessation of feeding , reduced or delayed responses to external stimuli , a distinct posture , and compensation following deprivation ( Van Buskirk and Sternberg , 2007; Raizen et al . , 2008; Schwarz et al . , 2012; Iwanir et al . , 2013; Cho and Sternberg , 2014 ) . The C . elegans homolog of the circadian clock protein PERIOD is required for synchronization of lethargus , and its mRNA levels track the developmental/molting cycle ( Jeon et al . , 1999; Allada et al . , 2001; Tennessen et al . , 2006; Monsalve et al . , 2011 ) . Additional conserved signaling pathways that exhibit functional similarities in mammalian , insect , and nematode sleep include the epidermal growth factor ( EGF ) ( Kramer et al . , 2001; Snodgrass-Belt et al . , 2005; Foltenyi et al . , 2007; Van Buskirk and Sternberg , 2007; Zimmerman et al . , 2008 ) , the cyclic GMP-dependent protein kinase PKG ( Van Buskirk and Sternberg , 2007; Raizen et al . , 2008; Langmesser et al . , 2009 ) , cAMP-dependent signaling ( Hendricks et al . , 2001; Graves et al . , 2003; Raizen et al . , 2008 ) , Gs signaling , and genes acting downstream of dopamine signaling ( Singh et al . , 2014 ) . Homeostatic regulation within C . elegans lethargus was previously examined by manually depriving the animals of quiescence . After a deprivation period of 30 min during lethargus , the onset of long response latencies to chemical stimuli was accelerated . In addition , mechanical stimulation for 60 min at the time that the onset of lethargus was expected resulted in increased subsequent peak quiescence ( Raizen et al . , 2008 ) . Recently , quiescent behavior and homeostatic rebound were also seen when a sleep-like state was induced anachronistically in adult animals , suggesting that developmental factors are not essential for neuromodulation during C . elegans sleep ( Cho and Sternberg , 2014 ) . C . elegans can locomote forward or backward by propagating dorsoventral body bends from anterior to posterior or vice versa , respectively . Alternatively , they move in a variety of non-directional manners collectively referred to as dwelling ( Gray et al . , 2005; von Stetina et al . , 2006; Gallagher et al . , 2013; Gjorgjieva et al . , 2014 ) . During lethargus , C . elegans prominently exhibit quiescence—the complete absence of dynamic muscle contraction . Alternating bouts of locomotion and quiescence comprise the simple architecture of C . elegans sleep ( Raizen et al . , 2008; Iwanir et al . , 2013 ) . In a previous study , we have shown that the durations of these bouts are correlated ( Iwanir et al . , 2013 ) , but the mechanisms underlying this process of routine stabilization were not examined . In this study , we analyze the behavioral responses of sleeping nematodes under undisturbed , weakly disturbed , and strongly disturbed conditions . To do so , we continuously assayed the locomotion of C . elegans from the mid fourth intermolt stage ( L4int ) , through the fourth lethargus stage ( L4leth ) , and into the mid young adult stage ( YA ) . We found that weak photo- or mechano-stimulation transiently skewed the dynamics of bouts while preserving the characteristic pairwise correlations . Thus , under unperturbed or weakly perturbed conditions , homeostatic compensation manifested as a transient extension of quiescence bouts ( and shortening of motion bouts under some conditions ) in response to prolonged motion . This form of compensation under low noise conditions , termed micro-homeostasis ( Iwanir et al . , 2013; Nelson and Raizen , 2013 ) , required the function of the neuropeptide Y ( NPY ) receptor homolog , NPR-1 . In contrast , strong stimuli induced a qualitatively different homeostatic response: the animals moved continuously for several minutes , after which quiescence monotonically returned to its baseline level . Compensation for the motion induced by a strong stimulus manifested as an upshift in the baseline fraction of time spent in quiescence , rather than a transient extension of quiescence bouts . The homeostatic responses to strong stimuli required the function of the DAF-16/FOXO in neurons ( see also Driver et al . , 2013 ) but not the function of NPR-1 . Conversely , micro-homeostasis was not abolished in daf-16 mutants . In addition , we show that neuropeptidergic signaling is not strictly required for maintaining high levels of mean quiescence during lethargus . The loss of function of UNC-31/CAPS , a calcium-dependent activator protein required for dense core vesicle exocytosis ( Avery et al . , 1993; Charlie et al . , 2006 ) , resulted in a minor reduction of overall quiescence . In contrast , quiescence was strongly suppressed by the loss of the subsets of mature neuropeptides that were processed by the EGL-3 proprotein convertase or the EGL-21 carboxypeptidase E ( CPE ) ( Kass et al . , 2001; Jacob and Kaplan , 2003; Husson et al . , 2006 , 2007 ) . As previously suggested ( Stawicki et al . , 2013 ) , this apparent discrepancy can be resolved: collectively , our data indicate that a balance between inhibitory and excitatory contributions from different peptides modulates the duration of bouts of quiescence . Our findings support a model in which locomotion during lethargus is coupled to a measure of increased sleep pressure . Quiescence serves to ameliorate this pressure and homeostatic regulation dynamically maintains an appropriate quiescence baseline . Interestingly , the homeostatic routine stabilization of motion and quiescence in low-noise environments is mechanistically distinct from homeostatic responses following strong , stressful , disruptions . To our knowledge , the analysis presented here is the first to identify this distinction . Homeostatic regulation of lethargus was previously examined using manually delivered strong mechanical stimuli , after which baseline levels of responsiveness were regained in 4 min ( Raizen et al . , 2008 ) . However , even undisturbed animals compensate for spontaneous prolonged motion with prolonged quiescence during lethargus ( Iwanir et al . , 2013 ) . Therefore , a mechanism that dynamically stabilizes lethargus behavior may be invoked by motion in quiet or weakly noisy environments . If so , weak stimuli should transiently skew the bout architecture by elongating motion bouts and causing a subsequent ( compensatory ) extension of quiescence bouts . To test this , we first exposed wild-type animals at the fourth intermolt larval stage , L4int , to pulses of blue light of intensities ranging from 0 . 3–100 mW/cm2 and measured their responses using the frame subtraction method . In brief , this method consists of digitally recording the behavior of the animals and assessing the levels of motion and quiescence based on the number of pixels that change their brightness between consecutive frames ( see Nagy et al . , 2014 ) . The observed responses depended on the light intensity and the duration of the stimulus , and we determined that a 15 s pulse of light at an intensity of 20–40 mW/cm2 evoked weak , reproducible responses ( Figure 1—figure supplements 1 , 2 ) . Interestingly , we noted that 5 s pulses failed to produce a sharp response specifically during lethargus . This suggested that the animals were less responsive during lethargus and that reduced responsiveness could be assayed separately from delayed responsiveness ( Raizen et al . , 2008 ) . The response of L4int larvae and post lethargus young adult ( YA ) animals to weak blue light stimuli consisted of elevated levels of locomotion , which persisted for 15–25 s after the end of the pulse , followed by a 2 min decline back to baseline locomotion levels . In contrast , during L4leth the average level of locomotion crossed its baseline 1 min after it peaked , proceeded to fall below it for 2–3 additional min ( p < 0 . 01 ) , and only then stabilized at baseline levels ( Figure 1A ) . The transient trough in locomotion resulted from an increase in the fraction of time the animals were quiescent rather than from slower motion ( Figure 1B ) . 10 . 7554/eLife . 04380 . 003Figure 1 . Motion plays a causal role in determining the duration of subsequent quiescence during lethargus . ( A ) Wild-type animals at the mid L4int , late L4int , L4leth , and YA stages were exposed to 30 s light stimuli at an intensity of 20 mW/cm2 . All stimuli were initiated at t = 0 . Outside lethargus , locomotion monotonically decayed to baseline levels in 2 min . During lethargus , the peak in locomotion was followed by a trough prior to returning to baseline . Insets: the responses during lethargus shown on a semi-log scale . ( B ) The fractions of quiescence were calculated for 1 min intervals centered at the times of the peak and trough of the L4leth responses , as well as for their respective pre-stimulus baselines . Plots and bars depict mean ± s . e . m obtained from datasets of N = 40–50 animals per condition . Asterisks indicate p < 0 . 001 . ( C ) Survival curves of quiescence and motion bouts of wild-type animals exposed to a 30 s , 20 mW/cm2 , blue light stimulus during the first hour of L4leth . Bouts were identified using the frame subtraction method and control data were obtained from the same animals , but 8 min after the stimulus ( non-stimulated control animals were also assayed , analyzed the same way , and found to be indistinguishable from this control group ) . Mean ± s . e . m , N > 200 bouts for each condition . ( D ) The dynamics of bouts obtained from a posture-based analysis following a 15 s , 20 mW/cm2 , blue light stimulus . Left and right panels correspond to the first and second halves of L4leth , respectively . See also Figure 1—figure supplements 1–3 . Plots depict mean ± s . e . m , smoothed using a 30 s running window average . N = 40 animals . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 00310 . 7554/eLife . 04380 . 004Figure 1—figure supplement 1 . Calibration of weak blue light stimuli . Responses to blue light stimuli were defined as the peak value of overall motion , as measured using the frame subtraction method , normalized by the baseline average motion during a 1-min period prior to the stimulus . Left: L4int larvae were exposed to 15 s pulses of blue light at different intensities . Locomotion responses increased as a function of the light intensity in the 2–100 mW/cm2 range . The plot depicts mean ± s . e . m responses , N = 20–30 animals . Shaded area emphasizes the range of stimuli used throughout the rest of the manuscript . Right: the total amount of motion induced by the stimulus was defined as the area under the response peak . Three durations of a 20 mW/cm2 light stimulus were assayed ( N > 200 trials per condition , error bars depict s . e . m , p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 00410 . 7554/eLife . 04380 . 005Figure 1—figure supplement 2 . Responses to weak light stimuli . Wild-type animals at the mid L4int , late L4int , L4leth , and YA stages were exposed to 5 ( left ) , 15 ( middle ) , and 30 s ( right ) light stimuli at an intensity of 20 mW/cm2 . All stimuli were initiated at t = 0 . A sharp peak in locomotion in response to a 5 s stimulus was observed outside lethargus , but not during lethargus . Both 15 s and 30 s stimuli evoked a transient increase in locomotion . Outside lethargus , locomotion monotonically decayed to baseline levels in 2 min . In contrast , during lethargus the peak in locomotion was followed by a trough prior to returning to baseline . Insets: the responses during lethargus shown on a semi-log scale . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 00510 . 7554/eLife . 04380 . 006Figure 1—figure supplement 3 . Responses during quiescence and motion . The fraction of quiescence during 1 min intervals centered at a time point prior to the onset of the stimulus ( baseline ) , at the peak of the locomotion response , and at the trough of the response . Dark or light bars correspond to data obtained from instances when the animals were quiescent or motile , respectively , at the time of the stimulus onset . Responses assayed using the frame subtraction method appeared indistinguishable based on the type of bout at the time of stimulus onset . Error bars depict s . e . m . Plots and bars in all panels depict mean ± s . e . m obtained from datasets of N = 40–50 animals per condition . Asterisks indicate p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 006 The presence of a weak light stimulus terminated bouts of quiescence prematurely and extended bouts of motion ( Figure 1C , p < 0 . 01 in both cases ) . Identical responses were observed whether the onset of the stimulus interrupted a bout of quiescence or motion ( Figure 1—figure supplement 3 ) . The increase in the fraction of time spent in quiescence after the stimulus was removed could have been caused by an extension of quiescence bouts , shortening of motion bouts , or both . To distinguish between these possibilities , we turned to an accurate and computationally intensive behavioral analysis . This previously described approach was based on continuous measurements of the dynamics of body posture at high temporal and spatial resolutions ( Iwanir et al . , 2013; Nagy et al . , 2014 ) . Using this analysis and a 15 s weak light stimulus , we measured the durations of bouts of motion and quiescence after the stimulus was turned off . We found that , during the first half of lethargus , the compensatory response was comprised of an increase and a decrease in the durations of quiescence and motion bouts , respectively ( Figure 1D , p < 0 . 01 ) . During the second half of lethargus , a compensatory increase in the durations of quiescence bouts was still observed . Taken together , these findings revealed that motion during non- or weakly-interrupted lethargus , but not during the L4int or YA stages , caused a compensatory transient increase in quiescence . Posture-based analysis allowed for improved measurements of pairwise correlations between durations of bouts of motion and subsequent bouts of quiescence in undisturbed animals ( Figure 2A , R = 0 . 47 ± 0 . 03 , p < 0 . 001 ) ( Iwanir et al . , 2013 ) . This approach revealed that these correlations gradually decayed as lethargus progressed ( Figure 2A ) . Moreover , it enabled us to compare groups of motion bouts that contained different qualities of motion despite having similar overall durations . We could thus address the question of whether vigorous or directed motion in and of itself might affect the subsequent bout of quiescence . 10 . 7554/eLife . 04380 . 007Figure 2 . Vigorous or directed motion extends the duration of subsequent quiescence during lethargus . ( A ) Posture-based analysis improved the measurement of pairwise correlations between the durations of motion bouts and those of subsequent quiescence bouts in undisturbed wild-type animals ( R = 0 . 47 ± 0 . 03 , N = 3609 bouts from 40 animals , p < 0 . 05 ) . As a guide to the eye , motion bouts were grouped according to their durations in 2 s wide bins . The mean ± s . e . m duration of the subsequent quiescence bouts for each bin was plotted and these mean values were fitted to a line . In addition , pairwise correlation coefficients were calculated for each 15 min interval of L4leth separately . As a guide to the eye , linear fits to the binned data are depicted . In all cases , the errors were defined as the 95% confidence intervals and the number of bouts is given in parentheses . ( B–C ) The overall levels of motion ( B ) and the fraction of directed motion ( C ) during a motion bout have a significant effect on subsequent quiescence . Overall motion was defined as the mean time derivative of the absolute values of 18 angles along the body and directed motion was defined as either forward or backward locomotion , as opposed to dwelling ( Nagy et al . , 2014 ) . Left: the median values of the overall vigor of motion ( B ) and the fraction of directed motion ( C ) as a function of the duration of the motion bouts ( binned in 2 s bins ) . Middle ( right ) : the durations of motion ( quiescence ) bouts calculated separately for the group of bouts that was above or below the median of its respective bin . The durations of quiescence bouts differed significantly between the two groups . N = 40 animals , error bars depict s . e . m , p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 007 To compare between groups of motion bouts of equal durations , we binned the bouts recorded during the first 90 min of L4leth of non-stimulated wild-type animals in 2 s wide bins . For each bin , we calculated the median vigor of locomotion as measured by the rate of change of body-curvature ( Figure 2B , left panel ) . The motion bouts ( from each bin ) were then separated into two groups: those exhibiting higher-than-median or lower-than-median vigor with respect to their bin of origin . Each of the two groups therefore contained bouts of all durations and , importantly , the average duration of a motion bout was the same in both groups ( Figure 2B , middle panel ) . Having controlled for the mere durations , we found a significant effect of the level of locomotion on the duration of the subsequent quiescence bout ( Figure 2B , right panel ) . A similar analysis , performed exclusively on bouts that contained directed motion , considered the separation between the two groups based on the fraction of the bout spent in directed motion and produced similar results ( Figure 2C ) . We thus conclude that enhanced or directed locomotion during a bout of a given duration positively affects subsequent quiescence . Dwelling behavior during a motion bout appears similar to dwelling behavior outside of lethargus . However , although locomotory responses were shown to be delayed during lethargus ( Raizen et al . , 2008; Cho and Sternberg , 2014 ) , they were not previously examined in detail . We asked whether responses to a weak stimulus during a bout of motion were distinct from responses during a bout of quiescence , from responses outside of lethargus , or from both . To examine behavioral responses to external perturbations throughout this study , we used a recurrent stimulus assay: animals were repeatedly exposed to a stimulation regime of brief , widely spaced , photo- or mechano-stimuli . The duration of each individual stimulus was 0 . 4 or 15 s , depending on the type of assay , and the spacing between consecutive stimuli was 15 min . Animals were continuously assayed for 10 hr from the L4int stage to the mid YA stage . A diagram outlining the design of these assays is depicted in Figure 3A . Since multiple 15-min cycles were aligned and averaged , the resulting data had periodic boundaries . For instance , the same 1-min period could be referred to as the 15th minute after the stimulus or the 1 min just prior to the stimulus ( see Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 04380 . 008Figure 3 . A posture-based analysis of locomotion responses to weak light stimuli . ( A ) A diagram describing the repeated stimulus assay , in which a generic brief stimulus ( vertical lines ) was repeatedly delivered at 15 min intervals ( long horizontal arrows ) . Each assay started at the mid L4int stage , continuously progressed through L4leth ( shaded area ) , and ended at the mid YA stage . For the purpose of illustration , the blue and red lines symbolize tentative probabilities of forward locomotion and quiescence , respectively . Baseline behavior was measured during the 5-min period starting 10 min after a stimulus , or equivalently , 5 min prior to the subsequent stimulus . The beginning of the first baseline period is depicted by a dashed vertical line . ( B ) The fraction of forward locomotion , backward locomotion , dwelling , and quiescence before , during , and after a weak ( 15 s , 20 mW/cm2 blue light ) stimulus provided at the L4int ( left ) , L4leth ( middle ) , and YA ( right ) stages . A compensatory post-stimulus enhancement of quiescence , as well as enhanced reversals during the stimulus , and a rising propensity for forward locomotion after the stimulus was turned off were uniquely observed during lethargus . Insets: the fraction of forward locomotion before and after the offset of the stimulus ( top ) and the fraction of backward locomotion before and after the onset of the stimulus . Shading denotes the presence of the light stimulus . All fractions were calculated from the 7 . 5-s period ( half of the duration of the stimulus ) , the scale bars represent a fraction of 0 . 5 , and asterisks denote p < 0 . 05 . ( C ) The data from the middle panel of ( B ) plotted separately for the first , second , and third hours of L4leth . Enhanced quiescence was observed in all three cases , although it was less prominent during the third hour . Plots in panels ( B , C ) depict mean ± s . e . m and the number of stimuli assayed is noted in parentheses for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 00810 . 7554/eLife . 04380 . 009Figure 3—figure supplement 1 . The averaged behavior data have periodic boundaries . Each sketched closed curve represents the average fraction of quiescence as a function of time for a particular set of experimental conditions . Since repeated 15-min cycles were aligned and averaged , the mean behavioral dynamics have periodic boundaries . Graphically , this can be represented by plotting the data on a circular time axis , where the onset of the stimulus is denoted by an asterisk . During lethargus , the immediate response to a stimulus lasted no more than 3 min . After this short-term response was complete , the fraction of quiescence returned to a steady state that was characteristic of the conditions of the experiment . The fraction of time spent is quiescence during this steady state was defined as the baseline level of quiescence for the relevant experimental conditions . In this framework , baselines were compared between different conditions . Left: responses to strong stimuli . The immediate response to a strong stimulus is depicted by the yellow curve and the baselines for periodically stimulated and undisturbed animals are depicted by brown and grey curves , respectively . Elevation of the steady state quiescence—as compared to the mean quiescence of undisturbed animals—constituted the sustained response to the presence of repeated strong stimuli . This elevation is emphasized by a double-headed arrow . Right: responses to weak stimuli . The immediate response to a weak stimulus is depicted by the orange curve and the baselines for periodically stimulated and undisturbed animals are depicted by blue and grey curves , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 009 In our hands , during lethargus , the transient changes in behavioral dynamics that constituted the short-term response to a stimulus were limited to a period of 3 min immediately following stimulation . After this short-term response was complete , behavioral dynamics returned to a steady state characteristic of the conditions of the experiment . Consequently , baseline behavior for each set of experimental conditions was defined as the steady state measured during a 5 min period starting 10 min after a stimulus and 5 min prior to the subsequent stimulus ( labeled explicitly in Figure 3A and Figure 3—figure supplement 1 , and depicted as t = −5…0 min in subsequent panels ) . Outside of lethargus , the onset of a weak light stimulus evoked a sharp rise in the propensity for forward locomotion , while backward locomotion was suppressed ( L4int ) or unchanged ( YA ) . From the time of the offset of the stimulus , forward locomotion monotonically declined and reversals returned to baseline levels ( L4int ) or were briefly elevated ( YA ) . The probability of forward locomotion decayed to its baseline value in 3 min as the baseline balance between directed motion and dwelling was re-established ( Figure 3B left and right panels ) . During L4leth , the onset of the stimulus evoked a sharp rise in the propensities for both forward and backward locomotion . The offset of the stimulus did not reverse the increasing propensity for moving forward . Rather , forward locomotion persisted for 20 s after the light was turned off and subsequently fell below its steady state value while quiescence levels exceeded their baseline ( Figure 3B middle panel ) . Similar features were observed for responses throughout L4leth ( Figure 3C ) , and regardless of whether the onset of the stimulus occurred during a motion or a quiescence bout ( data not shown ) . Thus , responses to weak stimuli revealed similar locomotory responses during bouts of motion and quiescence and differentiated both types of bouts from the L4int and YA stages . The compensatory extension of quiescence bouts after a weak light stimulus was distinct from previously reported responses to manually delivered strong mechanical stimuli ( Raizen et al . , 2008; Driver et al . , 2013 ) . To test whether the modulation of bout duration was specifically evoked by light , we assayed animals that were exposed to a mechanical stimulus: vibrations at a frequency of 1 kHz ( Nagy et al . , 2014 ) . The strength of the stimulus was tuned by varying its duration . Outside lethargus , a 0 . 4 s stimulus elicited a transient increase in reversals followed by a brief enhancement of the propensity for forward locomotion , while a 15 s stimulus elicited a similar initial recoil followed by an enhancement of forward locomotion that lasted for 10 min ( Figure 4A , B , left ) . We thus refer to the short stimulus as weak and the longer stimulus as strong . 10 . 7554/eLife . 04380 . 010Figure 4 . A posture-based analysis of locomotion responses to weak and strong mechanical stimuli . ( A ) A weak mechanical stimulus ( 0 . 4 s of 1 kHz vibrations ) produced a reversal followed by a small elevation of forward locomotion in L4int larvae ( left ) and a brief reversal followed by enhanced quiescence during L4leth ( right ) . Inset: the first quiescence bout after the stimulus was longer than subsequent bouts ( p < 0 . 05 ) . ( B ) A strong mechanical stimulus ( 15 s of 1 kHz vibrations ) produced reversals followed by a prolonged ( 10 min ) elevation of forward locomotion in L4int larvae ( left ) and a brief reversal followed by elevated levels of directed motion for 4–5 min during L4leth ( right ) . Notably , quiescence returned to its baseline value without transiently exceeding it . ( C ) Mean baseline fraction of quiescence was measured during the baseline period ( see Figure 3A ) . The baseline fraction of quiescence was significantly higher in strongly stimulated animals as compared to unstimulated and weakly stimulated animals . Weak light I and II labels refer to stimulus strengths of 20 and 40 mW/cm2 blue light , respectively . Plots in panels ( A , B ) depict mean ± s . e . m and the error bars in panel ( C ) depict ±s . e . m and asterisks denote p < 0 . 05 . The number of stimuli assayed is noted in parentheses for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 010 During L4leth , weak mechanical stimuli induced transient backward locomotion , followed by enhanced quiescence . Specifically , the first bout of quiescence after the recoil was elongated , and the architecture of locomotion and quiescence returned to baseline 1 min after the stimulus was delivered ( Figure 4A , right ) . In contrast , the strong stimulus disrupted the architecture of behavior during lethargus: it was followed by several minutes of enhanced motion and a monotonous relaxation to baseline quiescence levels . Upon return to baseline , quiescence bouts were not transiently extended such that a peak in quiescence was not observed . However , compensation took on a different form: the overall level of baseline quiescence was elevated ( Figure 4B , C ) . This overall elevation of quiescence was consistent with previously reported compensation after strong stimulation ( Raizen et al . , 2008; Driver et al . , 2013 ) . These results suggested that there were two regimes of disruption and compensation . Weak perturbations resulted in a transient modulation of bout durations that did not disrupt ( and could even enhance ) the characteristic correlations of the bouts architecture . In contrast , a strong perturbation abrogated the routine dynamics of bouts for several minutes and increased the baseline fraction of quiescence thereafter . Neuropeptide Y ( NPY ) and its receptors have been implicated in the regulation of sleep ( albeit in different manners ) in humans , rats , fruit flies , and nematodes ( Antonijevic et al . , 2000; Tóth et al . , 2007; Dyzma et al . , 2010; Van den Pol , 2012; Choi et al . , 2013; Nagy et al . , 2014 ) . To test their role in mediating micro-homeostasis , we assayed animals carrying two mutant alleles of the C . elegans NPY receptor homolog gene , npr-1 . In our hands , overall quiescence in animals carrying the npr-1 ( ky13 ) allele , a glutamine to ochre nonsense mutation at codon 61 , was only mildly different from wild-type ( Nagy et al . , 2014 ) . However , bout correlations in these mutants were significantly reduced . The npr-1 ( ad609 ) allele induced similarly reduced bout correlations and a more pronounced defect in the durations of quiescence bouts throughout lethargus ( Figure 5A ) . 10 . 7554/eLife . 04380 . 011Figure 5 . NPR-1 is required for micro-homeostasis but not for homeostatic responses to strong stimuli . ( A ) Undisturbed behavior of npr-1 mutants . Left: the fraction of quiescence of wild-type animals and npr-1 ( ad609 ) mutants during L4leth ( shaded area ) . The fraction of quiescence of npr-1 mutants was recently published ( Nagy et al . , 2014 ) and plotted here for comparison . Plots depict mean ± s . e . m , the numbers of animals assayed are denoted in parentheses . Middle and right: pairwise bout correlations and plots of binned bouts ( see Figure 2A for details ) . Pairwise correlations were significantly reduced in npr-1 mutants ( p < 0 . 05 ) . All correlations are given with 95% confidence intervals and error bars depict ±s . e . m . The number of bouts in each case is denoted in parentheses . ( B ) L4int , late L4int , L4leth , and YA npr-1 mutants were exposed to weak ( 15 s , 20 mW/cm2 light ) stimuli . All stimuli were initiated at t = 0 . In npr-1 mutants assayed using frame subtraction , a trough did not follow the transient increase in locomotion before returning to baseline . Insets: the responses during lethargus shown on a semi-log scale . For each strain , the quiescence fraction was calculated during 1 min intervals centered at the times of the peak and trough of the L4leth responses , as well as for their respective pre-stimulus baselines . Quiescence was not enhanced following the peak in locomotion in npr-1 mutants . Plots and bars depict mean ± s . e . m obtained from datasets of N = 50–60 animals per condition . Asterisks and double asterisks denote p < 0 . 05 and p < 0 . 01 , respectively . ( C ) A posture-based analysis of behavior of L4leth npr-1 mutants: the fraction of forward locomotion , backward locomotion , dwelling , and quiescence before , during , and after a weak ( 15 s , 20 mW/cm2 , blue light ) stimulus . The data were aligned by the time of the onset of the stimulus and then averaged . Plots depict mean ± s . e . m . In agreement with the frame subtraction measurements , the compensatory enhancement of quiescence fraction shortly after the stimulus was nearly abolished in npr-1 mutants . N = 14 and 13 animals ( ky13 and ad609 ) . ( D ) A posture-based analysis of bout dynamics of npr-1 mutants following a weak stimulus . N = 14 and 13 animals , plots depict mean ± s . e . m , smoothed using a 30 s running window . ( E ) The mean baseline fractions of quiescence during the 5 min intervals prior to each stimulus tested . Similar to wild-type , baseline quiescence fraction was significantly higher in strongly stimulated animals as compared to non-stimulated and weakly stimulated npr-1 mutants . See also Figure 1—figure supplements 1–3 . Error bars depict ±s . e . m and asterisks denote p < 0 . 05 . The number of stimuli assayed is noted in parentheses for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 011 We next assayed the responses of npr-1 mutants to weak blue light stimuli . Using the frame subtraction method , we could not detect significant compensation following the excess motion induced by the stimulus in either of the two mutant strains ( Figure 5B ) . The posture-based analysis confirmed their severe defect in modulation of bout durations ( Figure 5C , D ) . The overall activity and , in particular , the initial response of npr-1 mutants to the weak stimulus were similar to wild-type . This indicated that the mutants were not defective in sensing the stimulus or in their locomotory capabilities but specifically in their ability to compensate for a weak disturbance . In contrast , npr-1 mutants exhibited wild-type-like compensation following strong mechanical stimuli: when animals carrying either of the two mutant alleles were exposed to a strong mechanical stimulus , their baseline fraction of quiescence was elevated as compared to non stimulated or weakly stimulated animals ( Figure 5E ) . Thus , in addition to the phenotypic differences described above , homeostatic compensation during undisturbed or weakly disturbed lethargus was affected by NPR-1 , while homeostatic compensation for strong stimuli was not . We concluded that the routine stabilization of lethargus behavior in low-noise environments and the homeostatic compensation for stressful disturbances were mechanistically separable . These findings are consistent with a model in which NPR-1 modulates quiescence during lethargus in response to spontaneous or induced mild variations in locomotion . NPR-1 is a predicted neuropeptide receptor and the FMRFamide-like neuropeptides encoded by flp-18 and flp-21 were shown to be two of its ligands . ( De Bono and Bargmann , 1998; Kubiak et al . , 2003; Rogers et al . , 2003; Kim and Li , 2004 ) . We therefore asked whether peptidergic release from dense core vesicles ( DCVs ) was required for micro-homeostasis . To answer this question , we assayed the loss of function of UNC-31 , the sole C . elegans ortholog of mammalian calcium-dependent activator protein for secretion ( CAPS ) required for DCV exocytosis ( Avery et al . , 1993; Charlie et al . , 2006 ) . To confirm that the observed phenotype was explained by the mutation of interest , we tested a strong loss of function allele , unc-31 ( e169 ) , and a putative null allele , unc-31 ( e928 ) , ( Charlie et al . , 2006; Speese et al . , 2007 ) . Under undisturbed conditions , the quiescence bouts of unc-31 mutants were shorter than wild-type , but the overall amount of quiescence was only weakly reduced in these mutants ( Figure 6A , B ) . Moreover , unc-31 mutants did not exhibit paralysis or anachronistic quiescence outside of lethargus and their overall locomotory behavior during lethargus was similar to wild-type ( Nagy et al . , 2013 and data not shown ) . Nevertheless , pairwise correlations between subsequent bouts in these mutants were abolished ( Figure 6B ) . This could indicate that the absence of a group of functional neuropeptides impaired the dynamic extension of quiescence bouts in response to variations in durations and compositions of motion bouts . Alternatively , the quiescence bouts of unc-31 mutants may be too short to sustain detectable correlations . We favor the first explanation for two reasons . First , the Hawaiian strain ( a wild isolate of C . elegans ) exhibited quiescence bouts that were comparable in duration to those of unc-31 mutants but nevertheless maintained wild-type correlations during minutes 45–120 from the onset of L4leth ( Figure 6—figure supplements 1 , 2 ) . Second , when bout pairs containing longer quiescence bouts were excluded from the wild-type dataset , such that the mean duration of the remaining quiescence bouts equaled that of unc-31 mutants , the pairwise correlation between the remaining bouts was reduced to R = 0 . 2 ± 0 . 03 , but not abolished . 10 . 7554/eLife . 04380 . 012Figure 6 . UNC-31/CAPS is not required for establishing a high fraction of quiescence during lethargus but is required for micro-homeostasis . ( A ) Left: the fraction of quiescence of wild-type animals and unc-31 , egl-3 , and egl-21 mutants during L4leth ( shaded area ) . Quiescence was strongly reduced by the loss of function of EGL-3 or EGL-21 , but not UNC-31 . Right: the mean durations of bouts of quiescence of the same wild-type and mutant animals during the 15-min period of L4leth . Plots and bars depict mean ± s . e . m , the numbers of animals assayed are denoted in parentheses . ( B ) Pairwise bout correlations and plots of binned bouts in undisturbed animals ( see Figure 2A for details ) . Pairwise correlations were abolished in unc-31 , egl-3 , and egl-21 mutants . All correlations are given with 95% confidence intervals ( p < 0 . 05 ) and error bars depict ±s . e . m . The number of bouts in each case is denoted in parentheses . ( C ) A posture-based analysis of behavior of L4leth unc-31 mutants: the fraction of forward locomotion , backward locomotion , dwelling , and quiescence before , during , and after a weak ( 15 s , 20 mW/cm2 , blue light ) stimulus . See also Figure 6—figure supplements 1 , 2 . ( D ) A posture-based analysis of bout dynamics of unc-31 mutants following a weak stimulus . The duration of the motion induced by the weak stimulus was shorter than that of wild-type animals , and the compensatory enhancement of quiescence was weaker . N = 11 and 12 animals ( e169 and e928 ) , plots depict mean ± s . e . m , smoothed using a 30 s running window . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 01210 . 7554/eLife . 04380 . 013Figure 6—figure supplement 1 . Micro-homeostasis in undisturbed Hawaiian wild-isolates . Hawaiian animals display overall quiescence similar to those of unc-31 mutants . However , while pairwise correlations between consecutive bouts are abolished in unc-31 mutants , this is not the case in Hawaiian animals ( see Figure 7A–B for details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 01310 . 7554/eLife . 04380 . 014Figure 6—figure supplement 2 . Micro-homeostasis in undisturbed Hawaiian wild-isolates . The durations of bouts of quiescence , as well as pairwise correlations , were measured for Hawaiian animals and unc-31 mutants as a function of time during L4 lethargus . The detailed comparison reveals that the duration of bouts cannot in and of themselves account for the strength of the pairwise correlations ( see Figure 7A for details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 014 We next assayed the responses of unc-31 mutants to weak ( light ) stimuli . Animals carrying the unc-31 ( e928 ) null mutation , as well as animals carrying the unc-31 ( e169 ) loss of function mutation , exhibited a diminished ability to prolong quiescence bouts in response to prolonged motion . The stronger defect was observed in unc-31 ( e928 ) mutants ( Figure 6C , D ) . Collectively , these findings suggest that peptidergic signaling plays a key role in regulating micro-homeostasis . In addition to unc-31 , we assayed mutants in which neuropeptide processing was disrupted due to the loss of function of: ( i ) the proprotein convertase required for preprocessing of many , but not all , neuropeptides , EGL-3 ( Kass et al . , 2001; Husson et al . , 2006 ) , or ( ii ) the carboxypeptidase E ( CPE ) required to complete the processing of the majority of non insulin-like neuropeptides , EGL-21 ( Jacob and Kaplan , 2003; Husson et al . , 2007 ) . Consistent with previous reports ( Turek et al . , 2013 ) , overall quiescence during lethargus was significantly reduced in egl-3 mutants , individual quiescence bouts were very short , and ( as expected ) correlations between bout durations were abolished . The loss of function of EGL-21 resulted in an identical phenotype , demonstrating that the phenotype was caused by the mutations of interest ( Figure 6A , B ) . These results stood in contrast to the mild change in overall quiescence observed in unc-31 mutants , and this apparent discrepancy is discussed below . The FMRFamide-related neuropeptides FLP-18 and FLP-21 were shown to be ligands of NPR-1 , as well as two additional receptors ( Rogers et al . , 2003; Cohen et al . , 2009 ) . In addition , FLP-18 ( but not FLP-21 ) was shown to act synergistically , in an inhibitory fashion , in the homeostatic response to motoneuron imbalance ( see discussion and Stawicki et al . , 2013 ) . In our hands , flp-21 mutants did not exhibit defective micro-homeostasis . We used posture analysis to assay unperturbed flp-18 ( gk3036 ) mutants and the frame subtraction method to assay flp-18 ( gk3036 ) and flp-18 ( db99 ) mutants in the presence of weak perturbations ( Cohen et al . , 2009 ) . The overall quiescence fraction and the durations of quiescence bouts of flp-18 mutants were comparable to those of npr-1 mutants ( Figure 7A and data not shown ) . However , the correlations between subsequent bouts in undisturbed flp-18 ( gk3036 ) mutants were intermediate between the wild-type and npr-1 values: 0 . 33 ± 0 . 06 , 0 . 47 ± 0 . 03 , and 0 . 20 ± 0 . 07 , respectively ( p < 0 . 05 , Figure 7B ) . When stimulated with blue light , both flp-18 alleles were associated with defective compensatory responses , and the defect was more pronounced in flp-18 ( db99 ) mutants ( Figure 7C , D ) . 10 . 7554/eLife . 04380 . 015Figure 7 . FLP-18 plays a role in modulating bout durations in the presence of weak disturbances . ( A ) Posture analysis of undisturbed flp-18 ( gk3063 ) mutants revealed wild-type-like overall quiescence but reduced correlations between subsequent bouts . R = 0 . 33 ± 0 . 06 , N = 12 animals . These correlations were significantly different ( p < 0 . 05 ) from those of wild-type and npr-1 mutants shown in Figures 2A and 5A , respectively . ( B ) Frame subtraction analysis of flp-18 mutants during L4leth in the presence of weak blue light stimuli ( 15 s , 20 mW/cm2 ) . All stimuli were initiated at t = 0 . The dynamics of locomotion revealed defects in the ability of flp-18 mutants to compensate for the motion induced by the stimulus with enhanced quiescence . Left: the locomotion responses during lethargus of each of the two alleles tested and its wild-type control group shown on a semi-log scale . Shaded area denotes mean ± s . e . m . Asterisks denote that during the trough in locomotion , the fraction of quiescence of the mutant allele was significantly lower than that of its respective wild-type control ( p < 0 . 01 ) . Right: for each strain , the quiescence fraction was calculated during 1 min intervals centered at the times of the peak and trough of the L4leth responses , as well as for their respective pre-stimulus baselines . Plots and bars depict mean ± s . e . m obtained from datasets of N = 40–50 animals per condition . Asterisks and double asterisks denote p < 0 . 05 and p < 0 . 01 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 01510 . 7554/eLife . 04380 . 016Figure 7—figure supplement 1 . A fluorescent reporter of FLP-18 in VC motor neurons and head neurons . Top: sample images of the same animal during the late L4int stage ( left ) and the first half of L4leth ( right ) . Arrows point to VC motoneurons in which expression of the Pflp-18::flp-18::SL2::gfp reporter was visibly upregulated . Arrowheads point to head neurons in which changes in expression were not detected . Bottom: the mean fluorescence , before and during lethargus , from reporter expressed in VC motoneurons ( left ) and head neurons ( right ) . When VC neurons fluorescence prior to lethargus was low ( dark grey ) , it increased more than twofold during the first half of lethargus ( p < 0 . 05 ) . When VC neurons fluorescence prior to lethargus was high ( light grey ) , it did not change significantly afterward . The number of animals assayed is denoted in parentheses , error bars depict ±s . e . m , and asterisks denote p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 01610 . 7554/eLife . 04380 . 017Figure 7—figure supplement 2 . Bout correlations in undisturbed flp-18 mutants . Pairwise bout correlations in wild-type animals and flp-18 mutants during the first and second halves of L4leth . The behavior of the mutants differed from wild-type only during the first half of L4leth , corresponding to the period of upregulation of the expression reporter . Error bars depict 95% confidence intervals and asterisks denote p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 017 If FLP-18 plays a role in micro-homeostasis then its production , secretion , or both may be temporally correlated with lethargus . To test this , we examined the temporal dynamics of expression during the L4int and L4leth stages of the Pflp-18::flp-18::SL2::gfp reporter , which contains the upstream promoter region and the entire genomic locus of flp-18 ( Cohen et al . , 2009; Stawicki et al . , 2013 ) . As previously reported , expression was observed in several head and ventral cord ( VC ) neurons . We measured the total GFP fluorescence in head or VC neurons separately . Expression in head neurons was constant prior to the onset of and during L4leth ( Figure 7—figure supplement 1 ) . Surprisingly , the reporter expression in VC neurons differed between two sub-populations of animals . When low levels of fluorescence were initially detected during L4int , reporter fluorescence was enhanced more than twofold during the first half of L4leth . In contrast , initially high fluorescence levels were not further enhanced . Expression levels of the reporter in the VC neurons of the two sub-populations were similar during the second half of L4leth ( Figure 7—figure supplement 1 ) . The absence of a peak in fluorescence during lethargus in the initially strongly fluorescent sub-population may have resulted from non-physiological effects of overexpression . Alternatively , it may be the case that the shift in flp-18 expression or secretion can precede the onset of lethargus or be conditioned on ambient levels during late L4int . Since Pflp-18::flp-18::SL2::gfp expression peaked during the first half of lethargus , we examined the timing of the defect in bout correlations in flp-18 mutants with respect to the onset of lethargus . The positive pairwise bout correlations in flp-18 mutants were found to be smaller than wild-type during the first hour of lethargus , but not during the second hour , corresponding to the observed period of upregulation in expression of the reporter ( Figure 7—figure supplement 2 ) . Collectively , these findings suggest that both FLP-18 and its known receptor , NPR-1 , regulate micro-homeostasis during lethargus . Prolonged and stressful deprivation of quiescence during lethargus causes the translocation of DAF-16 , a FOXO transcription factor that activates stress responses , into the nucleus . Moreover , daf-16 mutants were shown to be defective in their behavioral response to prolonged deprivation ( Lin et al . , 1997; Henderson and Johnson , 2001; Driver et al . , 2013 ) . Although micro-homeostasis responses occur on a timescale that is too short to be regulated by changes in transcription , repeated weak stimuli may still be stressful . To test the roles of DAF-16 in regulating homeostasis during lethargus , we assayed daf-16 ( mu86 ) ( Libina et al . , 2003 ) mutants under no- , weak- , and strong-stimulus conditions . These mutants were similar to wild-type in their total fraction of quiescence , their initial responses to weak stimuli and subsequent compensation , their responses outside of lethargus to weak and to strong stimuli , and their initial responses during lethargus to strong stimuli . When not disturbed , the quiescence bouts of daf-16 mutants were shorter than wild-type ( data not shown ) and their pairwise correlations between subsequent bouts were smaller , but not abolished ( Figure 8A–C ) . A second mutant allele , daf-16 ( mgDf50 ) ( Ogg et al . , 1997 ) , exhibited similar behavior under unstimulated conditions ( data not shown ) . Thus , micro-homeostasis during C . elegans lethargus was mostly independent of DAF-16/FOXO signaling . 10 . 7554/eLife . 04380 . 018Figure 8 . Homeostatic responses to strong stimuli , but not micro-homeostasis , require DAF-16 . ( A ) Left: the fraction of quiescence of wild-type animals and daf-16 mutants during L4leth ( shaded area ) . Plots depict mean ± s . e . m , the numbers of animals assayed are denoted in parentheses . Right: pairwise bout correlations shown with a plot of binned bouts ( see Figure 2A for details ) . Pairwise correlations were reduced in the mutant , although less so than in npr-1 mutants ( p < 0 . 05 ) . All correlations are given with 95% confidence intervals and error bars depict ±s . e . m . The number of bouts in each case is denoted in parentheses . ( B ) A posture-based analysis of responses of L4int and L4leth daf-16 mutants to strong stimuli ( 15 s , 1 kHz vibrations ) : the fraction of forward locomotion , backward locomotion , dwelling , and quiescence before , during , and after the stimulus . ( C ) Left: frame subtraction based analyses of responses of L4leth daf-16 mutants to weak stimuli ( 15 s , 20 mW/cm2 , blue light ) . Inset: the response of daf-16 mutants during L4leth on a semi-log scale . Middle: the fraction of quiescence during 1 min intervals centered at the times of the peak and trough of the L4leth responses , as well as for their respective pre-stimulus baselines . All stimuli were initiated at t = 0 . N = 50–60 animals . Plots and bars depict mean ± s . e . m , asterisks denote p < 0 . 001 . Right: a posture-based analysis of bout dynamics of daf-16 mutants following a weak stimulus . Plots depict mean ± s . e . m , smoothed using a 30 s running window average . N = 12 animals . The compensatory enhancement of quiescence bouts shortly after the stimulus , as assayed by both methods , was similar to wild-type . ( D ) The mean baseline fractions of quiescence of daf-16 mutants in undisturbed animals and in the presence of weak and strong stimuli . In contrast to wild-type , baseline quiescence fraction was indistinguishable between the different conditions . Expression of daf-16 in neurons , but not in body-wall muscles , restored the homeostatic response of daf-16 mutants to strong mechanical stimuli . Error bar depicts ±s . e . m . The number of stimuli assayed is noted in parentheses for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 04380 . 018 In contrast , the homeostatic compensation in our strong stimulus assay was completely abolished in both daf-16 mutants ( Figure 8D ) . To test where the function of daf-16 was required , the function of daf-16 was rescued under the control of the daf-16 native promoter ( Pdaf-16 ) , a pan-neuronal promoter ( Punc-119 ) , and a body-wall muscle promoter ( Pmyo-3 ) ( Driver et al . , 2013 ) . Homeostatic compensation for strong disturbances was restored when daf-16 was expressed under its native promoter or in neurons , but not in muscles ( Figure 7D ) . These findings differ from the reported role of DAF-16 in sleep homeostasis , assayed using response latencies to a noxious chemical , where rescue in muscles but not in neurons restored wild-type-like latencies ( Driver et al . , 2013 ) . However , adult locomotion quiescence in daf-2 mutants ( an insulin/IGF-1 receptor homolog ) was dependent on the function of DAF-16 in neurons ( Gaglia and Kenyon , 2009 ) . Thus , DAF-16 may act in multiple tissues to regulate different aspects of the homeostatic response in C . elegans sleep . We noted that the baseline level of quiescence in undisturbed animals varied between the different transgenic strains ( Figure 8D ) . Broad expression of a rescue gene , or even a fluorescent reporter , often results in subtle changes in locomotion and quiescence that our assays are able to detect . Nevertheless , the data raised the possibility of a ceiling effect for quiescence in these experiments . Two observations suggest that , plausibly , this is not the case: ( i ) similar differences in undisturbed baseline quiescence were observed between the two npr-1 mutant alleles , yet both strains exhibited compensation for strong stimuli ( Figure 5E ) ; and ( ii ) undisturbed baseline quiescence in all daf-16 strains was similar to npr-1 ( ky13 ) mutants and lower than wild-type . Therefore , we favor the interpretation that the function of DAF-16 was required in neurons in our assays . Importantly , the opposing phenotypes of npr-1 and daf-16 mutants show that micro-homeostasis , routinely used to stabilize bout architecture in weakly noisy environments , is genetically distinct from the homeostatic responses that strong and stressful disturbances invoke . Using tunable photo- and mechano-stimulation , we have identified two distinct categories of disruptions to C . elegans sleep and characterized the corresponding responses . We have shown that during lethargus , motion plays a causal role in modulating the duration of subsequent quiescence . Under low noise conditions , micro-homeostasis manifested as a dependence of the duration of quiescence bouts on the duration and nature of recently preceding motion . The dynamic extension of quiescence bouts depended on the function of an NPY receptor-like protein ( NPR-1 ) . However , this did not require DAF-16/FOXO , perhaps because transcriptional level control is typically too slow to respond dynamically on timescales of 10s of seconds ( Yosef and Regev , 2011 ) . Since biological mechanisms naturally function in a continuous range of conditions , it was both expected and observed that similar mechanisms regulated the compensatory responses in the presence of weak or no stimuli . However , homeostasis in the presence of strong stimuli was behaviorally and mechanistically distinct . A strong disturbance resulted in the temporary disruption of normal bout dynamics followed by a compensatory upshift of baseline levels of quiescence . These responses did require the function of DAF-16/FOXO but not of NPR-1 . Neuropeptides have been proposed to regulate quiescence during lethargus , but their roles were not examined in detail ( Van Buskirk and Sternberg , 2007; Van Buskirk and Sternberg , 2010; Nelson et al . , 2013; Turek et al . , 2013 ) . The apparent discrepancy between the quiescence phenotypes of egl-3/egl-21 and unc-31 mutants resembles their seemingly contradicting roles in homeostatically ameliorating convulsions caused by cholinergic overexcitation and can be similarly rationalized ( Stawicki et al . , 2013 ) . The mature neuropeptides processed by EGL-3 and EGL-21 are but a subset of the components of dense core vesicles , such that excitatory and inhibitory neuropeptides could act in a combinatorial manner to affect quiescence . Resuming sleep after a strong or a mild disruption are both common experiences . Subjectively , the two are easily distinguishable , and in both cases the resulting changes to the architecture of sleep reflect homeostatic regulation . Broadly , homeostatic control ensuring adequate sleep amount and quality is a key criterion for sleep-like states ( Tobler , 1983; Campbell and Tobler , 1984; Sehgal and Mignot , 2011; Nelson and Raizen , 2013; Tononi and Cirelli , 2014 ) . Homeostasis in mammalian sleep can be readily observed under disturbed or undisturbed conditions . For instance , the spectral power density associated with slow wave sleep ( in the 0 . 75–4 . 0 Hz range ) decays exponentially during an undisturbed sleep period , while extending the duration of wakefulness enhances it ( Franken et al . , 1991; Kecklund and Åkerstedt , 1992 ) . Nevertheless , the sleep literature generally regards sleep homeostasis as a single mechanism ( Borbély , 1982; Daan et al . , 1984; Hendricks et al . , 2000; Saper et al . , 2005; Andretic et al . , 2008; Mackiewicz et al . , 2008; Cirelli , 2009; Crocker and Sehgal , 2010; Wang et al . , 2011; Brown et al . , 2012; Nelson and Raizen , 2013; Porkka-Heiskanen , 2013 ) . To our knowledge , responses to weak disturbances to sleep were not previously carefully analyzed , and the distinction between routine stabilization and compensation for stressful agitation was not examined in detail . Despite recent findings in genetically tractable invertebrate models , the understanding of mechanisms that regulate sleep homeostasis remains incomplete ( Andretic et al . , 2008; Cirelli , 2009; Sehgal and Mignot , 2011; Driver et al . , 2013; Shi et al . , 2014 ) . NPY was implicated in the regulation of sleep in humans , rats , fruit flies , and nematodes ( Antonijevic et al . , 2000; Tóth et al . , 2007; Dyzma et al . , 2010; Van den Pol , 2012; Choi et al . , 2013; He et al . , 2013; Nagy et al . , 2014 ) . In C . elegans , the NPY receptor homolog NPR-1 affects a range of responses to external stimuli , as well as innate behaviors such as social feeding and quiescence ( De Bono and Bargmann , 1998; De Bono et al . , 2002; Davies et al . , 2004; Chang et al . , 2006; Macosko et al . , 2009; McGrath et al . , 2009; Choi et al . , 2013; Nagy et al . , 2014 ) . Interestingly , NPR-1 was found to play a major role in both lethargus micro-homeostasis ( this study ) and the homeostatic response to a motoneuron imbalance . In the latter case , NPR-1 was required to compensate for cholinergic overexcitation and GABAergic inhibition that were caused by a gain-of-function in a neuronal nicotinic acetylcholine receptor ( Stawicki et al . , 2013 ) . We hypothesize that these two types of homeostatic responses are closely linked , and further studies will be required to conclusively determine if this is the case . Recent years have seen a rise in the appreciation of the importance and abundance of peptidergic modulation of neuronal function ( Li and Kim , 2008; Bargmann , 2012; Marder , 2012; Taghert and Nitabach , 2012; Holden-Dye and Walker , 2013 ) . In C . elegans , peptidergic regulation was shown to affect quiescence during lethargus ( Nelson et al . , 2013; Turek et al . , 2013 ) . The apparent discrepancy between the phenotypes of egl-3/egl-21 and unc-31 mutants suggests that quiescence may be regulated by the combinatorial action of excitatory and inhibitory neuropeptides and that this combinatorial regulation promotes responsive bout dynamics . Our findings are consistent with a model in which activity during lethargus generates a ‘pressure’ which is ameliorated during periods of quiescence . A particular balance of inhibitory and excitatory neuropeptides may be required for keeping a record of and/or for the process of alleviating this pressure . Finally , responses to external stimuli during lethargus were different from responses during the L4int and YA stages . In contrast , responses were similar whether the onset of the stimulus coincided with quiescence or motion during lethargus . This suggests that bouts of motion are not analogous to brief intervals of wakefulness . Rather , C . elegans sleep may progress through two alternating micro-states . C . elegans strains were maintained and grown according to standard protocols ( Brenner , 1974 ) . The following strains were used: wild-type strain N2 , Hawaiian CB4856 , CB169 unc-31 ( e169 ) , CB928 unc-31 ( e928 ) , CX4148 npr-1 ( ky13 ) , DA609 npr-1 ( ad609 ) , MT1541 egl-3 ( n729 ) , MT1241 egl-21 ( n611 ) , CF1038 daf-16 ( mu86 ) , GR1307 daf-16 ( mgDf50 ) , NQ440 daf-16 ( mgDf50 ) ; qnIs42[Punc-119::GFP::daf-16; Pmyo-2::mCherry] , NQ441 daf-16 ( mgDf50 ) ; qnIs45[Pdaf-16::GFP::daf-16; Pmyo-2::mCherry] , NQ145 daf-16 ( mgDf50 ) ; qnEx38[Pmyo-3::GFP::daf-16; Pmyo-2:mCherry] , VC2016 flp-18 ( gk3063 ) , AX1410 flp-18 ( db99 ) , AX1444 dbIs[Pflp-18::flp-18::sl2::gfp] . Motion and quiescence were identified using previously described methods ( Nagy et al . , 2014 ) . Briefly , animals were grown at 20°C on standard NGM plates seeded with Escherichi coli OP50 bacteria . Mid to late L4 individuals were sealed into individual ‘artificial dirt’ chambers filled with an overnight OP50 culture concentrated 10-fold and resuspended in NGM medium ( Singh et al . , 2011 ) . Animals were imaged at 2 frames per second at a 1 . 2× magnification for frame subtraction experiments or 10 frames per second at a 4 . 2× magnification for posture-based analysis using a CCD camera ( Prosilica GC2450 , Allied Vision Technologies , Stadtroda , Germany ) . Motion and quiescence were determined as previously described ( Iwanir et al . , 2013; Nagy et al . , 2014 ) . Frame subtraction data were obtained from the raw images using custom Matlab script ( Mathworks Inc . , Natick MA ) and quiescence was scored when no pixel changed its greyscale value beyond a threshold value ( Husson et al . , 2007 ) between consecutive frames . The precise analysis of animal behavior , based on the identification of the body posture , required high spatial and temporal resolution data . Image analysis and secondary data analysis were performed as previously described using a custom suite of machine vision tools , called PyCelegans , and custom Matlab scripts , respectively ( Nagy et al . , 2013 , 2014 ) . In brief , we identified the body midline in each frame , as well as the positions of the head and the tail . Each midline was divided into 20 equal intervals and the dynamics of the angles between these intervals were used to identify quiescence and directed locomotion states . The onset of lethargus was identified by visual inspection of quiescence data . We note that typical C . elegans behavioral assays provide a throughput of 100–1000 animals per day . In contrast , the detailed and computationally intensive posture-based analysis produced a detailed and an accurate account of behavior over 10 hr at a throughput of 3–5 animals per day . Blue light ( λ = 475 ± 15 nm ) was supplied by a Luxeon Star 7-LED assembly with a diffused optic array driven by a 700 mA FlexBlock driver . The LED assembly was mounted to the scopes approximately 7 cm from the sample location . Light intensity was measured at the location of the animals . The timing of light stimuli was controlled using LabView ( National Instruments Inc . , Austin TX ) . Mechanical stimuli were generated using 50 mm piezo buzzer elements ( Digikey part no . 668-1190-ND ) as previously described ( Nagy et al . , 2014 ) . The timing and duration of the stimuli were controlled using a custom Matlab script . An external stimulus was provided every 15 min throughout the course of each experiment . Animals resumed baseline behavior dynamics after no more than 5 min after each individual stimuli and no habituation was observed in the responses to the repeated stimuli . The Pflp-18::flp-18::SL2::gfp reporter strain was a kind gift from the de Bono lab ( Cohen et al . , 2009 ) . Late L4int larvae were placed in an artificial dirt microfluidic device filled with an overnight OP50 culture concentrated 10-fold and resuspended in NGM medium ( Lockery et al . , 2008 ) . Epi-fluorescence images of the freely behaving animals were acquired for 5 s every 15 min , for 8–9 hr , at a magnification of 20× and a frame rate of 4 frames per second . Regions of interest containing the neurons were identified by visual inspection . Fluorescence was quantified as the sum of pixel intensities that were higher than one standard deviation above the mean of the background pixel intensity . The background was calculated from a region of the body of the animal that was proximal to the neuron of interest but did not contain it . Under these conditions , no photo-bleaching was detected . Data analysis was performed using custom Matlab ( Mathworks Inc . , Natick , MA ) scripts . For comparisons in summary statistics panels , significance was calculated using a one-way ANOVA test . Post-hoc correction for multiple comparisons was performed using the Bonferroni adjustment . Correlation coefficients are presented with 95% confidence intervals ( Matlab statistical toolbox ) . Corresponding p-values are the probabilities of obtaining the observed correlation by chance , when the true correlation is zero . To graphically demonstrate pairwise correlations between durations of bouts in Figures 1E , 4B , 6A , and 7C , we grouped all bouts of motion in order of ascending duration in bins of 2 s and used a linear fit as a guide for the eye . The correlation coefficients were calculated using the original pairs of bout durations ( as opposed to the binned data ) . Bout correlations of wild-type animals , flp-18 mutants , and npr-1 mutants during the period 15–90 min from the onset of L4leth were compared by applying Fisher's z-transformation and calculating the 95% confidence interval for the difference of the correlation coefficients as described in Zou ( 2007 ) .	The regenerative properties of sleep are required by all animals , with even the simplest animal , the nematode Caenorhabditis elegans , displaying a sleep-like state called lethargus . During development , nematodes must pass through four larval stages en route to adulthood , and the end of each stage is preceded by a period of lethargus lasting 2 to 3 hr . Human sleep is divided into distinct stages that recur in a prescribed order throughout the night . Nematodes , on the other hand , simply experience alternating periods of activity and stillness as they sleep . Nevertheless , in both species , any disruptions to sleep automatically lead to adjustments of the rest of the sleep cycle to compensate for the disturbance and to ensure that the organism gets an adequate amount of sleep overall . To date , it has been assumed that a single mechanism is responsible for adjusting the sleep cycle after any disturbance , regardless of its severity . However , Nagy , Tramm , Sanders et al . now show that this is not the case in C . elegans . Sleeping nematodes that were lightly disturbed by exposing them to light or to vibrations—causing them to briefly increase their activity levels—compensated for the disturbance by lengthening their next inactive period . By contrast , worms that were vigorously agitated by stronger vibrations showed a different response: the alternating pattern of stillness and activity was disrupted for several minutes , followed by an overall increase in the length of time spent in the stillness phase . Experiments using genetically modified worms revealed that these two responses involve distinct molecular pathways . A signaling molecule called neuropeptide Y affects the response to minor sleep disruptions , whereas a transcription factor called DAF-16/FOXO is involved in the corresponding role after major disruptions . Given that neuropeptide Y has already been implicated in sleep regulation in humans and flies , it is not implausible that similar mechanisms may occur in response to disturbances of our own sleep .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Homeostasis in C. elegans sleep is characterized by two behaviorally and genetically distinct mechanisms
Resistance mutations against one drug can elicit collateral sensitivity against other drugs . Multi-drug treatments exploiting such trade-offs can help slow down the evolution of resistance . However , if mutations with diverse collateral effects are available , a treated population may evolve either collateral sensitivity or collateral resistance . How to design treatments robust to such uncertainty is unclear . We show that many resistance mutations in Escherichia coli against various antibiotics indeed have diverse collateral effects . We propose to characterize such diversity with a joint distribution of fitness effects ( JDFE ) and develop a theory for describing and predicting collateral evolution based on simple statistics of the JDFE . We show how to robustly rank drug pairs to minimize the risk of collateral resistance and how to estimate JDFEs . In addition to practical applications , these results have implications for our understanding of evolution in variable environments . The spread of resistance against most antibiotics and the difficulties in developing new ones has sparked considerable interest in using drug combinations and sequential drug treatments to treat bacterial infections , as well as cancers ( Pál et al . , 2015 ) . Treatments where the drugs are chosen so that resistance against one of them causes the pathogen or cancer population to become sensitive to the other—a phenomenon known as collateral sensitivity—can eliminate the population before multi-drug resistance emerges ( Pál et al . , 2015; Pluchino et al . , 2012 ) . The success of a multi-drug treatment hinges on knowing which drugs select for collateral sensitivity against which other drugs . This information is obtained empirically by exposing bacterial and cancer-cell populations to drugs and observing the evolutionary outcomes ( Roemhild et al . , 2020; Jensen et al . , 1997; Imamovic and Sommer , 2013; Lázár et al . , 2018; Maltas and Wood , 2019; Batra et al . , 2021; Sanz-García et al . , 2020; Schenk et al . , 2015; Lázár et al . , 2013; Barbosa et al . , 2019; Hernando-Amado et al . , 2020; Kim et al . , 2014; Jahn et al . , 2021; Kavanaugh et al . , 2020; Laborda et al . , 2021; Oz et al . , 2014; Munck et al . , 2014 ) . Prior studies have largely focused on various empirical questions related to the evolution of collateral sensitivity and resistance , such as identifying their genetic basis ( Lázár et al . , 2014; Roemhild et al . , 2020; Maltas and Wood , 2019; Hernando-Amado et al . , 2020; Kavanaugh et al . , 2020; Laborda et al . , 2021 ) , understanding how collateral outcomes depend on treatment design ( e . g . sequential versus combination ) ( Lázár et al . , 2014; Munck et al . , 2014; Bergstrom et al . , 2004; Batra et al . , 2021; Sanz-García et al . , 2020; Schenk et al . , 2015; Lázár et al . , 2013; Kim et al . , 2014; Jahn et al . , 2021 ) , or testing whether collateral sensitivity is an evolutionarily stable outcome ( Barbosa et al . , 2019 ) . However , one important feature of these experimental studies has received little attention , namely , the fact that different experiments often produce collateral sensitivity profiles that are inconsistent with each other ( e . g . Imamovic and Sommer , 2013; Oz et al . , 2014; Barbosa et al . , 2017; Maltas and Wood , 2019 ) . Some inconsistencies can be explained by the fact that resistance mutations vary between bacterial strains , drug dosages , etc . ( Mira et al . , 2015; Barbosa et al . , 2017; Das et al . , 2020; Pinheiro et al . , 2021; Card et al . , 2021; Gjini and Wood , 2021 ) . However , wide variation in collateral outcomes is observed even between replicate populations ( Oz et al . , 2014; Barbosa et al . , 2017; Maltas and Wood , 2019; Nichol et al . , 2019 ) . This variation suggests that bacteria and cancers have access to multiple resistance mutations with diverse collateral effects and that replicate populations accumulate different resistance mutations due to the intrinsic randomness of the evolutionary process ( Jerison et al . , 2020; Nichol et al . , 2019 ) . However , the diversity of collateral effects among resistance mutations has rarely if ever been systematically measured . Moreover , few existing approaches for designing robust multi-drug treatments have modelled this mutational diversity explicitly within the population genetics context ( Nichol et al . , 2019; Maltas and Wood , 2019 ) . Yet , a theory grounded in population genetics could help us understand how the expected collateral outcomes and the uncertainty around these expectations depend on evolutionary parameters and how these expectations and uncertainties change over time . Here , we develop such a theory . Collateral sensitivity and resistance are specific examples of the more general evolutionary phenomenon , pleiotropy , which refers to any situation when one mutation affects multiple phenotypes ( Wagner and Zhang , 2011; Paaby and Rockman , 2013 ) . In case of drug resistance evolution , the direct effect of resistance mutations is to increase fitness in the presence of one drug ( the ‘home’ environment ) . In addition , they may also provide pleiotropic gains or losses in fitness in the presence of other drugs ( the ‘non-home’ environments ) leading to collateral resistance or sensitivity , respectively . Classical theoretical work on pleiotropy has been done in the field of quantitative genetics ( Lande and Arnold , 1983; Rose , 1982; Barton , 1990; Slatkin and Frank , 1990; Jones et al . , 2003; Johnson and Barton , 2005 ) . In these models , primarily developed to understand how polygenic traits respond to selection in sexual populations , pleiotropy manifests itself as a correlated temporal change in multiple traits in a given environment . The question of how new strongly beneficial mutations that accumulate in asexual populations evolving in one environment affect its fitness in future environments is outside of the scope of these models . The pleiotropic consequences of adaptation have also been explored in various ‘fitness landscape’ models ( e . g . Connallon and Clark , 2015; Martin and Lenormand , 2015; Harmand et al . , 2017; Wang and Dai , 2019; Maltas et al . , 2021; Nichol et al . , 2019; Tikhonov et al . , 2020 ) . In particular , Nichol et al . , 2019 specifically addressed the problem of diversity of collateral resistance/sensitivity outcomes in the context of a combinatorially complete fitness landscapes of four mutations in the TEM β-lactamase gene . They found that different in silico populations adapting to the same antibiotic often arrive at different fitness peaks which results in different levels of collateral resistance/sensitivity against other drugs . They observed qualitatively similar variability in the collateral outcomes among replicate populations of the bacterium Escherichia coli evolving in the presence of cefotaxime ( CTX ) , although it is unclear whether different populations indeed arrived at different fitness peaks . In general , the fitness landscape approach helps us understand how evolutionary trajectories and outcomes depend on the global structure of the underlying fitness landscape . However , applying this approach in practice is challenging because the global structure of fitness landscapes is unknown and notoriously difficult to estimate , even in controlled laboratory conditions . Here , we take a different approach which is agnostic with respect to the global structure of the fitness landscape . Instead , we assume only the knowledge of the so-called joint distribution of fitness effects ( JDFE ) , that is , the probability that a new mutation has a certain pair of fitness effects in the home and non-home environments ( Jerison et al . , 2014; Martin and Lenormand , 2015; Bono et al . , 2017 ) . The JDFE is a natural extension of the DFE , the distribution of fitness effects of new mutations , often used in modeling evolution in a single environment ( King , 1972; Ohta , 1987; Orr , 2003; Kassen and Bataillon , 2006; Eyre-Walker and Keightley , 2007; Martin and Lenormand , 2008; MacLean and Buckling , 2009; Kryazhimskiy et al . , 2009; Levy et al . , 2015 ) . Like the DFE , the JDFE is a local property of the fitness landscape which means that it can be , at least in principle , estimated by using a variety of modern high-throughput techniques ( e . g . Qian et al . , 2012; van Opijnen et al . , 2009; Chevereau et al . , 2015; Levy et al . , 2015; Blundell et al . , 2019; Aggeli et al . , 2021 ) . The downside of this approach is that the JDFE can change over time as the population traverses the fitness landscape ( Good et al . , 2017; Venkataram et al . , 2020; Aggeli et al . , 2021 ) . However , in the context of collateral drug resistance and sensitivity , we are primarily interested in short time scales over which the JDFE can be reasonably expected to stay approximately constant . The rest of the paper is structured as follows . First , we use previously published data to demonstrate that E . coli has access to drug resistance mutations with diverse collateral effects . This implies that , rather than treating collateral effects as deterministic properties of drug pairs , we should think of them probabilistically , in terms of the respective JDFEs . We then show that a naive intuition about how the JDFE determines pleiotropic outcomes of evolution can sometimes fail , and a mathematical model is therefore required . We develop such a model , which reveals two key ‘pleiotropy statistics’ of the JDFE that determine the dynamics of fitness in the non-home condition . Our theory makes quantitative predictions in a variety of regimes if the population genetic parameters are known . However , we argue that in the case of drug resistance evolution the more important problem is to robustly order drug pairs in terms of their collateral sensitivity profiles even if the population genetic parameters are unknown . We develop a metric that allows us to do so . Finally , we provide some practical guidance for estimating the pleiotropy statistics of empirical JDFEs in the context of ranking drug pairs . We begin by demonstrating that the JDFE is a useful concept for modeling the evolution of collateral antibiotic resistance and sensitivity . If all resistance mutations against a given drug had identical pleiotropic effects on the fitness of the organism in the presence of another drug , the dynamics of collateral resistance/sensitivity could be understood without the JDFE concept . On the other hand , if different resistance mutations have different pleiotropic fitness effects , predicting the collateral resistance/sensitivity dynamics requires specifying the probabilities with which mutations with various home and non-home fitness effects arise in the population . The JDFE specifies these probabilities . Therefore , for the JDFE concept to be useful in the context of collateral resistance/sensitivity evolution , we need to show that resistance mutations against common drugs have diverse collateral effects in the presence of other drugs . To our knowledge , no data sets are currently publicly available that would allow us to systematically explore the diversity of collateral effects among all resistance mutations against any one drug in any organism . Instead , we examined the fitness effects of 3883 gene knock-out mutations in the bacterium Escherichia coli , measured in the presence of six antibiotics ( Chevereau et al . , 2015 ) , as well as the fitness effects of 4997 point mutations in the TEM-1 β-lactamase gene measured in the presence of two antibiotics ( Stiffler et al . , 2015 ) . For four out of six antibiotics used by Chevereau et al . , 2015 , we find between 12 ( 0 . 31 % ) and 170 ( 4 . 38% ) knock-out mutations that provide some level of resistance against at least one of the antibiotics ( false discovery rate ( FDR ) ∼25%; Figure 1 , Figure 1—source data 1; see Materials and methods for details ) . Plotting on the x-axis the fitness effect of each knock-out mutation in the presence of the drug assumed to be applied first ( i . e . the home environment ) against its effect in the presence of another drug assumed to be applied later ( i . e . the non-home environment , y-axis ) , we find mutations in all four quadrants of this plane , for all 12 ordered drug pairs ( Figure 1 , Figure 1—source data 1 ) . Similarly , we find diverse collateral effects among mutations within a single gene ( Figure 1—figure supplement 1; see Materials and methods for details ) . Since both data sets represent subsets of all resistance mutations , we conclude that E . coli likely have access to resistance mutations with diverse pleiotropic effects , such that a fitness gain in the presence of any one drug can come either with a pleiotropic gain or a pleiotropic loss of fitness in the presence of other drugs . Therefore , the JDFE framework is suitable for modeling the evolution of collateral resistance/sensitivity . In the next section , we formally define a JDFE and probe our intuition for how its shape determines the fitness trajectories in the non-home environment . For any genotype g that finds itself in one ( ‘home’ ) environment and may in the future encounter another ‘non-home’ environment , we define the JDFE as the probability density Φg⁢ ( Δ⁢x , Δ⁢y ) that a new mutation that arises in this genotype has the selection coefficient Δ⁢x in the home environment and the selection coefficient Δ⁢y in the non-home environment ( Jerison et al . , 2014 ) . For concreteness , we define the fitness of a genotype as its Malthusian parameter ( Crow and Kimura , 1972 ) . So , if the home and non-home fitness of genotype g are x and y , respectively , and if this genotype acquires a mutation with selection coefficients Δ⁢x and Δ⁢y , its fitness becomes x+Δ⁢x and y+Δ⁢y . This definition of the JDFE can , of course , be naturally extended to multiple non-home environments . In principle , the JDFE can vary from one genotype to another . However , to develop a basic intuition for how the JDFE determines pleiotropic outcomes , we assume that all genotypes have the same JDFE . We discuss a possible extension to epistatic JDFEs in Appendix 1 . The JDFE is a complex object . So , we first ask whether some simple and intuitive summary statistics of the JDFE may be sufficient to predict the dynamics of the non-home fitness of a population that is adapting in the home environment . Intuitively , if there is a trade-off between home and non-home fitness , non-home fitness should decline; if the opposite is true , non-home fitness should increase . Canonically , a trade-off occurs when any mutation that improves fitness in one environment decreases it in the other environment and vice versa ( Roff and Fairbairn , 2007 ) . Genotypes that experience such ‘hard’ trade-offs are at the Pareto front ( Shoval et al . , 2012; Li et al . , 2019 ) . For genotypes that are not at the Pareto front , some mutations that are beneficial in the home environment may be beneficial in the non-home environment and others may be deleterious . In this more general case , trade-offs are commonly quantified by the degree of negative correlation between the effects of mutations on fitness in the two environments ( Roff and Fairbairn , 2007; Tikhonov et al . , 2020 ) . Thus , we might expect that evolution on negatively correlated JDFEs would lead to pleiotropic fitness losses and evolution on positively correlated JDFEs would lead to pleiotropic fitness gains . To test this intuition , we generated a family of Gaussian JDFEs that varied , among other things , by their correlation structure ( Figure 2; Materials and methods ) . We then simulated the evolution of an asexual population on these JDFEs using a standard Wright-Fisher model ( Materials and methods ) and tested whether the trade-off strength , measured by the JDFE’s correlation coefficient , predicts the dynamics of non-home fitness . Figure 2 shows that our naive expectation is incorrect . Positively correlated JDFEs sometimes lead to pleiotropic fitness losses ( Figure 2D , I ) , and negatively correlated JDFEs sometimes lead to pleiotropic fitness gains ( Figure 2B , G ) . Even if we calculate the correlation coefficient only among mutations that are beneficial in the home environment , the pleiotropic outcomes still do not always conform to the naive expectation , as the sign of the correlation remains the same as for the full JDFEs in all these examples . There are other properties of the JDFE that we might intuitively expect to be predictive of the pleiotropic outcomes of adaptation . For example , among the JDFEs considered in Figure 2 , it is apparent that those with similar relative probability weights in the first and fourth quadrants produce similar pleiotropic outcomes . However , simulations with other JDFE shapes show that even distributions that are similar according to this metric can also result in qualitatively different pleiotropic outcomes ( Figure 2—figure supplement 1 ) . Overall , our simulations show that JDFEs with apparently similar shapes can produce qualitatively different trajectories of pleiotropic fitness changes ( e . g . compare Figure 2A , F and B , G or Figure 2D , I and E , J ) . Conversely , JDFEs with apparently different shapes can result in rather similar pleiotropic outcomes ( e . g . compare Figure 2B , G and E , J or Figure 2A , F and D , I ) . Thus , while the overall shape of the JDFE clearly determines the trajectory of pleiotropic fitness changes , it is not immediately obvious what features of its shape play the most important role , particularly if the JDFE is more complex than a multivariate Gaussian . In other words , even if we have perfect knowledge of the fitness effects of all mutations in multiple environments , converting this knowledge into a qualitative prediction of the expected direction of pleiotropic fitness change ( gain or loss ) does not appear straightforward . Therefore , we next turn to developing a population genetics model that would allow us to predict not only the direction of pleiotropic fitness change but also the expected rate of this change and the uncertainty around the expectation . To systematically investigate which properties of the JDFE determine the pleiotropic fitness changes in the non-home environment , we consider a population of size N that evolves on a JDFE in the ‘strong selection weak mutation’ ( SSWM ) regime , also known as the ‘successional mutation’ regime ( Orr , 2000; Desai and Fisher , 2007; Kryazhimskiy et al . , 2009; Good and Desai , 2015 ) . We consider an arbitrary JDFE without epistasis , that is a situation when all genotypes have the same JDFE Φ⁢ ( Δ⁢x , Δ⁢y ) . We explore an extension to JDFEs with a simple form of epistasis in Appendix 1 . We assume that mutations arise at rate U per individual per generation . In the SSWM limit , a mutation that arises in the population either instantaneously fixes or instantaneously dies out . Therefore , the population is essentially monomorphic at all times , such that at any time t we can characterize it by its current pair of fitness values ( Xt , Yt ) . If a new mutation with a pair of selection coefficients ( Δ⁢x , Δ⁢y ) arises in the population at time t , it fixes with probability π⁢ ( Δ⁢x ) =1-e-2⁢Δ⁢x1-e-2⁢N⁢Δ⁢x ( Kimura , 1962 ) in which case the population’s fitness transitions to a new pair of values ( Xt+Δ⁢x , Yt+Δ⁢y ) . If the mutation dies out , an event that occurs with probability 1-π⁢ ( Δ⁢x ) , the population’s fitness does not change . This model specifies a continuous-time two-dimensional Markov process . In general , the dynamics of the probability density p⁢ ( x , y , t ) of observing the random vector ( Xt , Yt ) at values ( x , y ) are governed by an integro-differential forward Kolmogorov equation , which is difficult to solve ( Materials and methods ) . However , if most mutations that contribute to adaptation have small effects , these dynamics are well approximated by a diffusion equation which can be solved exactly ( Materials and methods ) . Then p ( x , y , t ) is a normal distribution with mean vector ( 1 ) m ( t ) = ( x0y0 ) + ( r1r2 ) NUbt and variance-covariance matrix ( 2 ) σ2 ( t ) = ( D11D12D12D22 ) NUbt , where are r1 and r2 , given by Equation 7 and Equation 8 in Materials and methods , are the expected fitness effects in the home and non-home environments for a mutation fixed in the home environment , and D11 , D12 , and D22 , given by Equation 9–Equation 11 in Materials and methods , are the second moments of this distribution . Here , Ub=U∫−∞∞dη∫0∞dξΦ ( ξ , η ) is the total rate of mutations beneficial in the home environment , and x0 and y0 are the initial values of population’s fitness in the home and non-home environments . Equation 1 and Equation 2 show that the distribution of population’s fitness at time t in the non-home environment is entirely determined by two parameters , r2 and D22 , which we call the ‘pleiotropy statistics’ of the JDFE . The expected rate of fitness change in the non-home environment depends on the pleiotropy statistic r2 , which we refer to as the expected pleiotropic effect . Thus , evolution on a JDFE with a positive r2 is expected to result in pleiotropic fitness gains and evolution on a JDFE with a negative r2 is expected to result in pleiotropic fitness losses . Equation 2 shows that the variance around this expectation is determined by the pleiotropy variance statistic D22 . Since both the expectation and the variance change linearly with time ( provided r2≠0 ) , the change in the non-home fitness in any replicate population will eventually have the same sign as r2 , but the time scale of such convergence depends on the ‘collateral resistance risk’ statistic c=r2/D22 ( Materials and methods ) . This observation has important practical implications , and we return to it in the Section ‘Robust ranking of drug pairs’ . These theoretical results suggest a simple explanation for the somewhat counter-intuitive observations in Figure 2 . We may intuitively believe that evolution on negatively correlated JDFEs should lead to fitness losses in the non-home environment because on such JDFEs mutations with largest fitness benefits in the home environment typically have negative pleiotropic effects . However , such mutations may be too rare to drive adaptation . At the same time , the more common mutations that do typically drive adaptation may have positive pleiotropic effects , in which case the population would on average gain non-home fitness , as in Figure 2B . Our theory shows that to predict the direction of non-home fitness change , the frequency of beneficial mutations with various pleiotropic effects and the strength of these effects need to be weighted by the likelihood that these mutations fix . The expected pleiotropic effect r2 accomplishes this weighting . We tested the validity of Equation 1 and Equation 2 by simulating evolution in the SSWM regime on 125 Gaussian JDFEs with various parameters ( Materials and methods ) and found excellent agreement ( Figure 3A and B ) . However , many microbes likely evolve in the ‘concurrent mutation’ regime , that is , when multiple beneficial mutations segregate in the population simultaneously ( Desai and Fisher , 2007; Lang et al . , 2013 ) . As expected , our theory fails to quantitatively predict the pleiotropic fitness trajectories when N⁢Ub§gt;1 ( Figure 3—figure supplement 1 ) . However , the expected rate of change of non-home fitness and its variances remain surprisingly well correlated with the pleiotropy statistics r2 and D22 across various JDFEs ( Figure 3—figure supplement 1 ) . In other words , we can still use these statistics to correctly predict whether a population would lose or gain fitness in the non-home environment and to order the non-home environments according to their expected pleiotropic fitness changes and variances . We will exploit the utility of such ranking in the next section . We next sought to expand our theory to the concurrent mutation regime . A key characteristic of adaptation in this regime is that mutations whose fitness benefits in the home environment are below a certain ‘effective neutrality’ threshold are usually outcompeted by superior mutations and therefore fix with lower probabilities than predicted by Kimura’s formula ( Schiffels et al . , 2011; Good et al . , 2012 ) . Good et al . , 2012 provide an equation for calculating the fixation probability π⁢ ( Δ⁢x ) for a mutation with home fitness benefit Δ⁢x in the concurrent mutation regime ( Equation ( 6 ) in Good et al . , 2012 ) . Thus , by replacing 2⁢ξ ( the approximate fixation probability in the SSWM regime ) in Equation 8 and Equation 11 with π⁢ ( ξ ) , we obtain the adjusted pleiotropy statistics r2* and D22* for the concurrent mutation regime ( see Materials and methods for details ) . Note that in contrast to r2 and D22 , the adjusted statistics r2* and D22* depend on the population genetic parameters N and Ub . To test how well these statistics predict the dynamics of fitness in the non-home environment , we simulated evolution on the same 125 JDFEs using the full Wright-Fisher model with a range of population genetic parameters that span the transition from the successional to the concurrent mutation regimes for 1 , 000 generations . We find that r2* quantitatively predicts the expected rate of non-home fitness change , with a similar accuracy as Good et al . , 2012 predict the rate of fitness change in the home environment , as long as N⁢Ub§gt;1 ( Figure 3C , E and G; compare with Figure 3—figure supplement 1A , C , E ) . D22* also predicts the empirically observed variance in non-home fitness trajectories much better than D22 , although this relationship is more noisy than between mean fitness and r2* ( Figure 3D , F and H; compare with Figure 3—figure supplement 1B , D , F ) . Some of this noise can be attributed to sampling , as we estimate both the mean and the variance from 300 replicate simulation runs , and the variance estimation is more noisy . Even in the absence of sampling noise however , we do not expect that D22* would predict the non-home fitness variance perfectly because our theory does not account for the autocorrelation in the fitness trajectories that arise in the concurrent mutation regime but not in the successive mutation regime ( see Appendix D in Desai and Fisher , 2007 ) . To our knowledge , a rigorous analytical calculation for ensemble variance in fitness even in the home environment is not yet available . Overall , our theory allows us to quantitatively predict the dynamics of non-home fitness in a range of evolutionary regimes if the JDFE and the population genetic parameters N and Ub are known . However , neither the full JDFE nor the population genetic parameters will likely be known in most practical situations , such as designing a drug treatment for a cancer patient . In the next section , we address the question of how to robustly select drug pairs for a sequential treatment , assuming that the pleiotropy statistics r2 and D22 are known but the population genetic parameters are not . In the Section ‘Measuring JDFEs’ , we provide some guidance on how the JDFE can be measured . Consider a hypothetical scenario where a drug treatment is being designed for a patient with a tumor or a bacterial infection . In selecting a drug , it is desirable to take into account not only the standard medical considerations , such as drug availability , toxicity , etc . , but also the possibility that the treatment with this drug will fail due to the evolution of resistance . Therefore , it may be prudent to consider a list of drugs pairs ( or higher-order combinations ) , ranked by the propensity of the first drug in the pair to elicit collateral resistance against the second drug in the pair . All else being equal , the drug deployed first should form a high-ranking pair with at least one other secondary drug . Then , if the treatment with the first drug fails , a second one can be deployed with a minimal risk of collateral resistance . Thus , we set out to develop a metric for ranking drug pairs according to this risk . Clearly , any drug pair with a negative r2 is preferable over any drug pair with a positive r2 , since the evolution in the presence of the first drug in a pair with r2§lt;0 is expected to elicit collateral sensitivity against the second drug in the pair but the opposite is true for drug pairs with r2§gt;0 . It is also clear that among two drug pairs with negative r2 , a pair with a more negative r2 and lower D22 is preferable over a pair with a less negative r2 and higher D22 because evolution in the presence of the first drug in the former pair will more reliably lead to stronger collateral sensitivity against the second drug in the pair . The difficulty is in how to compare and rank two drug pairs where one pair has a more negative r2 but higher D22 . Our theory shows that the chance of emergence of collateral resistance monotonically increases with the collateral risk statistic c=r2/D22 ( see Materials and methods ) . Thus , we propose to rank drug pairs by c from lowest ( most negative and therefore most preferred ) to highest ( least negative or most positive and therefore least preferred ) . To demonstrate the utility of such ranking , consider four hypothetical drug pairs with JDFEs shown in Figure 4A . The similarity between their shapes makes it difficult to predict a priori which one would have the lowest and highest probabilities of collateral resistance . Thus , we rank these JDFEs by their c statistic . To test whether this ranking is accurate with respect to the risk of collateral resistance , we simulate the evolution of a Wright-Fisher population in the presence of the first drug in each pair for 600 generations and estimate the probability that the evolved population has a positive fitness in the presence of the second drug , that is , the probability that it becomes collaterally resistant ( Figure 4B ) . We find that our a priori ranking corresponds perfectly to the ranking according to this probability , evidenced by the consistently higher collateral resistance risk for JDFEs with higher c ( Figure 4B ) . Interestingly , the top ranked JDFE does not have the lowest expected pleiotropic effect r2 . Nevertheless , the fact that the pleiotropic variance statistic D22 for this JDFE is small ensures that the risk of collateral resistance evolution is the lowest . This 1–1 rank correlation holds more broadly , for all 125 Gaussian JDFEs and all population genetic parameters considered in the previous section ( Figure 4C ) as well as for the empirical TEM β-lactamase JDFEs ( Figure 4—figure supplement 1 ) . Overall , we find that we can use the collateral risk statistic c to robustly rank drug pairs according to the risk of collateral resistance evolution , irrespective of the population genetic parameters . So far , we assumed that the parameters of the JDFE on which the population evolves are known . In reality , they have to be estimated from data , which opens up at least two practically important questions . The first question is experimental . From what types of data can JDFEs be in principle estimated and how good are different types of data for this purpose ? We can imagine , for example , that some properties of JDFEs can be estimated from genome sequencing data ( Jerison et al . , 2020 ) or from temporally resolved fitness trajectories ( Bakerlee et al . , 2021 ) . Here , we focus on the most direct way of estimating JDFE parameters , from the measurements of the home and non-home fitness effects of individual mutations . The experimental challenge with this approach is to sample those mutations that will most likely contribute to adaptation in the home environment ( see ‘Discussion’ for an extended discussion of this problem ) . Below , we propose two potential strategies for such sampling: the Luria-Delbrück ( LD ) method and the barcode lineage tracking ( BLT ) method . The second question is statistical: how many mutants need to be sampled to reliably rank drug pairs according to the risk of collateral resistance ? We evaluate both proposed methods with respect to this property . The idea behind the LD method is to expose the population to a given drug at a concentration above the minimum inhibitory concentration ( MIC ) , so that only resistant mutants survive ( Pinheiro et al . , 2021 ) . This selection is usually done on agar plates , so that individual resistant mutants form colonies and can be isolated . The LD method is relatively easy to implement experimentally , but it is expected to work only if the drug concentration is high enough to kill almost all non-resistant cells . In reality , resistant mutants may be selected at concentrations much lower than MIC ( Andersson and Hughes , 2014 ) . Furthermore , mutants selected at different drug concentrations may be genetically and functionally distinct ( Lindsey et al . , 2013; Pinheiro et al . , 2021 ) and therefore may have statistically different pleiotropic profiles . As a result , mutants sampled with the LD method may not be most relevant for predicting collateral evolution at low drug concentrations , and other sampling methods may be required for isolating weakly beneficial mutations . Isolating individual weakly beneficial mutations is more difficult because by the time a mutant reaches a detectable frequency in the population it has accumulated multiple additional driver and passenger mutations ( Lang et al . , 2013; Nguyen Ba et al . , 2019 ) , all of which can potentially have collateral effects . One way to isolate many single beneficial mutations from experimental populations is by using the recently developed barcode lineage tracking ( BLT ) method ( Levy et al . , 2015; Venkataram et al . , 2016 ) . In a BLT experiment , each cell is initially tagged with a unique DNA barcode . As long as there is no recombination or other DNA exchange , any new mutation is permanently linked to one barcode . A new adaptive mutation causes the frequency of the linked barcode to grow , which can be detected by sequencing . By sampling many random mutants and genotyping them at the barcode locus , one can identify mutants from adapted lineages even if they are rare ( Venkataram et al . , 2016 ) . As a result , BLT allows one to sample mutants soon after they acquire their first driver mutation , before acquiring secondary mutations . To evaluate the quality of sampling based on the LD and BLT methods , we consider the following hypothetical experimental setup . K beneficial mutants are sampled from each home environment ( with either one of the methods ) , and their home and non-home fitness ( Xi , Yi ) are measured for each mutant i=1 , … , K . Since we are ultimately interested in ranking drug pairs by their risk of collateral resistance , we estimate the collateral risk statistic c^ from these fitness data for each drug pair and use c^ to rank them ( see Materials and methods for details ) . We compare such a priori ranking of 125 hypothetical drug pairs with Gaussian JDFEs used in previous sections with their a posteriori ranking based on the risk of collateral resistance observed in simulations . To model the LD sampling method on a given JDFE , we randomly sample K mutants whose home fitness exceeds a certain cutoff . To model a BLT experiment , we simulate evolution in the home environment and randomly sample K beneficial mutants segregating at generation 250 ( see Materials and methods for details ) . We find that the c^-ranking estimated with either LD or BLT methods captures the a posteriori ranking surprisingly well , even when the number of sampled mutants is as low as 10 per drug pair ( Figure 5 ) . Given that the JDFEs with adjacent ranks differ in c by a median of only 0 . 65% , the strong correlations shown in Figure 5 suggest that even very similar JDFEs can be differentiated with moderate sample sizes . As expected , this correlation is further improved upon increased sampling , and it is insensitive to the specific home fitness threshold that we use in the LD method ( Figure 5—figure supplement 1 ) . We conclude that estimating JDFE parameters is in principle feasible with a modest experimental effort , at least for the purpose of ranking drug pairs . We have shown that many resistance mutations against multiple drugs in E . coli exhibit a diversity of collateral effects . If this is true more generally , it implies that there is an unavoidable uncertainty in whether any given population would evolve collateral resistance or sensitivity , which could at least in part explain previously observed inconsistencies among experiments . We quantified the diversity of pleiotropic effects of mutations with a joint distribution of fitness effects ( JDFE ) and developed a population genetic theory for predicting the expected collateral outcomes of evolution and the uncertainty around these expectations . In the successional mutations regime , our theory shows that the average rate at which fitness in the non-home environment is gained or lost during adaptation to the home environment is determined by the pleiotropy statistic r2 given by Equation 8 . How strongly the non-home fitness in any individual population deviates from this ensemble average is determined by the pleiotropy variance statistic D22 given by Equation 11 . Importantly , r2 and D22 are properties of the JDFE alone , that is , they do not depend on the parameters of any specific population . In the concurrent mutations regime , the expected rate of non-home fitness gain or loss and the associated variance are reasonably well predicted by the adjusted pleiotropy statistics r2* and D22* . Unlike r2 and D22 , the adjusted statistics depend on the population size N and the rate of beneficial mutations Ub . To quantitatively predict the probability of evolution of collateral drug resistance in practice would require the knowledge of both the JDFE for the focal bacterial or cancer-cell population in the presence of the specific pair of drugs and its in vivo population genetic parameters . Since estimating the latter parameters is very difficult , it appears unlikely that we would be able to quantitatively predict the dynamics of collateral effects , even if JDFEs were known . A more realistic application of our theory is that it allows us to rank drug pairs according to the risk of collateral resistance even when the population genetic parameters are unknown . Such robust ranking can be computed based on the collateral risk statistic c=r2/D22 , a property of the JDFE but not of the evolving population . Drug pairs with positive values of c have a higher chance of eliciting collateral resistance than collateral sensitivity and should be avoided; drug pairs with more negative values of c have a lower risk of collateral resistance evolution than those with less negative values . We have validated our theory in silico , but how well it would work in vivo ( in the clinic ) or even in vitro ( in the lab ) is as of yet unclear . A direct way to validate the theory empirically would be to estimate JDFEs for a model organism , such as E . coli , in a number of drug pairs , rank these pairs according to our collateral rank statistic and then test this ranking by evolving replicate populations and measuring the empirical distributions of collateral resistance/sensitivity outcomes . To the best of our knowledge , the antibiotic resistance JDFEs among genome-wide mutations have not yet been measured . One could in principle use existing gene knock-out data , such as those obtained by Chevereau et al . , 2015 ( Figure 1 ) , or the data from deep mutational scanning experiments , such as those obtained by Stiffler et al . , 2015 ( Figure 1—figure supplement 1 ) , to estimate JDFEs . However , these experiments estimate fitness only for certain subsets of mutations ( gene knock-outs or point mutations within a single gene , respectively ) . Since resistance may arise via other types of mutations ( Nichol et al . , 2019 ) , these data would give us at best an incomplete picture of actual JDFEs . Our results suggest that JDFEs can be reasonably well estimated by sampling resistance mutants at drug concentrations above MIC or by employing the barcode lineage tracking method . Another obstacle is that , even though many researchers have experimentally evolved various microbes in the presence of drugs , most experiments have maintained too few replicate populations to accurately measure the variation in collateral outcomes of evolution . The study by Nichol et al . , 2019 , with 60 replicates , is a notable exception . In short , a rigorous test of our theory requires new data on the shapes of whole-genome JDFEs as well as higher throughput evolution experiments . What the most effective ways of measuring JDFEs are and whether it will be possible to measure JDFE in vivo are open questions . We speculate that the answers will depend on the shapes of the empirical JDFEs because some shapes may be more difficult to estimate than others . For example , if empirical JDFEs resemble multivariate Gaussian distributions , then we can learn all relevant parameters of such JDFE by sampling a handful of random mutants and measuring their fitness in relevant environments . One can also imagine more complex JDFEs where mutations beneficial in the home environment have a dramatically different distribution of non-home fitness effects than mutations that are deleterious or neutral in the home environment . In this case , very large samples of random mutations would be necessary to correctly predict the pleiotropic outcomes of evolution , so that methods that preferentially sample beneficial mutations may be required . We have considered two such methods , which are experimentally feasible . We have shown that both of them perform extremely well on Gaussian JDFEs in the sense that as few as 10 mutants per drug pair are sufficient to produce largely correct ranking of hypothetical drug pairs . However , it may be difficult to apply these methods in vivo , in which case JDFEs may have to be estimated in the lab , with selection pressures reproducing those in vivo as accurately as possible . Our model relies on two important simplifications . It describes the evolution of an asexual population where all resistance alleles arise from de novo mutations . In reality , some resistance alleles in bacteria may be transferred horizontally ( Sun et al . , 2019 ) . Understanding collateral resistance evolution in the presence of horizontal gene transfer events would require incorporating JDFE into other models of evolutionary dynamics ( e . g . Neher et al . , 2010 ) . Another major simplification is in the assumption that the JDFE stays constant as the population adapts . In reality the JDFE will change over time because of the depletion of the pool of adaptive mutations and because of epistasis ( Good et al . , 2017; Venkataram et al . , 2020 ) . How JDFEs vary among genetic backgrounds is currently unknown . In Appendix 1 , we have shown that our main results hold at least in the presence of a simple form of ‘global’ epistasis . Empirically measuring how JDFEs vary across genotypes and theoretically understanding how such variation affects the evolution of pleiotropic outcomes are important open questions . While we were primarily motivated by the problem of evolution of collateral drug resistance and sensitivity , our theory is applicable more broadly . The shape of JDFE must play a crucial role in determining whether the population evolves toward a generalist or diversifies into multiple specialist ecotypes . Previous literature has viewed this question primarily through the lense of two alternative hypotheses: antagonistic pleiotropy and mutation accumulation ( Visher and Boots , 2020 ) . Antagonistic pleiotropy in its strictest sense means that the population is at the Pareto front with respect to the home and non-home fitness , such that any mutation beneficial in the home environment reduces the fitness in the non-home environment ( Li et al . , 2019 ) . The shape of the Pareto front then determines whether selection would favor specialists or generalists ( Levins , 1968; Visher and Boots , 2020 ) . Alternatively , a population can evolve to become a home-environment specialist even in the absence of trade-offs , simply by accumulating mutations that are neutral in the home environment but deleterious in the non-home environment ( Kawecki , 1994 ) . More recently , it has been recognized that antagonistic pleiotropy and mutation accumulation are not discrete alternatives but rather extremes of a continuum of models ( Bono et al . , 2020; Jerison et al . , 2014; Jerison et al . , 2020 ) . The JDFE provides a mathematical way to describe this continuum . For example , strict antagonistic pleiotropy can be modeled with a JDFE with zero probability weight in the first quadrant and a bulk of probability in the fourth quadrant . A mutation accumulation scenario can be modeled with a ‘+’-like JDFE where all mutations beneficial in the home environment are neutral in the non-home environment ( i . e . concentrated on the x-axis ) and all or most mutations neutral in the home environment ( i . e . those on the y-axis ) are deleterious in the non-home environment . Our theory shows that in fact all JDFEs with negative r2 lead to loss of fitness in the non-home environment and therefore can potentially promote specialization . While our theory provides this insight , further work is needed to understand how JDFEs govern adaptation to variable environments . This future theoretical work , together with empirical inquiries into the shapes of JDFEs , will not only advance our ability to predict evolution in practical situations , such as drug resistance , but it will also help us better understand the origins of ecological diversity . According to Equation 1 and Equation 2 , both the expected non-home fitness and its variance change linearly with time , so that at time t the mean is Z=c⁢N⁢Ub⁢t standard deviations above y0 ( if r2§gt;0 ) or below y0 ( if r2§lt;0 ) , where c=r2/D22 . In other words , if r2§gt;0 , the bulk of the non-home fitness distribution eventually shifts above y0 , and if r2§lt;0 , it shifts below y0 . All else being equal , a larger value of \|c\| implies faster rate of this shift . The interpretation of these observations in terms of collateral resistance/sensitivity is that adaptation in the presence of the first drug will eventually lead to collateral resistance against the second drug if r2§gt;0 and to collateral sensitivity if r2§lt;0 . Furthermore , all else being equal , collateral sensitivity evolves faster and the chance of evolving collateral resistance is smaller for drug pairs with more negative c ( i . e . larger \|c\| ) . Thus , we use c to order drug pairs from the most preferred ( those with the most negative values of c ) to least preferred ( those with least negative or positive values of c ) . We carried out two types of simulations , SSWM model simulations and full Wright-Fisher model simulations . We model the LD sampling method by randomly drawing mutants from the JDFE until the desired number K of mutants whose home fitness exceeds the focal threshold are sampled . We estimate the c statistic from the pairs of home and non-home fitness effects Xi and Yi of these i=1 , … , K sampled mutants . To do so , we first estimate r2 and D22 as r^2=1/K⁢∑i=1KXi⁢Yi and D^22=1/K⁢∑i=1KXi⁢Yi2 . We then calculate c^=r^2/D^22 . For the BLT sampling method , we simulate the Wright-Fisher model as described above for N=106 and U=10-4 for 250 generations . At generation 250 , we randomly sample existing beneficial mutants proportional to their frequency in the population without replacement ( i . e . the same beneficial mutation is sampled at most once ) . Sampling more than ∼50 distinct beneficial mutants from a single population becomes difficult because there may simply be not enough such mutants or some of them may be at very low frequencies . Therefore , if the desired number of mutants to sample exceeds 50 , we run multiple replicate simulations and sample a maximum of 100 distinct beneficial mutants per replicate until the desired number of mutants is reached . We then estimate the c statistics as with the LD method .	Drugs known as antibiotics are the main treatment for most serious infections caused by bacteria . However , many bacteria are acquiring genetic mutations that make them resistant to the effects of one or more types of antibiotics , making them harder to eliminate . One way to tackle drug-resistant bacteria is to develop new types of antibiotics; however , in recent years , the rate at which new antibiotics have become available has been dwindling . Using two or more existing drugs , one after another , can also be an effective way to eliminate resistant bacteria . The success of any such ‘multi-drug’ treatment lies in being able to predict whether mutations that make the bacteria resistant to one drug simultaneously make it sensitive to another , a phenomenon known as collateral sensitivity . Different resistance mutations may have different collateral effects: some may increase the bacteria’s sensitivity to the second drug , while others might make the bacteria more resistant . However , it is currently unclear how to design robust multi-drug treatments that take this diversity of collateral effects into account . Here , Ardell and Kryazhimskiy used a concept called JDFE ( short for the joint distribution of fitness effects ) to describe the diversity of collateral effects in a population of bacteria exposed to a single drug . This information was then used to mathematically model how collateral effects evolved in the population over time . Ardell and Kryazhimskiy showed that this approach can predict how likely a population is to become collaterally sensitive or collaterally resistant to a second antibiotic . Drug pairs can then be ranked according to the risk of collateral resistance emerging , so long as information on the variety of resistance mutations available to the bacteria are included in the model . Each year , more than 700 , 000 people die from infections caused by bacteria that are resistant to one or more antibiotics . The findings of Ardell and Kryazhimskiy may eventually help clinicians design multi-drug treatments that effectively eliminate bacterial infections and help to prevent more bacteria from evolving resistance to antibiotics . However , to achieve this goal , more research is needed to fully understand the range collateral effects caused by resistance mutations .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology" ]	2021	The population genetics of collateral resistance and sensitivity
Recurrent connections are thought to be a common feature of the neural circuits that encode memories , but how memories are laid down in such circuits is not fully understood . Here we present evidence that courtship memory in Drosophila relies on the recurrent circuit between mushroom body gamma ( MBγ ) , M6 output , and aSP13 dopaminergic neurons . We demonstrate persistent neuronal activity of aSP13 neurons and show that it transiently potentiates synaptic transmission from MBγ>M6 neurons . M6 neurons in turn provide input to aSP13 neurons , prolonging potentiation of MBγ>M6 synapses over time periods that match short-term memory . These data support a model in which persistent aSP13 activity within a recurrent circuit lays the foundation for a short-term memory . As animals pursue their goals , their behavioral decisions are shaped by memories that encompass a wide range of time scales: from fleeting working memories relevant to the task at hand , to short-term and long-term memories of contingencies learned hours , days , or even years in the past . Working memory is thought to reflect persistent activity generated within neural networks , including recurrent circuits ( Wang , 2001 ) . In contrast , short-term memory ( STM ) and long-term memory ( LTM ) involves changes in synaptic efficacy due to functional and structural modification of synaptic connections ( Kandel , 2001 ) . However , the neural circuit mechanisms involved in the formation , persistence and transitions between these distinct forms of memory are not fully known . A robust form of memory in Drosophila is courtship memory , which can last from minutes to days , depending on the duration and intensity of training ( Siegel and Hall , 1979; McBride et al . , 1999 ) . Naïve Drosophila males eagerly court both virgin females , which are generally receptive , and mated females , which are not ( Manning , 1967; Wolfner , 2003 ) . However , upon rejection by mated females , they become subsequently less likely to court other mated females ( Tompkins , 1984 ) . This selective suppression of courtship towards mated females , called courtship conditioning , can be attributed to the enhanced sensitivity of experienced males to an inhibitory male pheromone deposited on the female during mating , cis-vaccenyl acetate ( cVA ) ( Keleman et al . , 2012 ) . Olfactory memory in insects relies on the function of a central brain structure called the mushroom body ( MB ) ( de Belle and Heisenberg , 1994; Heisenberg et al . , 1985 ) . The principal MB cells , the cholinergic Kenyon cells ( KCs ) ( Barnstedt et al . , 2016 ) , receive input from sensory pathways in the dendritic calyx region and from dopaminergic neurons ( DANs ) in the axonal lobes of the MB . These MB lobes are compartmentalized , with each compartment innervated by specific classes of DANs and MB output neurons ( MBONs ) ( Aso et al . , 2014a; Mao and Davis , 2009 ) . MBONs receive input from both KCs and DANs ( Takemura et al . , 2017 ) . We previously established that short-term courtship conditioning is mediated by the aSP13 class of DANs ( also known as the PAM-γ5 neurons , [Aso et al . , 2014a] ) , which innervate the MBγ5 compartment . The activity of aSP13 neurons is essential for courtship conditioning in experienced males and sufficient to induce conditioning in naïve males ( Keleman et al . , 2012 ) . Here we demonstrate that courtship memory also requires the corresponding MBγ KCs and the MBγ5 MBONs , the glutamatergic M6 neurons ( also known as MBON-γ5β’2a neurons [Aso et al . , 2014a] ) . Furthermore , we present evidence that MBγ , M6 , and aSP13 neurons form a recurrent circuit and that persistent activity of the aSP13 neurons mediates plasticity at the MBγ to M6 synapses that can last from minutes to hours . Consistent with this model , M6 activity is required not only for memory readout but also , like aSP13 , for memory formation . These data support a model in which persistent aSP13 activity within the MBγ>M6>aSP13 recurrent circuit lays the foundation for short-term courtship memory . We confirmed the involvement of MBγ and M6 neurons in courtship conditioning by chronically silencing them using cell-type specific GAL4 drivers ( Figure 1—figure supplement 1 ) to express tetanus toxin light chain ( UAS-TNT , an inhibitor of synaptic transmission; [Martin et al . , 2002] ) . Single males of each genotype were trained by first pairing them for 1 hr with a single mated female , and then testing their courtship towards a fresh mated female after a 30 min rest period . We used automated video analysis to derive a courtship index ( CI ) for each male , defined as the percentage of time over a 10 min test period during which the male courts the female . A suppression index ( SI ) was then calculated as the relative reduction in the mean courtship indices of trained ( CI+ ) versus naïve ( CI- ) populations: SI = 100* ( 1-CI+/CI- ) . Control flies expressing an inactive form of tetanus toxin ( UAS-TNTQ ) typically showed a SI of ~40–50% ( Figure 1A , B; Supplementary file 1 ) . By contrast , males in which M6 neurons or MBγ neurons were silenced with an inhibitory form of tetanus toxin ( UAS-TNT ) showed much less or no suppression ( Figure 1A , B; Supplementary file 1 ) . The DAN inputs to a given MB compartment are believed to modulate synaptic transmission from MB neurons to MBONs , primarily through their presynaptic inputs onto the KCs ( Kim et al . , 2007; Qin et al . , 2012 ) . Some studies have indicated that DANs enhance KC>MBON transmission ( Cohn et al . , 2015; Owald et al . , 2015; Plaçais et al . , 2013; Pai et al . , 2013 ) , whereas others have suggested that DANs depress these synapses ( Aso et al . , 2014a; Hige et al . , 2015; Owald et al . , 2015; Séjourné et al . , 2011; Hattori et al . , 2017; Lewis et al . , 2015 ) . The sign of modulation may therefore depend upon the context . We predicted that , if M6 is the relevant MBON for courtship conditioning , then artificial activation of M6 should suppress courtship . Moreover , if MBγ>M6 transmission is modified by training , then M6 activation should be equally potent in experienced and naïve males , whereas MBγ activation should be either more or less potent in experienced males , depending upon whether training potentiates or depresses MBγ>M6 synapses . We tested these predictions using the thermosensitive cation channel TrpA1 ( open at 32°C and closed at 20°C ) ( Rosenzweig et al . , 2005 ) to activate either MBγ or M6 cells . To measure the extent of courtship suppression we used unreceptive virgin females ( pseudomated females ) as testers , which do not elicit significant courtship suppression in experienced males ( Keleman et al . , 2012 ) . For each condition , we determined a SI as the percentage reduction in courtship activity towards these unreceptive virgins in 10 min assays performed at 32°C compared to 20°C: SI = 100* ( 1-CI32/CI20 ) . We found that MBγ activation was significantly more potent in experienced males than in naïve males , in which it had only a small effect on courtship ( Figure 1C; Supplementary file 1 ) . By contrast , M6 activation suppressed male courtship with equal potency in both naïve and experienced males ( Figure 1D; Supplementary file 1 ) . We conclude from these data that courtship experience with mated females potentiates synaptic transmission from MBγ to M6 cells . To examine synaptic transmission between MBγ and M6 neurons , we used optogenetics . We generated a step-function channelrhopodsin variant , SFOCatCh , that combines mutations to increase the off kinetics ( SFO = C128S/D156A , [Yizhar et al . , 2011] ) with a single amino acid substitution to enhance the conductance of divalent cations ( CatCh = L132C , [Kleinlogel et al . , 2011] ) . We validated SFOCatCh by whole-cell patch clamp recording in olfactory projection neurons ( Figure 2—figure supplement 1 ) . We used SFOCatCh in conjunction with GCaMP6s ( Chen et al . , 2013 ) to monitor calcium responses in whole explanted brains of naïve males . The combination of SFOCatCh and GCaMP6s temporally uncouples the optical inputs required for activity manipulation and calcium imaging . In all experiments with SFOCatCh and GCaMP6s reported here , we imaged calcium responses during three consecutive 4 s periods , each of which was preceded by a 100 ms pulse of green , blue , or green light , respectively , to turn SFOCatCh OFF , ON , or OFF again ( Figure 2A ) . This protocol thus provides a pre-stimulus baseline , a during-stimulus response , and a post-stimulus response . To assess whether and how dopamine modulates MBγ>M6 transmission , we repeated this OFF/ON/OFF protocol 3 times at 3 min intervals: first without dopamine , then with either 0 . 1 mM or 1 mM dopamine delivered for the first second of each imaging period through a perfusion pipette positioned at the γ5 compartment , and finally following dopamine washout ( Figure 2A ) . We could not detect any robust calcium response in either the dendrites ( Figure 2B–D ) or axon termini ( Figure 2—figure supplement 2 ) of M6 when we activated MBγ with SFOCatCh in the absence of exogenous dopamine . However , a strong dose-dependent calcium response was consistently observed in trials with dopamine during the SFOCatCh ON period . In contrast , little or no response was observed during either the SFOCatCh OFF periods ( Figure 2B–D , Figure 2—figure supplement 2 ) or the SFOCatCh ON period after dopamine washout ( Figure 2B , C , Figure 2—figure supplement 2 ) . We obtained similar results when we applied the dopamine receptor agonist apomorphine rather than dopamine ( Figure 2E ) , or used CsChrimson ( Klapoetke et al . , 2014 ) rather than SFOCatCh as the optogenetic activator ( Figure 2—figure supplement 3 ) . The response to dopamine and SFOCatCh was completely abolished by application of the nicotinic acetylcholine receptor antagonist mecamylamine ( Figure 2F ) , which blocks synaptic transmission from KCs to MBONs ( Barnstedt et al . , 2016 ) . Together , these data indicate that cholinergic synaptic transmission from MBγ to M6 cells is initially weak but can be acutely potentiated by dopamine . Whereas we could not detect a strong calcium response in M6 MBONs upon MBγ activation in the absence of dopamine , others have observed calcium responses in various MBONs , including M6 , upon activation of KCs without application of dopamine or DAN stimulation ( Cohn et al . , 2015; Owald et al . , 2015 ) . We noted however that in our initial control experiments without dopamine , in which we sometimes performed multiple trials on the same sample , a calcium response could indeed be detected in the later trials . This suggests that stimulus history may account for some of the variability in MBON responses to KC stimulation in the absence of dopamine or DAN activation . To explore this possibility more rigorously , we activated MBγ neurons with SFOCatCh using the same OFF/ON/OFF protocol as before , now repeating the stimulus every minute . The initial stimuli , as previously observed in the trials without exogenous dopamine , did not elicit a detectable GCaMP6s response in M6 neurons . However , after 3–4 trials a significant calcium response was observed ( Figure 3A , B ) . This response increased upon each successive stimulation before reaching a plateau after approximately 20 trials . This response was blocked by the dopamine D1-type receptor antagonist SCH23390 ( Figure 3C ) , regardless of whether it was applied during the induction or plateau phase . This suggests that , upon repetitive stimulation of MBγ neurons , endogenous dopamine enables synaptic transmission to M6 neurons . The most likely source of this endogenous dopamine supply is the aSP13 neurons . Indeed , by shifting GCaMP6s from M6 to aSP13 , we confirmed that the aSP13 DANs respond in a similar manner as M6 to the repetitive activation of MBγ neurons ( Figure 3E , F ) To determine how long MBγ>M6 synapses remain potentiated after repetitive MBγ activation , we first induced potentiation with 30 pulses of MBγ activation at 1 min intervals , and then examined the response of M6 neurons to a single pulse of MBγ activation after 1 , 2 or 3 hr . Potentiation at MBγ>M6 synapses was barely diminished after 1 hr , but fell to about 50% of its initial level after 3 hr ( Figure 3D ) . The persistence of potentiation at MBγ>M6 synapses in these experiments is thus in line with the persistence of the courtship memory after a 30 min training period ( Keleman et al . , 2012 ) . Anatomically , DANs and MBONs innervating the same MB compartment , have the potential to form recurrent loops , with MBONs providing input to DANs ( Aso et al . , 2014a; Takemura et al . , 2017; Ichinose et al . , 2015; Eichler et al . , 2017; Owald et al . , 2015 ) . In particular , the axonal termini of M6 MBONs are closely apposed to the aSP13 dendrites ( Aso et al . , 2014a ) . We therefore tested whether activation of M6 neurons elicits a calcium response in aSP13 neurons by expressing SFOCatCh in M6 and GCaMP6s in aSP13 . Indeed , acute activation of M6 neurons produced a strong calcium response in aSP13 ( Figure 4A , B ) . This response was blocked by the NMDA receptor antagonist AP-5 ( Figure 4C ) , consistent with glutamatergic transmission from M6 cells . Activation of MBγ neurons with SFOCatCh also elicited a strong calcium response in aSP13 neurons ( Figure 4E , F ) that was also dependent on glutamatergic neurotransmission , as well as both cholinergic transmission and dopamine ( Figure 4G ) . Whereas the M6 response to MBγ activation was diminished in the post-stimulus OFF period in trials with dopamine ( Figure 2B , C ) , the response of aSP13 neurons to either M6 or MBγ activation persisted into the post-stimulus SFOCatCh OFF period ( Figure 4A , B , E and F ) . In each case , the calcium response in aSP13 gradually declined over a 2 min period ( Figure 4D and H ) . The persistent response of aSP13 neurons is not an intrinsic property of aSP13 neurons , since it was not observed when SFOCatCh was used to activate the aSP13 neurons themselves ( Figure 4—figure supplement 1 ) . Given that the response of aSP13 to MBγ or M6 activation is blocked by AP-5 , we infer that this persistent activity is induced by glutamatergic transmission from M6 cells . The persistent release of dopamine by aSP13 neurons for several minutes after stimulation could create a time window during which MBγ to M6 transmission is facilitated . Activation of MBγ neurons during this time window , as in our repetitive SFOCatCh activation experiments , would thus lead to further activation of M6 and aSP13 , thereby creating a recurrent feedback circuit . We lack a reliable tripartite genetic means to test directly whether silencing aSP13 neurons blocks the GCaMP6 response in M6 upon repetitive activation of SFOCatCh in MBγ . We could confirm , however , that this prolonged M6 response is blocked by AP-5 ( Figure 4—figure supplement 2 ) , which inhibits NMDA-type glutamate receptors and the persistent response of aSP13 ( Figure 4C ) . Activity of aSP13 neurons is strictly required during the training period of courtship conditioning ( Keleman et al . , 2012 ) . The data presented here suggest that activation of aSP13 during training could open a time window of a several minutes during which MBγ>M6 transmission is facilitated . In our training paradigm , males usually court mated females in a series of brief courtship on and off periods that could repetitively activate MBγ in the time window when aSP13 neurons are persistently activated and thus engage the recurrent MBγ>M6>aSP13 circuit , thereby potentiating MBγ>M6 transmission for a period of 2–3 hr . This leads to the somewhat counterintuitive prediction that M6 should not only act in memory retrieval , as generally assumed for MBONs , but should also be required for memory formation . To test this prediction , we conditionally silenced M6 neurons with a temperature-sensitive inhibitory form of dynamin ( UAS-shits ) , which blocks synaptic transmission at 32°C but not at 22°C ( Kitamoto , 2002 ) . Single males were trained and tested as before , but kept at 22°C except for a brief shift to 32°C either during training or during testing . Males in which M6 neurotransmission was blocked either during training or during testing had suppression indices indistinguishable from 0 . Thus , whereas synaptic transmission from aSP13 is only required during memory acquisition ( Keleman et al . , 2012 ) , M6 output is required during both acquisition and recall ( Figure 4I; Supplementary file 2 ) . We therefore propose that STM formation requires the MBγ>M6>aSP13 recurrent circuit , whereas readout occurs through other M6-dependent pathways . In this study we have identified and characterized a tripartite MBγ>M6>aSP13 recurrent circuit that is essential for courtship memory in Drosophila . Our behavioral and physiological data suggest the following model for the function of this feedback loop in short-term courtship memory . When a naïve male courts a mated female , the aSP13 and MBγ neurons may both be activated , perhaps in response to behavioral rejection and olfactory stimuli presented by the female , respectively . Dopamine released by aSP13 neurons potentiates transmission from MBγ to M6 neurons , which in turn provide a recurrent excitatory glutamatergic input back onto aSP13 neurons . Upon activation by M6 , aSP13 activity persists for several minutes , providing a short time window during which continued MBγ activity can further drive M6 and aSP13 . Thus sustained , aSP13 activity can lead to a longer-lasting accumulation of dopamine in the γ5 compartment , facilitating MBγ>M6 neurotransmission for up to 2–3 hr . The timescales for these physiological processes in ex vivo brain preparations broadly match the dynamics of courtship training and short-term memory formation . In our standard training paradigm , the male typically courts the female over several minutes , during which he performs a series of courtship bouts , each lasting for several seconds . As a result , a behavioral memory forms that lasts for several hours ( Keleman et al . , 2012 ) . Memory formation during training requires both M6 and aSP13 , consistent with the notion that it reflects activation of the recurrent circuit ( Figure 4 and [Keleman et al . , 2012] ) . Memory readout requires M6 but not aSP13 ( Figure 4 and [Keleman et al . , 2012] ) , and so evidently does not involve the recurrent circuit . We infer that M6 suppresses courtship through other , aSP13-independent , pathways , and that its ability to do so is independent of experience . The consequence of training is to provide MBγ neurons with access to this M6-dependent courtship suppression pathway ( Figure 1 ) . Two important open questions are , first , what mechanism underlies the persistent calcium response in aSP13 , and second , how does potentiation of MBγ>M6 synapses result in enhanced sensitivity to cVA , the hallmark of courtship memory ( Keleman et al . , 2012 ) . The persistent response in aSP13 is evidently not an intrinsic property of aSP13 , as it is not induced when aSP13 neurons themselves are activated . This observation would also likely exclude reciprocal excitation between aSP13 and other DANs ( Plaçais et al . , 2012 ) . Persistent aSP13 activity is induced in response to transient M6 activation , and is not associated with any persistent activity of M6 neurons themselves . Thus , it is also unlikely to involve feedback from aSP13 and M6 , although aSP13 >M6 synapses likely do exist ( Eichler et al . , 2017; Lin et al . , 2007 ) . One possibility is that aSP13 persistence reflects unusually prolonged activation of the glutamatergic M6 >aSP13 synapses , or perhaps lies within interposed but still unidentified circuit elements . Given that M6 neurons activate a courtship suppression pathway , the potentiation of MBγ>M6 neurotransmission may explain why MBγ activation suppresses courtship in trained but not naïve flies . But MBγ neurons likely do not specifically respond to cVA ( Caron et al . , 2013; Gruntman and Turner , 2013 ) , so this change alone cannot account for the enhanced sensitivity of trained flies to cVA . A small and variable subset of MB γneurons do receive input from the olfactory pathway that processes cVA , but cVA is not required during training ( Keleman et al . , 2012 ) and it is difficult to envision any other mechanism by which aSP13-dependent plasticity could be specifically restricted to the cVA-responsive MBγ neurons . It is formally possible that , despite the broad potentiation of MBγ output synapses upon training , it is only the contribution of the cVA-responsive MBγ neurons that drives courtship suppression when the male subsequently encounters as mated female . Alternatively , it has been suggested that M6 neurons encode a generic aversive signal ( Aso et al . , 2014b ) , and so specificity to cVA might instead arise in downstream circuits that selectively integrate M6 output with the innate cVA-processing pathway from the lateral horn . In this regard , it is interesting to note that other MBONs have been implicated in courtship learning ( Montague and Baker , 2016 ) or general aversion ( Aso et al . , 2014b ) , but M6 is the only MBON common to both . Late activation of the same aSP13 neurons in the time window of 8–10 hr after training is both necessary and sufficient to consolidate STM to LTM ( Krüttner et al . , 2015 ) . Thus , in the time window when STM would otherwise decay ( Keleman et al . , 2007 ) , reactivation of the same MBγ>M6>aSP13 recurrent circuit may instead consolidate it into LTM . The mechanism by which aSP13 neurons are reactivated is unknown , but is evidently dependent upon their activation within the MBγ>M6>aSP13 recurrent circuit during training . It will be interesting to find out how this late aSP13 reactivation mechanism might relate to the mechanism that underlies persistent aSP13 activity during training . In summary , our data suggest that a brief persistent activity of aSP13 neurons represents a neural correlate of courtship working memory , while the prolonged potentiation of MBγ>M6 synapses corresponds to STM . We propose that persistent activity of the dopaminergic neurons in the MBγ>M6>aSP13 feedback loop lays the foundation for formation of short-term courtship memory in Drosophila , and that later reactivation of the same recurrent circuit consolidates STM into LTM . Thus , in contrast to the prevailing view of memory progression in the Drosophila MB that distinct memory phases are located in different compartments or lobes ( Aso and Rubin , 2016; Davis , 2011; Pascual and Préat , 2001 ) , our data suggest that in the context of courtship conditioning , working memory , STM , and LTM all reside in the same γ5 compartment . Our conclusions do not preclude however , the involvement of other MB neurons in courtship memory ( Montague and Baker , 2016 ) as it is conceivable that modulation , potentially of the opposite sign , of the appetitive memory pathways could be critical for courtship learning ( Perisse et al . , 2016 ) . We therefore envision that distinct courtship memory types are not located in distinct circuits , but rather mediated by distinct processes within a common circuit . Encoding distinct memory phases within a common circuit may be an efficient mechanism for encoding memories for which the behavioral consequence is largely independent of timing and context ( Fusi et al . , 2005 ) . Flies for behavior experiments were reared in vials with standard cornmeal food at 25°C , or as indicated , at 60% humidity in a 12 hr:12 hr light:dark cycle . Flies for physiological experiments were reared on standard cornmeal food , supplemented with 500 μM all-trans-retinal , in dark . For behavioral and physiological experiments we used VT-Gal4 and VT-LexA lines obtained from the VT library , a collection of 2 kb enhancer fragments , generated following the strategy of ( Pfeiffer et al . , 2008 ) ( B . J . D . , unpublished data ) , UAS-Kir2 . 1 ( Nitabach et al . , 2002 ) , UAS-TNT/UAS-TNTQ ( Martin et al . , 2002 ) , UAS-Shits ( Kitamoto , 2002 ) , UAS-TrpA1 ( Rosenzweig et al . , 2005 ) , UAS-SFOCatCh ( VIE-260b ) ( B . J . D . , unpublished ) , 20xUAS-CsChrimson-tdTomato ( SuHwattp5 ) and LexAop2-opGCaMPs ( SuHwattp1 ) ( gift from Barret Pfeiffer ) , LexAop-IVS-GCaMP6s-p10 ( attp1 ) ( Chen et al . , 2013 ) . Pseudomated females were ( elav-Gal4/+UAS-SP/+ ) virgins ( Keleman et al . , 2012 ) . Courtship conditioning was performed as described ( Siwicki and Ladewski , 2003 ) . For training , solitary males ( aged for 5–6 days ) were placed in food chambers for 1 hr either with ( trained ) or without ( naïve ) a single mated female . After training each male was recovered , allowed to rest for 30 min and tested with a fresh mated female . Tests were performed in 10 mm diameter chambers and videotaped for 10 min ( JVC handyman , 30 GB HD ) . We used automated video analysis to derive a courtship index ( CI ) for each male , defined as the percentage of time over a 10 min test period during which the male courts the female . A MATLAB script ( permutation test ) ( Kamyshev et al . , 1999 ) was used to for statistical comparison of SIs between two groups . Briefly , the entire set of courtship indices for both naïve and trained flies were pooled and then randomly assorted into simulated naïve and trained groups of the same size as the original data . A SI was calculated for each of 100 , 000 randomly permutated data sets , and P values were estimated for the null hypothesis that learning equals 0 ( H0: SI = 0 ) or for the null hypothesis that experimental and control males learn equally well ( H0: SI = SIc ) . Fly brains and ventral nerve cords were dissected in Schneider’s insect medium and fixed in 2% paraformaldehyde ( PFA ) at room temperature for 55 min . Tissues were washed in PBT ( 0 . 5% Triton X-100 in phosphate buffered saline ( PBS ) ) and blocked using 5% normal goat serum ) before incubation with antibodies ( diluted in blocking solution in a volume of 200 μl per sample ) . Primary antibodies ( rabbit anti-GFP A-11122 from Molecular Probes at 2 μg/ml and mouse anti-Bruchpilot nc82 hybridoma supernatant from DSHB at 1 μg/ml ) were applied for 2–3 days . After a rinse and four 15 min washes in PBT , tissues were then incubated for 2–3 days with Alexa Fluor 488-conjugated goat anti-rabbit and Alexa Fluor 568-conjugated goat anti-mouse secondary antibodies ( Molecular Probes; 2 . 5 μg/ml and 5 μg/ml , respectively ) . Each of the antibody incubations were done for 4 hr at room temperature before placing the samples at 4 ˚C for the remainder of the incubation time . After a rinse and four 15 min washes in PBT , tissues were fixed with 4% PFA in PBS for 4 hr , followed by a rinse and four 15 min washes in PBT . Directly before mounting , tissues were rinsed and washed for 15 min in PBS to remove the Triton . The tissues were mounted on poly-L-lysine-coated cover slips and then dehydrated with 10 min ethanol baths of 30% , 50% , 75% , 95% and 3 × 100% followed by three 5 min washes in 100% xylene . Finally , mounted samples were embedded in xylene-based mounting medium ( DPX; Electron Microscopy Science , Hatfield , PA ) and dried for 2 days . Images were collected using an LSM710 confocal microscope ( Zeiss , Germany ) fitted with a Plan-Apochromat 20x/0 . 8 M27 objective . SFOCatCh was constructed from a synthetic ChR2 open reading frame with codon usage optimized for Drosophila , using mutagenic PCR to introduce the C128S and D156A substitutions to make it switchable ( Yizhar et al . , 2011 ) and the L132C mutation to increase cation conductance ( Kleinlogel et al . , 2011 ) . The resulting coding fragment was inserted into a modified UAS vector for site-specific insertion at the VIE-260b landing site . For ex vivo calcium imaging experiments , 5–7 days old naïve males were briefly anesthetized on ice and brains were dissected out in calcium free external saline ( ES ) containing: 103 mM NaCl , 3 mM KCl , 5 mM TES ( N-tris[hydroxymethyl]methyl-2-aminoethane sulfonic acid , a buffer chemical with peak performance around pH7 . 5 ) , 10 mM trehalose , 10 mM glucose , 26 mM NaHCO3 , 1 mM NaH2PO4 , 4 mM MgCl2 , 7 mM sucrose , pH 7 . 4 , 275 mOsm ( Gu and O'Dowd , 2006 ) . The brain explants were transferred into a custom-made imaging chamber and mounted with anterior side up . Brains were perfused with ES supplemented freshly with 2 mM calcium , at speed 2 mL/min , pre-saturated with mixture of 95% O2/5% CO2 . All two-photon imaging were performed using 40x N . A . 0 . 75 water-immersion objective ( N-Achroplan , Zeiss ) , on LSM 7 MP microscope ( Zeiss ) with a Ti:sapphire laser ( Chameleon Vision II , Coherent , Santa Clara , CA ) . GCaMP was excited at 900 or 920 nm and emission signals were collected by GaAsP photomultiplier tubes ( PMTs ) . Frame images ( 256 × 256 pixels ) were acquired at 5–10 Hz . The region of interest ( ROI ) covers the entire bilateral medial γ5 lobe in MB . For consistency , imaging focus was kept approximately at the same level in different animal guided by axon position of M6 or aSP13 . For SFOCatCh experiments , neurons were activated with whole field light from a mercury lamp ( X-cite 120 PC , Excelitas Technologies ) . Light was filtered by 38HE 470/40 nm and 43HE 550/25 nm ( Zeiss ) , and pulse duration was controlled by a TTL-triggered shutter ( Uniblitz , Rochester , NY ) . Light density was calculated by dividing light power to fields of view ( FOV ) of objective: 480 nm , 0 . 28 mW/mm2 , and 540 nm , 0 . 86 mW/mm2 . For CsChrimson experiments , LED ( pE-4000 , CoolLED ) was used to deliver 2 ms light pulse as stimulation . Light ( peak 635 nm ) trains were further filtered by 635/18 nm ( Semrock , Rochester , NY ) and delivered at 30 Hz for 1 s . Light density was calculated as 0 . 126 mW/mm2 when staying persistent during measurement at 635 nm . Dopamine ( DA ) solution was prepared freshly before each experiment . DA solution was back-filled into a glass electrode with fine tip ( ~3 μm ) shortly before each focal application . DA was injected ( 1 s , 5 p . s . i ) in the medial γ5 lobe by Picospritzer-III ( Parker , Cleveland , OH ) ( Cassenaer and Laurent , 2012 ) . We controlled for dopamine diffusion by co-loading a fluorescent dye ( Texas red 3000 , 0 , 1 mg/ml ) to the focal pipette and monitoring the dye distribution area during two-photon scanning . GCaMP imaging data was analyzed in a custom program modified from Sun et al . , 2016 . Fluorescence changes in intensity were calculated as ΔF/F , where F is the average signals of the five frames before each stimulation . ROIs were chosen contained single optical plate with the neural processes of interest . Peak ΔF/F represents mean ΔF/F in continuous five frames responses ( SFOCatCh ON ) acquired during LTP procedure . All data represented as mean ± s . e . m . All statistical analyses were performed in Graphpad Prism 7 . 0a . Data were analyzed by unpaired Student’s t-test or one-way ANOVA test with post hoc Tukey’s range tests . For ex vivo patch-clamp recording from projection neurons ( GH146-Gal4 > UAS SFOCatCh + UAS -mCD8::GFP ) in antennal lobe brain explants were prepared as described in Ca imaging section , and recordings were performed as previously described ( Gu and O'Dowd , 2006 ) . Electrodes ( 5–7 MΩ ) were filled with saline solution containing 140 mM potassium aspartate , 10 mM HEPES , 1 mM KCl , 4 mM MgATP , 0 . 5 mM Na3GTP , 1 mM EGTA , pH 7 . 3 , and 265 mOsm . Cell-attached recording was performed in voltage-clamp mode with 0 mV holding potential . Whole-cell patch-clamp recording was performed in current-clamp mode with resting membrane potential around 55–65 mV . Signals were acquired by MultiClamp 700B amplifier , digitized at 10 kHz , and low-pass filtered at 5 kHz .	Memories help to shape behavior , and can last from a few seconds to an entire lifetime . Working memory , in which information is temporarily held available for use in an ongoing task , is the most fleeting form of memory . It relies on persistent activation of a network of nerve cells or neurons that represent the information in question . Strengthening the connections between those neurons may result in a longer-lasting memory . But the mechanisms that support the formation of memories of different durations are not fully understood . Zhao et al . have now explored these mechanisms in the fruit fly by studying memory for courtship behavior . Inexperienced male fruit flies will attempt to court both virgin females and females who have recently mated . But the latter reject courtship attempts , and male fruit flies therefore learn to avoid them . This is known as courtship memory , and it relies on a network of neurons within a region of the fruit fly brain called the mushroom body . Within the mushroom body , dopamine neuron sends signals to a neuron called the Kenyon cell , which in turn sends signals to a mushroom body output neuron . The latter activates circuits responsible for decision-making and movement . But it also activates the dopamine neuron , thereby forming a recurrent circuit or loop . When the courtship is rejected , the dopamine neuron becomes persistently active , which generates a working memory of the experience . If the circuit is activated again during this period of persistent firing , the working memory may be converted into a longer-lasting memory . The results of Zhao et al . provide insights into the mechanisms by which memories form and undergo strengthening . They suggest that distinct processes within a single neural circuit give rise to memories of different durations . Recurrent loops are also present within the brains of mammals . Similar processes may thus support the formation and persistence of our own memories .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2018	Persistent activity in a recurrent circuit underlies courtship memory in Drosophila
In competitive situations , winning depends on selecting actions that surprise the opponent . Such unpredictable action can be generated based on representations of the opponent’s strategy and choice history ( model-based counter-prediction ) or by choosing actions in a memory-free , stochastic manner . Across five different experiments using a variant of a matching-pennies game with simulated and human opponents we found that people toggle between these two strategies , using model-based selection when recent wins signal the appropriateness of the current model , but reverting to stochastic selection following losses . Also , after wins , feedback-related , mid-frontal EEG activity reflected information about the opponent’s global and local strategy , and predicted upcoming choices . After losses , this activity was nearly absent—indicating that the internal model is suppressed after negative feedback . We suggest that the mixed-strategy approach allows negotiating two conflicting goals: 1 ) exploiting the opponent’s deviations from randomness while 2 ) remaining unpredictable for the opponent . Even the most powerful backhand stroke in a tennis match loses its punch when the opponent knows it is coming . Competitions that require real-time , fast-paced decision making are typically won by the player with the greatest skill in executing action plans and who are able to choose their moves in the least predictable manner ( Camerer et al . , 2015; Nash , 1950; Morgenstern and Von Neumann , 1953; Lee , 2008 ) . Yet , how people can consistently achieve the competitive edge of surprise is not well understood . The fundamental challenge towards such an understanding lies in the fact that our cognitive system is geared towards using memory records of the recent selection history to exploit regularities in the environment . However , as suggested by decades of research ( Wagenaar , 1972; Baddeley , 1966; Rapoport and Budescu , 1997; Arrington and Logan , 2004; Mayr and Bell , 2006 ) , these same memory records will also produce constraints on current action selection that can work against unpredictable behavior . One such memory-based constraint on unpredictable action selection is that people often tend to repeat the last-executed action plan . A considerable body of research with the ‘voluntary task-switching’ paradigm ( Arrington and Logan , 2004; Mayr and Bell , 2006 ) has revealed a robust perseveration bias , even when subjects are instructed to choose randomly between two different action plans on a trial-by-trial basis––a regularity that in competitions could be easily exploited by a perceptive opponent . Another important constraint is the win-stay/lose-shift bias , that is a tendency to repeat the most recently reinforced action and abandon the most recently punished action . Reinforcement-based action selection does not require an internal representation of the task environment and is therefore often referred to as ‘model-free’ . Previous work has revealed that reinforcement learning can explain some of the choice behavior in competitive situations ( Cohen and Ranganath , 2007; Erev and Roth , 1998; Lee et al . , 2012 ) . Yet , players who rely on reinforcement-based selection can also be counter-predicted by their opponent , or run the risk of missing regularities in their opponents’ behavior . Therefore , recent research indicates that when performing against sophisticated opponents , model-free choice can be replaced through model-based selection , where choices are based on a representation of task-space contingencies ( Gläscher et al . , 2010 ) , including beliefs about the opponent’s strategies ( Donahue et al . , 2013; Tervo et al . , 2014 ) . Model-based selection is consistent with the view of humans as rational decision makers ( Nash , 1950; Morgenstern and Von Neumann , 1953 ) , yet also has known limitations . For example , it depends on attentional and/or working memory resources that vary across and within individuals ( Otto et al . , 2013a ) . In addition , people are prone to judgement and decision errors , such as the confirmation bias , that get in the way of consistently adaptive , model-based selection ( Abrahamyan et al . , 2016 ) . In light of the shortcomings of both , model-free choice and model-based strategies it is useful to consider the possibility that in some situations , actors can choose in a memory-free and thus stochastic manner ( Donahue et al . , 2013; Tervo et al . , 2014 ) . Memory-free choice would establish a ‘clean-slate’ that prevents the system from getting stuck with a sub-optimal strategy and instead allows exploration of the full space of possible moves . Moreover , it reduces the danger of being counter-predicted by the opponent ( Walker and Wooders , 2001; Chiappori et al . , 2002 ) . At the same time , an obvious drawback of stochastic choice is that without a representation of the opponent , systematic deviations from randomness in the opponent’s behavior remain undetected and therefore cannot be exploited . In addition , just as is the case for model-based selection , stochastic selection puts high demands on cognitive control resources ( Baddeley et al . , 1998 ) and therefore it is not clear under which circumstances people can consistently ignore or suppress context representations in order to choose in a memory-free manner ( Rapoport and Budescu , 1992 ) . As the model-based and the memory-free strategy both come with strengths and limitations , one potential solution is that people use a simple heuristic to move back and forth between these two modes of selection . Specifically , positive feedback ( i . e . , wins on preceding moves ) could serve as a cue that the current model is adequate and should be maintained . In contrast , negative feedback might serve as a signal that the current model needs to be suspended in favor of a memory-free mode of selection that maximizes exploration and unpredictability . In the current work , we used an experimental paradigm that provides a clear behavioral signature of model-based versus memory-free choices as a function of preceding win versus loss feedback . We found that following win feedback , people tended to choose their next move both on the basis of recent history and a more global model of the opponent . However following losses , we did not simply see choice behavior revert back towards simple memory-driven biases . Rather choices were less determined by recent history and task context––in other words more stochastic . In addition , we present neural evidence that loss feedback literally ‘cleans the slate’ by temporarily diminishing the representation of the internal model ( Donahue et al . , 2013; Tervo et al . , 2014; Kolling et al . , 2016a ) . Our experimental situation marries the voluntary task-switching paradigm , where people switch between action rules , with a two-person , matching-pennies game . Different from the standard matching-pennies game , winning a trial was defined by whether or not players chose matching action rules , rather than simple response choices . Specifically , on each trial players saw a circle , either on the bottom or the top of a vertically arranged frame ( see Figure 1a ) . Participants chose on each trial between the ‘freeze’ rule , which keeps the circle at the same location , and for which responses had to be entered at two vertically arranged keys on the right side of the keyboard , or the ‘run’ rule , which moves the circle to the opposite end of frame , and for which responses had to be entered at two vertically arranged keys on the left side . On a given trial , for the freeze rule , participants had to press the one right-hand key that was spatially compatible to the current circle location; for the run rule they had to press the one left-hand key that was incompatible with the current circle position . Wins versus losses were signaled through a smiley or frowny face at the place of the post-response circle position . Players were assigned either the role of the ‘fox’ or the ‘rabbit’ . Foxes win a given trial when they ‘catch the rabbit’ , that is when they pick the same rule as the rabbit on that trial . Rabbits win when they ‘escape the fox’ , that is when they pick the rule not chosen by the fox . We exposed each player to a set of simulated opponents that differed in their average of switching rules from one trail to the next ( e . g . , 20% , 35% , 50% , 65% , and 80% ) . Otherwise , these simulated opponents made choices randomly , and did not respond in any manner to the player’s behavior . Variations in opponents’ switch rate provide a diagnostic indicator of both model-based and stochastic behavior ( Figure 1b ) . Specifically , a model-based agent should appreciate the fact that when playing against an opponent who switches frequently between run and freeze rules ( i . e . , p>0 . 5 ) , it is best to switch rules after a win ( i . e . , ‘following along with the opponent’ ) , but to stick with the same rule after a loss ( i . e . , ‘waiting for the opponent to come to you’ ) ; the opposite holds for opponents with a low switch rate ( i . e . , p<0 . 5 ) . Thus , model-based behavior would produce a combination of the filled green and red switch-rate functions in Figure 1b . In contrast , a memory-free agent should produce random behavior ( i . e . , a switch rate close to p=0 . 5 ) irrespective of the opponent’s strategies ( i . e . , the blue function in Figure 1b ) . Thus , our hypothesis of a feedback-contingent mix between model-based and stochastic behavior predicts an increase of players’ switch rate as a function of their opponents’ switch rate for post-win trials ( the filled green function in Figure 1b ) , but a switch rate close to p=0 . 5 irrespective of the opponent’s switch rate on post-loss trials ( the blue function in Figure 1b ) . Figure 1b also shows how the lower-level perseveration bias and the win-stay/lose-shift bias would affect the data pattern . In our rule-selection version of the matching-pennies game , each rule is associated with two specific separate response options , only one of which is ‘allowed’ for the currently chosen action rule ( i . e . , ‘freeze’ vs . ‘run’ rule ) . This enabled us to determine if an increase in stochasticity is specific to the generation of action choices , or alternatively due to an unspecific increase of information-processing noise . In the latter case , greater choice stochasticity should go along with more action errors . In this paradigm , participants can make two types of such errors: They can either fail to pick the set of response keys that is consistent with a chosen rule ( e . g . , right side keys when the intended rule is ‘run’ ) , but then execute the response that is consistent with the intended rule ( e . g . , incompatible response on the right side ) , or they could correctly pick the side that is consistent with the intended rule , but then execute the wrong response option ( e . g . , a compatible response on the left side ) . Without knowing subjects’ intended choice on a given trial , we cannot distinguish between errors types . However , either one of these can be interpreted as an action error that occurs independently of the choice between rules . If stochasticity affects information processing in an unspecific manner then we should find that such errors covary with choice stochasticity , both across conditions and across subjects . We also wanted to ensure that our main conclusions are not limited to the rule-selection paradigm . Therefore , we attempted to replicate our basic pattern of results in Experiment three in a standard matching-pennies paradigm with simple response choices ( but no way of distinguishing choice stochasticity from unspecific information-processing stochasticity ) . To test the prediction of loss-induced stochastic behavior , we cannot simply contrast the slopes of post-win and post-loss switch-rate functions . Such a comparison would not differentiate between a pattern of post-loss and post-win functions with the same slope but opposite signs ( as would be consistent with the model-based choice strategy , see Figure 1b ) and the predicted pattern of more shallow slopes following losses . Therefore , as a general strategy , we tested our main prediction by comparing slopes after selectively inverting the labels for the opponent switch-rate in the post-loss condition ( e . g . , 80% becomes 20% ) . This allows direct comparisons of the steepness of post-win and post-loss switch-rate functions . In the SI , we also present results from standard analyses . Our behavioral indicator of a mix between model-based and stochastic behavior is expressed in players’ switch rate , which can also be affected by the perseveration and win-stay/lose-shift bias ( see Figure 1b ) . In standard sequential-decision paradigms it is difficult to distinguish between stochastic behavior and low-level choice biases . Therefore , we attempted to obtain a realistic characterization of the various influences on choice behavior by using a simple choice model to predict the probability of switch choices pswitch: ( 1 ) Pswitch= exp ( os∗ ( ms − ( wl +1 ) ∗ . 5∗ sm ) −pe+wl∗−ss ) 1+ exp ( os∗ ( ms − ( wl +1 ) ∗ . 5∗ sm ) −pe+wl∗−ss ) with: os = ln ( pos/ ( 1- pos ) ) ; post-win: wl = 1 , post-loss: wl = −1; where pos is the opponent’s switch rate , which is translated into its log-odds form ( os ) ; wl codes for wins versus losses on trial n-1 . The parameter ms ( model strength ) represents the strength of the model of the opponent ( ms = 1 would indicate direct probability matching between the opponent’s and the player’s switch probability ) . The parameter sm ( strategy mix ) represents the degree to which the model-based choice is changed on post-loss relative to post-win trials; a negative sm parameter would indicate suppression of the model in favor of stochastic choice following losses . In addition , a positive pe ( persevertion effect ) parameter represents the tendency to unconditionally favor the previously selected choice , and a positive ss ( win-stay/lose-shift ) parameter expresses the strength of the win-stay/lose-shift bias . We present predictions from this model in Figure 2 , and report additional details of the modeling results in section Modeling Results . Experiment 1 . In this experiment , we establish the basic paradigm . As shown in the upper-left panel of Figure 2 , participants increased their switch rate as a function of their opponents’ switch rates following win trials . In contrast , on post-loss trials , the change in players’ switch rate ( as a function of their opponents’ switch rate ) was much smaller than on post-win trials and it was centered at p=0 . 5––a pattern that is consistent with largely stochastic choice . The condition with an opponent switch rate of p=0 . 5 most closely resembles previous studies that have reported a win-stay/lose-shift bias in competitive situations ( Cohen and Ranganath , 2007 ) . In fact , for this condition , we did find a significantly higher switch rate after losses than after wins , indicating that reward-based choices are one factor that affects choice . Following win trials , participants’ switch rate follows opponents’ switch rate when the opponents’ switch rate was low , but only in a muted manner when the opponents’ switch rate was high ( i . e . , the switch-rate function was less than 1 . 0 ) . We attribute this reluctance to fully endorse the model-based strategy to the influences of counteracting , lower-level , win-stay/lose-shift and perseveratory tendencies . Indeed , as will be described in greater detail in the section Modeling Results , results from applying our choice model to the data indicate that ( a ) a strong tendency towards model-based choices on post-win trials , ( b ) an increase of stochastic choice on post-loss trials , ( c ) a general perseveratory tendency , and ( d ) a win-stay/lose-shift bias all contribute to the overall choice behavior . The Figure 2—figure supplement 1–3 provide additional information about determinants of choice behavior and also of participants’ success rate . Experiment 2 . Following feedback from the previous trial , participants had only 300 ms to choose their move for the next trial in Experiment 1 . Therefore , one might argue that the observed stochastic choice is simply a result of negative feedback temporarily interfering with model-based selection ( Otto et al . , 2013a ) . To examine this possibility , we manipulated the inter-trial-interval ( ITI ) in Experiment 2 between 300 ms and 1000 ms . As shown in Figure 2 , this manipulation had no effect , indicating that stochastic choice is not due to loss-induced processing constraints . Experiment 3 . The fox/rabbit task was modeled after the voluntary task-switching to allow us to distinguish between choice stochasticity and more general increase of noise in the cognitive system . However , it is important to explore how the observed pattern might change with less complex response rules than used in this paradigm . We therefore implemented in Experiment 3 simple choices without any contingencies on external inputs ( i . e . , the fox wins when selecting the same up or down location as the rabbit , and vice versa ) . Here , we generally found a stronger expression of model-based choice following both losses and wins , and also much less perseveration bias ( Figure 2 ) . Yet the win-loss difference in switch-rate slopes remained just as robust as in the other experiments . Thus , the more complex actions that players had to choose from in Experiments 1 and 2 may have suppressed the overall degree of model-based action selection ( Otto et al . , 2013b ) . However , response rule complexity did not appear to affect the win/loss-contingent difference in the relative emphasis on model-based versus stochastic choices . Experiment 4a . It is possible that the observed pattern of results is specific to experimental situations with a strong variation in simulated , opponent switch rates . To examine the degree to which this pattern generalizes to a more natural , competitive situation , we used in Experiment 4a pairs of participants who competed with each other in real time , with one player of each dyad acting as fox , the other as rabbit ( see Figure 2—figure supplement 4 for a comparison between the competitive Experiment 4a and the non-competitive Experiment 4b ) . Obviously , the naturally occurring variation in switch rates was much lower than in the experiments using simulated opponents ( see distributions of individuals average switch rates in Figure 2 ) . Nevertheless , the estimated slopes linking players’ switch rates to opponents’ switch rates showed a very similar pattern as in the other experiments with simulated opponents . We applied our choice model both to the group-average switch rates for each experiment , and to the individuals-specific switch rates . Table 1 shows the estimated parameters for each of the four experiment with simulated opponents , as well as model fits ( R2 ) for the group-level data . We found that each of the four different factors ( i . e . , model strength , suppression of model/strategy mix , perseveration effect , and win-stay/lose-shift bias ) were relevant for characterizing participants’ choices . For condition-average data , the model strength parameter ( ms ) ranged between . 48 and . 87 , indicating that overall , the opponent’s switch rate affected the participant’s switch rate in an incentive-compatible manner . Average ms values below 1 . 0 indicate that participants overall engaged in ‘imperfect’ probability matching ( ms = l0 . 0 would indicate perfect probability matching; ms >1 . 0 would indicate a maximizing tendency ) . This pattern is consistent with the previous literature , which suggests that probability matching is the dominant , albeit suboptimal strategy in serial decision tasks ( James and Koehler , 2011; Gaissmaier and Schooler , 2008 ) . The individual-specific parameter estimates also allowed us to examine the degree to which the different influences on choice were tied to competitive success . To this end , we entered each individual’s , four parameter estimates as fixed-effect predictors into a two-level regression analysis with experiment as a random factor and overall success ( i . e . , probability of win trials ) as criterion variable . While on average , pe and ws indicated the expected perseveration and win-stay/lose-shift biases ( i . e . , pe <0 and wl <0 ) , there were substantial individual differences in these parameters , including individuals with alternation or win-shift/lose-stay biases ( i . e . , pe >0 and ss >0; Table 1 ) . Given that any bias implies a deviation from optimal performance , we coded these two parameters in absolute terms ( we obtained similar results with signed values ) . As shown in Table 2 , model strength had a highly robust positive effect on success , whereas either a perseveration or an alternation bias reduced the amount of money earned; no corresponding effect was found for the win-stay/lose-shift parameter . As would be expected , the main effect of the strategy-choice parameter was positive , implying that less stochastic behavior after losses produced greater overall success . In Experiment 4a , participants were paired up to play against each other . Thus , here we had to use the natural , within-session variability of the opponent of each player for the opponent switch rate variable in a trial-by-trial version of our choice model . Therefore , the pos parameter was calculated as a running average of the opponent’s switch rate within each block . The latest running average from block n-1 was used as a starting value for block n ( p=0 . 5 for the first block ) , allowing some carry-over of prior knowledge of the opponent’s previous-block behavior . Results from this model are shown in the bottom row of Table 1 . Not surprisingly , the model strength ( ms ) was substantially smaller than in the preceding experiments , but still significantly larger than 0 . The perseveration effect ( pe ) and the win-stay/lose-shift bias ( ss ) were roughly in a similar range as in the remaining experiments . Importantly , the suppression of the model parameter ( sm ) was also statistically significant and of about the same size as the model strength parameter ( ms ) , indicating that on post-loss trials the effect of the model is essentially eliminated . We again used a multi-level regression model with participants grouped within dyads to predict each participant’s success ( in terms of probability of win trials ) as a function of the four model parameters . As in the preceding model ( Table 2 ) , we again used absolute values from the perseveration and the win-stay/lose-shift scores in order to capture biases in in either direction ( very similar results would have been obtained with signed values ) . Results showed greater reliance on the model , a smaller tendency to disregard the model after losses ( i . e . , a less negative sm score ) , a smaller , absolute perseveration score , and a larger absolute win-stay/lose-shift bias all contributed to greater success ( Table 3 ) . Aside from the somewhat surprising result for the win-stay/lose-shift parameter , the overall pattern was qualitatively very similar to the results from the simulated-opponent experiments . Combined , the modeling results reveal that choices are influenced by low-level influences ( perseveration and win-stay/lose-shift bias ) as well by a model of the opponent’s strategy . Most importantly , we found that over and above these previously established influences , the parameter reflecting a loss-contingent reduction of the model had a robust influence on choice behavior . Moreover , the different parameters had distinct effects on individual differences in competitive success , with the loss-contingent reduction of the model ( i . e . , increase in stochasticity ) clearly representing a sub-optimal influence . As described in the overview section , different from standard choice paradigms ( Daw et al . , 2006; Muller et al . , 2019 ) , the current paradigm allows us to distinguish between stochasticity during the choice between action rules and general information-processing noise ( Kane et al . , 2017 ) . If the loss-induced choice stochasticity is due to a general increase in information processing noise then we should see that greater stochasticity goes along with more errors and possibly also with slower responses . Figure 3 shows each individual’s degree of model-based choice ( expressed in terms of absolute switch-rate slopes ) after loss and win trials and as a function of both RTs or error rates . In most experiments , there was a slight increase in error rates following loss trials ( see marks beneath the x-axis ) . However , across individuals , the substantial reduction in model-based choice after loss trials was not associated with a consistent increase in error rates or RTs . Likewise , in multilevel regression models with the absolute switch-rate slopes as dependent variable , the post-win/loss contrast remained highly robust after controlling for RTs and errors as within-subject fixed effects ( range of t-values associated with the post-win/loss predictor: 3 . 96-10 . 78 ) . So far , we have established that participants were more sensitive to their opponents’ global strategies ( i . e . , the average switch rates ) following win than following loss trials . Next , we examined the degree to which these win-loss differences generalized to players’ consideration of the recent history of their opponents’ and their own choices . To this end , we used multi-level logistic regression models with the switch/repeat choice as criterion . The models included the trial n-1 to n-3 switch/repeat decisions for opponents and for players , along with the opponents’ overall switch rate and were separately run for post-loss and post-win trials to generate the coefficients presented in Figure 4 . To directly compare the size of the coefficients , irrespective of their sign , we again reversed the labels , both for the opponents’ global switch rate , but also for both the opponent’s and the player’s n-1 to n-3 switch/repeat decisions ( e . g . , switch becomes repeat; see sections Analytic Strategy for Testing Main Prediction and History Analyses ) . Consistent with the prediction that switch/repeat choices following losses are less dependent on recent history , the coefficients for the opponents’ history and also the players’ own history were in most cases substantially lower after loss than after win trials . In figure supplements to Figure 4 , we also show the signed coefficients as well as the results of an alternative analysis that does not require the reversal of labels . Research with animal models and human neuroimaging work indicates that midfrontal brain regions , such as the anterior cingulate cortex are involved in action-relevant representations and in the gating between different modes of action selection ( Daw et al . , 2005; Behrens et al . , 2007; Holroyd and Coles , 2002 ) . Further , a large body of research suggests that midfrontal EEG activity in response to action feedback contains prediction error signals ( Cohen and Ranganath , 2007; Cavanagh et al . , 2012; Gehring et al . , 1993; Cavanagh and Frank , 2014; Cohen et al . , 2011; Luft , 2014 ) , which in turn are reflective of action-relevant expectancies ( i . e . , the current task model ) . Therefore , it is theoretically important to link our behavioral results to this broader literature . Specifically , it would be useful to show that ( a ) only on post-win , but not on post-loss trials , the midfrontal EEG signal contains information about the choice context/model , and ( b ) that the context information contained in the EEG signal is in fact predictive of upcoming choices . In Experiment 5 , we assessed EEG while participants played the fox/rabbit game against three different types of opponents ( 25% , 50% , 75% switch rate ) . The ITI was 700 ms to capture feedback-related EEG signals developing prior to the onset of upcoming trials . The behavioral results were consistent with the other experiments ( see Figures 2 , 3 and 4 ) . We conducted a two-step analysis of the EEG signal . In the first step , we tested the prediction that the mid-frontal EEG signal contains less information about the choice-relevant context after loss-feedback than after win-feedback . To this end , we regressed trial-to-trial EEG signals on A ) the opponent’s overall switch rate , B ) the opponent’s lag-1 switch/no-switch , C ) the player’s lag-1 switch/no-switch , and D ) the interaction between A ) and B ) , that is between the local and global switch expectancies . The latter term was included to capture the fact that if feedback-related EEG reflects expectancies about opponents’ switch rates , local switch expectancies may depend on the global switch-rate context ( Cavanagh and Frank , 2014 ) . The standardized coefficients shown in Figure 5a indicate the amount of information about each of the four context variables that is contained in the mid-frontal EEG signal . As apparent , the EEG signal showed a robust expression of the history/context variables following win feedback . Following loss feedback , context information is initially activated , but then appears to be suppressed compared to post-win trials , and trends towards zero at the end of the feedback period . Accordingly , coefficients were significantly larger in post-win trials than in post-loss trials , opponents’ overall switch rate: b = 0 . 07 , se = 0 . 01 , t ( 25 ) =5 . 22 , p<0 . 001 , opponents’ lag-1 switch/no-switch: b = 0 . 04 , se = 0 . 01 , t ( 25 ) =4 . 14 , p=0 . 001 , player’s lag-1 switch/no-switch: b = 0 . 01 , se = 0 . 009 , t ( 25 ) =1 . 19 , p=0 . 24 , interaction between opponents’ overall switch rate and lag-1 switch: b = −0 . 08 , se = 0 . 01 , t ( 25 ) =-7 . 22 , p<0 . 001 . Given that feedback is related to subject’s propensity of switching on the upcoming trial , it is in principle possible that these coefficients simply reflect preparation or increased effort for the upcoming switch . However , as we show in Figure 5—figure supplement 1 , controlling for the upcoming switch has negligible effects on results . These analyses also show that while there is detectable information about the upcoming switch/no-switch choice , the decodability of the upcoming choice is not modulated by win/loss feedback ( see also , 14 ) . Our analytic strategy deviates from the standard approach of analyzing the EEG signal in terms of feedback-locked , event-related potentials ( ERPs; see Figure 5b ) . We used our approach because we did not have a-priori predictions about how exactly the combination of different history/context variables would affect ERPs . More importantly , our regression-based approach naturally yields trial-by-trial indicators of the expression of context-specific information , which can be used in the second step of our analysis ( see below ) , and which would be difficult to obtain through standard ERP analyses . In Figure 5—figure supplement 3 , we also show that the ERP results are indeed generally consistent with a prediction-error signal that is more strongly modulated by the choice context after wins than losses . The conclusion that post-loss stochastic behavior occurs because context representations are suppressed , would be further strengthened by evidence that the information contained in the EEG signal is actually relevant for upcoming choices . Therefore , as the second step , we conducted a psychophysical interaction ( PPI ) analysis ( Friston et al . , 1997 ) . In a multi-level , logistic regression analysis , we predicted players’ trial n switch choices , based on 1 ) the set of four context variables from the preceding analysis for trial n-1 , 2 ) the trial n-1 residuals from the preceding analysis ( reflecting trial-by-trial variations in the EEG signal after controlling for the four context variables ) , ( Morgenstern and Von Neumann , 1953 ) and the corresponding four interactions between the residuals and the context variables . As shown in Table 4 , we found for post-win trials significant main effects for the residual EEG signal and all context variables . Most importantly , the residual EEG signal modulated how the upcoming choice was affected by the opponent’s n-1 switch/repeat . These results indicate that the information about context variables contained in the EEG signal is indeed relevant for choices . Given the reduced context representation following losses ( see Figure 5 ) , one might expect that there is not sufficient trial-by-trial variability in such information to influence choices . However , it is also possible that even the remaining variability still has predictive power . Therefore , the post-win/post-loss difference in predicting choices is theoretically less informative than the finding of reduced context representations in the EEG-signal per se and the fact that these representations generally predict upcoming choices . Nevertheless , Table 4 shows the results also for post-loss trials . The relationship between the EEG representation of the global , opponent switch rate and the upcoming choice was as strong as for post-win . For the local history variables , we found robust effects only following win trials , but not following loss trials . Note , that the pattern of identical signs for the EEG-behavior relationship across post-win and post-loss trials and the flipped signs for the opponent switch rate/EEG relationship ( Figure 5—figure supplement 2 ) is consistent with the reversal of the relationship between opponent , overall switch rate and player switch rate depending on win or loss feedback ( e . g . , Figure 2 ) . As a final step , we also examined if variations in the strength of history/context representations can account for individual differences in choice behavior . We derived for each individual and predictor , the average , standardized coefficient from the analysis presented in Figure 5 across the 300 ms to 700 ms interval . Separately for post-win and post-loss trials , we correlated these scores with two behavioral measurements: 1 ) individuals’ switch-rate functions as an indicator for model-based choice ( see Figure 2 ) and 2 ) the overall rate of winning . For post-loss trials , we again used opponent-related predictors with reversed labels ( see EEG Recording and Analysis section for details ) . Thus , for all analyses , more positive scores are indicative of individuals with more model-conform behavior . As shown in Figure 6 , coefficients from post-win EEG signals generally predicted the variability among individuals in the degree of model-based adaptation and the rate of winning ( except for coefficients of n-1 player’s switch ) . In contrast , such relationships were absent for post-loss trials . Here , we also found significant post-win/post-loss differences . For the switch-rate function slopes , post-win/post-loss differences were present for the opponents’ lag-1 switch/no-switch contrast , z ( 25 ) =2 . 87 , p=0 . 003 , and the interaction between opponent’s overall switch rate and lag-1 switch/no-switch choice , z ( 25 ) =4 . 26 , p<0 . 001 , but not for opponent’s overall switch rate , z ( 25 ) =1 . 38 , p=0 . 16 or the player’s lag-1 switch/no-switch , z ( 25 ) =-0 . 76 , p=0 . 45 . Similarly , for the overall rate of winning , we found significant differences for opponents’ lag-1 switch/no- switch choice , z ( 25 ) =2 . 33 , p=0 . 02 , and the interaction between opponents’ overall switch rate and the lag-1 switch/no-switch , z ( 25 ) =2 . 21 , p=0 . 02 , but again not for opponents’ overall switch rate , z ( 25 ) =1 . 54 , p=0 . 12 , and the player’s lag-1 switch/no-switch , z ( 25 ) =-1 . 18 , p=0 . 23 . Combined , these individual differences results suggest that the degree to which history/context variables are represented in the EEG signal following win feedback , predicts both individuals’ reliance on the model of the opponent and their overall competitive success . These relationships are largely absent on post-loss trials . While the absence of a post-win/post-loss difference would have been difficult to interpret for the reasons discussed in the context of the within-subject analysis ( i . e . , Table 4 ) , the fact that we find robust differences here is consistent with the conclusion that following loss-feedback , model-based representations are suppressed and therefore are less relevant for behavior . With its small sample size , this experiment was not designed as an individual differences study and therefore these exploratory results need to be considered with caution . However , confidence in the results is strengthened by the fact that they are largely consistent with the findings from the within-subject PPI analyses . After acceptance of this manuscript , we became aware of a manuscript by Hermoso-Mendizabal et al . ( 2019 ) . These authors report a study with rats that both in terms of experimental design and results is remarkably consistent with what we report here . In a serial choice task , rats exploited experimentally induced sequential regularities ( i . e . , high frequency of repetitions versus alternations ) following positive feedback , but temporarily reverted to almost completely stochastic choice behavior following a single , negative feedback trial . Subjects were University of Oregon students who participated after giving informed consent in exchange for monetary payment or course credits; Experiment 1: N = 56 ( 38 female ) , Experiment 2: N = 40 ( 28 female ) , Experiment 3: N = 44 ( 25 female ) , Experiment 4a: N = 100 ( 62 female ) , Experiment 4b: N = 38 ( 20 female ) , Experiment 5: N = 25 ( 13 female ) . Four subjects from Experiment 1 and three pairs from Experiment 4a were excluded , because the experimental session could not be completed . The entire study protocol was approved by the University of Oregon’s Human Subjects Review Board ( Protocol 10272010 . 016 ) . On each trial of the fox/rabbit game , players observed a circle either on the bottom or the top of a vertically aligned rectangle . They had to choose between one of two rules for responding to the circle location . The ‘freeze rule’ implied that the circle stayed at the same location and it required participants to press among two keys the one that was compatible with the circle location ( ‘2’ and ‘5’ on the number pad ) , using the right-hand index finger . The ‘run rule’ implied that the circle moved to the opposite location within the vertical box and participants had to press among a separate set of vertically aligned keys ( ‘1’ and ‘4’ on the number pad ) the key that was incompatible with the circle location ( Mayr and Bell , 2006 ) , using the left-hand index finger . Participants were asked to rest the index finger of each hand between the two relevant keys ( e . g . , between '1' and '4' for the left index finger ) at the beginning of each trial . On a given trial , the fox player won two cent per trial , when both players chose the same rule , whereas the rabbit player won when choices were different . Participants had to respond within a 2000 ms interval and after that interval , they received feedback presented for 200 ms with a smiley face indicating a win trial and a frowny face a loss trial . Both incorrect responses ( e . g . , a compatible response using the keys for the incompatible rule ) or late responses ( which were extremely rare ) counted as errors . In terms of feedback , all errors were treated in the same way as loss trials , that is a frowny face was presented at the end of the 2000 ms response interval . The inter-trial-interval ( ITI ) was 300 ms . Participants initially were exposed to a block of 80 practice trials in order to familiarize them with the response procedures . This block was performed without a competitor , but with the typical ‘voluntary switching’ instruction that asks subjects to change rules randomly on a trial-by-trial basis . Following practice , participants played the fox/rabbit game for ten different blocks of 80 trials each . Both the switch rate of the opponent varied on a block-by-block basis between 20% , 35% , 50% , 65% , and 80% , and also whether the player had the role of the fox or the rabbit . Except for the switch-rate constraint , the simulated opponent’s choices were completely random . The ten different combinations of opponent switch rate and player role were randomly distributed across blocks . Participants were instructed that the different simulated players represented common strategies that one might find in human players . At the beginning of each block , participants were notified that they would be facing a new , simulated opponent , and whether they played the role of the fox or the rabbit , but received no instruction about the specific strategies . This and the following experiments were programmed in Matlab ( Mathworks ) using the Psychophysics Toolbox ( Brainard , 1997 ) and presented on a 17-inch CRT monitor ( refresh rate: 60 Hz ) at a viewing distance of 100 cm . This experiment was identical to Experiment 1 in all aspects , only that here the ITI varied between 300 ms and 1000 ms randomly , on a trial-by-trial basis . This experiment was again identical to Experiment 1 , only that here the choice between two different response rules was replaced by a simple choice between two different key-press responses . Each trial was initiated by a circle appearing at the center of the vertically arranged stimulus rectangle . Using the vertically arranged ‘2’ or ‘5’ keys , participants had to shift the circle up or down within the rectangle . Matching moves between opponents implied a win for the fox and a loss for the rabbit player . In Experiment 4a participants were paired into fox/rabbit dyads and played in real-time on two computers within the same room , but without opportunity for direct communication . Here , participants played 7 blocks of 80 trials each , and stayed within the same role throughout the experiment . All other aspects were identical to Experiment 1 . Experiment 4b served as a non-competition control experiment . Here , participants were given the standard instruction for the voluntary task-switching paradigm , namely to select tasks as randomly as possible ( "simulating a series of coin tosses" ) . Also , trial-by-trial wins and losses were completely random and participants were informed that this was the case . Otherwise , this experiment was identical to Experiment 4a . This experiment was again identical to Experiment 1 , but optimized towards EEG recording . For this purpose , we used only three simulated opponents ( switch rates of 25% , 50% and 75% ) across 24 blocks of 80 trials each ( i . e . , 4 repetitions of 3 opponent strategies x fox/rabbit roles ) . The ITI was 700 ms to allow assessment of feedback-related EEG activity . To evaluate the predictability of the current choice by the recent choice history , we fitted multilevel logistic regression models predicting the switch vs . repeat choice by the player’s and opponent’s switch history from n-3 , n-2 , and n-1 trials , the overall switch probability of the opponent , whether trial n-1 was a win or a loss trial , and the interactions between win/loss and all history/context variables ( i . e . , 15 predictors in total ) . We estimated both fixed and random effects of all predictors . For Experiment 4 , the model had three levels in which trials were nested within players , which in turn were nested in dyads . For the other experiments , models included only the first two levels . The signature of model-based selection is that predictors representing the opponent’s switch rate ( e . g . , the overall switch probability and opponent’s switch history ) is positively related to the player’s switch probability on post-win trials and negatively on post-loss trials . The main prediction we wanted to test was that following post-win trials the predictive relationship is stronger than following post-loss trials . Therefore , we examined the interaction between the post-win/loss contrast and each opponent-related predictor after reversing the label of the predictor for post-loss trials ( e . g . , n–one opponent ‘switch’ is relabeled as ‘repeat’ , 80% overall switch rate becomes 20% ) . This allowed us to test the difference in the strength of the relationship , while ignoring the direction of the relationship . We had no a-priori prediction about the direction of the relationship between previous switch/repeat choices and the trial n switch/repeat choice . Nevertheless , for a conservative test of post-win/loss differences we again reversed the post-loss label in each case where there was an empirical flip in sign between post-loss and post-win coefficients . We also present in Figure 4—figure supplement 2 the results without reversing labels . In Experiment 5 , Electroencephalographic ( EEG ) activity was recorded from 20 tin electrodes held in place by an elastic cap ( Electrocap International ) using the International 10/20 system . The F3 , Fz , F4 , T3 , C3 , CZ , C4 , T4 , P3 , PZ , P4 , T5 , T6 , O1 , and O2 of the 10/20 system were used along with five nonstandard sites: OL midway between T5 and O1; OR midway between T6 and O2; PO3 midway between P3 and OL; PO4 midway between P4 and OR; and POz midway between PO3 and PO4 . The left-mastoid was used as reference for all recording sites . Data were re-referenced off-line to the average of all scalp electrodes . Electrodes placed ~1 cm to the left and right of the external canthi of each eye recorded horizontal electrooculogram ( EOG ) to measure horizontal saccades . To detect blinks , vertical EOG was recorded from an electrode placed beneath the left eye and reference to the left mastoid . The EEG and EOG were amplified with an SA Instrumentation amplifier with a bandpass of 0 . 01–80 Hz and were digitized at 250 Hz in LabView 6 . 1 running on a PC . We used the Signal Processing and EEGLAB ( Delorme and Makeig , 2004 ) toolboxes for EEG processing in MATLAB . Trials including blinks ( >60 μv , window size = 200 ms , window step = 50 ms ) , large eye movements ( >1° , window size = 200 ms , window step = 10 ms ) , and blocking of signals ( range = −0 . 01 μv to 0 . 01 μv , window size = 200 ms ) were rejected excluded from further analysis . Single-trial EEG signals were segmented into 1250 ms epochs starting from 200 ms before the onset of feedback . Thus , each epoch included 700 ms post-feedback periods and the initial 250 ms intervals of the next trials . Each electrode’s EEG signal was also pre-whitened by linear and quadratic trends across experimental trials and blocks . After baselining signals with data from the initial , 200 ms interval , EEG activity from electrodes Fz and Cz , was averaged . These electrodes were selected based on previous studies reporting a robust interaction between the feedback and the probability context during reinforcement learning ( Cohen et al . , 2007 ) . The resulting signal was regressed via multilevel modeling with two levels ( i . e . , trials nested within participants ) on context variables , as described in the Results section . For illustrative purposes , this was done on a time-point by time-point basis ( see Figure 5 ) . To conduct statistical tests of the post-win versus post-loss regression coefficients for the psychophysiological interaction analysis predicting choices , for the individual differences ( Figure 6 ) , and for the topographic maps ( Figure 5a ) , we averaged the EEG signal for an a-priori defined 300–700 ms interval from the onset of feedback up to the beginning of the next trial . This interval is based on the typical time-course of feedback effects reported in the literature ( Cohen et al . , 2007 ) . The difference between post-win/loss models was tested in the same manner as in the multilevel model for history effects , namely by inverting predictors of opponents’ history/context for post-loss trials ( see section Analytic Strategy for Testing Main Prediction and History Effects Analysis ) .	The best predictor of future behavior is past behavior , so the saying goes . And studies show that in many situations , we do have a tendency to repeat whatever we did last time , particularly if it led to success . But while this is an efficient way to decide what to do , it is not always the best strategy . In many competitive situations – from tennis matches to penalty shoot-outs – there are advantages to being unpredictable . You are more likely to win if your opponent cannot guess your next move . Based on this logic , Kikumoto and Mayr predicted that in competitive situations , people will toggle between two different decision-making strategies . When they are winning , they will choose their next move based on their beliefs about their opponent’s strategy . After all , if your opponent in a tennis match has failed to return your last three backhands , it is probably worth trying a fourth . But if an action no longer leads to success , people will switch tactics . Rather than deciding what to do based on their opponent’s strategy and recent behavior , they will instead select their next move more at random . If your tennis opponent suddenly starts returning your backhands , trying any other shot will probably produce better results . To test this prediction , Kikumoto and Mayr asked healthy volunteers to play a game against real or computer opponents . The game was based on the 'matching pennies' game , in which each player has to choose between two responses . If both players choose the same response , player 1 wins . If each player chooses a different response , player 2 wins . Some of the opponents used response strategies that were easy to figure out; others were less predictable . The results showed that after wins , the volunteers’ next moves reflected their beliefs about their opponent's strategy . But after losses , the volunteers’ next moves were based less on previous behaviors , and were instead more random . These differences could even be seen in the volunteers’ brainwaves after win and loss trials . As well as providing insights into how we learn from failures , these findings may also be relevant to depression . People with depression tend to switch away from a rationale decision-making strategy too quickly after receiving negative feedback . This can lead to suboptimal behavior patterns that make it more difficult for the person to recover . Future studies should explore whether the short-term decision-making strategies identified in the current study can also provide clues to these behaviors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	Balancing model-based and memory-free action selection under competitive pressure
Reading frame maintenance is critical for accurate translation . We show that the conserved eukaryotic/archaeal protein Mbf1 acts with ribosomal proteins Rps3/uS3 and eukaryotic Asc1/RACK1 to prevent frameshifting at inhibitory CGA-CGA codon pairs in the yeast Saccharomyces cerevisiae . Mutations in RPS3 that allow frameshifting implicate eukaryotic conserved residues near the mRNA entry site . Mbf1 and Rps3 cooperate to maintain the reading frame of stalled ribosomes , while Asc1 also mediates distinct events that result in recruitment of the ribosome quality control complex and mRNA decay . Frameshifting occurs through a +1 shift with a CGA codon in the P site and involves competition between codons entering the A site , implying that the wobble interaction of the P site codon destabilizes translation elongation . Thus , eukaryotes have evolved unique mechanisms involving both a universally conserved ribosome component and two eukaryotic-specific proteins to maintain the reading frame at ribosome stalls . Accurate translation of mRNA into protein depends upon precise , repetitive three base translocation of the ribosome to maintain the correct reading frame throughout a coding sequence . Reading frame maintenance is challenging because multiple movements of the tRNAs and mRNA as well as conformational changes within the ribosome itself are required to complete a single elongation cycle ( Noller et al . , 2017 ) . For instance , the tRNA acceptor stems move within the large subunit during formation of the hybrid state , while the joining of EF-G-GTP ( eEF2 in eukaryotes ) results in additional movement of tRNA , and finally completion of translocation , driven by Pi release , requires additional movements ( Belardinelli et al . , 2016; Brilot et al . , 2013; Noller et al . , 2017; Pulk and Cate , 2013; Ramrath et al . , 2013; Ratje et al . , 2010; Tourigny et al . , 2013; Zhou et al . , 2014 ) . To accomplish this cycle , many interactions between the tRNAs and ribosome are disrupted , and new interactions are created , but the relative position of the tRNA anticodon to the mRNA codon must be maintained throughout all of these events ( Noller et al . , 2017; Dever et al . , 2018; Rodnina , 2018 ) . Thus , it is critical that mechanisms exist to prevent slippage during these transitions . Reading frame maintenance is facilitated by structures within the ribosome as well as by tRNA modifications . Structural features that contribute to reading frame maintenance , inferred from analysis of prokaryotic translation intermediates , include a swivel of the 30S head relative to the 30S body to form a contracted mRNA tunnel downstream of the A site prior to translocation ( Jenner et al . , 2010; Schuwirth et al . , 2005 ) . In addition , during translocation , two conserved bases in the 16S rRNA intercalate into different positions of the mRNA to prevent slippage ( Zhou et al . , 2013 ) , while domain IV of EF-G contacts the codon and tRNA in the A/P site and h44 of 16S rRNA , likely coupling mRNA and tRNA movement ( Ramrath et al . , 2013; Zhou et al . , 2014 ) . tRNA modifications within the anticodon loop also assist in reading frame maintenance , inferred both from genetic and structural analyses . Mutants that affect several such modifications in both bacteria and eukaryotes result in increased frameshifting ( Atkins and Björk , 2009; Jäger et al . , 2013; Tükenmez et al . , 2015; Urbonavicius et al . , 2001; Waas et al . , 2007 ) . Moreover , a cross-strand base stacking interaction between a modified ms2i6A37 in an E . coli tRNAPhe and the mRNA codon is proposed to prevent slippage of P site tRNA on the mRNA ( Jenner et al . , 2010 ) . Thus , a number of mechanisms exist to prevent loss of reading frame . Nevertheless , ribosomes do move into alternative reading frames in response to specific sequences and structures in mRNA ( Atkins and Björk , 2009; Dever et al . , 2018; Dinman , 2012 ) . The existence of such events has implied that ribosomal plasticity with respect to reading frame movement is an integral function of the translation machinery . The common feature of all frameshifting events in bacteria to humans is that the ribosome stalls ( Dever et al . , 2018 ) . The stall can be mediated by combined effects of the A and P site codons ( Farabaugh et al . , 2006; Gamble et al . , 2016 ) , by the presence of downstream structures , or by an upstream Shine-Dalgarno sequence in bacteria ( Caliskan et al . , 2014; Dinman , 2012 ) . Analysis of programmed frameshifting indicates that additional sequences or protein factors are frequently required to mediate efficient frameshifting ( Atkins and Björk , 2009; Dinman , 2012 ) . For instance , +1 programmed frameshifting events are frequently enhanced by stimulatory sequences , although the role of these sequences is not always clear ( Guarraia et al . , 2007; Taliaferro and Farabaugh , 2007 ) . The identification of mutants that either affect programmed frameshifting or suppress frameshift mutations has pointed to four key factors in reading frame maintenance . First , mutations of ribosomal proteins , particularly those that contact the P site tRNA can cause increased frameshifting . In bacteria , frameshifting mutations are suppressed by deletions within the C terminal domain of ribosomal protein uS9 , which contacts the P site tRNA anticodon loop ( Jäger et al . , 2013 ) . In the yeast Saccharomyces cerevisiae , programmed frameshifting in the L-A virus is affected by mutations in 5S rRNA or its interactors uL18 or uL5 , that also contact the P site tRNA ( Meskauskas and Dinman , 2001; Rhodin and Dinman , 2010; Smith et al . , 2001 ) . Frameshifting mutations are also suppressed by a mutation in the yeast RPS3 , although this mutation does not affect a tRNA contact ( Hendrick et al . , 2001 ) . Second , mutations in the basal translation machinery can also affect frameshifting . For instance , frameshifting mutations are suppressed by mutations in both EF-1α , which delivers tRNA to the ribosome ( Sandbaken and Culbertson , 1988 ) , and in SUP35 , encoding the translation termination factor eRF3 ( Wilson and Culbertson , 1988 ) . Third , miRNAs can affect the efficiency of programmed frameshifting , for instance at CCR5 in humans ( Belew et al . , 2014 ) . Fourth , mutations that affect proteins with previously unknown functions in translation can either alter programmed frameshifting or suppress frameshifting mutations . For instance , in yeast , frameshifting mutations are suppressed by mutations in MBF1 , encoding Multi-protein Bridging Factor 1 ( Hendrick et al . , 2001 ) , or in EBS1 ( Ford et al . , 2006 ) , while in the porcine virus PRRSV , the RNA binding protein nsp1β stimulates both −1 and −2 frameshifting events ( Li et al . , 2014 ) . Thus , reading frame maintenance is modulated by ribosomal components , many of which contact the tRNAs , as well as by non-ribosomal proteins and miRNAs . However , the roles of many of these proteins are not understood . We set out to work out the mechanisms that maintain reading frame when eukaryotic ribosomes encounter a stall , the common feature of all frameshifting events . In bacteria , ribosome stalls due to limited availability or functionality of tRNA seem to suffice to cause frameshifting ( Gurvich et al . , 2005; Seidman et al . , 2011 ) . In wild type yeast , ribosomes stall at CGA codon repeats , which inhibit translation due to wobble decoding of CGA by its native tRNAArg ( ICG ) ( Letzring et al . , 2010; Letzring et al . , 2013 ) . Although frameshifting was detected at several underrepresented heptanucleotide sequences in yeast ( Shah et al . , 2002 ) , it was not detected at CGA codon repeats ( Wolf and Grayhack , 2015 ) . Instead , eukaryotes have evolved new pathways to regulate inefficient translation events , such as the Ribosome Quality Control ( RQC ) pathway , in which these stalled ribosomes undergo ubiquitination of ribosomal proteins , followed by dissociation of the subunits , and recruitment of the RQC Complex , which mediates CAT tailing and degradation of the nascent polypeptide ( Bengtson and Joazeiro , 2010; Brandman and Hegde , 2016; Brandman et al . , 2012; Defenouillère et al . , 2013; Joazeiro , 2017; Kostova et al . , 2017; Shao et al . , 2013; Shen et al . , 2015; Verma et al . , 2013; Wilson et al . , 2007; Juszkiewicz and Hegde , 2017; Matsuo et al . , 2017; Simms et al . , 2017b; Sitron et al . , 2017; Sundaramoorthy et al . , 2017 ) . The ribosomal protein Asc1/RACK1 mediates these events ( Brandman et al . , 2012; Kuroha et al . , 2010 ) ; in the absence of Asc1 , ribosomes fail to engage the RQC ( Sitron et al . , 2017 ) , and also undergo substantial frameshifting at CGA codon repeats ( Wolf and Grayhack , 2015 ) . However , Asc1 sits on the outside of the ribosome at the mRNA exit tunnel and likely functions as scaffold for recruitment of other proteins , such as the E3 ubiquitin ligase Hel2/mammalian ZNF598 and Slh1 ( Juszkiewicz and Hegde , 2017; Matsuo et al . , 2017; Simms et al . , 2017b; Sitron et al . , 2017; Sundaramoorthy et al . , 2017 ) . Based on the location of Asc1 and the precedent that Asc1 recruits other proteins to abort translation , we considered it likely that Asc1 cooperates with additional proteins to mediate reading frame maintenance at CGA codon repeats and set out to find such factors . Here , we provide evidence that the Multi-protein Bridging Factor 1 ( Mbf1 ) and ribosomal proteins Rps3 and Asc1 ( homolog of human Rack1 ) work together to prevent translational slippage at CGA codon repeats . Frameshifting results from inactivation of MBF1 , or from mutations in amino acids in Rps3 located on an exposed surface of the protein near the mRNA entry site . Asc1 was previously known to mediate recruitment of the RQC , mRNA cleavage and mRNA decay at similar stall sites ( Ikeuchi and Inada , 2016; Kuroha et al . , 2010; Letzring et al . , 2013; Sitron et al . , 2017 ) , as well as reading frame maintenance ( Wolf and Grayhack , 2015 ) . We provide evidence that Asc1 and Mbf1 cooperate to mediate reading frame maintenance at similar sites , acting on a common set of substrates , including the seven most slowly translated codon pairs in yeast ( Gamble et al . , 2016 ) . We examined the precise frameshift at one of these inhibitory pairs , CGA-CGG , purifying the frameshifted polypeptide , followed by analysis with mass spectrometry . We found that frameshifting occurs in the +1 direction at the CGA codon and moreover , that frameshifting is modulated by the competition between the in-frame and +1 frame tRNAs . We considered it likely that proteins other than Asc1 worked to prevent frameshifting at CGA codon repeats for two reasons . First , Asc1 binds on the outside of the ribosome , not in the decoding center ( Rabl et al . , 2011 ) , and thus is not positioned in any obvious way to assist with reading frame maintenance . Second , Asc1 recruits other proteins , Hel2 and Slh1 , to recruit the RQC ( Brandman and Hegde , 2016; Joazeiro , 2017; Sitron et al . , 2017 ) , and thus is likely to work with other proteins in reading frame maintenance . Thus , we set out to identify genes responsible for reading frame maintenance at CGA codon repeats . To isolate mutants that frameshift due to translation of CGA codon repeats , we set up a selection in which a +1 frameshift ( caused by 6 adjacent CGA codons ) was required to express the URA3 gene . The native URA3 gene was placed in the +1 reading frame downstream of an N-terminal domain of GLN4 encoding amino acids 1–99 ( GLN4 ( 1-99 ) ) , followed by 6 CGA codons and one additional nucleotide upstream of the URA3 coding region ( Figure 1A ) . This strain exhibits an Ura- phenotype , due to the low levels of frameshifting in an otherwise wild-type background , allowing for an Ura+ selection to obtain mutants with increased frameshifting . As an independent secondary screen for frameshifting mutants due to CGA codon repeats , we integrated a modified version of the RNA-ID reporter with GLN4 ( 1-99 ) followed by 4 CGA codons and one additional nucleotide upstream of the GFP coding region into the ADE2 locus ( Dean and Grayhack , 2012; Wolf and Grayhack , 2015 ) . Thus , GFP expression was dependent upon frameshifting efficiency ( Figure 1A ) . To avoid obtaining mutations in the ASC1 gene , the selection strain also contained a second copy of the ASC1 gene on a plasmid . ( Figure 1A ) . We selected Ura+ mutants from 40 independent cultures each of MATa and MATα parents and then analyzed three Ura+ mutants from each culture by flow cytometry to measure GFP and RFP expression . Most mutants ( 60% of MATα mutants and 80% of MATa mutants ) showed elevated expression of GFP , and we studied those that exhibited relatively high levels of frameshifting , >30% of that in an asc1Δ mutant ( Figure 1B ) . Most mutants ( 43 of 48 examined ) were recessive and mapped to a single complementation group , based on growth of diploids on media lacking uracil ( Figure 1—figure supplement 1A ) , although four dominant mutants were also identified . To confirm that inhibitory decoding of CGA codon repeats is required for frameshifting in these mutants , we showed that introduction of an anticodon-mutated exact match tRNAArg ( UCG ) * suppressed the Ura+ phenotype of one mutant ( Figure 1C ) . We have shown previously that expression of this exact match tRNAArg ( UCG ) * results in efficient decoding of CGA codons and suppresses their inhibitory effects on gene expression ( Letzring et al . , 2010 ) . Thus , the Ura+ , GFP+ phenotype of this mutant was due to frameshifting that occurs when the ribosome translates CGA codon repeats inefficiently . We demonstrated that mutations in the yeast gene MBF1 , Multi-protein Bridging Factor 1 , were responsible for the defects in reading frame maintenance in recessive high GFP mutants . We identified the mutated gene by complementation of the Ura+ phenotype of the P25 recessive mutant with two plasmids from a library that contains 97 . 2% of the entire yeast genome ( Figure 1—figure supplement 1B ) ( Jones et al . , 2008 ) . The complementing plasmids share a single ORF , MBF1 . We confirmed that mutations in the MBF1 gene are responsible for frameshifting in three ways . First , a plasmid with only the MBF1 gene complemented the frameshifting Ura+ phenotype of two mutants ( Figure 1—figure supplement 2A ) . Second , deletion of MBF1 in the parent selection strain ( Figure 1A ) converted that strain from GFP- to GFP+ , similar to deletion of ASC1 ( Figure 1—figure supplement 2B ) . Third , 19/19 mutants tested contain mutations in the MBF1 gene . Point mutations isolated in our selection are located at conserved residues near the junctions between two domains ( Figure 1D ) . MBF1 is a highly conserved gene in eukaryotes and archaea , generally less than 160 amino acids with an N-terminal Mbf1-specific domain ( that differs between archaea and eukaryotes ) and a conserved cro-like helix-turn-helix ( HTH ) domain ( Figure 1D , Figure 1—figure supplement 3A ) . Mbf1 , which was initially identified as a transcription co-activator in Bombyx mori ( Li et al . , 1994; Takemaru et al . , 1997 ) , has been implicated in a similar function in yeast , in this case , interacting with the Gcn4 , transcription regulator of the general amino acid control pathway ( Takemaru et al . , 1998 ) . In testing sensitivity to 3-aminotriazole ( 3-AT ) ( Hilton et al . , 1965; Schürch et al . , 1974 ) , a phenotype of gcn4 mutants due to inability to induce expression of HIS3 , we found that two frameshifting point mutants ( mbf1-K64E and mbf1-I85T ) exhibit no growth defect even on high concentrations of 3-AT ( Figure 1—figure supplement 3B ) . Moreover , deletion of GCN4 does not affect frameshifting at CGA codon repeats in an asc1Δ mutant ( Wolf and Grayhack , 2015 ) . Thus , it is unlikely that the defect in reading frame maintenance in our mbf1 mutants is related to GCN4 . However , Mbf1 has also been implicated in translation , based on isolation of mutations in yeast MBF1 that suppress frameshifting mutations ( Hendrick et al . , 2001 ) , and the weak association of the archaeal homolog with ribosomes ( Blombach et al . , 2014 ) , but there is no information on its molecular role in translation . To identify the mutated gene ( s ) in our dominant mutants , we performed whole genome sequencing in two MATα mutants and found that each mutant contains a single amino acid change ( S104Y and G121D ) in RPS3 . Similarly , the two dominant MATa mutants also contain mutations in the RPS3 gene ( L113F and a duplication of N22 to A30 ) . RPS3 encodes a universally conserved ribosomal protein , a core component of the mRNA entry tunnel with a eukaryotic-specific C-terminal extension that interacts with Asc1 ( Rabl et al . , 2011 ) . One known mutation in RPS3 ( K108E ) affects reading frame maintenance ( Hendrick et al . , 2001 ) , while others affect different aspects of translation , from initiation to quality control ( Dong et al . , 2017; Graifer et al . , 2014; Limoncelli et al . , 2017; Takyar et al . , 2005 ) . The three residues S104 , L113 and G121 implicated in reading frame maintenance in our study , as well as K108 , are all found in two α-helices near the mRNA entry tunnel of the ribosome; these residues reside on the surface of the ribosome and could interact with mRNA or proteins outside of the ribosome ( Figure 2A ) . Moreover , the identity of all four of these residues is conserved in eukaryotes , but different in bacteria and archaea ( Graifer et al . , 2014 ) . We initially examined the effect of the RPS3-K108E mutation on both frameshifting and in-frame expression downstream of CGA codon repeats , and found that this mutation allows frameshifting but does not affect in-frame expression . We chose the K108E mutation because it is known to have only minor effects on the polysome to monosome ratio ( Dong et al . , 2017 ) , consistent with few nonspecific effects on translation . We introduced modified RNA-ID reporters into rps3Δ::bleR strains in which the only source of RPS3 is a plasmid-borne copy ( either wild type or K108E ) . As described previously , since the expression of GFP and RFP is driven by the bi-directional GAL1 , 10 promoter , we use the ratio of GFP/RFP to reduce noise and cell type specific differences in induction of this promoter ( Dean and Grayhack , 2012 ) . Neither the RPS3 mutant nor a mbf1Δ mutant had a substantial effect on GFP/RFP fluorescence ( protein ) , mRNA or protein/mRNA of reporters with CGA or AGA codon repeats in-frame ( Figure 2B ) . We did note relatively minor , but compensatory effects , of the mutants on both GFP and RFP mRNAs ( a 15–30% reduction in mbf1Δ mutants and a similar increase in the RPS3–K108E mutant ) ( Figure 2—figure supplement 1A ) . The RPS3-K108E and mbf1Δ mutants each caused substantially increased frameshifted GFP/RFP protein and protein/mRNA in the construct with four CGA codons , but had only small effects on GFP/RFP mRNA; no frameshifting was seen with four AGA codons ( Figure 2B; Supplementary file 1 ) . If Mbf1 and Rps3 proteins function in independent pathways to prevent frameshifting , we expected that RPS3-K108E mbf1Δ double mutants would frameshift more efficiently than either single mutant . Instead , we found that the double mutant RPS3-K108E mbf1Δ exhibited only a slight increase in frameshifted GFP/RFP protein; this increase is likely due to a slight increase in GFP/RFP mRNA relative to either single mutant , resulting in nearly identical protein/mRNA from the mbf1Δ and RPS3-K108E mbf1Δ mutants ( Figure 2B , Figure 2—figure supplement 1B ) . We also examined effects of combining MBF1 mutants with other RPS3 mutants , S104Y and G121D from our selection , to determine epistasis . In these cases again , each single mutant exhibited frameshifting and the double mutants exhibited similar amounts of frameshifted GFP/RFP to that in the mbf1Δ strain , although even an additive effect would be easily detectable ( Figure 2C ) . Thus , we think it is likely that Mbf1 and the two α-helices in the N-terminal Rps3 protein have related roles in reading frame maintenance . If Mbf1 and these two α-helices in Rps3 mediate a common function , then frameshifting in either RPS3-S104Y or G121D mutants might be suppressed by overproduction of MBF1 . We found that introduction of additional copies of the MBF1 gene into either of these mutants resulted in reduced expression of frameshifted GFP ( Figure 2D ) . Frameshifted GFP was reduced to 30% in the S104Y mutant and to 60% in the G121D mutant ( Figure 2D ) . Similarly , growth on media lacking uracil was severely compromised in the RPS3-S104Y mutant when MBF1 was expressed on a multi-copy plasmid , relative to an empty vector control ( Figure 2—figure supplement 1C ) , although both strains grow equally well on SD-Leu media . These observations are consistent with the idea that Mbf1 and Rps3 play similar roles in reading frame maintenance and support the idea that these RPS3 mutations reduce Mbf1 function . Since Asc1 is also required for reading frame maintenance at CGA codon repeats ( Wolf and Grayhack , 2015 ) , we examined the relationship between MBF1 and ASC1 on both frameshifting and in-frame expression , comparing GFP/RFP in the single mutants to that in the asc1Δ mbf1Δ double mutant . Asc1 is known to affect expression ( both in-frame and +1 frame ) downstream of four CGA codons ( Letzring et al . , 2013; Wolf and Grayhack , 2015 ) , but we previously noted that inhibitory effects of CGA codons are mediated by CGA codon pairs ( Gamble et al . , 2016; Letzring et al . , 2010 ) . Therefore , we compared the effects of these mutants on a set of reporters with three CGA-CGA ( or AGA-AGA ) codon pairs flanked by two non-Arg codons ( Figure 3A , Figure 3—figure supplement 1A , Supplementary file 2 ) to effects on a set with four adjacent CGA ( or AGA ) codons ( Figure 3—figure supplement 1B ) . Neither the upstream gene nor the arrangement of CGA codons affected the results . As expected based on a previous report ( Sitron et al . , 2017 ) , deletion of ASC1 resulted in increased protein and mRNA levels of reporters with in-frame CGA-CGA codon pairs . By contrast , deletion of MBF1 did not affect protein or mRNA levels substantially ( Figure 3A , Figure 3—figure supplement 1A and B , Supplementary file 2 ) . Thus , Asc1 clearly has a unique role in regulating mRNA and RQC recruitment at CGA codon pairs , but overall expression , measured as protein/mRNA , of all in-frame reporters is similar in the wild type , asc1Δ , mbf1Δ , and asc1Δ mbf1Δ double mutants ( Figure 3A ) . For the in-frame reporters , the relationship between GFP/RFP fluorescence and mRNA is linear; none of these mutants affect RFP mRNA ( Figure 3—figure supplement 1C and D ) . However , the increase in mRNA in an asc1Δ mutant does not explain the increase in frameshifted GFP/RFP protein in this mutant . That is , the 2 . 7-fold increase in GFP/RFP mRNA from the +1 reporter in an asc1Δ mutant ( relative to the wild type ) cannot account for the >50 fold increase in frameshifted GFP/RFP fluorescence ( Figure 3A ) . Thus , an asc1Δ mutant clearly exhibits a defect in reading frame maintenance . If increased frameshifted protein/mRNA in an asc1Δ mutant is due to a failure of the Mbf1 pathway , then we expected that asc1Δ mbf1Δ double mutants would frameshift with similar efficiency to the mbf1Δ mutant . In fact , the level of frameshifted GFP/RFP protein per mRNA was very similar in the single mbf1Δ mutant to that in the double asc1Δ mbf1Δ mutant ( Figure 3A , Figure 3—figure supplement 1D , Supplementary file 2 ) . These results are most consistent with a single pathway of reading frame maintenance , which Asc1 influences . We confirmed that the +1 GFP signal detected in our mutants was due to frameshifting rather than another aberrant translation event by directly measuring both the size and amount of GFP fusion protein . The amount of full-length GFP protein in the Western blot corresponds to the GFP/RFP values obtained from flow analysis ( Figure 3B ) indicating that +1 GFP/RFP signal in our mutants is due to frameshifting . If the defect in the asc1Δ mutant that results in frameshifting is due to a failure of the Mbf1 pathway , then overproduction of Mbf1 in the asc1Δ mutant might suppress frameshifting in this mutant . We found that expression of MBF1 on a multi-copy plasmid did suppress frameshifting in the asc1Δ strain to 1/3 that seen with an empty vector , but did not affect in-frame expression ( Figure 3C ) . The overproduction of Mbf1 was not complementing a reduced abundance of Mbf1 in this mutant . We did not detect a reduction in Mbf1-HA ( which complements the mbf1Δ mutant ) in the asc1Δ strain ( Figure 3—figure supplement 2A ) , although asc1 mutants generally exhibit a defect in expression of small proteins ( Thompson et al . , 2016 ) . We also considered that mbf1 mutants might require additional Asc1 protein , but additional copies of ASC1 did not suppress frameshifting in an mbf1Δ mutant ( Figure 3—figure supplement 2B ) . Thus , the frameshifted GFP/RFP fluorescence in the asc1Δ strain is likely a result of both an increase in mRNA and a defect in the Mbf1 pathway . We infer that Mbf1 and Asc1 contribute in distinct ways to the response to CGA codon pairs , but we do not know if Asc1 also has a direct role in the reading frame maintenance pathway . To address the mechanism of frameshifting and to understand the relationship between Asc1 and Mbf1 , we set out to identify the protein and sequence requirements for efficient frameshifting . We began by examining frameshifting at 12 of 17 inhibitory codon pairs all of which cause reduced expression and many of which exhibit high ribosome occupancy , indicative of slow translation ( Gamble et al . , 2016 ) , a common feature of many frameshifting sites . We found that Mbf1 and Asc1 act on the same inhibitory codon pairs . Frameshifting occurs with high efficiency at three codon pairs ( CGA-CGA , CGA-CGG , and CGA-CCG ) in the mbf1Δ mutant ( Figure 4A ) , the only three pairs at which Asc1 substantially modulates in–frame expression levels relative to synonymous optimal reporters ( Figure 4B ) . As might be expected , frameshifted GFP/RFP for CGA-CCG and CGA-CGA is greater in the asc1Δ mbf1Δ mutant than in the mbf1Δ single mutant ( Figure 4A , Figure 4—figure supplement 1A , Supplementary file 2 ) . Surprisingly , this is not true for the CGA-CGG construct at which frameshifting is remarkably high ( ~76% based on data from Figure 4—figure supplement 2C ) . We address the source of this high frameshifting below and the lack of synergy in the Discussion . Lower levels of frameshifting were also detected at 4 additional pairs ( CGA-AUA , CGA-CUG , CGA-GCG , and CUC-CCG ) in the asc1Δ mbf1Δ mutant , but not in the mbf1Δ single mutant ( Figure 4A , Figure 4—figure supplement 1B ) . Analysis of these seven codon pairs indicates that frameshifting was detected at the seven most slowly translated codon pairs in the yeast genome [based on analysis in ( Gamble et al . , 2016 ) ] and at every inhibitory pair with CGA in the 5’ position , consistent with slow decoding of CGA in the P site ( Tunney et al . , 2018 ) . To test the idea that Asc1 and Mbf1 prevent frameshifting at any slowly translated sequence , we measured frameshifting at a sequence which forms a secondary structure to slow down translation and induces no-go mRNA decay ( Doma and Parker , 2006; Harigaya and Parker , 2010; Passos et al . , 2009 ) . In accordance with this idea , frameshifted GFP/RFP protein from both +1 and −1 constructs was detectable in wild type , greater in each single mutant and even greater in the asc1Δ mbf1Δ mutant ( Figure 4C ) . By contrast , these mutants did not affect frameshifting efficiency at the programmed frameshift site for TY1 ( Figure 4—figure supplement 1C ) ( Belcourt and Farabaugh , 1990 ) . Thus , Mbf1 and Asc1 regulate reading frame maintenance at a translational pause ( no-go site ) , but do not enhance frameshifting at site in which translational slippage is encoded . To define the source of efficient frameshifting in the CGA-CGG reporter , which has 3 CGA-CGG pairs ( Figure 4A ) , we initially determined that the first CGA-CGG codon pair was responsible for highly efficient frameshifting ( Figure 4—figure supplement 2A and B ) . To define the sequence requirements for efficient frameshifting , we varied the sequences surrounding this single CGA-CGG pair and measured frameshifted GFP/RFP in the various mutants . Either of two changes to the sequence downstream of the CGA-CGG pair ( one a point mutation and another a codon insertion ) eliminated efficient frameshifting in all three mutant strains ( Figure 4D ) . Furthermore , altering the two nucleotides downstream of the first codon pair in the three codon pair reporter reduced frameshifted GFP/RFP in all three mutants and also restored the synergistic dependence on MBF1 and ASC1 ( Figure 4—figure supplement 2C and D ) . By contrast , none of three upstream changes ( to the single CGA-CGG reporter ) substantially reduced frameshifting ( Figure 4D ) . Thus , the CGA-CGG-C 7-mer is required for efficient frameshifting . To find out if frameshifting can occur at single CGA-CGG pairs in other sequence contexts , we tested sequences from seven yeast genes in our reporters , including six codons on either side of CGA-CGG pair . Frameshifted GFP/RFP was detected in asc1Δ mbf1Δ mutants in all cases ( Figure 4E , Figure 4—figure supplement 2E ) . In particular , the two native sequences with CGA-CGG-C resulted in frameshifted GFP/RFP in the single mutants ( Figure 4E , Figure 4—figure supplement 2E ) . Thus , CGA-CGG-C is likely a frameshifting sequence , and contexts that allow frameshifting have not been eliminated from native genes . To understand how frameshifting occurs , we wanted to define the direction and position of the actual frameshift . The high efficiency of frameshifting at the CGA-CGG-CAC sequence provided a useful tool to study frameshifting since there is only a short potential frameshifting sequence ( a single inhibitory codon pair ) . We inserted this sequence with its neighboring codons from the RNA-ID reporter into a construct for purification of the frameshifted polypeptide ( Figure 5A ) . The construct was designed such that the protein could be purified either with an upstream affinity tag ( GST ) to yield all polypeptides or with a downstream affinity tag ( Strep II or the ZZ domain of IgG ) to yield only frameshifted polypeptides . Treatment with LysC , which cleaves after lysine was expected to yield a 16 or 17 amino acid peptide for analysis by mass spectrometry , depending upon the mechanism of frameshifting . If frameshifting occurred in the local region near the CGA-CGG codon pair , there are four possible events that could all give rise to +1 GFP signal . Ribosomes could frameshift in the +1 direction with either the CGA or the CGG in the P site , yielding the RGTT or the RRTT sequences shown in Figure 5B . Alternatively , ribosomes could undergo −2 frameshifting at either codon , yielding the peptides RDGTT or RRGTT ( Figure 5B ) . In yeast , −2 frameshifting was observed upon expression of the mammalian antizyme ( Matsufuji et al . , 1996 ) and −2 frameshifting also occurs in PRRSV virus ( Fang et al . , 2012 ) . We purified the frameshifted protein , as well as an in-frame control protein with the sequence expected for a −2 frameshift at CGG ( Figure 5C ) and subjected them to mass spectrometry . The frameshifted protein yielded the peptide VTNLRGTTWSHPQFEK , the expected peptide from a +1 frameshift beginning with the CGA codon in the P site of the ribosome . Thus , we infer that frameshift occurs with CGA in the P site , yielding only one Arg amino acid on the nascent peptide , then switches to a glycine codon GGC . To determine if aminoacyl tRNA amounts affect frameshifting , we compared the effects of additional copies of specific Arg and Gly tRNAs on frameshifting in the asc1Δ mbf1Δ double mutant . We found that introduction of additional copies of the gene encoding tRNAArg ( CCG ) , which decoded the in-frame CGG codon , severely reduced frameshifting ( Figure 5D ) , as expected if arg-tRNAArg ( CCG ) competes with gly-tRNAGly ( GCC ) for the A site . Similarly , we found that addition of extra copies of tRNAGly ( GCC ) which decodes +1 frame GGC codon significantly increased frameshifting in our original CGA-CGG-CAC context , as might be expect if the GGC codon is used ( Figure 5D ) . Additional copies of tRNAArg ( ICG ) , tRNAAsp ( GUC ) , tRNAHis ( GUG ) , tRNASer ( AGA ) had little or no effect , as expected since none of the codons decoded by these tRNAs should be occupying the A site during frameshifting . These results indicate that the frameshifting occurs within the single CGA-CGG-CAC sequence and is modulated by the concentration of aminoacyl tRNAs decoding the out-of-frame codon . We have uncovered a eukaryotic specific system that maintains the reading frame when ribosomes stall . Reading frame maintenance of stalled ribosomes is achieved in two ways: by direct inhibition of frameshifting; and by aborted translation coupled with mRNA decay . The system is composed of two proteins that lack bacterial homologs , the archaeal/eukaryotic Mbf1 protein and the eukaryotic ribosomal protein Asc1/RACK1 , as well as one universally conserved ribosomal protein Rps3 . Mutations in any of these proteins result in increased frameshifting at CGA codon repeats . Moreover , the Rps3 residues in which mutations affect reading frame maintenance are specifically conserved in eukaryotes ( and differ in archaeal and bacterial Rps3 ) , consistent with a eukaryotic-specific mechanism . We suggest that when ribosomes stall ( Simms et al . , 2017b ) , two distinct sets of events occur: Asc1 triggers a set of responses that result in aborted translation , recruitment of the RQC complex and mRNA decay ( Brandman et al . , 2012; Defenouillère et al . , 2013; Shao et al . , 2013; Simms et al . , 2017b; Sitron et al . , 2017; Verma et al . , 2013 ) ; while Mbf1 and Rps3 cooperate at the stalled ribosomes to prevent frameshifting ( Figure 6A ) . In the absence of Mbf1 and Asc1 , ribosomes frameshift efficiently , even at a single CGA-CGG pair in some cases , including sequences found in the native yeast genome . Frameshifting on the CGA-CGG codon pair occurs in the +1 direction , with the CGA codon in the P site of the ribosome and is modulated by availability of in-frame and +1 frame A site tRNAs . The coordinated activities of both the Asc1 and Mbf1/Rps3 pathways are likely important to maintain the reading frame , since the absence of either pathway results in increased frameshifting . We think both pathways are likely engaged by similar stalls , since we noted evidence that Asc1 regulates expression ( either in-frame or out-of-frame ) at every sequence at which Mbf1 acted . However , there may be differences in the extent to which each pathway contributes at particular stall sites . For instance , at sequences at which frameshifting occurs rapidly ( i . e . CGA-CGG-C ) , Mbf1 may play the critical role; in the absence of Mbf1 , ribosomes frameshift before they can be captured by the Asc1 pathway . Generally , some fraction of ribosomes are removed from the translating pool with concomitant mRNA decay , while ribosomes that remain stalled are kept in-frame by Mbf1/Rps3 ( Figure 6A ) . The Asc1-mediated events couple mRNA decay to abortive translation ( Sitron et al . , 2017 ) , thus effectively reducing the number and duration of ribosome stalls . The integration of RNA decay with translation is extensive: cleavage of mRNA occurs upstream of many ribosome stalls including at CGA repeats ( Chen et al . , 2010; Doma and Parker , 2006; Guydosh and Green , 2014; Letzring et al . , 2010; Simms et al . , 2017b ) and is modulated by Asc1 ( Ikeuchi and Inada , 2016; Kuroha et al . , 2010 ) . mRNA decay is triggered by many problems in translation , such as nonsense mediated decay ( NMD ) , no-go and NonStop decay [See ( Shoemaker and Green , 2012; Simms et al . , 2017a ) and by slow translation ( Presnyak et al . , 2015; Radhakrishnan et al . , 2016 ) . For reading frame maintenance , it seems likely that both reducing the number of stalled ribosomes ( by aborting translation ) and removing the mRNA are important . Deletion of ASC1 in a mbf1Δ mutant results in an increase in frameshifted GFP/RFP protein that is directly proportional to the increase in mRNA . However , asc1Δ mutants exhibit a 50-fold increase in frameshifting relative to wild type cells with less than a 3-fold increase in mRNA levels . One likely explanation for frameshifting in the asc1Δ mutant is that Mbf1 becomes limiting due to an increase in the number of stalled ribosomes; this idea is strengthened by the observation that overproduction of Mbf1 suppressed frameshifting in the asc1Δ mutant . Alternatively , Asc1 might also play a direct role in the reading frame maintenance function . Asc1 interacts with the C terminal region of Rps3 and could affect the conformation of Rps3 or its interaction with Mbf1 . Mbf1 was found in the vicinity of Asc1 ( Opitz et al . , 2017 ) lending credence to the idea that Asc1 could interact with Mbf1 ( although such an interaction cannot be obligatory ) . We think Rps3 and Mbf1 inhibit frameshifting in a coordinated manner , perhaps due to their interactions with mRNA or to Mbf1’s interaction with the ribosome . The role of Rps3 in this process is likely to involve interactions with either the incoming mRNA or proteins external to the ribosome . The RPS3 mutations that affect frameshifting map to residues ( S104 , L113 , G121 , K108 ) on two α-helices or their connecting loop right next to the entering mRNA . Although this section of Rps3 is involved in helicase activity and initiation selectivity ( Dong et al . , 2017; Takyar et al . , 2005 ) , the residues mutated in frameshifting selections were not specifically those involved in these activities . Instead , these residues all sit on the solvent side of the ribosome and could form an interface interacting with mRNA or mRNA-bound proteins . The role of Mbf1 is likely mediated by interactions with either or both of the mRNA ( Beckmann et al . , 2015; Klass et al . , 2013 ) and the ribosome ( Blombach et al . , 2014; Opitz et al . , 2017 ) . The apparent RNA binding domain maps to the less conserved N terminal domain ( Klass et al . , 2013 ) , while ribosome binding activity of the archaeal homolog of Mbf1 maps to its C-terminal HTH domain and the linker at the N terminus of this domain , which are both conserved with eukaryotes ( Blombach et al . , 2014 ) ; our frameshifting mutations cluster in the conserved linker region of Mbf1 . Moreover , Mbf1 is sufficiently abundant with ~85 , 000 molecules per cell to participate in general translation cycles , although it is less abundant than core ribosomal proteins ( ~200 , 000 ) ( Kulak et al . , 2014 ) . There are two reasonable models to account for the role of Mbf1 and Rps3 in reading frame maintenance ( Figure 6B ) . The first model is that Mbf1 has a loose association with mRNA and is recruited to the leading stalled ribosome by an interaction with Rps3; the interactions with the ribosome and the mRNA at the stall site could restrict mRNA movement in the ribosome . Based on structures of prokaryotic ribosomes caught in translocation , mRNA flexibility may occur in ribosomes lacking an A site tRNA due to few contacts with the region of mRNA near the A site ( Zhou et al . , 2013 ) , or due to a failure of two rRNA pawls that lock the mRNA in a translocating ribosome ( Zhou et al . , 2013 ) , or due to defects in the interactions with elongation factor 2 ( Zhou et al . , 2014 ) . In the absence of Mbf1 , the ribosome stall might allow sufficient time for mRNA flexibility , resulting in frameshifting . The second model , which is based on the observation that ribosome collisions trigger no-go decay ( Simms et al . , 2017b ) , is that Mbf1 is recruited to colliding ribosomes to buffer the collision effects; in this case Asc1 and Rps3 might both participate in Mbf1 recruitment . Mbf1 could prevent ribosome collision-mediated movement of the leading ribosome on the mRNA . Frameshifting occurs by a mechanism that involves the interplay between the two adjacent codons , in which I•A wobble interaction in the P site in conjunction with competition between tRNAs entering the A site results in the frameshift , consistent with a model proposed by Baranov et al . ( Baranov et al . , 2004 ) . Two lines of evidence support this mechanism . First , we demonstrated that , in the asc1Δ mbf1Δ double mutant , ribosomes frameshift at a single CGA-CGG codon pair ( in a particular context ) when the CGA codon occupies the P site . We infer that CGA codon in the P site is generally important for frameshifting , because six of the seven codon pairs on which ribosomes frameshift are CGA-NNN and the three efficient pairs are CGA-CNN . The wobble interaction between the CGA codon and tRNA could weaken the interaction between mRNA and the ribosome , which in turn could slow down the elongation cycle . Second , we found that frameshifting is influenced by the abundance of the in-frame and out-of-frame tRNAs for next position , which implies that the frameshift occurs after translocation of the CGA from the A site to the P site . We speculate that the flexibility of the wobble base pair interaction between inosine and other nucleotides could actively facilitate the acceptance of out-of-frame A site tRNA . For instance , we consider that a rare instance in which the A base in CGA is bulged out might be stabilized by the very strong I•C interaction , increasing the time available to accept the out-of-frame tRNA . The eukaryotic specific reading frame maintenance activity , involving Mbf1 and ribosomal proteins Rps3 and Asc1 , is likely to be important for translation accuracy in the yeast genome . Mutations in either RPS3 or MBF1 suppressed frameshifting mutations in several native yeast genes ( Hendrick et al . , 2001 ) . Moreover , mutations in MBF1 and ASC1 resulted in detectable frameshifting in a set of native gene sequences with only a single inhibitory codon pair flanked by six adjacent codons on each side , although it is apparent that the frameshifting potential within a particular sequence is not simply due to the presence of a single inhibitory codon pair . These results confirmed that Mbf1 with Rps3 and Asc1 play a critical role in maintaining the reading frame during normal translation cycles . It is still unknown why this eukaryote-specific reading frame maintenance system evolved and why it is important to eukaryotes , but not bacteria . Strains , plasmids , and oligonucleotides used in these studies are listed in Supplementary file 3 , Tables 1-3 . Parents for all yeast strains described in this study were either BY4741 ( MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ) or BY4742 ( MATαhis3Δ1 leu2Δ0 lys2Δ0 ura3Δ0 ) ( Open Biosystems ) . The GLN4 ( 1-99 ) - ( CGA ) 6+1-URA3 reporter used in the selection was constructed with PCR-amplified DNAs ( using oligonucleotides OJYW085 , 086 , 041 , 089 , 095 and 099 ) , assembled by Ligation Independent Cloning ( LIC ) methods ( Alexandrov et al . , 2004; Aslanidis and de Jong , 1990 ) and then integrated into the CAN1/YEL063C locus on the chromosome V , selecting for canavanine-resistance; constructs were checked by sequencing of genomic PCR fragments . RNA-ID reporters were constructed as described previously and integrated at the ADE2 locus , using selection with MET15 marker in MATa strains or S . pombe HIS5 marker in MATα strains ( Dean and Grayhack , 2012; Gamble et al . , 2016; Wolf and Grayhack , 2015 ) . Yeast strains bearing MBF1 deletions were constructed by amplification of the kanR cassette in the yeast strain from the corresponding knockout strain in the systematic deletion collection ( Open Biosystems ) ( Giaever et al . , 2002 ) . The MATa yeast strain bearing a deletion of RPS3 was constructed by amplification of bleR cassette ( Gueldener et al . , 2002 ) ( oligos OW443 and OW445 ) and integration of this DNA into a strain bearing an URA3 [RPS3] covering plasmid ( pEAW433 ) . Yeast strains bearing deletions of ASC1 marked with the S . pombe HIS5 marker ( AW768 ) , which have been described previously ( Wolf and Grayhack , 2015 ) , were constructed and maintained in the presence of a plasmid born copy of ASC1 on a 2µ , URA3 plasmid . To obtain the asc1Δ strain from the selection parent strain , the ASC1 gene was deleted by a bleR cassette obtained by PCR amplification with oligos OW125 and OW126 . Plasmids bearing the MBF1 gene were constructed by amplification of chromosomal MBF1 gene from −580 in 5’ UTR to +300 in 3’ UTR with oligos OJYW124 and OJYW125 , followed by cloning into the 2µ , LEU2 vector ( pAVA0577 ) and into the CEN , LEU2 vector ( pAVA0581 ) to create pEJYW203 and pEJYW176 respectively . The chromosomal HA-tagged MBF1 was constructed by PCR amplification of HA-kanR sequence from pYM45 ( Euroscarf ) ( Janke et al . , 2004 ) with oligos OJYW130 and OJYW132 , bearing homology to MBF1 , followed by integration into the MBF1 locus . This MBF1-HA KanR cassette from −580 in 5’UTR to +300 in 3’UTR of MBF1 ( +1992 including KanR sequences ) was amplified from the chromosome with oligos OJYW157 and OJYW158 , cloned into the XmaI and NheI sites in Bluescript as pEJYW279 . The mbf1 point mutations K64E and I85T were individually introduced into the plasmid pEJYW279 to make pEJYW302 and pEJYW307 respectively . The mbf1-K64E cassette was directly PCR-amplified from the mutant strain YJYW290-P38 with oligos OJYW157 and OJYW158 followed by digestion with XmaI and BamHI and integration into these two sites on pEJYW279 . The mbf1-I85T mutation was introduced by PCR amplification from MBF1-HA cassette with OJYW170 , which contains the mutation , and OJYW166 , followed by integration into pEJYW279 between BamHI and AatII sites . Reconstructed mbf1 point mutants were introduced into YJYW2566 ( BY4741 , HIS3+ ) with XmaI/NheI digested pEJYW302 and pEJYW307 selecting with KanR marker . The plasmid template for in vitro transcription of GFP and RFP fragments ( pEJYW407 and pEJYW409 ) was constructed by PCR-amplifying pEAW315 with oligos OJYW295/OJYW296 ( for GFP ) or OJYW297/OJYW299 ( for RFP ) followed by digestion with SphI and XmaI and integration into these two sites on pSP73 ( Promega , cat . # P2221 ) . Plasmids expressing tRNAs were obtained from Phizicky and Grayhack lab stocks ( Guy et al . , 2012; Han et al . , 2015; Letzring et al . , 2010 ) . Ura+ mutants were selected from 40 independent cultures of each MATa and MATα parent strains ( YJYW289 , YJYW329 ) , and then were analyzed by flow cytometry to measure GFP and RFP expression . Ura+ GFP+ mutants , indicative of increased frameshifting efficiency , were selected for further study , with an emphasis on mutants that exhibited higher levels of frameshifting , that is GFP/RFP >4 , ( 28% MATα and 66% MATa mutants ) . Diploids between 12 MATa mutant and 20 MATα mutants were created by mating in YPD for 2 hr at 25°C and selection on SD-Lys-Leu-His media for diploid cells , followed by streaking for single colonies . Then overnights of the resultant diploids and their haploid parents were spotted on SD-Leu and SD-Leu-Ura plates , which were grown at 30°C . To identify the relevant mutation in YJYW290-P25 , we obtained the Leu- derivative of this mutant ( YJYW315 ) by screening replica plated single colonies from an overnight in YPD on YPD and SD-Leu plates . The Ura+/FOA-sensitive phenotype of this mutant was complemented with a genomic tiled library ( Jones et al . , 2008 ) , selecting for FOA-resistant cells . First , 17 pools of DNA , each of which contained 96 plasmids ( Jones et al . , 2008 ) , were transformed individually with >1000 colonies per plate . Transformants of each pool were then scraped and saved in 2 ml YPD +8% DMSO . These saves were plated based on their OD600 ( 2 × 107 cells/OD600 x ml ) to obtain approximately 5 , 000 cells on SD-Leu and 50 , 000 cells on SD-Leu +0 . 5 xFOA . For 16 of 17 pools , there were no colonies on the FOA plates , while transformants of pool 15 had 330 FOA-resistant colonies with 1404 colonies on SD–Leu plate , corresponding to FOA-resistance for 2 . 3% cells . The plasmids responsible for FOA-resistance was identified by complementing with plasmids from individual rows and columns in this pool as described above , followed by complementation with individual plasmids . Two plasmids from this pool conferred FOA-resistance and share a single gene , MBF1 . The MBF1 gene in 19 recessive mutants was amplified from their genomic DNA with oligos OJYW124 and OJYW125 , followed by sequencing to confirm the mutated residues . Whole genome sequencing on two dominant MATα mutants was performed to identify the mutated genes . For each strain , ~30 OD600 yeast cells were harvested and re-suspended in 1 ml prep buffer ( 2% Triton X-100 , 1% SDS , 100 mM NaCl , 10 mM Tris-Cl pH 8 . 0 , 1 mM EDTA ) with ~1 . 5 g Zirconia/Silica beads ( from BioSpec , catalog# 11079105z ) and 1 ml PCA pH 8 . 0 . The suspension was then vortexed at top speed for 3 min and mixed with 1 ml TE pH 8 . 0 , followed by centrifugation in prespun PLG tubes ( from 5prime , catalog# 2302830 ) . Nucleic acids in the aqueous layer were ethanol precipitated with 5 ml 100% ethanol , followed by freezing on dry ice and centrifugation for 20 min at 4 , 000 rpm at 4°C . The pellet was re-suspended in 200 µl TE and incubated at room temperature for 1 hr with 0 . 2 µg/µl RNaseA to remove RNA contamination , followed by addition of 200 µl 1 M Tris-Cl pH 8 . 0 , 2 µl of 5 mg/ml glycogen and 400 µl PCA , and centrifugation for 2 min at top speed at 4°C . The aqueous layer ( ~360 µl ) was precipitated with 720 µl 100% ethanol and frozen on dry ice for 15 min; resulting pellets were re-suspended in 100 µl TE pH 8 . 0 and 100 µl 1 M Tris-Cl pH 8 . 0 , followed by precipitation again with 400 µl 100% ethanol . The DNA pellet was then washed with 500 µl 70% ethanol and finally re-suspended in 50 µl sterile ddH2O . Whole genome sequencing was performed by the UR Genomics Research Center resulting in RPS3 mutations in these two MATα mutants . Mutations in two MATa dominant mutants were then identified by amplification of RPS3 cassette with oligos OJYW159 and OJYW210 , followed by sequencing . Appropriate control strains ( previously studied ) and 2–4 independent isolates of each strain being tested were grown overnight at 30°C in media indicated , diluted to obtain OD600 of 0 . 5 , then serially diluted 10-fold twice; 2 µl diluted cells were then spotted onto the indicated plates and grown at different temperatures for at least two days . To examine mutants in either RPS3 or ASC1 , reporters were introduced into sets of strains bearing an URA3 covering plasmid with either RPS3 or ASC1 , depending upon the chromosomal deletion . All sets of strains in a given panel contained the same URA3 plasmid . Prior to analysis of GFP expression , strains were streaked on FOA containing plates , then single colonies were grown for analysis by flow cytometry . Yeast strains bearing the modified RNA-ID reporters were grown overnight at 30°C in YP media ( for strains without plasmid ) or appropriate synthetic drop-out media ( for strains with plasmid ) containing 2% raffinose + 2% galactose + 80 mg/L Ade . The cell culture was diluted in the morning such that to the culture had a final OD600 between 0 . 8–1 . 0 . Analytical flow cytometry and downstream analysis were performed for four independent isolates of each strain ( Outliers were rejected using a Q test with >90% confidence level ) as previously described ( Dean and Grayhack , 2012 ) . Each flow experiment was also performed with proper controls including a GFP- , RFP+ strain . The GFP/RFP value from this control strain was subtracted from all tested strains on the same day to show signals above background ( negative values are set to 0 ) . P values were calculated using a one-tailed or two-tailed homoscedastic t test in Excel , as indicated in the source data for relevant figures . Western analysis of the GFP fusion proteins in the modified RNA-ID reporter and Mbf1 protein in yeast strains were performed with anti-HA antibody as described previously ( Gelperin et al . , 2005 ) . mRNA measurements with reverse transcription ( RT ) reaction and quantitative PCR were performed as described previously ( Gamble et al . , 2016 ) with one significant difference . Quantification of mRNA was performed using in vitro transcribed GFP and RFP mRNA fragments , synthesized from linearized plasmid pEJYW407 and pEJYW409 using RiboMAX Large Scale RNA Production System-T7 ( Promega , cat . # P1300 ) . The synthesis reaction was followed by DNase treatment to remove the DNA template and by elution through MicroSpin G-25 columns ( GE , cat . # 27-5325-01 ) to remove unincorporated nucleotides . The synthesized RNA sample was analyzed by ultraviolet light absorbance at 260 nm on a nanodrop to determine the concentration and by electrophoresis to assess integrity . Each qPCR plate contained 5-point 1:5 dilution standard curves for both GFP and RFP , which were optimized to ensure that all samples fall into the linear range of the curves . For each tested strain , three biological replicates were analyzed . To purify the frameshifted peptide from yeast , a LEU2 plasmid containing either in-frame or +1 frame protein purification constructs were transformed into the asc1Δ mbf1Δ strain ( YJYW378 ) . Two independent transformants ( FOA treated ) of each construct were grown overnight in SD-Leu media and transferred into 80 ml S-Leu + 2% raffinose media in the morning . After reaching an OD600 of 0 . 8–1 . 2 , expression of the GST-StrepII-ZZ construct was induced by addition of 40 ml 3xYP + 6% galactose and growth was continued for 10 hr . Cells were collected by centrifugation and cell pellets were quick frozen on dry ice . The cell pellets were re-suspended in 1 ml extraction buffer ( 50 mM Tris-Cl pH 7 . 5 , 1 mM EDTA , 4 mM MgCl2 , 5 mM DTT , 10% Glycerol , 1 M NaCl , 2 . 5 µg/ml leupeptin , 2 . 5 µg/ml pepstatin ) and lysed with bead beating ( 10 repeats of 20 s beating followed by 1 min on ice ) , essentially as described previously ( Quartley et al . , 2009 ) . The cell lysate was collected from the bead beating tubes by puncturing the bottom with a hot needle and blowing with low pressure air . Solid contents were removed by centrifugation before the remaining lysate was divided into half and purified on either GSH or Streptactin resin . For GST purification: the cell lysate was first diluted with equal volume No Salt Wash Buffer ( 50 mM Tris-Cl pH 7 . 5 , 4 mM MgCl2 , 5 mM DTT , 10% Glycerol ) to bring the salt to 0 . 5 M NaCl . GSH resin [Glutathione sepharose-4B from GE , catalog# 17-0756-01; pre-washed with Wash Buffer ( No Salt Wash Buffer + 0 . 5 M NaCl ) ] ( 50 µl/ ml of lysate ) was added to the diluted cell lysate and the mixture was nutated for 3 hr at 4°C . The resin was separated from the liquid by centrifugation at low speed ( <3 , 000 rpm ) and washed twice with 0 . 5 ml Wash Buffer followed by 20 min nutation . The bound protein products were then eluted by nutating for 40 min with 100 µl Elution Buffer ( Wash buffer + 20 mM NaOH + 25 mM glutathione ) ; the elution step was repeated to increase the yield . For Strep purification: the cell lysate was diluted with 5x volumes No Salt Wash Buffer ( 100 mM Tris-Cl pH 7 . 5 , 1 mM EDTA , 2 . 5 µg/ml leupeptin , 2 . 5 µg/ml pepstatin ) to bring the salt to 150 mM NaCl . MagStrep ‘type3’ XT beads [from IBA , cat# 2-4090-002; pre-washed with Wash Buffer ( No Salt Wash Buffer + 150 mM NaCl ) ] were added to the diluted cell lysate ( 80 µl/ 3 . 3 ml diluted cell lysate ) . After nutating for 2 hr at 4°C , resins were separated from liquid using a magnetic separator , then the resin was washed with 1 ml Wash Buffer three times without additional incubation . The bound protein products were then eluted by adding 50 µl Elution Buffer ( Wash buffer + 50 mM biotin ) and nutating for 10 min , followed by separation using magnetic separator; the elution step was repeated to increase the yield . The elution samples from both GST and Strep purification were analyzed by SDS-PAGE , followed by staining with Coomassie Blue . The bands from Strep purification of both in-frame and +1 frame constructs were excised and analyzed on the Q Exactive Plus Mass Spectrometer in the Mass Spectrometry Resource Center of the University of Rochester Medical Center .	Proteins perform all the chemical reactions needed to keep a cell alive; thus , it is essential to assemble them correctly . They are made by molecular machines called ribosomes , which follow a sequence of instructions written in genetic code in molecules known as mRNAs . Ribosomes essentially read the genetic code three letters at a time; each triplet either codes for the insertion of one of 20 building blocks into the emerging protein , or serves as a signal to stop the process . It is critical that , after reading one triplet , the ribosome moves precisely three letters to read the next triplet . If , for example , the ribosome shifted just two letters instead of three – a phenomenon known as “frameshifting” – it would completely change the building blocks that were used to make the protein . This could lead to atypical or aberrant proteins that either do not work or are even toxic to the cell . For a variety of reasons , ribosomes will often stall before they have finished building a protein . When this happens , the ribosome is more likely to frameshift . Cells commonly respond to stalled ribosomes by recruiting other molecules that work as quality control systems , some of which can disassemble the ribosome and break down the mRNA . In budding yeast , one part of the ribosome – named Asc1 – plays a key role in recruiting these quality control systems and in mRNA breakdown . If this component is removed , stalled ribosomes frameshift more frequently and , as a result , aberrant proteins accumulate in the cell . Since the Asc1 recruiter protein sits on the outside of the ribosome , it seemed likely that it might act through other factors to stop the ribosome from frameshifting when it stalls . However , it was unknown if such factors exist , what they are , or how they might work . Now , Wang et al . have identified two additional yeast proteins , named Mbf1 and Rps3 , which cooperate to stop the ribosome from frameshifting after it stalls . Rps3 , like Asc1 , is a component of the ribosome , while Mbf1 is not . It appears that Rps3 likely stops frameshifting via an interaction with the incoming mRNA , because a region of Rps3 near the mRNA entry site to the ribosome is important for its activity . Further experiments then showed that the known Asc1-mediated breakdown of mRNAs did not depend on Mbf1 and Rps3 , but also assists in stopping frameshifting . Thus , frameshifting of stalled ribosomes is prevented via two distinct ways: one that directly involves Mbf1 and Rps3 and one that is promoted by Asc1 , which reduces the amounts of mRNAs on which ribosomes frameshift . These newly identified factors may provide insights into the precisely controlled protein-production machinery in the cell and into roles of the quality control systems . An improved understanding of mechanisms that prevent frameshifting could eventually lead to better treatments for some human diseases that result when these processes go awry , which include certain neurological conditions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2018	Multi-protein bridging factor 1(Mbf1), Rps3 and Asc1 prevent stalled ribosomes from frameshifting
Animals integrate the different senses to facilitate event-detection for navigation in their environment . In vertebrates , the optic tectum ( superior colliculus ) commands gaze shifts by synaptic integration of different sensory modalities . Recent works suggest that tectum can elaborate gaze reorientation commands on its own , rather than merely acting as a relay from upstream/forebrain circuits to downstream premotor centers . We show that tectal circuits can perform multisensory computations independently and , hence , configure final motor commands . Single tectal neurons receive converging visual and electrosensory inputs , as investigated in the lamprey - a phylogenetically conserved vertebrate . When these two sensory inputs overlap in space and time , response enhancement of output neurons occurs locally in the tectum , whereas surrounding areas and temporally misaligned inputs are inhibited . Retinal and electrosensory afferents elicit local monosynaptic excitation , quickly followed by inhibition via recruitment of GABAergic interneurons . Multisensory inputs can thus regulate event-detection within tectum through local inhibition without forebrain control . Sensorimotor circuits have been studied in a wide range of biological organisms in pursuit of identifying the operational principles that govern the integration of sensory information from different modalities for the generation of goal-directed behavior . The optic tectum ( superior colliculus in mammals ) , has received particular attention for its distinct role in orienting behavior , i . e . the control of orienting and avoidance gaze movements ( Dean et al . , 1989; Moschovakis et al . , 1996; Basso and Wurtz , 1997; Sparks , 2002 ) , through the integration of different sensory modalities ( which are species-dependent ) like vision , auditory and electroreception ( Bodznick and Northcutt , 1981; Meredith and Stein , 1986; Wallace et al . , 1996 , Gingras et al . , 2009 ) . Although combining multiple sensory inputs has been proposed to increase the reliability of event detection in the environment ( Ernst and Banks , 2002; Fetsch et al . , 2009 ) , little is known about the neural mechanisms underlying this integration . Studies of the superior colliculus have established a set of empirical principles that place constraints on the spatial and temporal dimensions underlying multisensory integration ( Stein and Stanford , 2008 ) . In particular , extracellular activity correlated to gaze shift execution would increase with spatiotemporally congruent cues from two senses or decrease with spatially disparate and/or temporally asynchronous cues ( Meredith et al . , 1987; Meredith and Stein , 1996; Kadunce et al . , 1997; Recanzone , 2003 ) . Conceptual and computational models of multisensory integration have proposed the existence of an inhibitory mechanism to account for these effects ( Rowland et al . , 2007; Alvarado et al . , 2008; Ursino et al . , 2009; Ohshiro et al . , 2011; Miller et al . , 2015 ) , although they have so far remained hypothetical . The goal of this study is to determine the cellular and synaptic mechanisms embedded within the optic tectum that control multisensory integration . In many vertebrates , vision and electroreception are the spatial senses that are used to localize predators and prey in their immediate environment , whereas other species also rely on auditory , somatosensory , infrared , echolocation and/or magnetic systems . These modalities are represented within the optic tectum and are used for orienting and avoidance behaviors ( Hartline et al . , 1978; Semm and Demaine , 1986; Valentine and Moss , 1997 ) in vertebrates extending from lampreys to primates ( Wurtz and Albano , 1980; Nieuwenhuys and Nicholson , 1998; Saitoh et al . , 2007; Jones et al . , 2009; Asteriti et al . , 2015; Kardamakis et al . , 2015 ) . We have previously shown in the lamprey that site-specific stimulation across the deep layer of the optic tectum gives rise to eye-head gaze shifts of given amplitude and direction , thus , showing the existence of a motor map ( Saitoh et al . , 2007 ) . We now show that visual and electroreceptive inputs are integrated in the same deep layer neurons of the optic tectum , which provide the output to different brainstem centers . Projecting output neurons , as well as local interneurons , receive monosynaptic excitatory input from both sensory afferent pathways and also disynaptic inhibition triggered by the same afferents . This applies if the two signals are activated from the same point in space , whereas signals from surrounding areas provide only inhibition ( Kardamakis et al . , 2015 ) . The membrane properties of the tectal output neurons ensure the temporal integration of bimodally triggered excitatory and inhibitory currents . Due to the highly conserved organization of the optic tectum ( Nieuwenhuys and Nicholson , 1998; Saitoh et al . , 2007; Jones et al . , 2009; Asteriti et al . , 2015; Kardamakis et al . , 2015 ) , we anticipate that the mechanisms of integration of two senses on single output neurons , as demonstrated here , may also apply to other vertebrates . The full projection patterns of tectal efferents to the brainstem and incoming visual and electroreceptive afferents arising from the retina and the octavolateral area in the intact brain was examined ( Nieuwenhuys and Nicholson , 1998; Jones et al . , 2009; Ronan and Northcutt , 1987 ) ( Video 1 ) . For this , we used a tracer injection into the optic tectum followed by passive CLARITY-optimized light-sheet microscopy ( Tomer et al . , 2014 ) . Tectal efferents arise from neurons in the deep layer ( DL ) and project to the brainstem , where they make direct synaptic contacts onto the somata of reticulospinal ( RS ) neurons in the middle rhombencephalic reticulospinal nucleus ( MRRN ) . These projecting neurons in the deep layer of the optic tectum ( or superior colliculus ) control gaze movements ( Sparks , 2002; Robertson et al . , 2006 , Saitoh et al . , 2007; Jones et al . , 2009; Kardamakis et al . , 2015 ) and will herein be referred to as ‘output neurons’ . 10 . 7554/eLife . 16472 . 003Video 1 . Sensory inputs and motor output in the lamprey optic tectum in the intact brain using COLM . The visual input through the optic tract ( OpT ) to the optic tectum ( OT ) , and the electrosensory afferents from the octavolateral area ( OLA ) are shown after a neurobiotin injection in the OT , in a cleared brain using the method CLARITY . The motor output can be also followed from the deep layer to the reticulospinal cells in the brainstem . The brain is shown from a dorsal view and in the bottom-left corner a schematic in a sagittal view can be seen . A moving bar indicates the approximate region shown at each moment . Areas of interest are annotated through the duration of the movie . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 003 The optic tract , which carries retinal information , and the octavolateral tract , which carries electrosensory information to the optic tectum can be seen distinctly in Video 1 . By injecting two different tracers into the contralateral optic nerve and the contralateral octavolateral area ( Figure 1A-i ) , we revealed that the two sensory modalities have spatially segregated termination zones with minimal overlap ( retina in red , octavolateral in green; n = 3 , Figure 1A-ii ) . Retinal input targets the superficial layer , while electroreceptive afferents project to the intermediate layer . The deep layer ( DL , visible with the Nissl stain; in blue ) contains neurons that have dendritic arbors extending through the intermediate and into the superficial layers ( Figure 1A-iii ) , as revealed by intracellular staining of output neurons that were prelabeled following retrograde tracer injections into the MRRN . Effectively , they exhibit the optimal morphological structure to support sampling of layer-specific inputs carrying visual and electrosensory information . 10 . 7554/eLife . 16472 . 004Figure 1 . Integration of vision and electroreception in the deep layer of the lamprey optic tectum . ( A ) Inset i: Schematic of the lamprey brain showing the visual ( blue ) and electrosensory ( red ) afferents targeting the optic tectum ( OT ) . Inset ii: Photomicrograph of the optic tectum in a transversal view showing the retinal afferents reaching the most superficial layers ( red ) , and the octavolateral fibers innervating the intermediate layers ( green ) . Inset iii: Morphology of an output neuron in the deep layer retrogradely labeled following a tracer injection in the middle rhombencephalic reticulospinal nucleus ( MRRN ) and filled intracellularly with Neurobiotin while performing whole-cell recordings . Output cells extend their dendrites to the intermediate and superficial layers where the electrosensory and the visual inputs enter and terminate , respectively . Abbreviations: SL , superficial layer; IntL , intermediate layer; DL , deep layer . Scale bars: Inset ii , 100 µm; Inset iii , 50 µm . ( B ) Experimental settings for performing extracellular recordings during multisensory integration in the optic tectum . Dorsal view of the preparation , including the brain , the eyes and electrosensory areas ( depicted by the skin patches; for more information see Bodznick and Preston [1983] ) , while driving output activity with light and electrical stimuli that are spatiotemporally aligned in the immediate surrounding . Abbreviations: rec: extracellular recording electrode . ( C ) Rectified local field potentials obtained from visual ( inset i ) , electrosensory ( inset ii ) and bimodal sensory activation ( inset iii ) . Upper traces show sensory stimulation before ( black ) , and after local application of 10 μM gabazine ( green ) . Horizontal dotted lines illustrate the level of peak activity during control . ( D ) Sensory response against stimulus duration ( 50–1000 μs ) for visual , electroreceptive and bimodal activation , with and without local inhibition . The integral under the curve of rectified local field potentials , as those shown in C , is plotted on the y-axis and normalized to the maximum bimodal response measured during control ( n = 13 ) . Paired t-test gave statistical significance as indicated ( p<0 . 01 , p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 00410 . 7554/eLife . 16472 . 005Figure 1—figure supplement 1 . Actual responses against the predicted arithmetic sum of unisensory responses . Responses for visual ( blue ) , electrosensory ( red ) and multisensory ( black ) , are compared to the predicted arithmetic sum of the unisensory responses in control conditions ( green line ) . Responses after the application of Gabazine are shown as dotted lines . Data point represent the numerical values of the integral under the curve after full-wave rectification of the neural activity ( see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 005 To test the mechanisms underlying the integration of the two sensory modalities in tectal output neurons , we first used an intact preparation that enabled extracellular monitoring of neural activity from the deep layer , while visual and electrosensory inputs could be stimulated separately or in combination with coordinated spatial and temporal alignment ( Figure 1B; see Materials and methods ) . Separate stimulation of visual inputs with brief pulses of light ( 500 ms duration ) or electroreceptive inputs activated by electrical pulses ( 30 ms duration ) led to bursts of activity in the output neurons ( Figure 1C-i , ii , control traces ) . When the two sensory inputs were delivered in an overlapping temporal sequence , there was a marked enhancement of deep layer neural activity ( Figure 1C-iii , control; n = 6 ) . To achieve an enhancement , the two stimuli had to be delivered to the same parts of the visual and electroreceptive fields , respectively . To quantify the impact of stimulus strength on bimodal integration , we systematically varied the extent of visual and electrosensory activation by means of local electrical microstimulation of the retina and the anterior lateral line nerve , respectively ( for further information , see Materials and methods ) . To mimic the spatial resolution of visual stimuli , we activated the retinal area that coincided with the receptive field center of our recording sites . Stimulation of the rostral branches of the anterior lateral line nerve would simulate electrosensory stimuli incoming from the frontal regions ( Ronan and Northcutt , 1987 ) . Varying stimulus durations ( Figure 1D ) ranging from 50 μs to 1 ms were used to drive increasing tectal responses in the deep layer , which were measured from their rectified activity . The lower curves in Figure 1D illustrate the normalized effect that stimulus duration had on visual , electrosensory and combined responses during physiological conditions ( n = 13 ) . Unisensory inputs generated up to ~55% of the maximal bimodal response across the entire stimulus range . During bimodal activation , the combined response significantly exceeded the unimodal responses ( visual-bimodal: p<0 . 01; electrosensory-bimodal: p<0 . 001 between 100–1000 μs; Figure 1D , black control trace ) . Recently , we have shown how on-receptive field visual responses in tectal output layer neurons can be suppressed , via the local inhibitory system , by the presence of multiple visual stimuli located at disparate positions in the visual field ( Kardamakis et al . , 2015 ) . We now show that this unisensory response suppression can also be achieved by using multiple stimuli of different sensory modalities ( Figure 2A ) . Once a tectal region responsive to a local electroreceptive stimulus was established ( red trace , Figure 2B ) , we were able to suppress the magnitude of the on-response ( red trace ) by delivering a stimulus to an off-response region of the retina ( black trace ) . To confirm that the lack of activity in response to the visual stimulation ( blue trace ) was not due to unsuccessful stimulation , we performed control recordings in other tectal regions , as well as in downstream brainstem regions . An average response reduction of ~75% was observed during spatially misaligned cross-modal sensory stimulation ( Figure 2C; red for electrosensory and black for bimodal stimulation; n = 5 ) . By contrast , we were able to suppress by a negligible amount of only ~3% ( data not shown ) when the opposite combination of sensory modalites were used , i . e . on-responses to visual and off-response to electrosensory stimuli . The underlying cause that gives rise to this asymmetry remains unclear . 10 . 7554/eLife . 16472 . 006Figure 2 . Spatially misaligned stimuli give rise to response reduction . ( A ) Using the experimental strategy described in Figure 1B , we applied spatially disparate visual and electrosensory stimuli while recording responses in the contralateral optic tectum . ( B ) Local field potentials in response to electrosensory stimulation ( red trace ) , were drastically reduced when a different region of the tectal map was visually stimulated ( blue trace ) . The responses when combining both sensory modalities are shown in black . The yellow traces on top show the rectified signals . ( C ) Plot showing the normalized responses for electrosensory activation ( red ) , before and after simultaneously stimulating a visual off region ( black ) . For each animal ( n = 5 ) , stimulation was repeated throughout 10 sweeps . Responses were quantified as the area under the rectified signal and normalized to the maximum . Paired t-test gave a statistical significance of *p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 006 To evaluate the role of local inhibition , we microinjected the GABAA-receptor antagonist gabazine ( 10 μM ) in the area of the recording site ( Figure 1B ) . We observed a marked increase in the magnitude of the response during drug application ( Figure 1C , bottom traces ) . The responses to visual , electrosensory or the combination of both were now virtually identical . Without inhibition , neurons in the deep layer became irresponsive to the following incoming stimuli for a refractory period that exceeded 30 s ( n = 6 ) . When gabazine was microinjected ( Figure 1C , D ) , the responses increased by approximately 80–100% compared to their original magnitude throughout the stimulus durations ( for visual: 85 . 5 ± 13 . 7%; electrosensory: 97 . 4 ± 14 . 4%; for bimodal: 82 . 4 ± 13 . 0 , Means ± SD; n = 13 ) and were accompanied with a significantly higher degree of variance . At the same time , the relationship between unimodal and bimodal responses quickly deteriorated ( no significant differences ) with visual inputs dominating bimodal activation . To visualize the extent of the enhancement of unisensory response during bimodal activation , we plotted the actual data in control conditions and after the application of Gabazine as a function of the predicted arithmetic sums of each unisensory response ( Figure 1—figure supplement 1 ) . Our data thus suggest that tectal inhibition is critical to ensure response enhancement and a stable relationship between response amplitudes to unimodal inputs . Given that the role of inhibition is critical for integrating the two unimodal inputs ( Figure 1C , D ) , a central issue was to determine the inhibitory mechanisms that shape the synaptic integration at the level of the single output neuron . To address this , we performed whole-cell and cell-attached recordings from these cells by exposing the deep layer of the optic tectum in a slightly oblique sagittal midbrain section ( ~500–600 μm ) that maintains the mesencephalon and parts of the diencephalon and rhombencephalon directly adjacent to the midbrain ( Figure 3A ) . We could then selectively stimulate both the optic tract and the fiber bundle that exits the octavolateral nucleus ( Figure 3A ) , while recording from tectal output neurons . 10 . 7554/eLife . 16472 . 007Figure 3 . Output neurons receive visual and electrosensory inputs . ( A ) ( Top ) Schematic of the lamprey brain in a dorsal view showing the sensory afferent input to the optic tectum . Dotted rectangle shows the brain region of interest ( sectioned and shown in sagittal view; Bottom ) illustrating the settings for performing cell-attached and whole-cell recordings in tectal output neurons while stimulating the visual ( via the optic tract , OpT ) and electrosensory ( via the octavolateral area , OLA ) inputs . Abbreviations: OpT , optic tract; OLA , octavolateral area; OT , optic tectum; stim: stimulation electrode; MRRN , middle rhombencephalic reticulospinal nucleus . ( B ) ( Left ) Tectal output cell recordings driven by visual ( OpT ) , electrosensory ( OLA ) and bimodal ( both ) inputs in cell-attached configuration ( stimulation train of 20 pulses at 10 Hz; low threshold 1–10 μA ) . Time-locked action potentials can be observed in response to an impulse from either sensory pathway with enhanced excitability occurring during bimodal activation . ( Right ) Quantification of spike responses to unimodal and bimodal activation . A probability of unity indicates that an output neuron responds to all given impulses throughout 10 sweeps ( n = 6 ) . Paired t-test gave a statistical significance of *p<0 . 001 . ( C ) ( Left ) Action potentials discharged in response to unimodal and bimodal input at stimulation intensity threshold ( T ) . ( Middle ) Bimodal inputs drive suprathreshold responses in output neurons , with unimodal inputs failing to elicit spiking by adjusting T to a 70% of its initial value ( n = 6 ) . ( Right ) Unimodal and bimodal inputs yield equalizing effects with a rapid discharge of action potentials ( at 0 . 7T ) when blocking inhibition with bath application of 10 μM of Gabazine ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 00710 . 7554/eLife . 16472 . 008Figure 3—figure supplement 1 . Unimodal and bimodal sensory activation with and without inhibition – whole-cell recordings . Repetitive stimulation ( 10 pulses at 10 Hz ) of OpT ( left traces , blue ) and OLA ( right traces , red ) yields subthreshold excitatory postsynaptic potentials in the same output neuron in the deepe layer . Pharmacological blockade of GABAergic inhibition with bath application of 10 μM gabazine results in a drastic response enhancement with neurons rapidly firing action potentials ( usually between 2–7 spikes; n = 3 ) . Thicker lines are average traces . These data also provide an intracellular account that was observed with our extracellular recordings ( in the lower traces of Figure 1C ) and cell-attached recordings ( right panel of traces in Figure 2C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 008 To evaluate the reliability of bimodal integration , we initially recorded responses in cell-attached mode to ensure that the cell membrane and cytoplasm would remain intact , while delivering sustained presynaptic stimulation of the sensory afferent pathways ( 20 pulses/10 Hz; Figure 3B , left ) . We then quantified the likelihood of their all-or-none responses to each impulse ( n = 6; Figure 3B , right ) . Bimodal activation resulted in a higher spiking probability of single stimulus-locked action potentials ( statistical differences of p<0 . 001 ) . Unimodal stimuli were usually able to drive output neurons to discharge one time-locked action potential with stimulation intensity thresholds ( 1xT ) ranging from 1–10 μΑ ( Figure 3C , left panel ) . Reducing T by 30% ( 0 . 7xT ) would elicit a suprathreshold response only when both pathways were engaged ( Figure 3C , middle panel ) . This combined action of weaker stimuli leading to stronger bimodal responses is a key feature underlying multisensory integration . This enhancement is completely lost in the absence of local GABAergic inhibition , where output neurons rapidly fired multiple action potentials even in response to normally ineffective stimuli as in the case of 0 . 7xT ( Figure 3C , right panel; see also Figure 1C and Figure 3—figure supplement 1 ) . To determine the integrative properties of tectal output cells , we first established their active membrane properties with whole-cell voltage measurements in response to applied current steps ( Figure 4A-i , ii ) . Notably , these cells have long membrane time constants ( 98 . 8 ± 11 . 5 ms , n = 32; see Materials and methods ) due to their high input resistance ( 0 . 95 ± 0 . 24 GOhm , n = 32 ) . This property facilitates the temporal summation of excitatory currents and ensures that distal visual inputs ( in superficial layer ) and more proximal electrosensory inputs ( in intermediate layer ) will yield functionally similar responses ( i . e . amplitude and temporal characteristics ) at the final integration site in the soma ( in the deep layer; see Figure 1C , top traces ) . Furthermore , output neurons also discharge action potentials with a regular firing pattern in response to stepwise increases of applied current , and display a continuous and linear frequency-current relationship in the lower range of current injections ( Figure 4A-iii; slope: 0 . 53 ± 0 . 05 spikes/pA , n = 32 ) . This was further corroborated in experiments in which the action potentials were blocked intracellularly by QX-314 ( blocking fast sodium channels ) and the suprathreshold membrane fluctuations in response to positive current steps were monitored ( Figure 4A-ii ) . All neurons that were recorded had a linear input-output relationship across a wide range of membrane potentials ( −120 to −20 mV; Figure 4A-iv ) and would , thus , smoothly encode the strength of similar sensory inputs into output firing frequency . 10 . 7554/eLife . 16472 . 009Figure 4 . Sensory excitation and tectal inhibition are integrated by output neurons . ( A ) Inset i . Voltage responses to depolarizing and hyperpolarizing 500 ms current steps of 2 pA per step , elicited from rest at −68 mV . Inset iii . Spike frequency is plotted against current level . Inset ii . Voltage responses to depolarizing and hyperpolarizing 500 ms current steps of 5 pA per step , elicited from rest at −63 mV in the presence of QX-314 intracellularly ( 3 mM ) . Voltage traces marked in red belong to the shallow slope in the I-V plot indicative of a drop in DC impedance . Inset iv: Plot of somatic potential against current level . ( B ) Whole-cell recordings of excitatory and inhibitory postsynaptic potentials evoked in output cells that were held at −70 mV and −20 mV , respectively , in response to repetitive stimulation of OpT ( top traces; blue ) and OLA ( bottom traces; red ) afferents at 10 Hz , in the presence of fast-sodium channel blocker QX-314 in the recording pipette . Average traces are shown as thicker lines . ( C ) ( Top ) Output neuron responses to sensory stimulation ( train of 4 pulses at 10 Hz ) of OpT ( blue traces ) , OLA ( red traces ) and bimodal ( black traces ) from rest at −69 mV . ( Bottom ) Output responses recorded when holding near threshold ( at −45 mV with positive current injection ) reveals evoked inhibitory postsynaptic potentials to single impulses that also persist during bimodal integration . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 009 To determine the synaptic underpinnings of multisensory integration , we performed whole-cell recordings while applying a train of stimuli ( at 10 Hz ) to the retinal ( OpT ) and octavolateral ( OLA ) afferent pathways ( Figure 4B ) . We found that excitatory postsynaptic potentials were evoked onto the same output neurons . A concurrent inhibitory postsynaptic potential was also revealed when the membrane potential was held at −20 mV ( upper range of the linear domain of their V-I curve ) in current clamp mode . This inhibition is feedforward and is produced mainly disynaptically by projections of tectal interneurons onto output neurons ( Kardamakis et al . , 2015; see below ) . In a subset of neurons ( 4/6 ) studied in cell-attached configuration ( as shown in Figure 3 ) , the membrane was later ruptured allowing us to enter into whole-cell configuration and measure the combined excitatory and inhibitory action of stimulating both visual and electrosensory pathways ( Figure 4C , black traces ) . Consistent with extracellular ( Figure 1C ) and cell-attached recordings ( Figure 3A ) , the output neurons reach threshold more reliably and faster when both sensory modalities are present ( black trace in Figure 4C; resting at −69 mV; top traces ) , even though unisensory stimuli can eventually drive spiking ( Figure 4C; retinal in blue; electrosensory in red ) . Notably , the inhibition persisted not only during unimodal but also during bimodal stimulation . This can be seen by the long-lasting inhibitory postsynaptic potential immediately following the first action potential , which is visible when holding the neurons near threshold ( Figure 4C , bottom traces ) . Together , these data indicate that the membrane properties of output neurons are able to support the smooth integration of sensory-evoked excitation coupled to triggered inhibition . To improve the isolation of synaptic conductances underlying this sequence between excitation and inhibition during unimodal and bimodal integration , we blocked action potentials with the fast sodium-channel blocker QX-314 . In voltage-clamp , the synaptic currents evoked by unimodal stimulation summate during bimodal stimulation ( Figure 5A ) . Whole-cell current recordings ( retinal are blue traces; octavolateral are red traces ) began with a direct excitatory current that was followed ~5–10 ms later by an intense inhibitory current ( see also Kardamakis et al . , 2015 ) , when the somatic potential was held at the reversal potentials for inhibition ( −65 mV ) and excitation ( 0 mV ) , respectively . When co-stimulating both afferent pathways , excitatory postsynaptic current and inhibitory postsynaptic current amplitudes increased in direct correspondence to the unimodal-evoked responses ( Figure 5A , B ) . On a trial-to-trial basis , the statistical difference in the magnitude of evoked EPSC and IPSC currents by bimodal and visual or electroreceptive afferent stimulation was significant ( visual-bimodal: p=0 . 008; electrosensory-bimodal: p=0 . 006; one-way ANOVA ) , but not between those evoked during visual and electrosensory ( p=0 . 99 ) . The same was true for the evoked IPSCs ( visual-bimodal: p=0 . 006; electrosensory-bimodal: p=0 . 0009; both visual and electrosensory against bimodal: p=0 . 24 ) . Notably , no statistical significance ( p=0 . 85 for EPSCs and p=0 . 92 for IPSCs ) was detected when comparing the mean differences between the sum of unisensory-evoked and bimodal-evoked currents ( Figure 5B; EPSCs , visual: 28 . 63 ± 5 . 6 pA , electrosensory: 28 . 4 ± 5 . 85 pA , both: 56 . 27 ± 9 . 11 , n = 12; IPSCs , visual: 24 . 04 ± 6 . 15 pA , electrosensory: 33 . 7 ± 14 . 82 pA , both: 55 . 22 ± 15 . 79 pA , n = 7 ) . As predicted by their membrane properties , output neurons are able to temporally integrate the synaptic currents evoked by individual unisensory inputs summate to generate an enhanced bimodal product . 10 . 7554/eLife . 16472 . 010Figure 5 . Excitatory and inhibitory postsynaptic currents and potentials evoked from visual and electrosensory inputs onto tectal output cells . ( A ) Inhibitory and excitatory postsynaptic currents ( EPSCs ) elicited by visual , electrosensory and bimodal stimulation in an output cell recorded in voltage-clamp at 0 mV to show inhibitory currents , and at −65 mV ( equilibrium for chloride-mediated GABAergic inhibition ) to show excitatory currents . Drop lines show the onsets of excitatory and inhibitory currents . ( B ) Quantification of peak amplitudes of postsynaptic currents elicited by OpT ( blue ) and OLA ( red ) stimulation summate linearly when compared to the bimodal stimulation ( grey ) in output cells ( n = 12 ) . Abbreviation: PSC , postsynaptic currents . ( C ) Quantification of postsynaptic potential ( PSP ) amplitudes in output cells evoked by sustained stimulation ( 10 pulses at 10 Hz ) of the OpT ( blue ) , OLA ( red ) and bimodal input ( black ) and recorded in current clamp mode . Values are normalized to the first PSP . Both excitatory postsynaptic potentials ( EPSPs , upper curves ) and inhibitory postsynaptic potentials ( IPSPs , lower curves ) decay during the first two to three impulses and reach steady-state thereafter . ( D ) Comparison of the first evoked EPSP and IPSP amplitudes ( obtained from traces as shown above ) . In current clamp , however , EPSP and IPSP amplitudes do not summate in unimodal and bimodal conditions . QX-314 was applied in the pipette during these recordings . Data represented as Means ± SEM . One-way ANOVA was used to determine the p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 010 This temporal integration of excitation and inhibition has a differential impact on the membrane potential . Figure 5C shows the quantification of excitatory postsynaptic potentials ( holding at −65 mV; a representative example of evoked EPSPs is shown in Figure 6B ) and inhibitory postsynaptic potentials ( holding at −20 mV ) elicited by separate and combined ( black ) stimulation of the octavolateral ( red ) and optic tract ( blue ) . After a brief transient phase ( usually after three pulses ) , both EPSPs and IPSPs peak constant values during further repetitive stimulation ( Figure 5C ) . Compared to unisensory stimulation , bimodal stimulation resulted not only in larger EPSPs ( visual-bimodal: p=0 . 0001; electrosensory-bimodal; p=0 . 006; visual: 12 . 8 ± 1 . 4 mV , electrosensory: 10 . 5 ± 0 . 9 mV , bimodal: 16 . 4 ± 1 . 5 mV , n = 24 ) , but also in larger IPSPs ( visual-bimodal: p=0 . 05; electrosensory-bimodal; p=0 . 02; visual: 11 . 7 ± 1 . 4 mV , electrosensory: 10 . 2 ± 1 . 2 mV , bimodal: 13 . 9 ± 1 . 2 mV , n = 24; Figure 5D ) . , The quick opposing action of feedforward inhibition curtails the temporal window that allows output neurons to integrate excitatory inputs , thus , restraining the impact that excitation has on membrane potential deflection . When bimodal inputs are used , excitatory currents summate and increase the resultant EPSP amplitude ( Figure 5C , D ) but are quickly quenched by an also stronger amount of inhibition . 10 . 7554/eLife . 16472 . 011Figure 6 . Dynamics between excitation and inhibition during unimodal and bimodal integration . ( A ) ( Top ) Phase plot showing the estimated change in synaptic conductance of an output neuron ( shown in B ) plotted against the reversal potential during OpT ( blue ) , OLA ( red ) and bimodal ( black ) afferent stimulation . Conductances were estimated at each time point for the interval ranging from 10 ms before the impulse until 100 ms after stimulation onset . ( Bottom ) Line graph highlighting the paired differences between reversal potential across cells . Lines link measurements across conditions performed on the same neuron . On average , there is no statistical difference between unimodal and bimodal integration ( n = 10 ) . ( B ) A representative example of the synaptic responses evoked by single impulses of the OpT , OLA and both from an output neuron resting at −65 mV . ( C ) Time course of the estimated underlying synaptic conductances for excitation and inhibition is shown ( average data from n = 10 ) . Drop lines show the onsets of excitation and inhibition ( which shows a delay with respect to excitation of ~5–10 ms ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 011 To capture the dynamics underlying bimodal stimulus integration , we measured the synaptic responses in current clamp at varying holding potentials ( usually at −65 , −45 , and −20 mV ) by using QX-314 in the recording pipette and then calculated the synaptic conductances during the evoked unisensory and bimodal responses using conventional linear methods ( Monier et al . , 2008 ) . The plot in Figure 6A shows how the overall synaptic conductance ( Gsyn; after subtraction of rest conductance ) varies during the first 100 ms of the synaptic response , as plotted against the estimated reversal potential ( Vrev ) . The overall shape of this relationship for visual ( blue dots ) , electrosensory ( red dots ) and bimodal ( black dots ) is very similar on a cell-to-cell basis . The few points with a reversal potential near zero represent the initial portion of the response when excitation dominates . However , the peak conductance is at a reversal potential near −48 to −50 mV . The peak conductance values for unimodal inputs were visual: 1 . 6 ± 0 . 2 nS , electrosensory: 2 . 1 ± 0 . 3 nS; n = 10 , while an increase of ~150–200% occurred during bimodal stimulation ( bimodal: 3 . 3 ± 0 . 6 nS ) . Strikingly , these values were observed for the corresponding reversal potentials , which were −48 . 2 ± 2 . 7 mV for visual inputs , −50 . 4 ± 2 . 2 mV for electrosensory and −48 . 2 ± 2 . 4 mV for combined inputs . These values are near the action potential threshold ( −43 . 7 ± 1 . 0 mV , n = 10; Figure 6A , bottom ) . When the total synaptic conductance is at a maximum , the reversal potential target values that are relatively invariant , despite the additional excitatory current when switching from unimodal to bimodal input . The latency between the onset of excitation and inhibition is time-locked ( Figure 6C ) for both unimodal and bimodal inputs , as well as the latency between their peaks . Figure 6B shows a representative example of the time course of an excitatory response pattern belonging to an output neuron ( in current clamp ) in response to the underlying excitatory ( Ge ) and inhibitory synaptic conductances ( Gi; population average of n = 10 , lower traces; see Figure 6C ) . Visual , electrosensory and bimodal peak values of Ge and Gi were found to be visual: Ge = 1 . 11 ± 0 . 09 nS , Gi = 0 . 89 ± 0 . 08 nS; electrosensory: Ge = 1 . 08 ± 0 . 07 nS , Gi = 1 . 13 ± 0 . 17 nS; bimodal: Ge = 1 . 96 ± 0 . 53 nS , Gi = 1 . 81 ± 0 . 31 nS; n = 10 ) . The latency between the onset of Gi and Ge was between 5–10 ms ( n = 10; see arrows in Figure 6C , and Kardamakis et al . [2015] ) , thus , providing a temporal window of opportunity for output neurons to integrate independent excitatory inputs from each sensory modality . This time lag is sufficient for transient stimuli to efficiently summate inputs towards spike threshold before the onset of the inhibition . The quenching effect of this feedforward inhibition can account for the mismatch between the evoked EPSP amplitudes to their underlying excitatory conductances during bimodal stimulation , i . e . the subadditive effect of combining both sensory inputs . Furthermore , the summating effect observed in the synaptic current measurements obtained in voltage clamp ( Figure 5A ) is also reflected in Ge and Gi ( Figure 6C ) . This close agreement suggests that non-linearities in the form of strong voltage-dependent conductances or shunting inhibition are not likely to play a significant role in the gain regulation of responses in output neurons during multisensory integration . To test how the timing of sensory inputs affects the amplitude of their excitatory responses , we systematically altered the stimulation onset of either sensory input from 0 to 50 ms ( Figure 7A , B ) . The maximal response was obtained when the onset of both sensory-evoked EPSPs was aligned so that excitation would be in-phase , i . e . latency of 0 ms ( Figure 7A , left & 6B , black trace ) , whereas amplitude reduction occurred when afferent stimulation onsets were temporally misaligned using an offset from 5 to 50 ms . For instance , when incoming afferent excitation coincided with a prior afferent-triggered inhibitory event , as in the case of a 5–10 ms delay ( Figure 7B , orange trace ) , an attenuation of the resultant EPSP was observed . Temporal summation was greater for sensory inputs that were separated further apart to the peak of any preceding inhibition , i . e in the decay phase ( typically for latencies > 20 ms; Figure 7B , green trace ) . To quantify this effect , we selectively stimulated each sensory afferent pathway ( with 10 pulses at 10 Hz; Figure 7A ) and normalized the combined amplitudes to the first temporally-aligned EPSP ( Figure 7C ) . The largest combined responses were generated when both stimuli were aligned . A maximal attenuation of 15% was achieved with a 5–10 ms latency ( see also Figure 7B , orange trace ) , matching the maximum peak of inhibition , with a subsequent increase with successively longer delays greater than 10 ms ( n = 10; Figure 7C ) . Thus , the sequence of sensory modality inputs did not have a differential impact on the attenuation level . Taken together , the timing of sensory-evoked inhibitory events regulates the responsiveness of output neurons . 10 . 7554/eLife . 16472 . 012Figure 7 . Temporal effects on bimodal integration . ( A ) The response profile of a representative output neuron during repetitive bimodal stimulation ( 4 pulses at 10 Hz ) of OpT and OLA . Traces from left to right show the excitatory postsynaptic potentials with offsets ranging from 0 ( temporal alignment ) , 10 , 30 and 50 ms . ( B ) Recordings of EPSPs evoked after variable activation of OpT and OLA with temporal offsets of: 0 ms ( aligned; black trace ) , 5 ms ( orange trace ) and 50 ms ( green trace ) . ( C ) Curve fit illustrating the variation of the combined excitatory postsynaptic potential amplitude against the temporal offset between the two sensory modalities . The graph was normalized to the maximal EPSP amplitude , which always occurred when inputs were temporally aligned ( i . e . , 0 ms ) . Average shown from n = 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 012 The key finding that bimodal inputs to the optic tectum can trigger inhibitory responses raised the question of the location of the cells of origin . By removing the diencephalon and ventral parts of the midbrain ( see Figure 3A ) , we excluded potential extrinsic sources of inhibition , thus , enabling us to limit the source of inhibition to within the tectum . As seen in Figure 8A-i , ii , GABAergic interneurons are strategically located within the stratum opticum - at the interface between the superficial and intermediate layers ( Isa et al . , 1998; Del Bene et al . , 2010; Kardamakis et al . , 2015 ) . Previously , we have shown that microstimulation of this population of tectal interneurons elicits monosynaptic inhibition onto output neurons in the presence of glutamate blockade ( 2 mM Kynurenic acid ) to eliminate excitatory transmission ( see Figure 2 in Kardamakis et al . , 2015 ) . To analyze the activation of these cells by visual and electrosensory inputs , we performed whole-cell recordings while microstimulating both sensory pathways . EPSPs were evoked ( Figure 8B ) after a train of stimuli ( at 10 Hz ) in cells held at −65 mV in current-clamp , both after electrosensory ( Figure 8B–i ) and visual stimulation ( Figure 8B-ii ) , showing that the same pool of interneurons is activated by both sensory modalities . Remarkably , sensory stimulation not only evoked excitation ( EPSPs , visual: 8 . 6 ± 1 . 5 mV , electrosensory: 5 . 6 ± 0 . 8 mV , n = 8 ) , but also a prominent inhibition ( when holding at −45 mV; IPSPs , visual: 9 . 8 ± 1 . 6 mV , electrosensory: 9 . 8 ± 1 . 7 mV , n = 8 ) . In addition , these interneurons have a strikingly different morphology ( Figure 8C ) compared to that of output neurons ( Figure 1A ) . Figure 8C illustrates six interneurons that have broad dendritic arbor extending in all planes ( over around 200 µm ) , and also bridge the layers of retinal and electroreceptive fiber termination ( as shown in Figure 1A ) . In half of the neurons that were stained ( 3/6 ) , it was possible to also see long range axonal projections , as required for the lateral inhibition . 10 . 7554/eLife . 16472 . 013Figure 8 . Visual and electrosensory inputs activate the same set of interneurons . ( A ) Inset i: A neural population located at the interface between the superficial and intermediate layers express GABA . These cells can be seen in more detail in inset ii . Scale bars: Inset i , 150 µm; Inset ii , 50 µm . ( B ) An example of a tectal interneuron that receives excitatory inputs from both sensory modalities ( visual and electrosensory ) along with triggered tectal inhibitory inputs . Here , EPSPs ( hold at −65 mV ) and IPSPs ( hold at −45 mV ) are shown in response to repetitive stimulation of the OLA ( red , bottom traces ) and OpT ( blue , top traces ) . The action potentials evoked by the initial impulse when holding at −45 mV are truncated . The morphology of this cell is shown in Figure 8—figure supplement 1 . ( C ) Six reconstructed morphologies of interneurons that were stained with neurobiotin , in the transversal ( inset i ) and sagittal ( inset ii ) dimensions . Scale bars: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 01310 . 7554/eLife . 16472 . 014Figure 8—figure supplement 1 . A representative example of the morphology of a tectal interneuron that was filled with neurobiotin while performing the whole-cell recordings shown in Figure 8B . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 014 To test if superficial layer interneurons are activated in a spatiotopic manner , like deep layer output cells , we used a preparation exposing the optic tectum to allow patch-clamp recordings from the superficial layer while keeping the retina and the optic nerve intact , so that local stimulation was possible ( Figure 9A , top; see Kardamakis et al . , 2015 for experimental procedure ) . By dividing the retina into four quadrants , we determined the one that gave rise to on-receptive field responses by using cell-attached ( Figure 9A , bottom ) and whole-cell recordings ( Figure 9B ) from tectal interneurons located in the stratum opticum . As with tectal output neurons , action potentials time-locked to the stimulus were recorded during on-receptive field activation , but disappeared usually after the second pulse ( in 5 out of 7 neurons ) due to synaptic depression . A representative example of the interplay between visually-evoked synaptic excitation and inhibition in tectal interneurons can be seen when entering whole-cell configuration ( Figure 9B ) with subthreshold EPSPs ( red trace ) accompanied by inhibitory postsynaptic potentials visible when depolarizing near threshold ( ~ −45 mV ) with positive somatic current injections ( blue trace ) . Voltage clamp recordings reveal the sequence of the inward excitatory ( red ) and outward inhibitory synaptic currents ( blue ) in response to the on-receptive field quadrant stimulation of the retina ( Figure 9C ) . This is further decomposed into the underlying synaptic conductance in Figure 9D , where temporal lag of inhibition with respect to excitation is reflected . 10 . 7554/eLife . 16472 . 015Figure 9 . Tectal interneurons receive diverse forms of local inhibition . ( A ) ( Top ) Schematic of the in vitro eye-brain preparation used to locally stimulate regions of the retina ( divided in four quadrants ) while performing whole-cell recordings in the superficial layer interneurons located in the stratum opticum . ( Bottom ) Cell-attached recording of a tectal interneuron in response to electrical microstimulation of the on-receptive field in the retina ( stimulation train of 8 pulses at 10 Hz with a recovery pulse; threshold ~50 μA , n = 7 ) . As with tectal output neurons ( see Figure 3B ) , one time-locked action potential can be observed in response to an impulse delivered to the preferred retinal quadrant , with depression quickly following the first or second impulse . ( B ) Whole-cell current clamp recording of an interneuron held at the reversal for chloride , −65 mV , and near threshold ~45 mV using positive somatic current injection . ( C ) Whole-cell voltage recordings of an interneuron clamped at −65 and 0 mV using QX-314 in the pipette solution to block action potential generation ( average of 10 sweeps ) . ( D ) Decomposition of the underlying excitatory ( Ge; red ) and inhibitory synaptic conductances ( Gi; blue ) arising from electrical microstimulation of the preferred quadrant . Drop lines indicate the onsets of excitation and inhibition . ( E–G ) Quantification of the maximal excitatory ( depolarization from −65 mV ) and maximal inhibitory amplitudes ( hyperpolarization from −20 mV ) in current clamp arising from selective activation of each of the four retinal quadrants ( dorsal , ventral , anterior and posterior ) using a train of 8 pulses at 10 Hz ( n = 10 ) . However , to avoid dependencies on the particular stimulated retinal area , we sorted the quadrants that give rise to the strongest excitatory component in descending order ( Q1 being the largest EPSP ) . Lines with matching colors on each side of the y-axis represent the postsynaptic potential amplitudes obtained from individual recorded interneurons ( positive values represent the total EPSP size and negative values represent the total IPSP size ) . We determined three subtypes of interneurons on the basis of their inhibitory response patterns . In E , neurons displayed inhibitory responses only when excitation was present , which usually was broad and spread into neighboring quadrants ( top: typical traces obtained from Q1 ) . In F , interneurons received weak synaptic inhibition ( top: typical build-up excitation obtained Q1 & Q2 ) . In G , interneurons received local excitation and widespread inhibition arising from quadrants that did not display excitation ( top: typical off-receptive inhibition arising from Q3 & Q4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16472 . 015 To identify the synaptic inputs that arise from the four retinal quadrants ( Q1 , Q2 , Q3 & Q4 ) , we measured the maximal amount of depolarization from −65 mV and hyperpolarization from −20 mV that was generated with a stimulation train of 8 pulses at 10 Hz . All interneurons that were recorded received excitation from usually one main quadrant and in some cases ( 6 out of 13 ) they also received from neighboring ones . Their inhibitory responses , however , displayed a range of variability by which we subdivided into three provisional groups: ( a ) local excitation and local inhibition ( Figure 9E ) , ( b ) local excitation and no inhibition ( Figure 9F ) and ( c ) local excitation and global inhibition ( Figure 9G ) . We sorted the responses that generated the largest PSPs by quadrant starting from Q1 through Q4 ( Q1 being the strongest ) . Despite the variable nature of visually-evoked synaptic inhibition onto tectal interneurons , the spatiotopic arrangement of their excitatory components is critical for the feedforward inhibition onto tectal output neurons . The tectal GABAergic system regulates the incoming excitatory sensory flow to deep layer output neurons and controls their response profile to stimuli of different modalities . It is responsible for the suppression of stimuli with spatial and temporal offsets . Recently , we have shown that on-receptive field retinal stimuli result in local excitation followed by feedforward inhibition , while off-field retinal stimulation yields only lateral inhibition , which is widespread ( Kardamakis et al . , 2015 ) . This inhibition is carried by short- and long-range connections spanning rostrocaudally and mediolaterally across the optic tectum ( Phongphanphanee et al . , 2014; Kardamakis et al . , 2015; Figure 8 ) . While this local inhibition can act independently for generating stimulus selection , it has also been suggested that exogenous sources are required for generating global inhibition via GABAergic neurons in the isthmic nucleus ( Mysore and Knudsen , 2013 ) . Here , we show that inhibition hard-wired in the optic tectum via interneurons , that are activated from both retinal and electroreceptive afferents in the same way as the output neurons , is essential for multisensory stimulus selection . Notably , the two types of sensory inputs not only evoke excitation in tectal interneurons , but also trigger recurrent inhibition . Our results suggest that spatiotopic organization in the optic tectum is columnar , with output neurons and interneurons receiving retinal excitation from the same quadrant , as suggested by their morphological features and the fact that the excitation is retinotopically arranged . However , the detailed arrangement of columnar structure and function has yet to be revealed . Sensory inputs provide monosynaptic drive to the deep layer output cells and to superficial layer interneurons . The latter , in turn , provide local and long-range inhibitory projections , therefore allowing for spatial sensory discrimination . This role of inhibition is observed in our extracellular recordings . Consistent with bimodal suppression in mammals ( Meredith and Stein , 1996; Kadunce et al . , 1997 ) , responses in on-field regions of tectum to electrosensory stimulation undergo a drastic reduction when delivering visual stimuli to other receptive fields of the tectal map . In the present study , we could not elicit the same sort of reduction in tectal areas responsive for visual stimulation when applying off-field electrosensory stimuli . One possibility is that electrosensory maps in the lamprey tectum are not as refined as the visual ones , and therefore inhibition is less effective suppressing off-regions . In this sense , it has been shown that electrosensory receptive fields in elasmobranchs tectum are much larger than the visual ones ( Bodznick , 1990 ) . We cannot exclude either that the way we perform electrosensory stimulation is not as local as the visual one , making it less effective to evoke a similar reduction . Consistent with our previous findings , these data suggest that inhibitory connections embedded within the tectal circuit can account for how normally effective stimuli can be attenuated by the presence of bimodal stimuli when they are spatial and/or temporally misaligned . Despite the powerful action of the GABAergic system , it does allow for response enhancement during bimodal integration . The output neurons yield responses of matching amplitudes to both types of sensory inputs , without any apparent domination by a particular modality . Two overlapping ( in space and time ) stimuli from these two modalities will produce a stronger response , than when only one modality is present . As a rule , two sensory modalities that provide stimuli spatially and temporally aligned are most likely related to the same external event . In this manner , the response enhancement caused by the integration of two senses gives rise to higher event detection reliability . When this occurs , bimodal inputs are able to increase the amount of depolarization in output neurons and increase the probability of evoking an action potential , and therefore , to activate their downstream targets and elicit a motor response . Tectal inhibition will continuously reset active areas to enable the constant monitoring of information flow from the environment . Without this inhibition , response enhancement will not be possible since tectal output cells rapidly saturate most likely due to a ceiling effect , making neurons incapable of further responding and thereby losing their response sensitivity , and hence , event detection reliability . Tectal inhibitory action can be interpreted as a gain modulatory mechanism whose effects are well captured by the conductance-based analysis of output neurons , and visible throughout a wide range of membrane potentials ( i . e . , varying from below to above spike threshold ) . Local circuit dynamics triggered by sensory inputs will allow for depolarizing synaptic events to reach the reversal potential of the synaptic response . We show that this value is invariant during unimodal or bimodal stimulation implying that a strong unimodal stimulus can bring the membrane potential to action potential threshold , as well as two weaker bimodal stimuli . When bimodal stimuli overlap , they will reinforce each other and generate larger depolarizing currents that will increase the likelihood that the output neurons reach spike threshold before inhibition fully develops . This enhancement most likely arises from sensory inputs that target the different dendritic compartments of output neurons ( i . e . , superficial vs intermediate layer ) , that were previously inactive during unimodal stimulation . Although our findings imply that tectum can perform stimulus selection without the external control from any other brain region , activity in the deep layer output neurons can be strongly influenced by forebrain structures such as the cortex and the basal ganglia . As in mammals , tectal output neurons in lamprey also receive the combination of inhibitory and excitatory input from the forebrain . The GABAergic output neurons of the basal ganglia output nuclei provide tonic inhibition at rest in both lamprey and primates , and can trigger movements by way of disinhibition of the tectal output neurons ( Wurtz and Hikosaka , 1986; Stephenson-Jones et al . , 2012; Kim and Hikosaka , 2015 ) . Pallium ( evolutionary forerunner of the mammalian cortex ) projects to the lamprey optic tectum and provides monosynaptic excitation to the output neurons in the deep layer , and its stimulation is able to elicit eye-head gaze movements ( Ocaña et al . , 2015 ) . In the forebrain control of tectum , it would seem likely that the basal ganglia disinhibition will be complementary to the excitation conveyed from pallium/cortex , and in both cases they target the soma level of the output neurons rather than the dendrites . Extracellular recordings in cats and primates have shown a supra-linear enhancement of deep layer collicular activity in response to overlapping weak multimodal stimuli ( Meredith and Stein , 1983; Wallace and Stein , 1994; Wallace et al . , 1996; Stein , 2005 ) . This amplification is thought not to arise from the intrinsic collicular circuitry , but from extrinsic cortico-collicular afferents from the anterior ectosylvian sulcus . When this area is inactivated the amplification vanishes ( Wallace and Stein , 1994; Jiang et al . , 2001; Alvarado et al . , 2009 ) . One mechanism would be to bypass tectal inhibition by selectively increasing excitation onto the output neurons , which are targeted by monosynaptic projections from cortex/pallium in both mammals and lamprey . Another potential mechanism that would introduce such non-linearities could be a selective decrease of inhibition by briefly ‘switching off’ the tectal GABAergic system . The variable classes of tectal interneurons may also provide an additional flexibility to manipulate sensorimotor processing within the optic tectum . The basic anatomical and functional organization of the optic tectum across vertebrates is highly conserved , from the earliest group of vertebrates that has evolved - the lamprey ( Nieuwenhuys and Nicholson , 1998; Saitoh et al . , 2007; Jones et al . , 2009; Asteriti et al . , 2015; Kardamakis et al . , 2015 ) , despite the particular sensory modalities that the different species depend on . The lamprey optic tectum consists of glutamatergic and GABAergic neurons activated by unimodal and bimodal inputs . The output neurons found in the deep layer are excitatory and project to the brainstem , and control gaze movements ( Saitoh et al . , 2007 ) . In lamprey , visual and electrosensory stimuli evoke direct excitation quickly followed by recruited inhibition onto these output neurons . These synaptic responses are amplified when spatiotemporally aligned bimodal inputs are present , thus , leading to a response enhancement . We have shown here that sensory-evoked inhibition scales the neural activity of output cells to allow for a robust spatiotemporal integration , capable of integrating the enormous amount of incoming sensory inputs during natural conditions . Local inhibition may be critical for spatiotemporal processing not only in lamprey but also in other vertebrates . Experiments were performed on 47 adult river lampreys ( Lampetra fluviatilis ) . During the investigation , every effort was made to minimize suffering and to reduce the number of animals used , in accordance with the Guide for the Care and Use of Institute of Laboratory Animal Research , National Research Council ( 1996 ) . After injecting the tracer Neurobiotin ( 20% , see above ) , brains were dissected and incubated with CLARITY monomer solution containing 1% acrylamide , 0 . 0125% bis-acrylamide , and 4% PFA and then polymerized at 37°C for 8 hr . The brains were passively cleared in SDS Borate Buffer ( pH 8 . 5 ) at 37°C for 2 weeks and then equilibrated in 0 . 2 M Borate Buffer ( pH 7 . 6 ) containing 0 . 1% Triton . The brains were incubated with Streptavidin Alexa Fluor 488 ( 1:100 ) in the same buffer at 37°C for 4 days . Sequential washing followed prior to equilibrating to a final concentration of 65% glycerol containing Anti-fade for imaging . Once mounted they were imaged using COLM methods described in Tomer et al . ( 2014 ) . The animals ( n = 3 ) were deeply anesthetized with tricaine methane sulfonate ( MS-222 ) ( 100 mg/L; Sigma ) diluted in fresh water . During the surgery and the injections , the entire animal was submerged in ice-cooled artificial cerebrospinal fluid ( aCSF ) solution containing the following ( in mM ) : 125 NaCl , 2 . 5 KCl , 2 CaCl2 , 1 MgCl2 , 10 glucose , and 25 NaHCO3 , saturated with 95% ( vol/vol ) O2/5% CO2 . An incision was performed directly above the octavolateral area to expose the brain . All injections were made with glass micropipettes ( borosilicate; o . d . = 1 . 5 mm , i . d . = 1 . 17 mm; Hilgenberg ) with a tip diameter of 10–20 μm . The micropipettes were fixed to a holder attached to an air supply and a Narishige micromanipulator . Fifty to 200 nL of Alexa Fluor 488-dextran ( 10 kDa; 12% ( wt/vol ) in saline; Molecular Probes ) were pressure injected unilaterally into the octavolateral area . Subsequently , an incision was performed in the primary spectacle , the lens was removed to expose the retina , and Neurobiotin [20% ( wt/vol ) in aCSF containing Fast Green to aid visualization of the injected tracer; Vector Laboratories] , was injected in the central retina , ipsilateral to the octavolateral area injection . Following injections , the dorsal skin and the spectacle were sutured , and the animal was returned to its aquarium for 48–72 hr to allow transport of the tracers . The brains were then dissected out and fixed in 4% formaldehyde and 14% saturated picric acid in 0 . 1 M phosphate buffer ( PB ) , pH 7 . 4 , for 12–24 h , and then cryoprotected in 20% ( wt/vol ) sucrose in PB for 3–12 hr . 20-μm-thick transverse sections were made using a cryostat , collected on gelatin-coated slides , and stored at −20°C until further processing . For detection of Neurobiotin , Cy2 conjugated streptavidin ( 1:1000; Jackson ImmunoResearch , PA ) and a deep red Nissl stain ( 1:1000; Molecular Probes ) were diluted in 1% bovine serum albumin ( BSA ) , 0 . 3% Triton X-100 in 0 . 1 M PB . All sections were mounted with glycerol containing 2 . 5% diazabicyclooctane ( Sigma-Aldrich ) . To label brainstem projecting neurons for patch-clamp experiments , tetramethylrhodamine-dextran ( 3 kDa; 12% in saline; Molecular Probes ) was pressure injected unilaterally into MRRN ( Middle Rhombencephalic Reticular Nucleus ) in the brainstem . Prelabeled brainstem-projecting cells and superficial layer interneurons were intracellularly injected with 0 . 3–0 . 5% Neurobiotin ( Vector Laboratories ) during patch-clamp recordings . Brain slices were fixed overnight in 4% formaldehyde and 14% picric acid in 0 . 1 M PB . Following a thorough rinse in PBS , the slices were incubated in streptavidin-Cy2 ( 1:1000 , Jackson ImmunoResearch ) in 0 . 3% Triton X-100 and 1% BSA in 0 . 1 M PB for two hours at room temperature . The slices were then rinsed in 0 . 01 M phosphate buffered saline ( PBS ) and mounted in glycerol containing 2 . 5% diazabicyclooctane ( DABCO; Sigma ) . Labeled cells were analyzed by either confocal or conventional fluorescence microscopy . Local field potentials ( LFPs ) were recorded using tungsten microelectrodes ( ~1–5 MΩ , a 4-channel MA 102 amplifier and a MA 103 preamplifier ( Elektroniklabor , Zoologie , University of Cologne ) , and digitized at 20 kHz using pClamp ( version 9 . 2 ) software . For natural sensory stimulation , the head was transected from animals deeply anesthetized with MS-222 ( 100 mg/L; Sigma ) and the dorsal skin and cartilage were removed to expose the brain . The skin with the electrosensory receptive organs in the rostrolateral part of the head was kept , and the muscles in the ventral part were removed . To avoid movements that could destabilize the preparation , the neuromuscular blocker α-bungarotoxin ( 12 . 5 μM , Sigma ) was locally injected into the muscles that could not be removed . Visual stimulation was performed with brief flashes of light ( 500 ms duration ) , using a 100 μm diameter-thick optic fiber connected to a standard LED light source . The optic fiber was connected to a borosilicate glass pipette painted black with nail polisher to avoid light spread , so that the light spot diameter was ~50 μm . The pipette was attached to a Narishige micromanipulator , and placed ~2 cm distance from the retina , and once the receptive field in the optic tectum ( OT ) was located by recording LFPs , an electric field was spatially aligned with the visual stimulus . To generate it , two copper wires ( used respectively as negative and positive poles ) were connected to a stimulus isolation unit ( MI401; Zoological Institute , University of Cologne ) , and submerged in the aCSF at a distance of ~5–10 cm from the preparation , keeping a small distance between poles to ensure a local stimulus ( ~1 cm ) . Electrosensory stimulation was presented as brief pulses ( 30 ms in duration ) with intensities between 10–100 μA . For electric stimulation of the retina and the anterior line nerve ( ALLN ) , the procedure was the same , but the cartilage of the otic capsule was removed in order to expose the ALLN . The stimulation was performed by using borosilicate glass microcapillaries connected to a stimulus isolation unit ( MI401; Zoological Institute , University of Cologne ) . The stimulation intensity was set to the threshold strength ( typically 10–100 μA ) necessary to evoke LFPs with a 50 μs duration stimulus , and consecutive increasing stimulation durations of 100 , 200 , 300 , 400 , 500 , 700 and 1000 μs were then applied . All experiments were performed in darkness to avoid interfering visual stimuli . A novel preparation was used to perform whole-cell current-clamp recordings , slicing a thick brain section exposing the deep layer of the optic tectum , keeping the retinal and octavolateral afferent tracts intact ( see Figure 2A ) . To expose the different layers of the optic tectum by sectioning , the entire preparation was first embedded in agar ( 4% dissolved in aCSF; Fluka ) . The agar block containing the brain was then cut in an oblique angle and glued to a metal plate , quickly transferred to ice-cold aCSF and sagittal-oblique slices were cut using a vibrating microtome ( Microm HM 650V; Thermo Scientific ) until the deep layer of tectum as well as the retinal and octavolateral tracts were exposed . The agar block was then mounted in a submerged recording chamber . Whole-cell voltage and current-clamp recordings were performed with patch pipettes made from borosilicate glass ( Hilgenberg ) using a vertical puller ( Model PP-830; Narishige ) . The resistance of recording pipettes was 7–10 MΩ when filled with intracellular solution of the following composition ( in mM ) : 130 potassium gluconate , 5 KCl , 10 phosphocreatine disodium salt , 10 HEPES , 4 Mg-ATP , 0 . 3 Na-GTP; ( osmolarity 265–275 mOsmol ) . The electrode solution also included in some cases 3 mM of triethylammonium bromide ( QX-314; Sigma ) to block action potentials . Bridge balance and pipette-capacitance compensation were adjusted for using a MultiClamp 700B patch amplifier and Digidata 1322 analog-to-digital converter under software control ‘PClamp’ ( Molecular Devices ) . Perfusion of the preparation was performed with aCSF at 6–8° . Stimulation of the octavolateral and retinal afferents was performed with borosilicate glass microcapillaries ( the same as for patch recordings ) , connected to a stimulus isolation unit ( MI401; Zoological Institute , University of Cologne ) . The stimulation intensity was set to one to two times the threshold strength ( typically 10–100 μA ) to evoke PSPs . To investigate the short-term dynamics of synaptic transmission , a stimulus train of ten pulses at 10 Hz was used ( Ericsson et al . , 2013 ) . To examine the effect on the bimodal integration of the temporal overlap between visual and electrosensory modalities , we performed patch-clamp recordings using asynchronous visual and electrosensory stimulations , delaying one stimulus respect to the other in steps of 5 , 10 , 20 , 30 , 40 and 50 ms . To temporally align visual and electrosensory responses , the EPSPs evoked by the first pulse of each sensory modality were analyzed online for every single cell to calculate the difference between the onsets . This difference was then used to adjust the timing for the different stimuli steps , which were programmed using a Master-8 programmable pulse generator ( AMP Instruments LTD ) . During extracellular recordings , the GABAA-receptor antagonist ( Gabazine; 10 μM; Tocris ) was locally applied in the deep layer of the optic tectum by pressure injection through a micropipette fixed to a holder , which was attached to an air supply and a Narishige micromanipulator . For patch-clamp recordings , gabazine was bath applied . For immunohistochemical detection of GABA , sections were incubated overnight with a mouse monoclonal anti-GABA antibody ( 1:5000; mAb 3A12; kindly donated by Dr . Peter Streit , Brain Research Institute , University of Zürich , Zürich , Switzerland ) . The sections were subsequently incubated for 2 hr at room temperature with a Cy3-conjugated donkey anti-mouse IgG ( 1:500; Jackson ImmunoResearch ) and , in those cases in which the cells were filled with neurobiotin , also with Cy2-conjugated streptavidin ( 1:1000; Jackson ImmunoResearch ) . All primary and secondary antibodies were diluted in 1% BSA and 0 . 3% Triton-X 100 in 0 . 1 M PB . Photomicrographs were taken with an Olympus XM10 digital camera mounted on an Olympus BX51 fluorescence microscope ( Olympus Sweden ) . Illustrations were prepared in Adobe Illustrator and Adobe Photoshop CS4 . Images were only adjusted for brightness and contrast . Confocal Z-stacks of optical sections were obtained using a Zeiss Laser scanning microscope 510 , and the projection images were processed using the Zeiss LSM software , ImageJ and Adobe Photoshop CS4 . For all electrophysiological recordings , data analysis was performed using custom written functions in Matlab ( see Source code 1 ) . For extracellular recordings , the integral under the curves were compared after fully rectifying the signals using trapezoidal numerical integration ( ‘trapz’ function ) . For patch-clamp recordings , subsequent PSPs often started on the decay phase of previous responses , so that to extract correct amplitudes the synaptic decay was either fitted by an exponential curve and subtracted or manually subtracted . For the estimation of the synaptic conductances , we used the direct extraction of the excitatory and inhibitory conductance based on solving the conductance model equation applied to the current-clamp data as explained in p . 328–329 in Monier et al . ( 2008 ) . In all recordings , these estimates were accurate because ( i ) the synaptic responses were composed primarily of ionotropic glutamatergic synapses ( primarily AMPA , see also Kardamakis et al . , 2015 ) and chloride-mediated GABAergic ( GABAA ) synaptic inputs , ( ii ) their reversal potentials were experimentally determined and found to be at 0 and −75 mV , respectively , and ( iii ) due to the linearity domain of their voltage-current relationships and similar time constants . For statistical analysis , we used two-sample unpaired and paired t-tests in Matlab . Throughout the figures , sample statistics are expressed as Means ± SEMs ( SEM; standard error ) , unless specified otherwise .	Many events occur around us simultaneously , which we detect through our senses . A critical task is to decide which of these events is the most important to look at in a given moment of time . This problem is solved by an ancient area of the brain called the optic tectum ( known as the superior colliculus in mammals ) . The different senses are represented as superimposed maps in the optic tectum . Events that occur in different locations activate different areas of the map . Neurons in the optic tectum combine the responses from different senses to direct the animal’s attention and increase how reliably important events are detected . If an event is simultaneously registered by two senses , then certain neurons in the optic tectum will enhance their activity . By contrast , if two senses provide conflicting information about how different events progress , then these same neurons will be silenced . While this phenomenon of ‘multisensory integration’ is well described , little is known about how the optic tectum performs this integration . Kardamakis , Pérez-Fernández and Grillner have now studied multisensory integration in fish called lampreys , which belong to the oldest group of backboned animals . These fish can navigate using electroreception – the ability to detect electrical signals from the environment . Experiments that examined the connections between neurons in the optic tectum and monitored their activity revealed a neural circuit that consists of two types of neurons: inhibitory interneurons , and projecting neurons that connect the optic tectum to different motor centers in the brainstem . The circuit contains neurons that can receive inputs from both vision and electroreception when these senses are both activated from the same point in space . Incoming signals from the two senses activate the areas on the sensory maps that correspond to the location where the event occurred . This triggers the activity of the interneurons , which immediately send ‘stop’ signals . Thus , while an area of the sensory map and its output neurons are activated , the surrounding areas of the tectum are inhibited . Overall , the findings presented by Kardamakis , Pérez-Fernández and Grillner suggest that the optic tectum can direct attention to a particular event without requiring input from other brain areas . This ability has most likely been preserved throughout evolution . Future studies will aim to determine how the commands generated by the optic tectum circuit are translated into movements .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Spatiotemporal interplay between multisensory excitation and recruited inhibition in the lamprey optic tectum
The bacterial plasma membrane is an important cellular compartment . In recent years it has become obvious that protein complexes and lipids are not uniformly distributed within membranes . Current hypotheses suggest that flotillin proteins are required for the formation of complexes of membrane proteins including cell-wall synthetic proteins . We show here that bacterial flotillins are important factors for membrane fluidity homeostasis . Loss of flotillins leads to a decrease in membrane fluidity that in turn leads to alterations in MreB dynamics and , as a consequence , in peptidoglycan synthesis . These alterations are reverted when membrane fluidity is restored by a chemical fluidizer . In vitro , the addition of a flotillin increases membrane fluidity of liposomes . Our data support a model in which flotillins are required for direct control of membrane fluidity rather than for the formation of protein complexes via direct protein-protein interactions . The shape of a bacterium is predominantly defined by the structure of its peptidoglycan . Although there is a great variety in bacterial shapes , the overall chemistry of peptidoglycan is very similar between bacteria and thus the shape of peptidoglycan is primarily determined by the temporal and spatial regulation of peptidoglycan synthesis . In rod-shaped bacteria , peptidoglycan synthesis is thought to be mediated by two protein assemblies , the elongasome and the divisome , that synthesise peptidoglycan along the long axis and across the division plane of the cell , respectively ( Typas et al . , 2012; Zhao et al . , 2017 ) . These complexes contain a set of proteins required for the final steps of synthesis and translocation of the peptidoglycan precursor , LipidII , from the inner to the outer leaflet of the cytoplasmic membrane , and proteins that incorporate LipidII into peptidoglycan . These include SEDS ( Shape , Elongation , Division and Sporulation ) proteins that can perform glycosyl transferase reactions ( Cho et al . , 2016; Meeske et al . , 2016; Taguchi et al . , 2019 ) , and Penicillin Binding Proteins ( PBPs ) that are divided in class A PBPs ( aPBPs ) that catalyse both glycosyl transferase and transpeptidase reactions , class B PBPs ( bPBPs ) that only catalyse transpeptidase reactions and low molecular weight PBPs that modify peptidoglycan , as well as hydrolases ( Zhao et al . , 2017; Morales Angeles and Scheffers , 2017 ) . Coordination of these complexes is linked to cytoskeletal elements , MreB ( -like proteins ) for the elongasome and FtsZ for the divisome . In models , the cytoplasmic membrane is often depicted as a passive environment in which these machineries are embedded . However , it is becoming clear that the structure of the membrane plays a critical role in the coordination of peptidoglycan synthesis ( Strahl and Errington , 2017 ) . Inward membrane curvature serves as a localisation trigger for MreB and the elongasome , and enhanced local synthesis at bulges straightens out the membrane sufficient to convert spherical cells to a rod shape ( Hussain et al . , 2018; Ursell et al . , 2014 ) . In Bacillus subtilis , the motion of MreB along the membrane is associated with elongasome activity ( Domínguez-Escobar et al . , 2011; Garner et al . , 2011 ) , and the velocity of MreB patches is related to growth rate ( Billaudeau et al . , 2017 ) , indicating that MreB motion can be used as a marker for elongasome activity . Interestingly , MreB localises to and organises regions of increased membrane fluidity ( RIF ) ( Strahl et al . , 2014 ) , which in turn is linked to the presence of LipidII , which favours a more fluid membrane and promotes local membrane disorder ( Ganchev et al . , 2006; Witzke et al . , 2016 ) . Inhibition of LipidII synthesis by genetic or chemical means results in a dissolution of membrane structures observed with the dye FM 4–64 and release of MreB from the membrane ( Domínguez-Escobar et al . , 2011; Garner et al . , 2011; Muchová et al . , 2011; Schirner et al . , 2015 ) . Next to RIFs , membrane regions of decreased fluidity have been identified in bacteria ( Strahl and Errington , 2017; Bramkamp and Lopez , 2015; Lopez and Koch , 2017 ) . These so-called functional membrane microdomains ( FMMs ) are thought to be organised by the bacterial flotillin proteins , are enriched in isoprenoid lipids ( García-Fernández et al . , 2017; López and Kolter , 2010 ) , and can be found in so-called Detergent Resistant Membrane ( DRM ) fractions of the membrane . Since the formulation of the FMM hypothesis , FMMs have been linked to many processes , such as protein secretion , biofilm formation , competence and cell morphology ( Mielich-Süss and Lopez , 2015; Mielich-Süss et al . , 2013; Bach and Bramkamp , 2013; Dempwolff et al . , 2012 ) . Cell morphology defects are linked to cell wall synthesis , and analysis of the protein content of Bacillus subtilis DRMs identified several PBPs , MreC and other proteins involved in cell wall metabolism as well as the two flotillins , FloA and FloT ( López and Kolter , 2010; Bach and Bramkamp , 2013; Yepes et al . , 2012 ) . FloA is constitutively expressed , whereas FloT is expressed primarily during stationary growth , cell wall stress and sporulation ( Schneider et al . , 2015a; Huang et al . , 1999; Nicolas et al . , 2012 ) . Super resolution microscopy showed that the flotillins and other proteins found in DRMs do not colocalise and have different dynamics ( Dempwolff et al . , 2016 ) , so it is unlikely that FMMs are regions in the membrane that offer a favourable environment in which these membrane proteins are continuously present and active . Recently , the hypothesis has been put forward that FMMs/flotillins form a platform for the formation of functional protein oligomers , as work in Staphylococcus aureus showed that multimerisation of Type 7 secretion systems and PBP2a depends on FMMs ( Lopez and Koch , 2017; García-Fernández et al . , 2017; Mielich-Süss et al . , 2017 ) . Here , we have analysed the role of flotillins in peptidoglycan synthesis in B . subtilis . Our results show that , at high growth rates , flotillins control membrane fluidity in a manner that is critical for peptidoglycan synthesis and MreB dynamics , but have no effect on PBP oligomerisation . This results in a new model for flotillin function in the physical organisation of membranes during fast growth . In previous studies , a double deletion of floA/floT was either reported to suffer severe shape defects and perturbed membrane structure ( Dempwolff et al . , 2012 ) , or to not have strong shape defects but with a change in the overall lipid ordering of the membrane ( Bach and Bramkamp , 2013 ) . We grew wild type and ΔfloAT strains and analysed exponentially growing cells . We did not observe striking shape defects but did see an increase in median cell length and distribution of cell lengths in the absence of flotillins ( Figure 1A , G ) . To look at effects on peptidoglycan synthesis , we labelled cells with HADA , a fluorescent D-Alanine analogue that reports on sites of active peptidoglycan synthesis ( Kuru et al . , 2012 ) , and with fluorescent vancomycin ( Van-FL ) , which labels LipidII and peptidoglycan containing pentapeptide side chains ( Daniel and Errington , 2003; Morales Angeles et al . , 2017 ) . This revealed a significant accumulation of peptidoglycan synthesis stains at division septa in the ΔfloAT strain ( Figure 1A–C ) . To look at membrane structure , cells were labelled with FM4-64 , Nile-Red and DiI-C12 , which are lipid dyes that accumulate in zones enriched in fluid lipids ( Strahl et al . , 2014 ) . Again , the stains accumulated at the septa in the ΔfloAT strain , which also showed some accumulation of FM4-64 and DiI-C12 in patches , suggesting that the more fluid regions of the membrane are coalescing into larger regions ( Figure 1A , D–F ) . The HADA , FM4-64 and Nile-Red measurements were repeated using a wild type strain expressing endogenous GFP , allowing simultaneous imaging of both strains on the same slide , and gave similar results , confirming that the observed signal increase is not due to variation between microcopy experiments ( Figure 1—figure supplement 1A , B ) . In this mixed-strain experiment , Nile-Red labelling at the lateral membrane was the same between wild type and ΔfloAT strains , indicating that there is no difference in dye diffusion between the strains ( Figure 1—figure supplement 1D ) . Inspection of the septa by electron microscopy revealed that there was no difference between the thicknesses of the septa between the wild type and ΔfloAT strain , ruling out that the increase in signal was due to formation of thicker septa ( Figure 1—figure supplement 1C ) . The shift of peptidoglycan synthesis to the division site could hint at stress in the overall peptidoglycan synthesis route . This was confirmed by growing cells at a sublethal concentration of fosfomycin , which limits synthesis of LipidII ( Kahan et al . , 1974 ) , but that does not impact growth rate at the concentration used . This resulted in bulging cells and some lysis , which was exacerbated in the ΔfloAT strain ( Figure 1—figure supplement 1F , G ) . It should be noted that the peptidoglycan synthesis stress caused by fosfomycin is not the same as the stress caused by the absence of flotillins , as the phenotypes of wild type cells with sublethal fosfomycin are quite distinct from ΔfloAT cells without fosfomycin . We ruled out that the peptidoglycan synthesis stress was caused by a change in the folding or complex formation by PBPs in the absence of flotillins , as there were no differences in the overall PBP-profiles of Bocillin-FL labelled wild type or flotillin deletion strains ( Figure 1—figure supplement 2A ) . PBP complex formation was analysed using a combination of Native-PAGE and SDS-PAGE with Bocillin-labelled membrane fractions ( Trip and Scheffers , 2016 ) and showed that various PBPs can be found in a high-MW complex ( notably PBPs 1 , 2 , 3 and 4 ) , but that complex formation is similar in the ΔfloAT strain ( Figure 1—figure supplement 2B ) . Also , none of the five functional GFP-PBPs examined changed their localisation in the ΔfloAT strain ( Figure 1—figure supplement 2C ) . Overall , the data suggest that in the absence of flotillins , peptidoglycan synthesis is affected and relatively increased at division septa , with a concomitant accumulation of membrane dyes that are indicative of higher membrane fluidity . We reasoned that a non-lethal defect in septal peptidoglycan synthesis could reveal more about the role of flotillins and constructed a flotillin mutant that lacks PBP1 , a bifunctional glycosyl transferase/transpeptidase that is required for efficient cell division ( Scheffers and Errington , 2004 ) . Simultaneous deletion of pbp1 , floA , and floT resulted in strong filamentation and delocalisation of peptidoglycan synthesis as well as membrane dyes to patches ( Figure 2A , Figure 2—figure supplement 1A ) . Deletion of single flotillin genes and PBP1 had similar , albeit less severe effects ( Figure 2—figure supplement 1B , C ) . To exclude the possibility that an alteration of peptidoglycan modification resulted in the delocalisation of HADA and Van-FL , we used D-Alanine-D-Propargylglycine ( D-Ala-D-Pra ) , a clickable dipeptide analogue which is exclusively incorporated into peptidoglycan via LipidII ( Sarkar et al . , 2016 ) . D-Ala-D-Pra incorporation was delocalised in the ∆pbp1ΔfloAT strain , indicating that peptidoglycan synthesis itself is delocalised ( Figure 2—figure supplement 1D ) . So far , our experiments were done with fast growing cells and Lysogeny Broth ( LB ) as the growth medium . Strikingly , none of the mutant strains had an apparent phenotype when cultivated in Spizizen’s minimal medium ( SMM , Figure 2B ) , and peptidoglycan synthesis and lipid dyes were no longer accumulating at division sites in the ΔfloAT strain ( Figure 2—figure supplement 2 ) . SMM has a higher Mg2+ concentration , which is known to rescue various cell shape mutations by inhibition of cell wall hydrolysis ( Dajkovic et al . , 2017 ) . However , the increase in Mg2+ was not sufficient to explain the reversal of phenotype as cells grown on LB supplemented with Mg2+ ( 6 mM , concentration in SMM , or 20 mM ) still displayed the elongated phenotype with delocalised peptidoglycan synthesis ( Figure 2—figure supplement 3 ) . This indicated that the phenotypes associated with the absence of flotillins are growth-rate and/or nutrient related . Next , we determined lipid packing order in the different strains using the fluorescent dye Laurdan , a reporter for flotillin-mediated lipid ordering ( Bach and Bramkamp , 2013 ) . LB-grown cells lacking flotillins displayed an increased generalised polarisation ( GP ) ( Bach and Bramkamp , 2013 ) , indicative of an overall increase in ordered lipid packing in the membrane , but the effect of flotillins on membrane ordering completely disappeared when cells were grown on SMM ( Figure 3 ) . The resolution obtained with Laurdan does not allow the detection of local differences in fluidity between the lateral membrane and the septa , but does report on overall lipid ordering . Overall , lipid order was increased in cells grown on SMM compared to LB ( Figure 3 ) , whereas the absence of PBP1 had no significant effect on membrane fluidity , also not when combined with flotillin deletions ( Figure 3 ) . The changes in lipid ordering were not due to changes in the overall fatty acid composition of the membranes - the ratios of C17/C15 side chains and iso/anteiso fatty acids , which are indicative of fluidity ( Strahl et al . , 2014 ) , were identical for wild type and ΔfloAT strains grown on LB , and very similar for cells grown on SMM ( Figure 3—figure supplement 1 ) . The GP values indicated that membranes are more ordered when cells are grown on minimal medium , and this suggests that the flotillin-associated increase in overall membrane fluidity is important for cell shape control at high growth rates . This was tested by growing the strains lacking flotillins and PBP1 on LB in the presence of benzyl alcohol , an extensively used membrane fluidiser that increases membrane hydration due to disordering of membrane structure ( Konopásek et al . , 2000 ) . Notably , the addition of benzyl alcohol increased membrane fluidity to similar extents in the wildtype and the mutant strains ( see Figure 3C ) , but did not affect the growth rates of the strains ( Figure 2—figure supplement 4 ) . The increase in membrane fluidity restored normal cell length and normal peptidoglycan synthesis patterns to the pbp1/floA/floT strain ( Figure 2C ) . In B . subtilis , the rate of growth and of peptidoglycan synthesis is linked to the speed of MreB movement – in minimal media , the speed of MreB patches is reduced compared to the speed in rich media ( Billaudeau et al . , 2017 ) . Analysis of the movement of a fully functional mRFPruby-MreB fusion ( Domínguez-Escobar et al . , 2011 ) by time lapse TIRF ( Total Internal Reflection Fluorescence ) microscopy , confirmed that MreB patch mobility is higher in cells grown on LB than in cells grown on SMM , with MreB speeds similar to those reported previously ( Billaudeau et al . , 2017; Figure 4 , Figure 4—videos 1 and 2 ) . Strikingly , in the absence of flotillins , MreB patch mobility was notably decreased in cells grown on LB , while in SMM grown cells MreB patch mobility was independent of the presence of flotillins ( Figure 4 , Figure 4—videos 3 and 4 ) . Fluidising the membrane with benzyl alcohol , which does not alter the growth rate , almost completely restored MreB mobility in LB grown cells ( Figure 4 , Figure 4—videos 5 and 6 ) . These results indicate that the MreB patch mobility is not only controlled by growth rate , but also by membrane fluidity . Thus , in fast growing cells with decreased membrane fluidity there is a decrease in elongasome mediated peptidoglycan synthesis , reflected by the reduction of MreB mobility . This fits with an observed increase in peptidoglycan synthesis at the division site which may act as a compensatory mechanism . To assess whether the influence of flotillins on membrane fluidity is direct , we determined the membrane fluidity of model membranes with purified flotillin using solid-state NMR ( ssNMR ) . 2H ssNMR is a biophysical tool that assesses lipid mobility in native-like model membranes on the atomic level , by monitoring the carbon-deuterium order parameter of a deuterated lipid along the acyl chain ( here POPC-d31 ) ( Molugu et al . , 2017; Legrand et al . , 2019 ) . We purified B . subtilis FloT and tested the impact of FloT on the membrane , when reconstituted in POPC-d31 liposomes ( Schematically depicted in Figure 5A ) . FloT decreases the spectral width of the 2H quadrupolar splitting , reflecting an increase in motion on the atomic scale ( Figure 5B ) . The 2H spectrum encodes the local order parameter SCD of the carbon-deuterium in absence and in presence of FloT . Strikingly , FloT has an important impact on the order parameter along the entire acyl chain . It is remarkable that the protein significantly decreases the order parameter SCD , reaching even the inner carbon atoms of the acyl chain , indicating a different packing behaviour and increased membrane fluidity upon interaction with FloT ( Figure 5B ) . The strong fluidising effect described for FloT is notably different from the effects other proteins have when reconstituted into liposomes , such as plant remorins ( Legrand et al . , 2019 ) or the membrane binding peptide of the nonreceptor tyrosine kinase Src ( Scheidt and Huster , 2009 ) . The anisotropic lineshape of the 31P spectra indicates that the membrane is in the lamellar phase as expected for POPC at the chosen temperature ( 298K ) ( Huster , 2014 ) . Upon interactions with FloT the lamellar phase remains intact with formation of a few smaller objects , indicating that the overall liposome structure is not affected and that its phase is maintained ( Figure 5—figure supplement 1 ) . Our data provide evidence that flotillins play a direct role in controlling membrane fluidity and that membrane fluidity is critical for peptidoglycan synthesis at certain growth conditions . In vitro , flotillins enhance the fluidity of a model membrane , and in vivo , the membranes of fast growing flotillin-mutant cells are less fluid even though the fatty acid composition in these cells is identical . Therefore , we propose that the effect of flotillins on membrane fluidity is direct , through a change in the packing behaviour of the lipids resulting in an efficient separation of states of liquid ordered and disordered lipid domains in the membrane bilayer ( Bach and Bramkamp , 2013 ) . We found that membrane fluidity is not solely a function of temperature , but also of growth conditions . In vivo , flotillins may also recruit specific , more rigid lipids , such as hopanoids and carotenoids ( Bramkamp and Lopez , 2015; García-Fernández et al . , 2017; López and Kolter , 2010 ) which have been found in association with FMMs , and whose synthesis could be growth condition dependent . The predominantly physical role in membrane organisation for flotillins fits with our observation that adding a chemical fluidiser is sufficient to restore MreB dynamics and cell shape to fast growing cells that lack flotillins . We propose that in fast-growing cells on rich medium , flotillin-mediated control of membrane fluidity is critical and sufficient to allow essential membrane bound processes , such as peptidoglycan synthesis , to proceed normally . A sufficiently fluid membrane is necessary for the efficient recruitment and movement of MreB , and provides a more favourable environment for the peptidoglycan precursor LipidII ( Hussain et al . , 2018; Ursell et al . , 2014; Schirner et al . , 2015 ) . It has recently been shown that modulation of either MreBCD or PBP1 levels is sufficient to alter the shape of B . subtilis cells ( Dion et al . , 2019 ) , underscoring the importance of both systems . In the absence of flotillins , the activity of the MreBCD component is strongly reduced – as evidenced by the reduction of MreB speed – and the overall rigidity of the membrane is increased . This results in a less favourable environment for the peptidoglycan precursor LipidII , which prefers more liquid , disordered membrane phases ( Ganchev et al . , 2006; Witzke et al . , 2016; Calvez et al . , 2019 ) . Our data indicate that the reduction in elongasome activity , which does not impact the growth rate itself , is compensated by increased peptidoglycan synthesis activity around division sites in flotillin mutants , which is sufficient to keep the overall cell shape intact , although cells are elongated . The accumulation of lipid dyes indicative of increased fluidity at division sites is in line with a recent study that showed phases of different fluidity in Streptococcus pneumoniae membranes , with more fluid membranes and LipidII localising at midcell ( Calvez et al . , 2019 ) where the membrane is most bent . Our findings are also in agreement with the recent observation that B . subtilis cells elongate and lose organisation of MreB when membrane fluidity is decreased by altering the membrane fatty acid composition ( Gohrbandt , 2019 ) ( H . Strahl , personal communication ) . It could very well be that the shift of fluidity towards the septum is only relative as the overall fluidity of the membrane is decreased in the absence of flotillins . This is yet to be determined , as the resolution of Laurdan imaging does not allow conclusive statements about local fluidity changes at the septum . The observation that reduced MreB mobility and therefore altered lateral cell wall synthesis lead to accumulation of Van-FL and HADA staining at the septum is not immediately conclusive . Since septal PG synthesis is MreB independent in B . subtilis , a direct effect of MreB seems unlikely . Rather , a reduction of overall membrane fluidity in a flotillin knock-out might impair LipidII dynamics within the membrane . MurG is the enzyme that catalyses the final step of LipidII synthesis . There are several reports that MurG localises to the septum in different organisms ( Aaron et al . , 2007; Mohammadi et al . , 2007 ) . Thus , it seems likely that the septum is a place of increased LipidII synthesis and a change in membrane fluidity would create problems for LipidII molecules to diffuse away from their insertion site , resulting in reduced lateral PG synthesis and MreB mobility . Alternatively , the reduced MreB mobility and reduced lateral PG synthesis lead to a reduced LipidII consumption at the lateral wall , and the excess LipidII is used by the septal PG synthesis machinery , thereby leading to an increase in midcell PG . It remains to be tested which of these possibilities is responsible for the observed phenotype . Nevertheless , in both cases the increase in cell wall staining at the septum would be indicative of a higher local synthesis activity . It may seem paradoxical that cells elongate when elongasome activity is reduced , but one has to remember that a large amount of the peptidoglycan synthesis contributing to elongation of bacterial cells is actually taking place at midcell , before the ingrowth of the septum ( Aaron et al . , 2007; Pazos et al . , 2018; Varma and Young , 2009 ) . A relative increase in peptidoglycan synthesis at future division sites makes the activity of PBP1 critical and explains why its deletion has such a dramatic effect in cells lacking flotillins . Restoring fluidity using a chemical fluidiser allows the MreBCD component to again efficiently drive peptidoglycan synthesis during elongation , which is sufficient to suppress the flotillin mutant phenotype . The net effect of this is that cells lacking PBP1 and flotillins grown with benzyl alcohol behave as cells that only lack PBP1 , which is quite similar to wild type . At low growth rates , there is no difference between wild type cells and cells lacking flotillins with respect to membrane fluidity , and the speed of MreB is similar between the two cell types . Thus the deletion of flotillins does not exacerbate the phenotype of cells lacking PBP1 . The reason for the change in membrane fluidity between cells grown on rich or minimal medium is not yet clear – it does not seem to be caused by a large shift in the fatty acid composition of the membranes . Various factors could play a role , such as the synthesis of specific lipids ( hopanoids , isoprenoids ) on either type of medium , but also protein crowding , which is higher in membranes of fast-growing cells than in slow-growing cells ( Szenk et al . , 2017 ) . It will be an important future challenge to establish the cause for this difference . An overall rigidification of the membrane may also lead to retardation of processes which require membrane modifications such as division and sporulation , which is indeed observed in B . subtilis flotillin mutants ( Dempwolff et al . , 2012; Donovan and Bramkamp , 2009 ) . One of the proposed roles for flotillin proteins is that they form a ‘platform’ that transiently interacts with membrane proteins that need to oligomerise into functional complexes ( Lopez and Koch , 2017 ) . We tested this hypothesis for B . subtilis PBPs by comparing their localisation and oligomerisation in wild type and flotillin mutant strains . Although we were capable of detecting a high MW complex containing various PBPs ( notably PBP1 , 2a , 2b , 3 and 4 ) , the complex was not dependent on the presence of flotillins . We also note that PBP1 was present in the complex , as well as present in a large smear in the first dimension native gel , which would explain why PBP1 was detected by mass spectrometry analysis of a native PAGE band containing FloA ( Schneider et al . , 2015a ) . PBP5 , on the other hand , was not part of the high MW complex , which fits with its role in processing of the terminal D-Ala from stem-peptides that have not been cross-linked , which it exercises over the entire surface of the cell ( Kuru et al . , 2012 ) . Although we cannot exclude that flotillins may affect PBPs that are not easily detected by Bocillin-FL , our results do not provide any evidence for a role for flotillins in the oligomerisation of PBPs in B . subtilis . This extends the finding of the Graumann lab that found either transient or no colocalisation between flotillins and other proteins present in DRM fractions ( Dempwolff et al . , 2016 ) . Although it is obvious that peptidoglycan synthesis is altered in the absence of flotillins , our data strongly suggest that the basis for this alteration is in the physical organisation of the membrane rather than inefficient formation of divisome or elongasome complexes in the absence of flotillins , because flotillin mutants strains show no synthetic phenotype on minimal medium , and the defects on rich medium can be reverted by chemically fluidising the membrane . In conclusion , our data provide a new model for flotillin function in the physical organisation of membranes during fast growth . The observation that flotillins differentially affect the membrane in different growth conditions also explains the diversity of phenotypes described for flotillin mutants in the literature . All B . subtilis strains used in this study are derived from strain 168 and are listed in the Key resources table . Construction of new strains was based on natural competence of B . subtilis ( Harwood and Cutting , 1990 ) . Gene integration or deletion was validated by colony PCR whereas the expression and localisation of the fluorescent fusions was additionally validated by microscopy . Cells were grown either in LB Lennox ( 5 g/L yeast extract; 5 g/L NaCl; 10 g/L tryptone ) ( Lennox , 1955 ) or Spizizen minimal medium ( SMM ) ( Anagnostopoulos and Spizizen , 1961 ) supplemented with 1% glucose , at 37°C and 200 rpm , unless indicated otherwise . Induction of the Pxyl promoter was triggered by addition of 0 . 2–0 . 5% xylose . Cell cultures were supplemented with spectinomycin ( 50 µg/ml ) , tetracycline ( 10 µg/ml ) , chloramphenicol ( 5 µg/ml ) , kanamycin ( 5 µg/ml ) , erythromycin ( 1 µg/ml ) , benzyl alcohol ( BnOH , 0 . 1% ) or magnesium sulphate ( MgSO4 , 6–20 mM ) when necessary . Growth experiments were performed either manually or automatically with a PowerWave 340 microplate reader ( BioTek Instruments , U . S . A ) . Strains were pre-cultured overnight in 3 ml LB or SMM medium at 37°C with shaking at 200 rpm . Next , stationary or late-exponentially cells were diluted with fresh LB or SMM medium ( supplemented when necessary ) , and cell densities ( OD600 ) were measured every 1 hr when monitored manually or every 10 min when monitored automatically , for a total time of 7–22 hr . For standard fluorescence microscopy , exponentially growing cells were immobilised on microscope slides covered with a thin film of 1% agarose ( w/v ) in water or the appropriate medium . For TIRFM , agarose pads were mounted using Gene Frames ( 1 . 7 × 2 . 8 cm chamber , 0 . 25 mm thickness , 125 µL volume ) from ThermoScientific . Standard fluorescence microscopy was carried out using an Axio Zeiss Imager M1 fluorescence microscope ( EC Plan-Neofluar 100x/1 . 30 Oil Ph3 objective ) equipped with an AxioCam HRm camera and an Nikon-Ti-E microscope ( Nikon Instruments , Tokyo , Japan ) equipped with Hamamatsu Orca Flash 4 . 0 camera . For Laurdan and TIRFM experiments , a Delta Vision Elite microscope ( Applied Precision , GE Healthcare ) equipped with an Insight SSI Illumination , an X4 Laser module , a CoolSnap HQ ( Zhao et al . , 2017 ) CCD camera and a temperature-controlled chamber set up at 37 ⁰C was used . Laurdan images were taken with an Olympus UplanSApo 100x/1 . 4 oil objective . TIRFM image series were taken using an Olympus UAPO N 100X/1 . 49 TIRF objective and a 561 nm laser ( 50 mW , 100% power ) . Data processing was performed with softWoRx Suite 2 . 0 Software . Peptidoglycan ( PG ) synthesis was assessed by labelling the cells with HADA ( 7-hydroxycoumarin 3-carboxylic acid-amino-D-alanine ) ( Kuru et al . , 2015 ) , Van-FL ( Daniel and Errington , 2003 ) or D-Ala-D-Pra ( Sarkar et al . , 2016 ) . HADA: synthesised as described ( Morales Angeles et al . , 2017 ) . Overnight cultures of B . subtilis strains were diluted 1:100 into fresh LB medium or LB medium supplemented with 0 . 1% ( w/v ) of benzyl alcohol ( BnOH ) , a membrane fluidiser . Cells were grown until exponential phase , a sample of 1 ml of culture was spun down for 30 s , 5000 × g and the cell pellet was resuspended in 25 µl of fresh pre-warmed LB or LB containing 0 . 1% ( w/v ) BnOH . HADA was added to a 50 µM final concentration . Cells were incubated for 10 min in the dark ( 37 ⁰C , 200 rpm ) and then washed twice in 1 ml PBS buffer ( 58 mM Na2HPO4 , 17 mM NaH2PO4 , 68 mM NaCl , pH 7 . 3 ) to remove the excess of unbounded HADA . Cells were spun down again and resuspended in 25 µl of the appropriate medium and 2 µl of cells were mounted on 1% agarose slides before visualisation . Visualisation of HADA patterns ( excitation: 358 nm/emission: 461 nm ) under fluorescence microscopy from two biological replicates and cell length measurements were taken from at least 100 cells each strain/treatment . Van-FL: a 1:1 mixture of vancomycin ( Sigma Aldrich ) and BODIPYFL Vancomycin ( Molecular Probes ) at a final concentration 1 µg/ml was used to label cells for 5–10 min at room temperature . D-Ala-D-Pra: synthesised as described ( Sarkar et al . , 2016 ) . 1 ml of cell cultures was pelleted and resuspended in 50 µl of PBS buffer . Dipeptide was added to a final concentration of 0 . 5 mM following with 5 min incubation at room temperature . Cells were fixed by adding 70% ethanol and incubated for minimum 2 hr in −20°C . Next , cells were washed twice with PBS in order to remove unattached peptides , and resuspended in 50 µl of PBS . The D-Ala-D-Pra was subsequently labelled via a click reaction with fluorescent azide ( 20 μM ) that was incubated for 15 min at room temperature with addition of copper sulphate ( CuSO4 , 1 mM ) , tris-hydroxypropyltriazolylmethylamine ( THPTA , 125 μM ) and ascorbic acid ( 1 . 2 mM ) . The sample was washed twice with PBS and resuspended in 50 µl of the same buffer . Laurdan ( 6-Dodecanoyl-N , N-dymethyl2-naphthylamine , Sigma-Aldrich ) was used to detect the liquid ordering in the membrane , as decribed ( Bach and Bramkamp , 2013 ) , with modifications . Cells were grown in LB or SMM medium until late exponential phase . Laurdan , dissolved in dimethyformamide ( DMF ) , was added at 10 µM final concentration and cells were incubated for 10 min in the dark at 37 °C , 200 rpm . Cells were then washed twice in PBS buffer supplemented with 0 . 2% ( w/v ) glucose and 1% ( w/v ) DMF , and resuspended in fresh prewarmed appropriate medium . Laurdan was excited at 360 ± 20 nm , and fluorescence emission was captured at 460 ± 25 nm ( exposure time: 500 ms ) and at 535 ± 25 nm ( exposure time: 1 s ) ( Strahl et al . , 2014 ) . The image analysis including the generation of GP maps was carried out using Fiji Software ( Schindelin et al . , 2012 ) in combination with the macro tool CalculateGP designed by Norbert Vischer ( http://sils . fnwi . uva . nl/bcb/objectj/examples/CalculateGP/MD/gp . html ) . The GP values were measured for at least 100 individual cells after background subtraction , from two biological replicates . B . subtilis cell membranes were probed with Nile Red ( 0 . 5 µg/ml ) , FM4-64 ( 0 . 5 µg/ml ) or DiI-C12 ( 2 . 5 µg/ml ) . To this end , an overnight culture was grown in appropriate antibiotic , diluted 1:100 in LB medium supplemented with DiI-C12 followed by growth until exponential phase . Membranes were probed with Nile Red or FM4-64 for 5 min at room temperature after reaching exponential phase . The stained cells were washed three times in prewarmed LB medium supplemented with 1% DMSO before visualisation under fluorescence microscopy . Time-lapse TIRFM movies were taken in two independent experiments for each strain and condition . To this end , overnight cultures of strains grown in LB medium supplemented with the appropriate antibiotic were diluted 1:100 in medium containing 0 . 5% ( w/v ) xylose and grown until exponential phase . All experiments were performed inside the incubation chamber set to 37 °C , no longer than 10 min after taking the sample . The cells were imaged over 30 s with 1 s inter-frame intervals in a continuous illumination and ultimate focus correction mode . The single particle tracking analyses and kymographs were done using Fiji Software ( Schindelin et al . , 2012 ) in combination with the MTrackJ ( Meijering et al . , 2012 ) and MicrobeJ plugins ( Ducret et al . , 2016 ) . Cells were grown until an OD600 of 0 . 4–0 . 5 , and washed twice with PBS . Next , samples were resuspended in 50 μl PBS containing Bocillin-FL ( 5 µg/ml ) and incubated at room temperature for 10 min . Subsequently cells were harvested , lysed by sonication and cell-free extracts were prepared . Samples , equalised for culture OD , were prepared with SDS-PAGE sample buffer and run on a 12% SDS-PAGE gel . Fluorescent bands were visualised using a Typhoon Trio ( GE Healthcare ) scanner . Membrane isolation was adapted from Schneider et al . , 2015b . Briefly , cells were grown until an OD600 of 0 . 4–0 . 5 , cell fractions were collected and resuspended in PBS with Lysozyme ( 1 µg/ml ) , EDTA ( 5 mM ) , 1/10 tablet cOmplete protease inhibitor ( Roche ) , and DNAse ( 5 µg/ml ) and incubated for 30 min on ice . Samples were sonicated , cell which did not lyze were spun down ( 8000 rpm , 2 min , 4°C ) , and the supernatant fraction was centrifuged at 4°C and 40000 rpm for 1 hr . The membrane pellet was dissolved in ACA750 buffer ( 750 mM aminocaproic acid , 50 mM Bis-Tris , pH 7 . 0 ) to a final protein concentration of 1 μg/μl . Membranes were solubilised overnight at 4°C in 1% ( w/v ) dodecylmaltoside ( DDM ) and either used directly or stored at −20°C . The experiment was performed as described ( Trip and Scheffers , 2016 ) . Samples were prepared by mixing sample buffer ( 0 . 1% Ponceau S , 42 . 5% Glycerol ) with solubilised membranes in a 1:3 ratio . Samples were resolved on a mini-PROTEAN TGX Stain-Free gradient gel ( 4–15% , BioRad ) using cathode ( 50 mM Tricine and 15 mM BisTris ) , and anode ( 50 mM BisTris pH 7 . 0 ) buffers . The Novex NativeMark Unstained Protein Standard marker was used as a Mw marker . A lane of interest was excised from the Native-PAGE gel and immobilised horizontally on top of a SDS-PAGE gel ( 5% stacking , 12% resolving ) . The excised fragment was flanked with a piece of Whatman paper soaked with PageRuler Prestained Protein Ladder . The gel fragment to be resolved in the second dimension was topped with a mix of 1% ( w/v ) LowTemperature agarose , 0 . 5% ( w/v ) SDS and bromophenol blue . After the agarose had solidified , standard SDS-PAGE electrophoresis was performed . Cultures were harvested by centrifugation and a small amount of pellet was placed on a copper dish . A 400 copper mesh grid and a 75 µm aperture grid was placed on top of the cells to create a thin layer . The sandwiched cells were plunged rapidly into liquid propane . Sandwiches were then disassembled and placed on frozen freeze-substitution medium containing 1% osmium tetroxide , 0 . 5% uranyl acetate and 5% water in acetone . Cells were dehydrated and fixed using the rapid freeze substitution method ( McDonald , 2014 ) . Samples were embedded in epon and ultrathin sections were collected on formvar coated and carbon evaporated copper grids and inspected using a CM12 ( Philips ) transmission electron microscope . For each strain 70 random septa were imaged with pixel resolution of 1 . 2 nm . Using ImageJ the cell wall thickness for each septum was measured at 4 places from which the average was taken . Each set of micrographs to be analysed was imaged with the same exposure time . For the septum intensity analysis of HADA , FM4-64 and Nile Red , the wildtype strain ( expressing GFP ) and the ΔfloAT strains were mixed , labelled and imaged on the same agarose pad . Intensity of the fluorescently labelled septa was measured using the ObjectJ macro tool PeakFinder ( https://sils . fnwi . uva . nl/bcb/objectj/examples/PeakFinder/peakfinder . html ) ( Vischer et al . , 2015 ) . A perpendicular line was drawn across the septal plane , the background intensity was removed resulting in a maximum peak intensity . The number of septa compared was indicated for every individual experiment . Populations were compared using the non-parametric Mann-Whitney test . The null hypothesis was tested with the p value of 0 . 05 . The statistical analyses and their graphical representation ( box plots ) were generated with GraphPad Prism 8 . 1 ( San Diego , California , USA ) . Box plots show the median and the interquartile range ( box ) , the 5th and 95th percentile ( whiskers ) . Laurdan fluorescence generalised polarisation , cell length and MreB speed statistical analyses were performed using Kruskal-Wallis with Dunn's multiple comparison post-hoc test . The fatty acid composition of B . subtilis wild-type cells and the flotillin/PBP mutants was analysed with gas chromatography as fatty acid methyl esters . Cells for the analyses were grown at 37°C in LB or SMM until mid-exponential ( OD600 ~0 . 5 ) , harvested ( 6000 rpm , 10 min , 4°C ) and washed with 100 mM NaCl . Next , the cells were freeze dried at −50°C , 0 . 012 mbar for a minimum of 18 hr . All analyses were carried out on biological duplicates by the Identification Service of the DSMZ , Braunschweig , Germany . FloT was essentialy purified as described ( Bach and Bramkamp , 2013 ) , in solubilised form , and stored in buffer A ( 50 mM Tris HCl pH 7 . 5 , 150 mM NaCl , 5 mM MgCl2 ) supplemented with 0 . 05% Triton X-100 . Liposomes containing POPC-d31 were prepared by mixing appropriate lipid powders in organic solvents ( chloroform/methanol , 2:1 ratio ) . Solvents were evaporated under a flow of N2 to obtain a thin lipid film . Lipids were rehydrated with ultrapure water before lyophilisation over night . The lipid powder was hydrated with an appropriate amount of buffer A with 10% glycerol and homogenised by three cycles of vortexing , freezing ( liquid nitrogen , −196°C , 1 min ) and thawing ( 40°C in a water bath , 10 min ) . This protocol generated a milky suspension of micrometer-sized multilamellar vesicles . FloT was solubilised in Buffer A supplemented with 0 . 05% Triton X-100 and added to preformed liposomes and incubated for 1 hr at room temperature . A dialysis step was then performed against Buffer A at 4°C under agitation to remove the detergent . Samples were centrifuged at 100 , 000 g at 4°C for 1 hr to pellet the proteoliposomes . 2H solid-state NMR spectra were recorded of liposomes in the presence or absence of FloT at a lipid/protein ratio of 25:1 at 298 K . 2H NMR spectroscopy experiments were performed using a Bruker Avance III 500 MHz WB ( 11 . 75 T ) spectrometer . They were recorded on 2H-labelled POPC at 76 . 77 MHz with a phase-cycled quadrupolar echo pulse sequence ( 90°x-t-90°y-t-acq ) . Acquisition parameters were as follows: spectral window of 500 kHz for 2H NMR spectroscopy , p/2 pulse width of 3 . 90 ms for 2H , interpulse delays ( t ) were of 40 µs , recycled delays of 1 . 3 s for 2H; 3000 and 8000 scans were used for 2H NMR spectroscopy on liposomes and liposomes with FloT , respectively . Spectra were processed using a Lorentzian line broadening of 300 Hz for 2H NMR spectra before Fourier transformation from the top of the echo . Samples were equilibrated for 30 min at a given temperature before data acquisition . All spectra were processed and analysed using Bruker Topspin 3 . 2 software . Spectral moments were calculated for each temperature using the NMR Depaker 1 . 0rc1 software [Copyright ( C ) 2009 Sébastien Buchoux] . Orientational order parameters ( SCD ) were calculated from experimental quadrupolar splittings ( DnQ ) as described in Huster , 2014 . For 31P ssNMR , we applied a static Hahn spin echo sequence at the 31P frequency of 162 MHz on a 400 MHz ( 9 . 4T ) Bruker Avance III HD spectrometer , with a 90◦ pulse of 8 μs , a delay of 40 μs , a recycle delay of 5 s , a spectral window of 400 ppm and a number of scans of 4000 and 3400 was used on liposomes and liposomes with FloT , respectively . Spectra were processed using a Lorentzian line broadening of 100 Hz .	Every living cell is enclosed by a flexible membrane made of molecules known as phospholipids , which protects the cell from harmful chemicals and other threats . In bacteria and some other organisms , a rigid structure known as the cell wall sits just outside of the membrane and determines the cell’s shape . There are several proteins in the membrane of bacteria that allow the cell to grow by assembling new pieces of the cell wall . To ensure these proteins expand the cell wall at the right locations , another protein known as MreB moves and organizes them to the appropriate place in the membrane and controls their activity . Previous studies have found that another class of proteins called flotillins are involved in arranging proteins and phospholipid molecules within membranes . Bacteria lacking these proteins do not grow properly and are unable to maintain their normal shape . However , the precise role of the flotillins remained unclear . Here , Zielińska , Savietto et al . used microscopy approaches to study flotillins in a bacterium known as Bacillus subtilis . The experiments found that , in the presence of flotillins , MreB moved around the membrane more quickly ( suggesting it was more active ) than when no flotillins were present . Similar results were observed when bacterial cells lacking flotillins were treated with a chemical that made membranes more ‘fluid’ – that is , made it easier for the molecules within the membrane to travel around . Further experiments found that flotillins allowed the phospholipid molecules within an artificial membrane to move around more freely , which increases the fluidity of the membrane . These findings suggest that flotillins make the membranes of bacterial cells more fluid to help cells expand their walls and perform several other processes . Understanding how bacteria control the components of their membranes will further our understanding of how many currently available antibiotics work and may potentially lead to the design of new antibiotics in the future .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "microbiology", "and", "infectious", "disease" ]	2020	Flotillin-mediated membrane fluidity controls peptidoglycan synthesis and MreB movement
A major constraint on the evolution of large body sizes in animals is an increased risk of developing cancer . There is no correlation , however , between body size and cancer risk . This lack of correlation is often referred to as 'Peto's Paradox' . Here , we show that the elephant genome encodes 20 copies of the tumor suppressor gene TP53 and that the increase in TP53 copy number occurred coincident with the evolution of large body sizes , the evolution of extreme sensitivity to genotoxic stress , and a hyperactive TP53 signaling pathway in the elephant ( Proboscidean ) lineage . Furthermore , we show that several of the TP53 retrogenes ( TP53RTGs ) are transcribed and likely translated . While TP53RTGs do not appear to directly function as transcription factors , they do contribute to the enhanced sensitivity of elephant cells to DNA damage and the induction of apoptosis by regulating activity of the TP53 signaling pathway . These results suggest that an increase in the copy number of TP53 may have played a direct role in the evolution of very large body sizes and the resolution of Peto's paradox in Proboscideans . Lifespan and maximum adult body size are fundamental life history traits that vary considerably between species ( Healy et al . , 2014 ) . The maximum lifespan among vertebrates , for example , ranges from over 211 years in the bowhead whale ( Balaena mysticetus ) to only 59 days in the pygmy goby ( Eviota sigillata ) whereas body sizes ranges from 136 , 000 kg in the blue whale ( Balaenoptera musculus ) to 0 . 5 g in the Eastern red-backed salamander ( Plethodon cinereus ) ( Healy et al . , 2014 ) . Similar to other life history traits , such as body size and metabolic rate or body size and age at maturation , body size and lifespan are strongly correlated such that larger species tend to live longer than smaller species ( Figure 1A ) . While abiotic and biological factors have been proposed as major drivers of maximum body size evolution in animals , maximum body size within tetrapods appears to be largely determined by biology ( Smith et al . , 2010; Sookias et al . , 2012 ) . Mammals , for example , likely share biological constraints on the evolution of very large body sizes with rare breaks in those constraints underlying the evolution of gigantism in some lineages ( Sookias et al . , 2012 ) , such as Proboscideans ( elephants and their and extinct relatives ) , Cetaceans ( whales ) , and the extinct hornless rhinoceros Paraceratherium ( ‘Walter’ ) . 10 . 7554/eLife . 11994 . 003Figure 1 . Body size evolution in vertebrates . ( A ) Relationship between body mass ( g ) and lifespan ( years ) among 2556 vertebrates . Blue line shows the linear regression between log ( body mass ) and log ( lifespan ) , R2 = 0 . 32 . ( B ) Body size comparison between living ( African and Asian elephants ) and extinct ( Steppe mammoth ) Proboscideans , Cetaceans ( Minke whale ) , and the extinct hornless rhinoceros Paraceratherium ( ‘Walter’ ) , and humans . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 003 A major constraint on the evolution of large body sizes in animals is an increased risk of developing cancer . If all cells have a similar risk of malignant transformation and equivalent cancer suppression mechanisms , organism with many cells should have a higher risk of developing cancer than organisms with fewer cells; Similarly organisms with long lifespans have more time to accumulate cancer-causing mutations than organisms with shorter lifespans and therefore should be at an increased risk of developing cancer , a risk that is compounded in large bodied , long-lived organisms ( Cairns , 1975; Caulin and Maley , 2011; Doll , 1971; Peto , 2015; Peto et al . , 1975 ) . There are no correlations , however , between body size and cancer risk or lifespan and cancer risk across species ( Leroi et al . , 2003 ) , this lack of correlation is often referred to as ‘Peto’s Paradox’ ( Caulin and Maley , 2011; Peto et al . , 1975 ) . Epidemiological studies in wild populations of Swedish roe deer ( Capreolus capreolus ) and beluga whales ( Delphinapterus leucas ) in the highly polluted St . Lawrence estuary , for example , found cancer accounted for only 2% ( Aguirre et al . , 1999 ) and 27% ( Martineau et al . , 2002 ) of mortality , respectively , much lower than expected given body size of these species ( Caulin and Maley , 2011 ) . Among the mechanisms large , long lived animals may have evolved that resolve Peto’s paradox are a decrease in the copy number of oncogenes , an increase in the copy number of tumor suppressor genes ( Caulin and Maley , 2011; Leroi et al . , 2003; Nunney , 1999 ) , reduced metabolic rates leading to decreased free radical production , reduced retroviral activity and load ( Katzourakis et al . , 2014 ) , increased immune surveillance , and selection for 'cheater' tumors that parasitize the growth of other tumors ( Nagy et al . , 2007 ) , among many others . Naked mole rats ( Heterocephalus glaber ) , for example , which have very long lifespans for a small-bodied organism evolved cells with extremely sensitive contact inhibition likely acting as a constraint on tumor growth and metastasis ( Seluanov et al . , 2009; Tian et al . , 2013 ) . Similarly long-lived blind mole rats ( Splanx sp . ) evolved an enhanced TP53-signaling and necrotic cell death mechanisms that also likely constrains tumor growth ( Ashur-Fabian et al . , 2004; Avivi et al . , 2007; Avivi et al . , 2005; Gorbunova et al . , 2012; Manov et al . , 2013 ) . Thus , while some of the mechanisms that underlie cancer resistance in small , long-lived mammals have been identified , the mechanisms by which large bodied animals evolved enhanced cancer resistance are unknown . Here we use evolutionary genomics and comparative cell biology to explore the mechanisms by which elephants , the largest extant land mammal ( Figure 1B ) , have evolved enhanced resistance to cancer . We found that the elephant genome encodes a single TP53 gene and 19 TP53 retrogenes , several of which are transcribed and translated in elephant tissues . Comparison of the African and Asian elephant TP53 gene copy number with the copy number in the genome of the extinct American mastodon , woolly mammoth , and Columbian mammoth indicates that copy number increased relatively rapidly coincident with the evolution of large body-sizes in the Proboscidean lineage . Finally , we show that elephant cells have an enhanced response to DNA-damage that is mediated by a hyperactive TP53 signaling pathway and that this augmented TP53 signaling is dependent upon TP53 retrogenes and can be transferred to the cells of other species through exogenous expression of elephant TP53 retrogenes . These results suggest that the origin of large body sizes , long lifespans , and enhanced cancer resistance in the elephant lineage evolved at least in part through reinforcing the anti-cancer mechanisms of the major ‘guardian of the genome’ TP53 . We characterized TP53 copy number in 61 Sarcopterygians ( Lobe-finned fishes ) with draft or completed genomes , including large , long-lived mammals such as the African elephant ( Loxodonta africana ) , Bowhead ( Balaena mysticetus ) and Minke ( Balaenoptera acutorostrata scammoni ) whales . We found that all Sarcopterygian genomes encoded a single TP53 gene and that some lineages also contained a few TP53 retrogenes ( TP53RTG ) , including marsupials , Yangochiropteran bats , and Glires , in which ‘processed’ TP53 pseudogenes have previously been reported ( Ciotta et al . , 1995; Czosnek et al . , 1984; Hulla , 1992; Tanooka et al . , 1995; Weghorst et al . , 1995; Zakut-Houri et al . , 1983 ) . We also identified a single TP53RTG gene in the lesser hedgehog tenrec ( Echinops telfairi ) , which had been previously reported ( Belyi et al . , 2010 ) , rock hyrax ( Procavia capensis ) , and West Indian manatee ( Trichechus manatus ) . The African elephant genome , however , encoded 19 TP53RTG genes ( Figure 2A ) , 14 of which retain potential to encode truncated proteins ( Table 1 ) . 10 . 7554/eLife . 11994 . 004Figure 2 . Expansion of the TP53RTG gene repertoire in Proboscideans . ( A ) TP53 copy number in 61 Sarcopterygian ( Lobe-finned fish ) genomes . Clade names are shown for lineages in which the genome encodes more than one TP53 gene or pseudogene . ( B ) Estimated TP53/TP53RTG copy number inferred from complete genome sequencing data ( WGS , purple ) , 1:1 orthology ( green ) , gene tree reconciliation ( blue ) , and normalized read depth from genome sequencing data ( red ) . Whiskers on normalized read depth copy number estimates show the 95% confidence interval of the estimate . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 00410 . 7554/eLife . 11994 . 005Figure 2—figure supplement 1 . Reconciled TP53/TP53RTG gene trees . Reconciled TP53RTG gene trees for Columbian mammoth , woolly mammoth , Asian elephant , and American mastodon . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 00510 . 7554/eLife . 11994 . 006Table 1 . Summary information for African elephant TP53/TP53RTG genes . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 006Gene Id ENSEMBL Id Scaffold ( loxAfr3 ) Chromosome ( loxAfr4 ) Coding potential ORF size TP53 ENSLAFG0000000748347Chr 11Yes392aaTP53RTG1 ENSLAFG00000025553175UnmappedNoN/ATP53RTG2 N/A217UnmappedYes134aaTP53RTG3 ENSLAFG00000027474406UnmappedYes79aaTP53RTG4 N/A627UnmappedYes134aaTP53RTG5 ENSLAFG00000027348221UnmappedYes162aaTP53RTG6 N/A76Chr 27Yes123aaTP53RTG7 N/A208UnmappedNoN/ATP53RTG8 ENSLAFG00000027820294UnmappedYes210aaTP53RTG9 ENSLAFG00000027669786UnmappedNoN/ATP53RTG10 ENSLAFG00000030555221UnmappedYes210aaTP53RTG11 N/A281Yes203aaTP53RTG12 ENSLAFG00000028299825UnmappedYes180aaTP53RTG13 ENSLAFG00000032042458UnmappedNoN/ATP53RTG14 ENSLAFG00000026238928UnmappedYes210aaTP53RTG15 ENSLAFG00000027365656UnmappedYes210aaTP53RTG16 ENSLAFG00000030880378UnmappedNoN/ATP53RTG17 ENSLAFG00000028692552UnmappedYes111aaTP53RTG18 N/A498UnmappedYes111aaTP53RTG19 ENSLAFG00000032258342UnmappedYes210aa To trace the expansion of TP53RTG gene family in the Proboscidean lineage with greater phylogenetic resolution , we used three methods to estimate the minimum ( 1:1 orthology ) , average ( normalized read depth ) , and maximum ( gene tree reconciliation ) TP53/TP53RTG copy number in the Asian elephant ( Elephas maximus ) , extinct woolly ( Mammuthus primigenius ) and Columbian ( Mammuthus columbi ) mammoths , and the extinct American mastodon ( Mammut americanum ) using existing whole genome sequencing data ( Enk et al . , 2014 , 2013; Rohland et al . , 2010; Wilkie et al . , 2013 ) . As expected , we identified a single canonical TP53 gene in these species and estimated the TP53RTG copy number in the Asian elephant genome to be 12–17 , approximately 14 in both the Columbian and woolly mammoth genomes , and 3–8 in the 50 , 000–130 , 000 year old American mastodon genome ( Figure 2Band Figure 2—figure supplement 1 ) . These data indicate that large-scale expansion of the TP53RTG gene family occurred in the Proboscidean lineage and suggest that TP53RTG copy number was lower in ancient Proboscideans such as the mastodon , which diverged from the elephant lineage ~ 25 MYA ( Rohland et al . , 2010 ) , than in recent species such as elephants and mammoths . Several mechanisms may have increased the TP53RTG copy number in the Proboscidean lineage including serial retrotransposition from the TP53 gene , serial retrotransposition from the TP53 and one or more daughter transcribed retrogenes , repeated segmental duplications of chromosomal loci containing TP53RTG genes , or some combination of these mechanisms . Consistent with copy number expansion through a single retrotransposition event followed by repeated rounds of segmental duplication , we found that each TP53RTG retrogene was flanked by nearly identical clusters of transposable elements ( Figure 3A ) and embedded within a large genomic region with greater than 80% sequence similarity ( Figure 3B ) . Next we used progressiveMAUVE to align the 18 elephant contigs containing TP53RTG retrogenes and found that they were all embedded within large locally collinear blocks that span nearly the entire length of some contigs ( Figure 3C ) , as expected for segmental duplications . 10 . 7554/eLife . 11994 . 007Figure 3 . TP53RTG copy number increased through segmental duplications . ( A ) Organization of the TP53 and TP53RTG loci in African elephant . The TP53/TP53RTG gene tree is shown at the left . The location of homologous transposable elements that flank the TP53RTG genes are shown and l Abegglen ed . TP53RTG genes with intact start codons are labeled with arrows , stop codons are shown in red . ( B ) Multiple sequence alignment ( MUSCLE ) of elephant TP53RTG containing contigs . The location of the TP53RTG genes is shown with a blue bar . Sites are color coded according to their conservation ( see inset key ) . ( C ) ProgressiveMAUVE alignment of elephant TP53RTG containing contigs . Colored boxes shown the location of collinear blocks , lines connect homologous collinear blocks on different contigs . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 007 Our observation that TP53RTG genes expanded through segmental duplications suggests they may have a tree-like phylogenetic history that preserves information about when in the evolution of Proboscideans the duplicates occurred . Therefore we assembled a dataset of TP53/TP53RTG orthologs from 65 diverse mammals and jointly inferred the TP53/TP53RTG gene tree and duplication dates in a Bayesian framework to determine if TP53/TP53RTG copy number was correlated with body size evolution in Proboscideans . For comparison , we also inferred the phylogenetic history of TP53/TP53RTG genes using maximum likelihood and an additional Bayesian method . We found all phylogenetic inference methods inferred that the TP53RTG genes from elephant , hyrax , and manatee formed a well-supported sister clade to the canonical genes from these species , whereas the tenrec TP53 and TP53RTG genes formed a separate well-supported clade ( Figure 4A and Figure 4—figure supplement 1 ) . These data indicate that retrotransposition of TP53 occurred independently in tenrecs and in the elephant , hyrax , and manatee stem-lineage ( Paenungulata ) , followed by expansion of TP53RTG genes in the Proboscidean lineage . 10 . 7554/eLife . 11994 . 008Figure 4 . TP53RTG copy number is correlated with body size evolution in Proboscideans . ( A ) Time calibrated Bayesian phylogeny of TP53/TP53RTG genes . TP53RTG genes are shown in blue , the 95% highest posterior density ( HPD ) of estimated divergence dates are shown as red bars , nodes with a posterior probability ( PP ) ( PP ) > 0 . 95 are labeled with closed circles whereas nodes with a PP ≤ 0 . 95 . 95 are labeled with open circles . The period corresponding to the expansion of the TP53RTG gene repertoire is shown in a grey . ( B ) TP53/TP53RTG copy number ( blue ) and Proboscidean body size ( red ) increases through time are correlated . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 00810 . 7554/eLife . 11994 . 009Figure 4—figure supplement 1 . TP53/TP53RTG gene trees . Maximum likelihood gene trees with aBayes , Chi-Square , aLRT , and SH-like branch supports . Bayesian tree inferred by MrBayes . Numbers along branches indicate node supports from each method . Note that the full dataset included 65 diverse species and only the Paeunngulate clade is shown . TP53 gene trees were inferred using a general time reversible model ( GTR ) , empirical nucleotide frequencies ( +F ) , a proportion of invariable sites estimated from the data ( +I ) , four gamma distributed rate categories ( +G ) , and using the best of NNI and SPR branch moves during the topology search ( for the maximum likelihood tree ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 009 Based on our time-calibrated phylogeny , we inferred that the initial retrotransposition of the TP53 gene in the Paenungulata stem-lineage occurred approximately 64 MYA ( 95% HPD = 62 . 3–66 . 2 MYA; Figure 4A ) . This was followed by a period of ~25 million years during which no further retrotranspositions or segmental duplications were fixed in the genome , however , the TP53RTG gene family rapidly expanded after ~40 MYA ( 95% HPD = 30 . 8–48 . 6 MYA; Figure 4A ) . To correlate TP53/TP53RTG copy number and the origin of large body sizes in Proboscideans we estimated TP53/TP53RTG copy number through time and gathered data on ancient Proboscidean body sizes from the literature ( Evans et al . , 2012; Smith et al . , 2010 ) . We found that the increase in TP53/TP53RTG copy number in the Proboscidean lineage and Proboscidean body size evolution closely mirrored each other ( Figure 4B ) . If expansion of the TP53RTG gene repertoire played a role in the resolution of Peto’s paradox during the evolution of large bodied Proboscideans , then one or more of the TP53RTG genes should be transcribed . Therefore we generated RNA-Seq data from Asian elephant dermal fibroblasts , African elephant term placental villus and adipose tissue , and used previously published RNA-Seq data from Asian elephant PBMCs ( Reddy et al . , 2015 ) and African elephant fibroblasts ( Cortez et al . , 2014 ) to determine if TP53RTG genes were transcribed . We found that the TP53 and TP53RTG12 genes were robustly transcribed in all samples , whereas TP53RTG3 and TP53RTG18 transcripts were much less abundant ( Figure 5A ) . To confirm that the African and Asian elephant TP53RTG genes were transcribed , we designed a set of PCR primers specific to the TP53 and TP53RTG genes that flank a diagnostic 15–30 bp deletion in TP53RTG genes ( Figure 5—figure supplement 1 ) and used RT-PCR to assay for expression in Elephant fibroblast cDNA generated from DNase treated RNA . Consistent with transcription of the TP53RTG genes , we amplified PCR products at the expected size for the TP53 and TP53RTG transcripts but did not amplify PCR products from negative control ( no reverse transcriptase ) samples ( Figure 5B ) . Sanger sequencing of the cloned PCR products confirmed transcription of TP53RTG12 and TP53RTG18/19 ( Figure 5—figure supplement 1 ) in African elephant and TP53RTG12 and TP53RTG13 in Asian elephant fibroblast . We note that we used a Poly-T primer for cDNA synthesis , thus the amplification of TP53RTG transcripts indicates that these transcripts are poly-adenylated . 10 . 7554/eLife . 11994 . 010Figure 5 . TP53RTG12 is transcribed and translated . ( A ) Transcription of elephant TP53 and TP53RTG genes in dermal fibroblasts , white adipose , and placental villi . RNA-Seq data are shown as mean transcripts per million ( TPM ) with 95% confidence intervals of TPM value . Blue bars show TPM estimates from ‘end-to-end’ read mapping and gray bars shown ‘local’ read mapping . ( B ) qRT-PCR products generated Asian ( left , blue sqaure ) and African ( right , light blue square ) elephant fibroblast cDNA using primers specific to TP53 and TP53RTG12 . cDNA was generated from DNaseI-treated RNA . No reverse transcriptase ( no RT ) controls for each qPCR reaction are shown , end point PCR products are shown . ( C ) Coverage of mapped reads from Asian ( dark blue ) and African ( light blue ) elephant fibroblast RNA-Seq data across the region of scaffold_885 encoding the TP53RTG12 gene . The location of TP53RTG12 exons predicted from geneid and GENESCAN are shown in blue introns are shown as lines with arrows indicating the direction of transcription . Gray bars show the location of transposable elements around the TP53RTG12 gene , darker gray indicates high sequence similarity to the consensus of each element . PCR tiles across this region are shown for African ( Lox . ) and Asian ( Ele . ) elephants , PCR primers generating amplicons are shown in red . The inferred TP53RTG12 transcript is shown below . one kb scale shown from position one of African elephant ( Broad/loxAfr3 ) scaffold_825 . ( D ) Relative luciferase ( Luc . ) expression in Asian and African fibroblasts transfected with either the promoterless pGL4 . 10[luc2] luciferase reporter vector ( empty vector ) , pGL4 . 10 containing the RTE_LA consensus sequences ( Consensus ) , pGL4 . 10 containing the RTE_LA from Asian elephant ( Ele . RTE_LA ) , or pGL4 . 10 containing the RTE_LA from African elephant ( Lox . RTE_LA ) . Results are shown as fold difference in Luc . expression standardized to empty vector and Renilla controls . n = 16 , Wilcoxon P-values . ( E ) Western blot of total cell protein isolated from South African Rock hyrax , Asian elephant ( Elephas ) , and African elephant ( Loxodona ) dermal fibroblasts . − , control cells . + , cells treated with 50 J/m2UV-C and the proteasome inhibitor MG-132 . The name and predicted molecular weights of TP53 isoforms are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 01010 . 7554/eLife . 11994 . 011Figure 5—figure supplement 1 . PCR and Sanger sequencing confirm TP53RTG12 is transcribed in elephant fibroblasts . ( A ) Multiple sequence alignment ( MSA ) of African elephant TP53 and TP53RTG1-TP53RTG19 . Numbering ( shown at top ) is relative to the canonical TP53 transcript in African elephant ( ENSLAFT00000007484 ) , sequence conservation in this region is shown in Bits , as a logo , and percent identity ( green , 100% identity of paralogs; yellow , < 100% identity of paralogs ) . Bases in each gene are color-coded ( blue , C; red , A; yellow , G; and T , green ) . The TP53RTG-specific deletion is shown as a black line , and the location of the PCR primers than span the deletion are shown as arrows . ( B ) Chromatogram from Sanger sequencing of the TP53RTG12 qRT-PCR amplicon confirm the TP53RTG12 is transcribed in African elephant fibroblasts . Bases that flank the TP53RTG-specific deletion are highlighted in red . ( C ) Chromatogram from Sanger sequencing of the TP53RTG12 qRT-PCR amplicon confirm the TP53RTG12 is transcribed in Asian elephant fibroblasts . Bases that flank the TP53RTG-specific deletion are highlighted in red . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 01110 . 7554/eLife . 11994 . 012Figure 5—figure supplement 2 . Unedited Western blots shown in Figure 5E . ( A ) Western blot of total cell protein isolated from African elephant and hyrax fibroblasts . − , control cells . + , cells treated with 50 J/m2UV-C and MG-132 . 1° and 2° , probed with primary antibody against TP53 and secondary HRP conjugated antibody . 2° only , probed only with secondary HRP conjugated antibody ( B ) Western blot of total cell protein isolated from Asian elephant and hyrax fibroblasts . − , control cells . + , cells treated with 50 J/m2UV-C and MG-132 . 1° and 2° , probed with primary antibody against TP53 and secondary HRP conjugated antibody . 2° only , probed only with secondary HRP conjugated antibodyDOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 012 Most retrogenes lack native regulatory elements such as promoters and enhancers to initiate transcription , thus transcribed TP53RTG genes likely co-opted existing regulatory elements or evolved regulatory elements de novo . To identify putative transcriptional start sites and promoters of the highly expressed TP53RTG12 gene we used geneid and GENESCAN to computationally predict exons in the African elephant gene and mapped the African and Asian elephant fibroblast RNA-Seq data onto scaffold_825 , which encodes the TP53RTG12 gene . We found that both computational methods predicted an exon ~2 kb upstream of the ENSEMBL annotated TP53RTG12 gene , within an RTE-type non-LTR retrotransposon ( RTE1_LA ) that we annotated as Afrotherian-specific ( Figure 5C ) . Consistent with this region encoding a transcribed 5’-UTR , a peak of reads mapped within the predicted 5’-exon and within the RTE1_LA retrotransposon ( Figure 5C ) . We attempted to identify the transcription start site of the TP53RTG12 gene using several 5’-RACE methods , however , we were unsuccessful in generating PCR products from either African Elephant fibroblast or placenta cDNA , or Asian elephant fibroblast cDNA . Therefore , we designed a set of 34 PCR primers tiled across the region of scaffold_825 that encodes the TP53RTG12 gene and used these primers to amplify PCR products from African and Asian Elephant fibroblast cDNA generated from DNase treated RNA . We then reconstructed the likely TP53RTG12 promoter , transcription start site , and exon-intron structure from the pattern of positive PCR products . These data suggest that the major transcription initiation site of TP53RTG12 is located within a RTE1_LA class transposable element ( Figure 5C ) . Next we tested the ability of the African and Asian elephant RTE1_LA elements and the RTE1_LA consensus sequence ( as a proxy for the ancestral RTE1_LA sequence ) to function as a promoter in transiently transfected African and Asian elephant fibroblasts when cloned into the promoterless pGL4 . 10[luc2] luciferase reporter vector . We found that the African and Asian elephant RTE1_LA elements increased luciferase expression 3 . 03-fold ( t-test , p=2 . 41 × 10–8 ) and 1 . 37-fold ( t-test , p=2 . 60 × 10–4 ) , respectively , compared to empty vector controls ( Figure 5D ) . However , luciferase expression from the pGL4 . 10[luc2] vector containing the RTE1_LA sequence was not significantly different than the empty vector control in either Asian ( 0 . 96-fold; t-test , p=0 . 61 ) or African elephant fibroblasts ( 0 . 95-fold; t-test , p=0 . 37; Figure 5D ) . These data indicate that transcription of TP53RTG12 likely initiates within a RTE1_LA-derived promoter , but that the ability of this RTE1_LA element to function as a promoter is not an ancestral feature of RTE1_LA elements . To determine if TP53RTG transcripts are translated , we treated African elephant , Asian elephant , and hyrax cells with 50 J/m2 UV-C ( to stabilize TP53 ) and the proteasome inhibitor MG-132 ( to block protein degradation ) , and assayed for TP53/TP53RTG proteins by Western blotting total cell protein with a polyclonal TP53 antibody ( FL-393 ) that we demonstrated recognizes Myc-tagged TP53RTG12 . We identified bands in both African and Asian elephant and hyrax total cell protein at the expected size for the full length p53 , Δ133 p53β/γ , and p53ψ-like isoforms of the TP53 protein ( Khoury and Bourdon , 2010 ) as well as high molecular weight bands corresponding to previously reported SDS denaturation resistant TP53 oligomers ( Cohen et al . , 2008; Ottaggio et al . , 2000 ) and ( poly ) ubiquitinated TP53 congugates ( Sparks et al . , 2014 ) ( Figure 5E ) . We also identified an elephant-specific band at the expected size for the TP53RTG12 ( 19 . 6 kDa ) and TP53RTG19 ( 22 . 3 kDa ) proteins , suggesting that the TP53RTG12 and TP53RTG19 transcripts are translated in elephant fibroblasts ( Figure 5Eand Figure 5—figure supplement 2 ) . Our observation that TP53RTG genes are expressed suggests that elephant cells may have an altered TP53 signaling system compared to species without an expanded TP53/TP53RTG gene repertoire . To directly test this hypothesis we transiently transfected primary African elephant , Asian elephant , South African Rock hyrax ( Procavia capensis capensis ) , East African aardvark ( Orycteropus afer lademanni ) , and Southern Three-banded armadillo ( Tolypeutes matacus ) dermal fibroblasts with a luciferase reporter vector containing two TP53 response elements ( pGL4 . 38[luc2p/p53 RE/Hygro] ) and Renilla control vector ( pGL4 . 74[hRluc/TK] ) . Next we used a dual luciferase reporter assay to measure the activation of the TP53 pathway in response to treatment with three DNA damage inducing agents ( mitomycin C , doxorubicin , or UV-C ) or nutlin-3a , which inhibits the interaction between MDM2 and TP53 and thus promotes TP53 signaling . We found that elephant cells generally up-regulated TP53 signaling in response to lower doses of each drug and UV-C than closely related species without an expanded TP53 gene repertoire ( Figure 6A ) , indicating elephant cells have evolved an enhanced TP53 response . 10 . 7554/eLife . 11994 . 013Figure 6 . Elephant cells have enhanced TP53 signaling and are hyper-responsive to DNA damage . ( A ) Relative luciferase ( Luc . ) expression in African elephant , Asian elephant , hyrax , aardvark , and armadillo fibroblasts transfected with the pGL4 . 38[luc2p/p53 RE/Hygro] reporter vector and treated with either mitomycin c , doxorubicin , nutlin-3a , or UV-C . Data are shown as fold difference in Luc . expression 18 hr after treatment standardized to species paired empty vector and Renilla controls . n = 12 , mean±SD . ( B ) Relative capsase-3/7 ( Cas3/7 ) activity in African elephant , Asian elephant , hyrax , aardvark , and armadillo treated with either mitomycin c , doxorubicin , nutlin-3a , or UV-C . Data are shown as fold difference in Cas3/7 activity 18 hr after treatment standardized to species paired untreated controls . n = 12 , mean±SD . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 013 To determine the consequences of an enhanced TP53 response we treated primary African and Asian elephant , hyrax , aardvark , and armadillo dermal fibroblasts with mitomycin C , doxorubicin , nutlin-3a , or UV-C and measured cell viability ( live-cell protease activity ) , cytotoxicity ( dead-cell protease activity ) , and the induction apoptosis ( caspase-3/7 activation ) using an ApoTox-Glo Triplex assay . Consistent with the results from the luciferase assay , we found that lower doses of mitomycin C or doxorubicin induced apoptosis in elephant cells than the other species ( Figure 6B ) and that the magnitude of the response was greater in elephant than other species ( Figure 6A ) . Similarly UV-C exposure generally induced more elephant cells to undergo apoptosis than other species ( Figure 6A ) . A striking exception to this trend was the response of elephant cells to the MDM2 antagonist nutlin-3a , which elicited a strong TP53 transcriptional response ( Figure 6A ) but did not induce apoptosis ( Figure 6B ) . Thus we conclude that elephant cells generally upregulate TP53 signaling and apoptosis at lower levels of DNA-damage than other species , but are resistant to nutlin 3 a induced apoptosis . To test if TP53RTG genes are necessary for the enhanced TP53-dependent DNA-damage response , we cotransfected African elephant fibroblasts with the pGL4 . 38[luc2p/p53 RE/Hygro] luciferase reporter vector , the pGL4 . 74[hRluc/TK] Renilla control vector , and either a TP53RTG-specific siRNA or a scrambled siRNA control ( Figure 7A ) . Next we used a dual luciferase reporter assay to measure the activation of the TP53 pathway in response to treatment mitomycin C , doxorubicin , UV-C , or nutlin-3a . As expected given our previous results , we found that African elephant fibroblasts transfected with control siRNA induced TP53 signaling in response to each treatment ( Figure 7A ) . In contrast , African elephant fibroblasts transfected with TP53RTG-specific siRNA had significantly lower luciferase expression , and thus reduced TP53 signaling , in response to either DNA-damaging agents ( mitomycin C , doxorubicin , UV-C ) or MDM2 antagonism ( nutlin-3a ) . TP53RTG knockdown also elevated baseline TP53 signaling ( Figure 7B ) . These data suggest that TP53RTG proteins have at least two distinct functions , inhibiting TP53 signaling in the absence of inductive signals and potentiation of TP53 signaling after the induction of DNA damage . 10 . 7554/eLife . 11994 . 014Figure 7 . TP53RTG genes are required for enhanced TP53 signaling and DNA-damage responses . ( A ) Expression of TP53RTG and TP53 transcripts in African elephant fibroblasts treated with an siRNA to knockdown the expression of TP53RTG genes ( red ) or a scrambled ( Control ) siRNA ( blue ) . Results are shown as fold-change in TP53RTG and TP53 transcript abundance relative to transcript abundance in scrambled siRNA control cells . The TP53RTG siRNA efficiently reduces the expression of TP53RTG transcripts , but does not reduce the expression of TP53 transcripts . ( B ) Relative luciferase ( Luc . ) expression in African elephant fibroblasts transfected with the pGL4 . 38[luc2p/p53 RE/Hygro] reporter vector and treated with an siRNA to knockdown the expression of TP53RTG genes ( red ) or a scrambled ( negative control ) siRNA ( blue ) , and treated with either mitomycin c , doxorubicin , nutlin-3a , or UV-C . Data is shown as fold difference in Luc . expression 18 hr after treatment standardized to Renilla controls and no treatment . n > 4 , mean±SD . , Wilcoxon p>0 . 01 . * , Wilcoxon p>0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 014 To test if TP53RTG12 is sufficient to mediate enhanced TP53 signaling and DNA-damage responses we synthesized the African elephant TP53RTG12 gene ( with mouse codon usage ) and cloned it into the mammalian expression vector pcDNA3 . 1 ( + ) /myc-His . We then transiently transfected mouse 3T3-L1 cells with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and used the pGL4 . 38[luc2P/p53 RE/Hygro] reporter system and ApoToxGlo assays to monitor activation of the TP53 signaling pathway and the induction of apoptosis in response to treatment with mitomycin C , doxorubicin , nutlin-3a , or UV-C . We found that heterologous expression of TP53RTG12 in mouse 3T3-L1 cells dramatically augmented luciferase expression from the pGL4 . 38[luc2P/p53 RE/Hygro] reporter vector in response to each treatment ( Figure 8A ) compared to empty vector controls , consistent with a enhancement of the endogenous TP53 signaling pathway . Similarly , expression of TP53RTG12 significantly augmented the induction of apoptosis in response to each treatment although the effect sizes were modest ( Figure 8B ) . These data indicate that TP53RTG12 acts via a trans-dominant mechanism to enhance the induction of apoptosis by endogenous TP53 and that TP53RTG12 is sufficient to recapitulate at least some of the enhanced sensitivity of elephant cells to DNA damage . Furthermore , our observation that transfection with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector augments TP53 signaling and apoptosis suggests that the TP53RTG12 protein rather than transcript is responsible for these effects because the TP53RTG12 transgene was synthetized with mouse codon usage and is only 73% ( 394/535 nts ) identical to the elephant TP53RTG12 gene . 10 . 7554/eLife . 11994 . 015Figure 8 . TP53RTG12 enhances TP53 signaling and DNA-damage responses . ( A ) Relative luciferase ( Luc . ) expression in mouse 3T3-L1 cells co-transfected with either the pGL4 . 38[luc2P/p53 RE/Hygro] Luc . reporter vector , TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector , or empty pcDNA3 . 1 ( + ) /myc-His and treated with either mitomycin c , doxorubicin , nutlin-3a , or UV-C . Data is shown as fold difference in Luc . expression 18 hr after treatment standardized to cells transfected with only pGL4 . 38[luc2P/p53 RE/Hygro] and Renilla controls . n = 12 , mean±SD . * , Wilcoxon p>0 . 001 . ( B ) Relative capsase-3/7 ( Cas3/7 ) activity in mouse 3T3-L1 cells transfected with either the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector or empty pcDNA3 . 1 ( + ) /myc-His and treated with either mitomycin c , doxorubicin , nutlin-3a , or UV-C . Data is shown as fold difference in Cas3/7 activity 18 hr after treatment standardized to mock transfected . n = 12 , mean±SD . * , Wilcoxon p>0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 015 TP53RTG proteins are unlikely to directly regulate TP53 target genes because they lack critical residues required for nuclear localization , tetramerization , and DNA-binding ( Figure 9A ) . Previous studies , for example , have shown the TP53 mutants lacking the tetramerization domain and C-terminal tail are unable to bind DNA or transactivate luciferase expression from a reporter vector containing TP53 response elements ( Kim et al . , 2012 ) . Similarly the p53Ψ isoform , which is truncated in the middle of the DNA binding domain and lacks the nuclear localization signal and oligomerization domain , is unable to bind DNA and is transcriptionally inactive ( Senturk et al . , 2014 ) . Unexpectedly , we found that the GFP-tagged TP53RTG12 protein was both cytoplasmic and nuclear localized in transfected African elephant fibroblasts ( Figure 9B ) , suggesting it interacts with another nuclear localized protein to enter the nucleus . Despite relatively strong nuclear localization , however , cells co-transfected with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and the pGL4 . 38[luc2P/p53 RE/Hygro] luciferase reporter vector did not have elevated luciferase expression compared to controls , suggesting that TP53RTG12 is transcriptionally inactive or requires a cofactor to regulate target genes ( Figure 9C ) . 10 . 7554/eLife . 11994 . 016Figure 9 . TP53RTG proteins are unlikely to directly regulate TP53 target genes . ( A ) Domain structure of TP53 and TP53RTG proteins . ( B ) Localization of TP53RTG12-GFP in African elephant fibroblasts . ( C ) Relative luciferase ( Luc . ) expression in African elephant fibroblasts transfected with co-transfected with either the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and the pGL4 . 38[luc2P/p53 RE/Hygro] luciferase reporter vector , or empty pcDNA3 . 1 ( + ) /myc-His and the pGL4 . 38[luc2P/p53 RE/Hygro] luciferase reporter vector . Data are shown as fold difference in Luc . expression 48 hr after transfection standardized to Renilla and cells transfected with pcDNA3 . 1 ( + ) /myc-His and pGL4 . 38[luc2P/p53 RE/Hygro] . n = 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 016 While TP53RTG12 does not appear to directly regulate gene expression , many of the TP53RTG proteins ( including TP53RTG12 ) retain the MDM2 interaction motif in the transactivation domain and dimerization sites in the DNA binding domain ( Figure 9A ) . These data suggest at least two non-exclusive models of TP53RTG action: ( 1 ) TP53RTG proteins may act as ‘decoys’ for the MDM2 complex allowing the canonical TP53 protein to escape negative regulation ( Figure 10A ) ; and ( 2 ) TP53RTG proteins may protect canonical TP53 from MDM2 mediated ubiqutination , which requires tetramerization ( Kubbutat et al . , 1998; Maki , 1999 ) , by dimerizing with canonical TP53 and thereby preventing the formation of tetramers ( Figure 10F ) . 10 . 7554/eLife . 11994 . 017Figure 10 . TP53RTG12 interacts with TP53 but not MDM2 . ( A ) Decoy model of TP53RTG12 function . Under normal conditions TP53 is negatively regulated by the MDM complex which ubiquitinates TP53 tetramers leading to proteosomal degradation . Upon DNA damage TP53 is phosphorylated , preventing interaction with the MDM complex and activating downstream TP53 signaling . In elephant cells TP53RTG12 may dimerize with the MDM2 complex , allowing TP53 to escape negative regulation . ( B ) Logo of the MDM2 interaction motif from 99 mammalian TP53 proteins ( upper ) and TP53RTG proteins ( lower ) . ( C ) Structure model of the MDM2/TP53 ( left ) and MDM2/TP53RTG12 ( right ) interaction . MDM2 is shown in tan with hydrophobic residues that mediate the interaction with TP53 as spheres , TP53 in gray with W23 as a sphere , and TP53RTG12 shown in blue with G23 as a sphere . ( D ) Predicted effects of the W23G substitution ( ΔΔG ) on the stability of the MDM2/TP53RTG12 interaction estimated with mCSM , SDM , and DUET . ( E ) HEK-293 cells were transiently transfected with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and total cell protein immunoprecipitated with an α-MDM2 antibody . Co-immunoprecipitation of Myc-tagged TP53RTG12 and TP53 were assayed by Western blotting with α-Myc , α-TP53 , and α-MDM2 antibodies after chemically stripping the blot , respectively . Although TP53 was co-immunoprecipitated , Myc-tagged TP53RTG12 was not . ( F ) Rather than interfering with the interaction between the MDM complex and TP53 as in the ‘Decoy” model , TP53RTG12 may dimerize with canonical TP53 and block formation of TP53 tetramers . These TP53RTG12/TP53 dimers cannot be ubiquitinated generating a pool of TP53 proteins to rapidly respond to lower levels or DNA damage and stress than other species . ( G ) Logo of the dimerization domain of TP53 from 99 mammals ( upper ) and TP53RTG proteins ( lower ) . ( H ) Model of the TP53/TP53RTG12 interaction . Residues critical for dimerization are shown as spheres , note sites involved in dimerization are conserved in TP53RTG proteins . ( I ) Predicted effects of the W23G substitution ( ΔΔG ) on the stability of the MDM2/TP53RTG12 interaction estimated with mCSM , SDM , and DUET . ( J ) HEK-293 cells were transiently transfected with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and total cell protein immunoprecipitated with an α-Myc antibody . Co-immunoprecipitation of TP53 and MDM2 were assayed by serial Western blotting with α-Myc , α-TP53 , and α-MDM2 antibodies after chemically stripping the blot , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 01710 . 7554/eLife . 11994 . 018Figure 10—figure supplement 1 . Uncropped Western blots shown in Figure 8B . HEK-293 cells were transiently transfected with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector and total cell protein immunoprecipitated with either an α-Myc ( IP: RTG12-Myc ) or α-MDM2 ( IP: MDM2 ) antibody . Co-immunoprecipitation of Myc-tagged TP53RTG12 , TP53 , or MDM2 were assayed by Western blotting with α-Myc , α-TP53 , and α-MDM2 antibodies respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11994 . 018 The decoy model depends on the ability of TP53RTG proteins to physically interact with MDM2 . Previous crystallographic studies of the TP53/MDM2 interaction have shown that a trio of residues in TP53 ( F19 , W23 , and L26 ) insert deeply into a hydrophobic cleft in MDM2 , which stabilizes the interaction ( Kussie et al . , 1996 ) . We identified a W23G substitution in all TP53RTG proteins at a site that is invariant for tryptophan in TP53 proteins including African and Asian elephant TP53 ( Figure 10B ) , suggesting that TP53RTG proteins may be unable to physically interact with MDM2 . To infer the structural and functional consequences of the TP53RTG W23G substitution we generated a homology model of the elephant TP53RTG12/MDM2 complex using I-TASSER/ModRefiner ( Roy et al . , 2010; Xu and Zhang , 2011; Zhang , 2008 ) and the crystal structure of the MDM2/TP53 dimer as a template ( Kussie et al . , 1996 ) . We found that the TP53RTG12 transactivation domain was inferred to be a short α-helix ( Figure 10C ) and was very similar to the template structure ( RMSD: 1 . 756 ) , however , the W23G substitution is predicted to abolish crucial hydrophobic interactions between the amphipathic α-helix of TP53 and the hydrophobic cleft MDM2 . Indeed , three methods ( Pires et al . , 2014 ) inferred that the W23G substitution is destabilizing on the TP53RTG12/MDM2 interaction ( mCSM ΔΔG = −2 . 42 , SDM ΔΔG = −5 . 46 , DUET ΔΔG = −2 . 38 ) ( Figure 10D ) . To experimentally test for an interaction between TP53RTG12 and MDM2 we transiently transfected HEK-293 cells with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector , immunoprecipitated endogenous human MDM2 , and assayed for co-immunoprecipitation of TP53RTG12 by Western blotting . While we efficiently co-immunoprecipitated endogenous human TP53 we did not co-immunoprecipitate Myc-tagged TP53RTG12 ( Figure 10Eand Figure 10—figure supplement 1 ) , consistent with a lack of interaction between TP53RTG12 and MDM2 . Unlike the decoy model , the guardian model of TP53RTG function depends upon a physical interaction between TP53RTG and TP53 . The TP53 dimer is stabilized by hydrophobic and polar interactions including a shell of nonpolar interactions formed by P177 , H178 , M243 , and G244 and a stabilization network next to the nonpolar layer formed by charged residues from the two monomers ( R181 , E180 , and R174 ) . In addition , several polar and charged residues nearby but not within the dimerization interface contribute to the stability of the interacting monomers including D184 with R175 , most of which are conserved in TP53RTG proteins ( Figure 10G ) . To infer if derived residues in the TP53RTG12 dimerization interface might disrupt a physical interaction between TP53RTG12 and TP53 we generated a homology model of the TP53RTG12/TP53 dimer using I-TASSER/ModRefiner ( Roy et al . , 2010; Xu and Zhang , 2011; Zhang , 2008 ) and the crystal structure of the TP53 tetramer as a template ( Kitayner et al . , 2006 ) . We found that the TP53RTG12 dimerization interface was inferred to be a 2–4 residue α-helix ( Figure 10H ) that was nearly identical to the template structure ( RMSD: 0 . 605 ) , suggesting TP53RTG12-specific residues in dimerization interface are unlikely to disrupt the structure of the interface . Unlike the MDM2 interaction site , TP53RTG12-specific substitutions were predicted to maintain intermolecular hydrophobic interactions with TP53 ( Figure 10H ) . Consistent with maintenance of dimerization potential , the net ΔΔG of the derived amino acid substitutions in the TP53RTG12 dimerization interface were under 2 ( Figure 10I ) . To experimentally test for an interaction between TP53RTG12 and TP53 we transiently transfected HEK-293 cells with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector , immunoprecipitated TP53RTG12 with a Myc antibody , and assayed for co-immunoprecipitation of endogenous human TP53 by Western blotting . We found that Myc-tagged TP53RTG12 efficiently co-immunoprecipitated endogenous human TP53 ( Figure 10J ) . These data are consistent with a physical interaction between TP53RTG12 and TP53 but not between TP53RTG12 and MDM2 , supporting the guardian model . Previous studies have suggested that the TP53 gene family expanded in the elephant lineage ( Abegglen et al . , 2015; Caulin and Maley , 2011; Caulin et al . , 2015 ) , however , these studies did not establish the mechanism by which the TP52RTG gene family expanded . Several potential mechanisms could have increased TP52RTG copy number , including serial ( independent ) retrotranspostion from the parent TP53 gene , a single retrotranspostion event followed by repeated rounds of segmental duplication of TP52RTG containing loci , retrotransposition of expressed TP52RTG genes , or some combination of these models . Each model is associated with a distinct set of genomic ‘fingerprints’ . If copy number expanded through independent retrotransposition events , for example , TP53RTG encoding regions of the genome will not be homologous whereas the model of a single retrotranspostion event followed by repeated rounds of segmental duplication predicts that the TP53RTG encoding loci will be homolgous . Consistent with copy number expansion through a single retrotransposition event followed by repeated rounds of segmental duplication , we found that flaking regions of each TP52RTG locus were homologous and contained the same unique combination of transposable elements . Indeed , the 3’-end of each duplicate terminates at a ~5 kb long L1MB5 LINE element suggesting that transposable element mediated recombination may have played a role in promoting segmental duplication . If TP53RTG copy number expansion played a causal role in evolution of enhanced cancer resistance in elephants then the gene family must have expanded prior to or coincident with the evolution of increased body sizes in Proboscideans rather than after the evolution of large bodies but before the African and Asian elephant lineages diverged ~8 MYA ( Rohland et al . , 2010 , 2007 ) . Thus dating the expansion of the TP53RTG gene family is essential for determining if TP53RTG genes played a role in the evolution of enhanced cancer resistance in elephants . Previous studies , however , did not establish when TP53RTG copy number expanded in the evolution of Proboscideans . Fortunately our observation that copy number expansion occurred through segmental duplications allowed us to use molecular phylogenetic methods to date each duplication event . These data indicate that the initial TP53RTG retrotransposition event occurred in the Paenungulate stem-lineage ~ 64 MYA , followed the rapid expansion after ~40 MYA . We also found that the increase in TP53RTG copy number occurred coincident with the evolution of large bodies in the Proboscidean lineage , implicating copy number expansion in the resolution of Peto’s paradox . While Abegglen et al . ( 2015 ) found that elephant and human cells had different sensitivities to ionizing radiation , their taxon sampling did not allow for polarizing which species was different . Indeed , their taxon sampling does not allow for a Eutherian out-group . To determine if elephants have a derived sensitivity to genotoxic stress , we compared the response of fibroblasts from an African savannah elephant , an Asian elephant , and closely related out-group species to DNA damage inducing agents . Our out-group species included the South African Rock hyrax , the closest living relative of elephants , the East African aardvark , an Afrotherian from the sister lineage of the Paenungulates , and the Southern Three-banded armadillo , an Atlantogenatan from the sister lineage of the Afrotherians . Our results indicate that elephant cells are particularly sensitive to genotoxic tress , suggesting that this sensitivity evolved coincident with the evolution of large body sizes and an expanded TP53 gene repertoire in Proboscideans . Our observation that the elephant genome contains 19 TP53RTG genes raises numerous questions: Do these loci , for example , encode functional genes or pseudogenes and what processes underlie copy number expansion ? Answers to these questions are essential for understanding whether TP53RTG genes are casually related to the resolution of Peto’s paradox in Proboscideans or if they are irrelevant relicts of ancient transposition events , like so many other pseudogenes that riddle mammalian genomes . Unfortunately , it is difficult to answer these questions . Which TP53RTG loci encode functional genes and which encode pseudogenes is not easy to infer . Many duplicate genes are preserved because they evolve tissue-specific or developmental-stage specific expression patterns ( subfunctionalization ) , new functions ( neofunctionalization ) that resolve redundancy between duplicates , or reduced expression levels that preserves correct expression dosage . Exhaustively characterizing gene expression in elephant tissues is difficult because appropriate tissue samples are unavailable , therefore we are unable to definitively determine which TP53RTG loci are transcribed . Our RNA-Seq and RT-PCR/Sanger sequencing data indicate that at least TP53RTG12 , TP53RTG18/19 , and TP53RTG13 are transcribed in dermal fibroblasts . Abegglen et al . ( 2015 ) used RT-PCR and Sanger sequencing to show two distinct transcripts were expressed in elephant PBMCs , but they did not assign the loci to which these transcripts correspond . We analyzed the chromatograms shown in Abegglen et al . Figure 4 and found that the 185 bp product is likely a transcript from the TP553RTG14 gene and the 201 bp product is likely a transcript from the TP553RTG5 gene . Thus , our combined data suggest that at least five TP53RTG genes are transcribed . Furthermore we did not observe TP553RTG14 or TP553RTG5 expression in adipose , placenta , or fibroblasts suggesting that the expression of some TP53RTG genes is tissue-specific . The large number of TP53RTG loci in elephants combined with the observation that only five are transcribed suggests that this gene family may evolve by a birth and death process , in which new genes are created by duplication and some duplicates are maintained in the genome whereas others become nonfunctional or deleted similar to other large genes families such as histones ( González-Romero et al . , 2010; Rooney et al . , 2002 ) and venom genes ( Lynch , 2007 ) . Under this model selection acts to maintain a minimal number of functional copies ( functional copy number ) rather than total copy number , the increase in total copy number is driven by the total number of loci and the rates of duplication , loss , and fixation . Thus the overall increase in total TP53RTG copy number may be a selectively neutral process , driven simply by higher rates of duplication and/or fixation than loss . Our results demonstrate that elephant cells induce TP53 signaling and trigger apoptosis at lower thresholds of genotoxic stress than closely related species without an expanded TP53 repertoire and that this reduced sensitivity is dependent upon TP53RTG genes . Furthermore , heterologous expression of TP53RTG12 in mouse cells was sufficient to augment endogenous TP53 signaling and recapitulate an elephant-like sensitivity to genotoxic stress , indicating that TP53RTGs acts through a transdominant mechanism . While the mechanism of action is unclear , TP53RTG genes may augment TP53 signaling through several non-exclusive mechanisms including functioning as non-coding RNAs ( Poliseno et al . , 2010 ) , protein ‘decoys’ for the MDM2 complex that allowing canonical TP53 to escape negative regulation ( Abegglen et al . , 2015 ) , and protein ‘guardians’ that protect canonical TP53 from MDM2 mediated ubiqutination . Consistent with the guardian model , we found that Myc-tagged TP53RTG12 efficiently co-immunoprecipitated with endogenous TP53 but not with endogenous MDM2 . The lack of an interaction between TP53RTG12 and MDM2 likely results from a W23G substitution that is predicted to abolish a crucial hydrophobic interaction between the amphipathic α-helix of TP53 and the hydrophobic cleft MDM2 . The W23G substitution is found in all TP53RTG proteins but not African and Asian elephant TP53 , indicating that it occurred before the expansion of the TP53RTG gene family and likely prevents interaction of any TP53RTG protein and MDM2 . Previous studies have shown that tetramerization of TP53 is required for its efficient MDM2-mediated ubiquitination ( Kubbutat et al . , 1998; Maki , 1999 ) , suggesting that TP53RTG proteins may dimerize with and protect TP53 from ubiquitination thereby contributing to a standing pool of TP53 that is able to rapidly respond to DNA damage . Consistent with this mechanism , transgenic mice with an increase in TP53 copy number ( García-Cao et al . , 2002 ) or a hypomorphic Mdm2 allele have elevated basal TP53 activity and are resistant to tumor formation ( Mendrysa et al . , 2006 ) , indicating that shifting the TP53-MDM2 equilibrium away from TP53 degradation can directly promote cancer resistance . The ‘decoy’ model ( Abegglen et al . , 2015 ) has also been challenged because it would allow for activation of the TP53 signaling pathway in the absence of DNA damage ( Perez and Komiya , 2016 ) , which is generally lethal in animal models ( Hoever et al . , 1994; Lozano , 2010 ) . Thus if TP53RTG proteins allow TP53 to escape negative regulation by MDM2 , how do elephants tolerate elevated basal TP53 levels ? Clearly further studies are required to specifically test whether TP53RTG proteins interfere with the interaction between the MDM2 complex and TP53 , protect TP53 from ubiquitination , or have other functions . Our observation reveals that functional TP53 duplicates only occur in the elephant lineage ( and perhaps some bats ) suggests that increased TP53 dosage has a cost . Previous studies found that transgenic mice that overexpress Trp53 were cancer resistant but had major life history tradeoffs including slower pre- and post-natal growth rates and reduced size ( Maier et al . , 2004 ) , a shortened lifespan ( Maier et al . , 2004 ) , accelerated aging ( Tyner et al . , 2002 ) , and reduced fertility ( Allemand et al . , 1999; Maier et al . , 2004 ) , as well as developmental tradeoffs including reduced proliferation , cellularity , and atrophy across multiple organ and tissue systems ( Dumble et al . , 2007; Maier et al . , 2004 ) , defective ureteric bud differentiation , and small kidneys ( Godley et al . , 1996 ) . Thus , increases in TP53 copy number protects against cancer but appears to come with the cost of developmental delays , accelerated aging , and reduced fertility ( Campisi , 2003; Donehower , 2002; Ferbeyre and Lowe , 2002; Rodier et al . , 2007 ) . Reduced male fertility appears to be a particularly expensive cost of increased TP53 dosage . Allemand et al . , 1999 ) , for example , generated transgenic lines of mice with one ( MTp53-176 ) , two ( MTp53-112 ) , or 15 ( MTp53-94 ) extra copies of the Trp53 gene fused to the inducible promoter of the metallothionein I ( MT ) gene . They found that transgenic males with the highest Trp53 dosage were nearly infertile because the majority of developing spermatids underwent apoptosis before developing into mature sperm , males with intermediate dosage were subfertile and produced sperm with abnormal morphologies ( teratozoospermia ) indicative of defective terminal differentiation of postmeiotic cells , whereas males with the lowest dosage were fertile . Similarly , Maier et al . ( 2004 ) generated a transgenic line that overexpresses the p44 isoform of Trp53 ( Δ40 p53 ) and found that both males and females exhibited shortened reproductive lifespans . Males , however , were more severely affected than females with a catastrophic loss of sperm-producing cells and massive degeneration of the seminiferous epithelium leading to a 'Sertoli-cell only' phenotype . In contrast to transgenic mice that either overexpress full length TP53 or the Δ40 p53 isoform , transgenic 'super p53' mice with one ( p53-tg ) or two ( p53-tgb ) additional copies of the endogenous TP53 locus have an enhanced DNA-damage response and are tumor resistant , yet age normally and are fertile ( García-Cao et al . , 2002 , 2006; Matheu et al . , 2007 ) . Enhanced tumor suppression and normal aging is also observed in Mdm2puro/Δ7–12 transgenic mice , which have one hypomorphic and one null allele of Mdm2 , express ∼30% of the wild-type level of Mdm2 , and have constitutively high TP53 activity ( Mendrysa et al . , 2006 , 2003 ) . These results suggest that the costs of increased TP53 copy number are incurred above a threshold of about 2–3 extra copies and can be reduced by maintaining normal regulation of TP53 transcription and negative post-transcriptional regulation by MDM2 . Consistent with this model , deletion of either Mdm2 or Mdm4 rescues Trp53-dependent embryonic lethality in mice ( Finch et al . , 2002; Migliorini et al . , 2002; Parant et al . , 2001 ) and zebrafish ( Chua et al . , 2015 ) . Collectively , these data indicate that increased TP53 dosage comes with a cost , and suggest that Proboscideans evolved a mechanism that reduced this cost , broke a major developmental and evolutionary constraint on TP53 copy number , or perhaps evaded paying the cost of increased TP53 copy number altogether . TP53RTG genes , for example , are transcribed from a transposable element derived promoter that is evolutionarily younger than the retrogenes . Thus , the initial TP53RTG genes were unlikely to be transcribed and incur a cost . Similarly , our phylogenetic analyses indicates that copy number expanded after the initial TP53RTG gene acquired several mutations , including a premature stop codon that terminates the protein before the DNA-binding domain , which likely reduces or completely eliminates the costs associated with TP53 target gene regulation . Although more detailed evolutionary and comparative analyses are required to determine if TP53RTG genes incurred a cost , it is possible that the costs were minimized because functional TP53RTG genes evolved through non-functional intermediates , which accumulated loss of function mutations that minimized redundancy with TP53 . We used BLAT to search for TP53 genes in 61 Sarcopterygian genomes using the human TP53 protein sequences as an initial query . After identifying the canonical TP53 gene from each species , we used the nucleotide sequences corresponding to this TP53 CDS as the query sequence for additional BLAT searches within that species genome . To further confirm the orthology of each TP53 gene we used a reciprocal best BLAT approach , sequentially using the putative CDS of each TP53 gene as a query against the human genome; in each case the query gene was identified as TP53 . Finally , we used the putative amino acid sequence of the TP53 protein as a query sequence in a BLAT search . We thus used BLAT to characterize the TP53 copy number in Human ( Homo sapiens; GRCh37/hg19 ) , Chimp ( Pan troglodytes; CSAC 2 . 1 . 4/panTro4 ) , Gorilla ( Gorilla gorilla gorilla; gorGor3 . 1/gorGor3 ) , Orangutan ( Pongo pygmaeus abelii; WUGSC 2 . 0 . 2/ponAbe2 ) , Gibbon ( Nomascus leucogenys; GGSC Nleu3 . 0/nomLeu3 ) , Rhesus ( Macaca mulatta; BGI CR_1 . 0/rheMac3 ) , Baboon ( Papio hamadryas; Baylor Pham_1 . 0/papHam1 ) , Marmoset ( Callithrix jacchus; WUGSC 3 . 2/calJac3 ) , Squirrel monkey ( Saimiri boliviensis; Broad/saiBol1 ) , Tarsier ( Tarsius syrichta; Tarsius_syrichta2 . 0 . 1/tarSyr2 ) , Bushbaby ( Otolemur garnettii; Broad/otoGar3 ) , Mouse lemur ( Microcebus murinus; Broad/micMur1 ) , Chinese tree shrew ( Tupaia chinensis; TupChi_1 . 0/tupChi1 ) , Squirrel ( Spermophilus tridecemlineatus; Broad/speTri2 ) , Mouse ( Mus musculus; GRCm38/mm10 ) , Rat ( Rattus norvegicus; RGSC 5 . 0/rn5 ) , Naked mole-rat ( Heterocephalus glaber; Broad HetGla_female_1 . 0/hetGla2 ) , Guinea pig ( Cavia porcellus; Broad/cavPor3 ) , Rabbit ( Oryctolagus cuniculus; Broad/oryCun2 ) , Pika ( Ochotona princeps; OchPri3 . 0/ochPri3 ) , Kangaroo rat ( Dipodomys ordii; Broad/dipOrd1 ) , Chinese hamster ( Cricetulus griseus; C_griseus_v1 . 0/criGri1 ) , Pig ( Sus scrofa; SGSC Sscrofa10 . 2/susScr3 ) , Alpaca ( Vicugna pacos; Vicugna_pacos-2 . 0 . 1/vicPac2 ) , Dolphin ( Tursiops truncatus; Baylor Ttru_1 . 4/turTru2 ) , Cow ( Bos taurus; Baylor Btau_4 . 6 . 1/bosTau7 ) , Sheep ( Ovis aries; ISGC Oar_v3 . 1/oviAri3 ) , Horse ( Equus caballus; Broad/equCab2 ) , White rhinoceros ( Ceratotherium simum; CerSimSim1 . 0/cerSim1 ) , Cat ( Felis catus; ICGSC Felis_catus 6 . 2/felCat5 ) , Dog ( Canis lupus familiaris; Broad CanFam3 . 1/canFam3 ) , Ferret ( Mustela putorius furo; MusPutFur1 . 0/musFur1 ) , Panda ( Ailuropoda melanoleuca; BGI-Shenzhen 1 . 0/ailMel1 ) , Megabat ( Pteropus vampyrus; Broad/pteVam1 ) , Microbat ( Myotis lucifugus; Broad Institute Myoluc2 . 0/myoLuc2 ) , Hedgehog ( Erinaceus europaeus; EriEur2 . 0/eriEur2 ) , Shrew ( Sorex araneus; Broad/sorAra2 ) , Minke whale ( Balaenoptera acutorostrata scammoni; balAcu1 ) , Bowhead Whale ( Balaena mysticetus; v1 . 0 ) , Rock hyrax ( Procavia capensis; Broad/proCap1 ) , Sloth ( Choloepus hoffmanni; Broad/choHof1 ) , Elephant ( Loxodonta africana; Broad/loxAfr3 ) , Cape elephant shrew ( Elephantulus edwardii; EleEdw1 . 0/eleEdw1 ) , Manatee ( Trichechus manatus latirostris; Broad v1 . 0/triMan1 ) , Tenrec ( Echinops telfairi; Broad/echTel2 ) , Aardvark ( Orycteropus afer afer; OryAfe1 . 0/oryAfe1 ) , Armadillo ( Dasypus novemcinctus; Baylor/dasNov3 ) , Opossum ( Monodelphis domestica; Broad/monDom5 ) , Tasmanian devil ( Sarcophilus harrisii; WTSI Devil_ref v7 . 0/sarHar1 ) , Wallaby ( Macropus eugenii; TWGS Meug_1 . 1/macEug2 ) , Platypus ( Ornithorhynchus anatinus; WUGSC 5 . 0 . 1/ornAna1 ) , Medium ground finch ( Geospiza fortis; GeoFor_1 . 0/geoFor1 ) , Zebra finch ( Taeniopygia guttata; WashU taeGut324/taeGut2 ) , Budgerigar ( Melopsittacus undulatus; WUSTL v6 . 3/melUnd1 ) , Chicken ( Gallus gallus; ICGSC Gallus_gallus-4 . 0/galGal4 ) , Turkey ( Meleagris gallopavo; TGC Turkey_2 . 01/melGal1 ) , American alligator ( Alligator mississippiensis; allMis0 . 2/allMis1 ) , Painted turtle ( Chrysemys picta bellii; v3 . 0 . 1/chrPic1 ) , Lizard ( Anolis carolinensis; Broad AnoCar2 . 0/anoCar2 ) , X . tropicalis ( Xenopus tropicalis; JGI 7 . 0/xenTro7 ) , Coelacanth ( Latimeria chalumnae; Broad/latCha1 ) . Relative divergence ( duplication ) times of the TP53 retrogenes were estimated using the TP53 alignment described above and BEAST ( v1 . 7 . 4 ) ( Rohland et al . , 2010 ) . We used the general time reversible model ( GTR ) , empirical nucleotide frequencies ( +F ) , a proportion of invariable sites estimated from the data ( +I ) , four gamma distributed rate categories ( +G ) , an uncorrelated lognormal relaxed molecular clock to model substitution rate variation across lineages , a Yule speciation tree prior , uniform priors for the GTR substitution parameters , gamma shape parameter , proportion of invariant sites parameter , and nucleotide frequency parameter . We used an Unweighted Pair Group Arithmetic Mean ( UPGMA ) starting tree . To obtain posterior distributions of estimated divergence times , we use five node calibrations modeled as normal priors ( standard deviation = 1 ) to constrain the age of the root nodes for the Eutheria ( 104 . 7 MYA ) , Laurasiatheria ( 87 . 2 MYA ) , Boreoeutherian ( 92 . 4 MYA ) , Atlantogenatan ( 103 MYA ) , and Paenungulata ( 64 . 2 MYA ) ; divergence dates were obtained from www . timetree . org using the ‘Expert Result’ divergence dates . The analysis was run for 5 million generations and sampled every 1000 generations with a burn-in of 1 million generations; convergence was assessed using Tracer , which indicated convergence was reached rapidly ( within 100 , 000 generations ) . Proboscidean body size data were obtained from previously published studies on mammalian body size evolution ( Evans et al . , 2012 ) . We also observed that the genomic region surrounding each TP53RTG gene contained blocks of homolgous transposable element insertions , suggesting that these regions are segmental duplications . To confirm this observation , we used MUSCLE ( Edgar , 2004 ) to align an approximately 20 kb region surrounding each TP53RTG gene and found that conservation within this region was very high , again suggesting these regions are relatively recent segmental duplications . To identify if the contigs on which TP53RTG genes are located contained locally collinear blocks ( LCBs ) , as expected for segmental duplications , we aligned contigs using progressiveMAUVE ( Darling et al . , 2004 ) as implemented in Genious ( v6 . 1 . 2 ) . We generated a dataset of TP53 orthologs from 65 diverse mammals identified from GenBank , and included the TP53 genes and retrogenes we identified from the African elephant , hyrax , manatee , tenrec , cape elephant shrew , and armadillo genomes . Nucleotide sequences were aligned using the MAFFT algorithm ( Katoh and Standley , 2014 ) and the FFT refinement strategy implemented in the GUIDANCE webserver ( Penn et al . , 2010 ) . Alignment confidence was assessed with 100 bootstrap replicates and ambiguously aligned sites ( under the default GUIDANCE exclusion rule ) removed prior to phylogenetic analyses; lineage specific insertions and deletions were also removed prior to phylogenetic analyses . TP53 phylogenies were inferred using maximum likelihood implemented in PhyML ( v3 . 1 ) ( Guindon et al . , 2010 ) using a general time reversible model ( GTR ) , empirical nucleotide frequencies ( +F ) , a proportion of invariable sites estimated from the data ( +I ) , four gamma distributed rate categories ( +G ) , and using the best of NNI and SPR branch moves during the topology search . Branch supports were assessed using the aBayes , aLRT , SH-like , and Chi2-based ‘fast’ methods as well as 1000 bootstrap replicates ( Guindon et al . , 2010 ) . We also used MrBayes ( v3 . 2 . 2 x64 ) ( Huelsenbeck and Ronquist , 2001 ) for Bayesian tree inference using the GTR+I+G model; the analysis was run with two simultaneous runs of four chains for 4 million generations with the chains sampled every 1000 generations . We used Tracer ( v1 . 5 ) to assess when the chains reached stationarity , which occurred around generation 100 , 000 . At completion of the run the PRSF value was 1 . 00 indicating the analyses had reached convergence . To determine if TP53RTG genes were expressed , we generated RNA-Seq data from term placental villus and adipose tissue from African elephants ( Loxodonta africana ) and primary dermal fibroblasts from Asian elephants ( Elepas maximus ) . Briefly , sequencing libraries were prepared using standard Illumina protocols with poly ( A ) selection , and sequenced as 100 bp single-end reads on a HiSeq2000 . We also used previously published RNA-Seq data generated from primary fibroblasts ( isolated from ear clips ) from a male ( GSM1227965 ) and female ( GSM1227964 ) African elephant generated on an Illumina Genome Analyzer IIx ( 101 cycles , single end ) and a male ( GSM1278046 ) African elephant generated on an Illumina HiSeq 2000 ( 101 cycles , single end ) ( Cortez et al . , 2014 ) and combined these reads into a single dataset of 138 , 954 , 285 reads . Finally , we used previously published 100 bp single-end reads generated on a HiSeq2000 to identify TP53RTG transcripts from Asian elephant PBMCs ( SRX1423033 ) ( Reddy et al . , 2015 ) . Reads were aligned to a custom built elephant reference gene set generated by combining the sequences of the canonical TP53 gene and TP53RTG genes with the ENSEMBL African elephant ( Loxodonta africana ) CDS gene build ( Loxodonta_africana . loxAfr3 . 75 . cds . all . fa ) with Bowtie2 ( Langmead and Salzberg , 2012 ) . Bowtie two settings were: ( 1 ) both local alignment and end-to-end mapping; ( 2 ) preset option: sensitive; ( 3 ) Trim n-bases from 5' of each read: 0; and ( 4 ) Trim n-bases from 3' of each read: 0 . Transcript assembly and FPKM estimates were generated with Cufflinks ( version 0 . 0 . 7 ) ( Trapnell et al . , 2012 ) using aligned reads from Bowtie2 , non-default parameters included quartile normalization and multi-read correction . Finally , we transformed FPKM estimates into transcripts per million ( TPM ) , TPM= ( FPKM per gene/sum FPKM all genes ) x106 , and defined genes with TPM ≥ 2 as expressed ( Li et al . , 2010; Wagner et al . , 2012 , 2013 ) . The TP53RTG genes are 80 . 0–82 . 7% identical to TP53 at the nucleotide level , with 204–231 total nucleotide differences compared to TP53 . This level of divergence allows for many reads to be uniquely mapped to each gene , there will also be significant read mapping uncertainty in regions of these genes with few nucleotide differences . However , if read mapping uncertainty was leading to false positive mappings of TP53 derived reads to TP53RTG genes we would expect to observe the expression of many TP53RTG genes , rather the robust expression of a single ( TP53RTG12 ) gene . We also counted the number of uniquely mapped reads to each TP53RTG gene and TP53 . We found that 0–8 reads were uniquely mapped to most TP53RTG genes , except TP53RTG12 which had ~115 uniquely mapped reads across samples and TP53 which had ~3000 uniquely mapped reads across tissue samples . Thus we conclude that read mapping uncertainty has not adversely affected our RNA-Seq analyses . We further confirmed expression of TP53RTG12 transcripts in elephant cells through RT-PCR , taking advantage of differences between TP53RTG12 and TP53 sequences to design two aligned primer sets , one TP53-specific , the other TP53RTG12 specific . TP53RTG12 primers were: 5´ ggg gaa act cct tcc tga ga 3´ ( forward ) and 5´ cca gac aga aac gat agg tg 3´ ( reverse ) . TP53 primers were: 5´ atg gga act cct tcc tga ga 3´ ( forward ) and 5´ cca gac gga aac cat agg tg 3´ ( reverse ) . The TP53 amplicon is expected to be 251 bps in length , while deletions present in the TP53RTG12 sequence lead to a smaller projected amplicon size of 220 bps . Total RNA was extracted from cultured Loxodonta and Elephas fibroblasts ( RNAeasy Plus Mini kit , Qiagen ) , then DNase treated ( Turbo DNA-free kit , Ambion ) and reverse-transcribed using an olgio-dT primer for cDNA synthesis ( Maxima H Minus First Strand cDNA Synthesis kit , Thermo Scientific ) . Control RT reactions were otherwise processed identically , except for the omission of reverse transcriptase from the reaction mixture . RT products were PCR-amplified for 45 cycles of 94°/20 s , 56°/30 s , 72°/30 s using a BioRad CFX96 Real Time qPCR detection system and SYBR Green master mix ( QuantiTect , Qiagen ) . PCR products were electrophoresed on 3% agarose gels for 1 hr at 100 volts , stained with SYBR safe , and imaged in a digital gel box ( ChemiDoc MP , BioRad ) to visualize relative amplicon sizes . PCR products were also directly sequenced at the University of Chicago Genomics core facility , confirming projected product sizes and sequence identities . We used geneid v1 . 2 ( http://genome . crg . es/software/geneid/geneid . html ) to infer if the TP53RTG12 gene contained a non-coding exon 5’ to the predicted ATG start codon . For gene structure prediction we used the full-length scaffold_825 from African elephant ( Broad/loxAfr3 ) sequence and forced an internal exon where the TP53RTG12 gene is encoded in scaffold_825 . Geneid identified a putative exon 5’ to the TP53RTG12 gene from nucleotides 1761–1935 and an exon 3’ from nucleotides 6401–6776 . We also used GENESCAN ( http://genes . mit . edu/cgi-bin/genscanw_py . cgi ) to predict the location of exons in the full-length scaffold_825 sequence and identified a putative exon from nucleotides 1750–1986 . We next used Bowtie2 ( Langmead and Salzberg , 2012 ) to map African and Asian elephant fibroblast RNA-Seq data onto African elephant scaffold_825 with the default settings . The RTE1_LA non-LTR retrotransposon has previously been described from the African elephant genome , these elements are generally more than 90% identical to the consensus RTE1_LA sequence but less than 70% identical to other mammalian RTEs ( http://www . girinst . org/2006/vol6/issue3/RTE1_LA . html ) . These data suggest that the RTE_LA element has relatively recently expanded in the elephant genome . To determine the taxonomic distribution of the RTE1_LA element we used BLAT to search the lesser hedgehog tenrec ( Echinops telfairi ) , rock hyrax ( Procavia capensis ) , West Indian manatee ( Trichechus manatus ) , armadillo ( Dasypus novemcinctus ) , and sloth ( Choloepus hoffmanni ) genomes . We identified numerous copies of the RTE_LA element in the genome of the Afrotherians , but not the Xenarthran genomes . These data indicate that the RTE_LA element is Afrotherian-specific , rather than Elephant-specific . Elephant and Hyrax primary fibroblasts ( San Diego Zoo , 'Frozen Zoo' ) were grown to confluency in 10 cm dishes at 37°C/5% CO2 in a culture medium consisting of FGM/EMEM ( 1:1 ) supplemented with insulin , FGF , 6% FBS and Gentamicin/Amphotericin B ( FGM-2 , singlequots , Clonetics/Lonza ) . Culture medium was removed from dishes just prior to UV treatment and returned to cells shortly afterwards . Experimental cells were exposed to 50 J/m2UV-C radiation in a crosslinker ( Stratalinker 2400 , Stratagene ) , while control cells passed through media changes but were not exposed to UV . A small volume ( ~3 mL ) of PBS covered fibroblasts at the time of UV exposure . To inhibit TP53 proteolysis , MG-132 ( 10 µM ) was added to experimental cell medium 1 hr prior to UV exposure , and maintained until the time of cell lysis . 5 hr post-UV treatment , cells were briefly rinsed in PBS , then lysed and boiled in 2x SDS-PAGE sample buffer . Lysates were separated via SDS-PAGE on 10% gels for 1 hr at 140 volts , then electrophoretically transferred to PVDF membranes ( 1 hr at 85 volts ) . Membranes were blocked for 1 hr in 5% milk in TBST and incubated overnight at 4°C with rabbit polyclonal TP53 antibodies ( FL-393 , Santa Cruz Biotechnology , and ab131442 , Abcam ) . Blots were washed 3x in TBST , incubated with HRP-conjugated , anti-rabbit IgG 2° antibodies for 1 hr at RT , and washed four more times in TBST . Protein bands were visualized via enhanced chemiluminescence ( BioRad Clarity ) , and imaged in a digital gel box ( Chemidoc MP , BioRad ) . Western blots were replicated three independent times . Human HEK-293 cells were grown to 80% confluency in 20 cm dishes at 37°C/5% CO2 in a culture medium consisting of DMEM supplemented with 10% FBS . At 80% confluency , cells were transiently transfected with the TP53RTG12 pcDNA3 . 1 ( + ) /myc-His expression vector . After 16 hr the transfection media was removed and replaced with fresh DMEM , and the cells were incubated an additional 24 hr before harvesting . After removing DMEM and washing cells twice with PBS , 1 mL ice-cold lysis buffer ( 20 mM Tris , pH 8 . 0 , 40 mM KCl , 10 mM MgCl2 , 10% glycerol , 1% Triton X-100 , 1x Complete EDTA-free protease inhibitor cocktail ( Roche ) , 1x PhosSTOP ( Roche ) ) was added to each plate and cells were harvested by scraping with a rubber spatula . Cells were then incubated on ice for 30 min in 420 mM NaCl . The whole cell lysate was cleared by centrifugation at 10 , 000 rpm for 30 min at 4°C , and the supernatant was transferred to a clean microfuge tube . After equilibrating protein concentrations , 1 mL of sample was mixed with 40 mL of α-MDM2 or α-Myc antibody conjugated agarose beads ( Sigma ) pre-washed with TNT buffer ( 50 mM Tris-HCl , pH 7 . 5 , 150 mM NaCl , 0 . 05% Triton X-100 ) , and rotated overnight at 4°C . The following day , samples were treated with 50 U DNase ( Roche ) and 2 . 5 μg RNase ( Roche ) for 60 min at room temperature , as indicated . Samples were washed 3x with 1 mL wash buffer ( 150 mM NaCl , 0 . 5% Triton X-100 ) . After the final wash , agarose beads were resuspended in elution buffer ( 500 mM Tris pH 7 . 5 , 1 M NaCl ) , and boiled to elute immunoprecipitated complexes . Eluted protein was run on Bis-tris gels , probed with antibodies and visualized by Chemi-luminescence . Serial Westerns were performed for each antibody following chemical stripping and re-blocking . Antibodies were from Santa Cruz: MDM2 ( SMP14 ) sc-965 , lot #J2314; p53 ( Fl-393 ) sc-6243 , lot # D0215; c-Myc ( 9E10 ) sc-40 , lot # G2413 . Primary fibroblasts were grown to 80% confluency in T-75 culture flasks at 37°C/5% CO2 in a culture medium consisting of FGM/EMEM ( 1:1 ) supplemented with insulin , FGF , 6% FBS and Gentamicin/Amphotericin B ( FGM-2 , singlequots , Clonetics/Lonza ) . 104 cells were seeded into each well of an opaque bottomed 96-well plate , leaving a column with no cells ( background control ) ; each 96-well plate contained paired elephant and hyrax samples . Serial dilutions of Doxorubicin ( 0 uM , 0 . 5 uM , 1 . 0 uM , 5 uM , 10 uM and 50 uM ) , Mitomycin c ( 0 uM , 0 . 5 uM , 1 . 0 uM , 5 uM , 10 uM and 50 uM ) , and Nutlin-3a ( 0 uM , 0 . 5 uM , 1 . 0 uM , 5 uM , 10 uM and 50 uM ) and 90% culture media were added to each well such that there were four biological replicates for each condition . After 18 hr of incubation with each drug , cell viability , cytotoxicity , and caspase-3/7 activity were measured using the ApoTox-Glo Triplex Assay ( Promega ) in a GloMax-Multi+ Reader ( Promega ) . Data were standardized to no-drug control cells . For UV-C treatment , culture medium was removed from wells prior to UV treatment and returned to cells shortly afterwards . Experimental cells were exposed to 50 J/m2UV-C radiation in a crosslinker ( Stratalinker 2400 , Stratagene ) , while control cells passed through media changes but were not exposed to UV . A small volume ( ~30 uL ) of PBS covered fibroblasts at the time of UV exposure . Cell viability , cytotoxicity , and caspase-3/7 activity were measured using the ApoTox-Glo Triplex Assay ( Promega ) in a GloMax-Multi+ Reader ( Promega ) 6 , 12 , 28 . 5 and 54 . 5 hr after UV-C treatment . Data were standardized to no UV-C exposure control cells . ApoTox-Glo Triplex Assays were replicated three independent times . Primary fibroblasts were grown to 80% confluency in T-75 culture flasks at 37°C/5% CO2 in a culture medium consisting of FGM/EMEM ( 1:1 ) supplemented with insulin , FGF , 6% FBS and Gentamicin/Amphotericin B ( FGM-2 , singlequots , Clonetics/Lonza ) . At confluency , cells were trypsinized , centrifuged at 90 g for 10 min and resuspended in nucleofection/supplement solution and incubated for 15 min . After incubation 1ug of the pGL4 . 38[luc2p/p53 RE/Hygro] luciferase reporter vector and 100 ng of the pGL4 . 74[hRluc/TK] Renilla reporter vector were transiently transfected into an elephant and hyrax cells using the Amaxa Basic Nucleofector Kit ( Lonza ) using protocol T-016 . Immediately following nuleofection , 104 cells were seeded into each well of an opaque bottomed 96-well plate , leaving a column with no cells ( background control ) ; each 96-well plate contained paired elephant and hyrax samples . 24 hr after nucleofection cells were treated with either vehicle control , Doxorubicin , Mitomycin c , Nutlin-3a , or 50 J/m2UV-C . Luciferase expression was assayed 18 hr after drug/UV-C treatment cells , using the Dual-Luciferase Reporter Assay System ( Promega ) in a GloMax-Multi+ Reader ( Promega ) . For all experiments luciferase expression was standardized to Renilla expression to control for differences nucleofection efficiency across samples; Luc . /Renilla data is standardized to ( Luc . /Renilla ) expression in untreated control cells . Each luciferase experiment was replicated three independent times . siRNAs designed to specifically-target TP53RTG were validated via qRT-PCR using the two primer sets described earlier , which amplify either TP53RTG or canonical TP53 cDNAs , to confirm specificity and efficacy of knockdown . Sequences of the three TP53RTG-specific siRNAs used are as follows: ( 1 ) 5’-CAGCGGAGGCAGUAGAUGAUU-3’ , ( 2 ) 5’-GGCUCAAGGAAUAUCAGAAUU-3’ , ( 3 ) 5’-CAGCAGCGGAGGCAGUAGAUU-3’ ( Dharmacon ) . Loxodonta fibroblasts were transfected with siRNAs using Lipofectamine LTX , and tested 48–72 hr later for either TP53 response via luciferase assay or for cell viability/toxicity/apoptosis via ApoTox-Glo assay .	As time passes , healthy cells are more likely to become cancerous because more and more damaging mutations accumulate in the cell’s DNA . Assuming that all cells have a similar risk of acquiring mutations , larger and longer-lived animals – like elephants – should have a higher risk of cancer than smaller , shorter-lived animals – like mice . However , there does not appear to be any link between the size of an animal and its risk of developing cancer . Consequently , a key question in cancer biology is how very large animals protect themselves against these diseases . One gene that is often damaged during an animal’s lifetime is called TP53 . This gene normally produces a tumor suppressor protein that senses when DNA is damaged or a cell is under stress and either briefly slows the cell’s growth while the damage is repaired or triggers cell death if the stress is overwhelming . One way that large animals could reduce their risk of cancer is to have extra copies of TP53 or other genes that encode tumor suppressor proteins . Here Sulak et al . used an evolutionary genomics approach to study TP53 in 61 animals of various sizes , including several large animals such as African elephants and Minke whales . All of the animals studied had at least one copy of TP53 , and several had a few extra copies , known as TP53 retrogenes . African elephants – the largest living land mammal – had more retrogenes than any of the others with 19 in total . To investigate why African elephants have so many TP53 retrogenes , Sulak et al . also analyzed DNA from Asian elephants and several other closely related , but now extinct species , including the woolly mammoth . As expected , as species evolved larger body sizes they also evolved more TP53 retrogenes . Further experiments indicate that several of the TP53 retrogenes in African elephants are likely to be able to produce the tumor suppressor protein and that they contribute to elephant cells being better equipped to deal with DNA damage . The next step following on from this work will be to find out exactly how TP53 retrogenes help to protect animals from cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "cell", "biology" ]	2016	TP53 copy number expansion is associated with the evolution of increased body size and an enhanced DNA damage response in elephants
The cJun NH2-terminal kinase ( JNK ) signaling pathway is implicated in the response to metabolic stress . Indeed , it is established that the ubiquitously expressed JNK1 and JNK2 isoforms regulate energy expenditure and insulin resistance . However , the role of the neuron-specific isoform JNK3 is unclear . Here we demonstrate that JNK3 deficiency causes hyperphagia selectively in high fat diet ( HFD ) -fed mice . JNK3 deficiency in neurons that express the leptin receptor LEPRb was sufficient to cause HFD-dependent hyperphagia . Studies of sub-groups of leptin-responsive neurons demonstrated that JNK3 deficiency in AgRP neurons , but not POMC neurons , was sufficient to cause the hyperphagic response . These effects of JNK3 deficiency were associated with enhanced excitatory signaling by AgRP neurons in HFD-fed mice . JNK3 therefore provides a mechanism that contributes to homeostatic regulation of energy balance in response to metabolic stress . The regulation of energy balance ( food consumption and energy expenditure ) is important for health and survival . Sustained negative energy balance caused by cachexia and anorexia is associated with serious injury to multiple organ systems ( Aoyagi et al . , 2015; Mehler and Brown , 2015 ) . Similarly , sustained positive energy balance caused by hyperphagia results in obesity associated with severe metabolic disorders ( e . g . type 2 diabetes , cardiovascular disease , hepatitis , neurodegeneration and cancer ) that represent leading causes of morbidity and mortality ( Flegal et al . , 2013 ) . The homeostatic maintenance of energy balance is therefore critically important . It is established that the arcuate nucleus ( ARC ) in the hypothalamus plays a key role in the regulation of energy balance ( Cone , 2005 ) . AgRP neurons in the ARC mediate orexigenic signals , including neuropeptide Y ( NPY ) , agouti-related peptide ( AgRP ) , and γ-aminobutyric acid ( GABA ) that project to POMC neurons in the ARC and to secondary response neurons in many brain regions , including the lateral hypothalamus ( LH ) and the paraventricular nucleus ( PVN ) of the hypothalamus . In contrast , POMC neurons mediate anorexigenic signals , including cocaine and amphetamine regulated transcript ( CART ) and pro-opiomelanocortin ( POMC ) -derived α-melanocyte stimulating hormone ( α-MSH ) . POMC neurons project to many brain areas , including the PVN and LH in the hypothalamus where α-MSH acts as an agonist of the melanocortin receptors MC3R and MC4R on secondary response neurons to inhibit feeding and increase energy expenditure . Importantly , this action of α-MSH is antagonized by AgRP . Moreover , POMC neurons receive inhibitory GABAergic input from AgRP neurons . Consequently , AgRP and POMC neurons act together to balance food consumption , energy expenditure and nutrient homeostasis ( Cone , 2005 ) . AgRP and POMC neurons integrate signals from nutrients ( e . g . glucose and fatty acids ) and peripheral hormones ( e . g . leptin , insulin , ghrelin , and cytokines ) to mediate opposite actions regulating downstream neuroendocrine circuits linking internal and environmental stimuli with the coordinated control of homeostatic satiety ( Blouet and Schwartz , 2010; Varela and Horvath , 2012 ) . Thus , leptin activates POMC neurons ( Cowley et al . , 2001 ) and inhibits AgRP neurons ( Takahashi and Cone , 2005 ) leading to reduced food consumption and increased energy expenditure . These processes can be regulated by intracellular signaling networks , including the Janus kinase 2-signal transducer and activator of transcription 3 ( JAK2-STAT3 ) axis ( Bates and Myers , 2003 ) , Rho-associated coiled coil containing protein kinase 1 ( ROCK1 ) ( Huang et al . , 2012 ) , mechanistic target of rapamycin ( mTOR ) ( Mori et al . , 2009; Kocalis et al . , 2014 ) , adenosine monophosphate-activated protein kinase ( AMPK ) ( Claret et al . , 2007; Dagon et al . , 2012 ) , and phosphatidylinositol-4 , 5-bisphosphate 3-kinase ( PI3K ) ( Niswender et al . , 2003 ) , that contribute to the fine-tuning of energy balance . The anorexigenic hormone leptin plays a key role in the regulation of food consumption . Leptin can act directly on AgRP and POMC neurons , but leptin can also act on other neurons in several brain sub-regions , including mid-brain and brainstem nuclei ( Scott et al . , 2009; Patterson et al . , 2011 ) . Control of leptin signaling in these neurons is important for maintaining energy balance . For example , obesity causes an increase in the blood concentration of leptin , most likely because of increased adipose tissue mass . The increased leptin concentration can lead to tachyphylaxis and suppression of the anorexigenic actions of leptin ( Frederich et al . , 1995 ) . This mechanism enables homeostatic regulation of feeding behavior in response to metabolic stress . Whether this mechanism represents “leptin resistance” is unclear ( Myers et al . , 2010 ) because some biochemical aspects of leptin signaling are maintained in the obese state ( Ottaway et al . , 2015 ) . A requirement for leptin signaling may reflect the role of the leptin-stimulated JAK2-STAT3 pathway to increase expression of the negative regulator SOCS3 ( Allison and Myers , 2014 ) . Negative regulation of leptin signaling may also involve the tyrosine phosphatases PTPN1 and PTPN2 ( Bence et al . , 2006; Loh et al . , 2011 ) , reactive oxygen species ( Diano et al . , 2011 ) , the endoplasmic reticulum unfolded protein response ( Zhang et al . , 2008; Ozcan et al . , 2009 ) , autophagy ( Kaushik et al . , 2011 ) , and low-grade inflammation ( de Git and Adan , 2015 ) . The purpose of the study reported here was to test whether the cJun NH2-terminal kinase ( JNK ) signaling pathway regulates feeding behavior . Previous studies have established that the ubiquitously expressed JNK1 and JNK2 isoforms play an important role in the metabolic stress response of peripheral tissues ( Sabio and Davis , 2010 ) . However , loss-of-function studies have not identified a role for JNK in the control of food consumption . Here we demonstrate that the neuronal isoform JNK3 ( encoded by the Mapk10 gene ) plays a key role in the maintenance of energy balance during consumption of a high fat diet ( HFD ) by promoting leptin signaling . Mapk10 gene ablation studies identify AgRP neurons as a site of JNK3 function . JNK3 is therefore a key mediator of homeostatic regulation of energy balance in response to metabolic stress . Leptin is an anorexigenic hormone . Indeed , treatment of chow-fed mice with leptin suppressed feeding behavior and caused decreased body mass ( Figure 1A ) . In contrast , HFD-fed mice failed to respond to leptin ( Figure 1A ) . The mechanism that accounts for this observation is unclear , but may involve both decreased leptin signaling and reduced signaling by down-stream mediators ( e . g . MC4R ) . Tachyphylaxis may be a contributing factor and mutational analysis of leptin signaling components implicates functions of the leptin receptor , tyrosine phosphatases , reactive oxygen species , and SOCS3 ( Myers et al . , 2010 ) . 10 . 7554/eLife . 10031 . 003Figure 1 . JNK3 deficiency causes hyperphagia and obesity . ( A ) WT mice were fed ( 4 wk ) a chow diet ( CD ) or a high-fat diet ( HFD ) . The body mass change at 24 hr post-injection ( i . p . with solvent ( PBS ) or 2 . 5 mg/kg leptin ) was measured ( mean ± SEM; n=8; *p<0 . 001 ) . ( B ) WT and Mapk10-/- mice fed ( 12 wk ) a CD or a HFD were starved overnight . Phospho-JNK3 , JNK3 , and GAPDH in the hypothalamus were measured by immunoblot analysis . ( C , D ) The body mass gain of CD-fed and HFD-fed ( 12 wk ) mice was measured ( mean ± SEM; n=10~12 ) ( C ) . Fat and lean mass were measured by 1H-MRS analysis ( mean ± SEM; n=10~12 ) . ( D ) Statistically significant differences between WT and Mapk10-/- mice are indicated ( p<0 . 001 ) . ( E ) Paraffin embedded sections of epididymal white adipose tissue ( WAT ) , interscapular brown adipose tissue ( BAT ) , and liver were prepared from HFD-fed ( 12 wk ) WT and Mapk10-/- mice . The sections were stained with hematoxylin & eosin . Scale bar , 100 µm . ( F ) Food consumption by WT and Mapk10-/- mice fed a CD or a HFD ( 3 wk ) was measured ( mean ± SEM; n=6; p<0 . 01; *p<0 . 001 ) . ( G ) WT and Mapk10-/- mice fed a CD or a HFD ( 4 wk ) were fasted overnight and the blood concentration of leptin and insulin was measured ( mean ± SE; n=10~12; p<0 . 05 ) . ( H , I ) Glucose tolerance tests ( H ) and insulin tolerance tests ( I ) were performed on WT and Mapk10-/- mice fed a CD or a HFD ( 12 wk ) by measurement of blood glucose concentration ( mean ± SEM; n=10~12; p < 0 . 05; p < 0 . 01; *p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 00310 . 7554/eLife . 10031 . 004Figure 1—figure supplement 1 . JNK3 deficiency causes obesity without changes in energy expenditure . ( A ) Organ mass of CD-fed and HFD-fed ( 12 wk ) WT and Mapk10-/- mice was measured ( mean ± SEM; n=10~12; p<0 . 05; *p<0 . 001 ) . ( B ) CD-fed and HFD-fed ( 4 wk ) WT and Mapk10-/- mice were examined using metabolic cages to measure VO2 , VCO2 , and energy expenditure ( mean ± SEM; n=8; p<0 . 001 ) . ( C ) Blood lipids and lipoproteins in overnight starved CD-fed and HFD-fed ( 12 wk ) WT and Mapk10-/- mice were measured ( mean ± SEM; n=10~12; p<0 . 01; *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 00410 . 7554/eLife . 10031 . 005Figure 1—figure supplement 2 . Time course of the development of hyperphagia in HFD-fed JNK3-deficient mice . ( A ) Metabolic cage analysis of WT and Mapk10-/- mice fed a HFD . The amount of food consumed 7pm – 7am each day was measured ( mean ± SEM; n=6; p<0 . 05; p<0 . 01 ) . ( B ) Fat and lean body mass of WT and Mapk10-/- mice fed a HFD ( 3 days ) was measured by 1H-MRS analysis ( mean ± SEM; n=6 ) . No statistically significant differences were detected ( p>0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 00510 . 7554/eLife . 10031 . 006Figure 1—figure supplement 3 . Increased food consumption is required for obesity caused by JNK3 deficiency in HFD-fed mice . WT and Mapk10-/- mice were fed a HFD ad libitum . A second group of Mapk10-/- mice were pair-fed with the WT mice . The change in body mass was measured ( mean ± SEM; n= 6; p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 006 We considered the possibility that a stress-activated MAP kinase pathway may contribute to the regulation of leptin signaling in HFD-fed mice . It is established that feeding a HFD causes activation of the ubiquitously expressed isoforms JNK1 and JNK2 in peripheral tissues , including liver , muscle , and adipose tissue ( Sabio and Davis , 2010 ) . However , the regulation of JNK caused by feeding a HFD in the central nervous system is unclear because these ubiquitously expressed JNK isoforms in neurons are constitutively activated and are primarily localized to axons and dendrites ( Coffey et al . , 2000; Oliva et al . , 2006 ) . In contrast , the neuron-specific isoform JNK3 exhibits low basal activity and can be activated in the nucleus when neurons are exposed to environmental stress ( Yang et al . , 1997 ) . We therefore tested whether feeding a HFD caused activation of JNK3 . This analysis demonstrated that feeding a HFD caused JNK3 phosphorylation and activation in the hypothalamus ( Figure 1B ) . JNK3 in the central nervous system is therefore responsive to diet-induced metabolic stress . This JNK3 pathway represents a possible mediator of altered leptin signaling in HFD-fed mice . To examine the role of the JNK3 pathway , we investigated the effect of feeding a chow diet ( CD ) or a HFD to wild-type ( WT ) mice or Mapk10-/- ( JNK3-deficient ) mice . We found that Mapk10-/- mice gained similar body mass when fed a CD , but these mice gained significantly greater mass when fed a HFD compared with WT mice ( Figure 1C ) . 1H-MRS analysis demonstrated that the greater HFD-induced body mass was caused by increased fat and lean mass ( Figure 1D ) . Indeed , HFD-fed Mapk10-/- mice exhibited increased liver , skeletal muscle , heart , and adipose tissue mass compared with HFD-fed WT mice ( Figure 1—figure supplement 1A ) . Microscopic examination of tissue sections demonstrated increased hypertrophy of white and brown adipocytes and increased hepatic steatosis in HFD-fed Mapk10-/- mice compared with HFD-fed WT mice ( Figure 1E ) . We performed metabolic cage analysis to examine the mechanism of obesity promoted by JNK3 deficiency . These studies demonstrated that Mapk10 gene ablation selectively increased consumption of a HFD , but not a CD ( Figure 1F ) . Time course analysis demonstrated that the HFD-selective hyperphagia was observed within 2 days of consuming the HFD ( Figure 1—figure supplement 2A ) and was detected prior to the development of obesity ( Figure 1—figure supplement 2B ) . No significant changes in VO2 , VCO2 , or energy expenditure were detected in the HFD-fed mice ( Figure 1—figure supplement 1B ) . These data suggest that hyperphagia contributes to the increased obesity of HFD-fed Mapk10-/- mice compared with HFD-fed WT mice . We used a pair-feeding protocol to test whether the increased obesity of Mapk10-/- mice compared with WT mice was caused by greater food consumption . We found that WT and Mapk10-/- mice gained similar body mass when fed the same amount of food ( Figure 1—figure supplement 3 ) . These data demonstrate that hyperphagia accounts for the increased HFD-induced obesity of Mapk10-/- mice compared with WT mice . Consequences of the increased HFD feeding behavior of Mapk10-/- mice include increased hyperinsulinemia and hyperleptinemia ( Figure 1G ) , increased blood lipid concentrations ( Figure 1—figure supplement 1C ) , decreased glucose tolerance ( Figure 1H ) , and increased insulin resistance ( Figure 1I ) when fed a HFD , but not a CD . These data indicate that Mapk10-/- mice may exhibit increased HFD-induced insulin resistance . To test this hypothesis , we performed a hyperinsulinemic-euglycemic clamp study . No significant differences between CD-fed WT and Mapk10-/- mice were detected ( Figure 2A–F ) . In contrast , HFD-fed Mapk10-/- mice showed significantly reduced glucose infusion rate ( a measure of whole body insulin sensitivity ) , reduced glucose turnover , reduced whole body glycolysis , increased hepatic glucose production , and decreased hepatic insulin action compared with HFD-fed WT mice ( Figure 2A–F ) . These data demonstrate that Mapk10-/- mice exhibit a profound defect in glycemic regulation compared with WT mice when fed a HFD , but not a CD . 10 . 7554/eLife . 10031 . 007Figure 2 . JNK3 deficiency promotes and adipose tissue inflammation and insulin resistance . ( A-F ) Hyperinsulinemic-euglycemic clamps were performed on CD-fed or HFD-fed ( 3 wk ) WT and Mapk10-/- mice . Clamp hepatic glucose production ( A ) , hepatic insulin action ( B ) , glucose turnover ( C ) , glucose infusion rate ( D ) , whole body glycolysis ( E ) , and glycogen plus lipid synthesis ( F ) were measured ( mean ± SE; n=8; p<0 . 05; *p<0 . 001 ) . ( G-J ) Sections of epididymal WAT from HFD-fed ( 12 wk ) WT and Mapk10-/- mice were stained with hematoxylin & eosin or with an antibody to the macrophage protein F4/80 ( G ) . Macrophage infiltration was examined by measurement of the expression of Cd68 and Emr1 ( F4/80 ) mRNA ( H ) and also mRNA expressed by genes associated with M1-like ( I ) and M2-like ( J ) polarization by Taqman© assays ( mean ± SEM; n=10~12; p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 007 The increased adipose tissue mass of HFD-fed Mapk10-/- mice compared with HFD-fed control mice was associated with increased adipose tissue infiltration by F4/80+ macrophages ( Figure 2G ) . Indeed , gene expression analysis identified markedly increased expression of macrophage marker genes ( Emr1 ( F4/80 ) & Cd68 ) , increased expression of genes associated with M1-like macrophage polarization ( Ccl2 , Il1b , Il6 & Tnf ) , and decreased expression of genes associated with M2-like macrophage polarization ( Arg1 , Mgl2 , Mrc1 & Mrc2 ) in the adipose tissue of HFD-fed Mapk10-/- mice compared with HFD-fed control mice ( Figure 2H–J ) . These data indicate that JNK3 deficiency promotes increased adipose tissue inflammation in HFD-fed mice . It is likely that this increase in inflammation contributes to the glucose intolerant and insulin resistant phenotype of HFD-fed Mapk10-/- mice compared with HFD-fed WT mice ( Brestoff and Artis , 2015 ) . Low concentrations of leptin were detected in the blood when WT and Mapk10-/- mice were fed a CD ( Figure 1G ) . The blood leptin concentration was increased when these mice were fed a HFD and was significantly greater in HFD-fed Mapk10-/- mice compared with HFD-fed WT mice ( Figure 1G ) . These changes in the amount of leptin circulating in the blood correlate , as expected , with differences in obesity ( Friedman , 2014 ) . However , the hyperleptinemia and hyperphagia of HFD-fed Mapk10-/-mice is not consistent with the established anorexigenic function of leptin . This analysis suggested that leptin signaling may be suppressed in HFD-fed Mapk10-/- mice . To test this hypothesis , we examined the effect of treating mice with leptin . We found that intracerebroventricular administration of leptin decreased the body mass of WT mice , but not Mapk10-/- mice ( Figure 3A ) . Measurement of hypothalamic gene expression demonstrated that leptin decreased Agrp and Npy expression in WT mice , but not Mapk10-/- mice ( Figure 3B ) . In contrast , leptin caused increased Pomc and Socs3 gene expression in both WT and Mapk10-/- mice ( Figure 3B ) . These data indicate that Mapk10-/- mice exhibit a selective deficiency in leptin regulation of Agrp and Npy expression . To confirm this conclusion , we compared hypothalamic gene expression in CD-fed and HFD-fed mice . This analysis demonstrated increased Agrp and Npy expression in HFD-fed Mapk10-/- mice compared with HFD-fed WT mice ( Figure 3C ) . In contrast , no significant difference in Pomc and Socs3 gene expression between HFD-fed WT and Mapk10-/- mice was detected ( Figure 3C ) . These observations indicate that JNK3 deficiency caused a selective defect in leptin signaling . 10 . 7554/eLife . 10031 . 008Figure 3 . JNK3 deficiency causes a selective defect in AgRP neurons . ( A ) HFD-fed ( 4 wk ) WT and Mapk10-/- mice were treated by intracerebroventricular administration of 5 µg leptin or solvent ( Control ) . The change in body mass at 24 hr post-treatment was measured ( mean ± SEM; n=10~12; *p<0 . 001 ) . ( B ) WT and Mapk10-/- mice were treated without or with leptin ( 2h ) prior to measurement of hypothalamic gene expression by Taqman© assays ( mean ± SEM; n=10~12; p<0 . 05 ) . ( C ) Hypothalamic gene expression in CD-fed and HFD-fed ( 12 wk ) WT and Mapk10-/- mice was measured by Taqman© assay ( mean ± SEM; n=10~12; p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 008 To examine the mechanism of JNK3 function , we established floxed Mapk10 mice to investigate the neuron-specific effects of JNK3 on feeding behavior ( Figure 4—figure supplement 1 ) . We tested whether JNK3 in neurons that express the leptin receptor LEPRb regulates feeding behavior by investigating the effect of Mapk10 gene ablation specifically in LEPRb+ neurons . This analysis demonstrated that control Leprb-cre ( LepRWT ) mice and Leprb-cre Mapk10Loxp/LoxP ( LepR∆J3 ) mice gained similar body mass when fed a CD . However , HFD-fed LepR∆J3 mice gained significant more body mass than LepRWT mice ( Figure 4A and Figure 4—figure supplement 2A ) . 1H-MRS analysis showed that the difference in body mass was caused by increased fat mass ( Figure 4B ) . Metabolic cage analysis demonstrated that Mapk10 gene ablation in LEPRb+ neurons caused no change in CD food consumption , but caused increased HFD food consumption ( Figure 4C ) . This increase in HFD consumption was not associated with changes in VO2 , VCO2 , or energy expenditure ( Figure 4—figure supplement 2B ) . JNK3 in LEPRb+ neurons of HFD-fed mice therefore regulates feeding behavior , but not other aspects of energy balance . 10 . 7554/eLife . 10031 . 009Figure 4 . JNK3 deficiency in leptin-responsive neurons causes HFD-induced hyperphagia and obesity . ( A ) The total body mass gain of CD-fed and HFD-fed mice was examined ( mean ± SEM; n = 10~25; p<0 . 05; p<0 . 01 ) . JNK3 deficiency in LEPRb+ neurons was studied by comparing Leprb-cre control mice ( LepRbWT mice ) and Leprb-cre Mapk10LoxP/LoxPmice ( LepR∆J3 mice ) . ( B ) The fat and lean mass of CD-fed and HFD-fed ( 16 wk ) mice was measured by 1H-MRS analysis ( mean ± SEM; n = 8~10; p<0 . 001 ) . ( C ) Food consumption by CD-fed and HFD-fed ( 4 wk ) LepRWT and LepR∆J3 mice was examined ( mean ± SEM; n = 8; p<0 . 05 ) . ( D , E ) Glucose tolerance ( D ) and insulin tolerance ( E ) tests were performed using CD-fed and HFD-fed ( 12 wk ) LepRWT and LepR∆J3 mice ( mean ± SEM; n = 8~12; p<0 . 05; *p<0 . 01 ) . ( F-H ) CD-fed and HFD-fed ( 12 wks ) LepRWT and LepR∆J3 mice were fasted overnight and the blood concentration of glucose ( F ) , insulin ( G ) , and leptin ( H ) was measured ( mean ± SEM; n = 8~20; p<0 . 05p<0 . 01 ) . ( I ) Sections of epididymal WAT , interscapular BAT , and liver from CD-fed and HFD-fed ( 12 wk ) LepRWT and LepR∆J3 mice were stained with hematoxylin & eosin . Bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 00910 . 7554/eLife . 10031 . 010Figure 4—figure supplement 1 . Establishment of Mapk10LoxP/LoxPmice . ( A ) The strategy employed to target the Mapk10 locus ( exons 6–9 ) by homologous recombination is illustrated . Restriction sites and PCR amplimers are indicated . Abbreviations: NEO , neomycin resistance cassette; and DT , diptheria toxin cassette . ( B ) The integrity of 5’ and 3’ LoxP sites of the Mapk10LoxP allele were verified by PCR analysis of genomic DNA using amplimers 1F & 2R ( 5’ site ) and 3F & 5R ( 3’ site ) . ( C ) Comparison of hypothalamic expression of selected genes in WT and Mapk10LoxP/LoxPmice by Taqman© assays . No significant differences ( p>0 . 05 ) between WT and Mapk10LoxP/LoxPmice were detected ( mean ± SEM; n=3 ) . ( D ) PCR genotype analysis of DNA isolated from different tissues of WT mice , Mapk10LoxP/LoxPmice , Agrp-cre Mapk10LoxP/LoxPmice ( Agrp∆J3 ) , Pomc-cre Mapk10LoxP/LoxPmice ( Pomc∆J3 ) , and LepRb-cre Mapk10LoxP/LoxPmice ( LepR∆J3 ) using amplimers 1F & 2R ( upper panel ) and 1F & 4R ( lower panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 01010 . 7554/eLife . 10031 . 011Figure 4—figure supplement 2 . JNK3 deficiency in leptin-responsive neurons causes obesity . ( A ) Organ mass of CD-fed and HFD-fed ( 16 wk ) LepRb-cre ( LeprWT ) and LepRb-cre Mapk10Loxp/LoxP ( Lepr∆J3 ) mice was measured ( mean ± SEM; n=10~12; p<0 . 01; **p<0 . 001 ) . ( B ) CD-fed and HFD-fed ( 4 wk ) LeprWT and Lepr∆J3 mice were examined using metabolic cages to measure VO2 , VCO2 , and energy expenditure ( mean ± SEM; n=8; p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 011 The increased feeding behavior of HFD-fed ( but not CD-fed ) LepR∆J3 mice was associated with decreased glucose tolerance ( Figure 4D ) , increased insulin resistance ( Figure 4E ) , increased blood glucose concentration ( Figure 4F ) , increased hyperinsulinemia ( Figure 4G ) , and increased hyperleptinemia ( Figure 4H ) . White and brown adipose tissue ( WAT & BAT ) in HFD-fed LepR∆J3 mice exhibited increased adipocyte hypertrophy compared with HFD-fed LepRWT mice ( Figure 4I ) . Moreover , JNK3 deficiency in LepRb+ neurons caused increased HFD-induced hepatic steatosis ( Figure 4I ) . To identify a LepRb+ neuronal sub-population relevant to JNK3-regulated HFD feeding behavior , we examined Mapk10 gene ablation in selected neurons within the hypothalamus . Gene expression analysis demonstrated that JNK3 was required for HFD-induced regulation of Agrp and Npy , but not Pomc ( Figure 3 ) . This analysis indicated that AgRP neurons rather than POMC neurons may play an important role in JNK3-regulated feeding behavior in HFD-fed mice . To test this hypothesis , we examined the phenotype of Agrp-cre Mapk10Loxp/LoxP ( Agrp∆J3 ) mice and Pomc-cre Mapk10Loxp/LoxP ( Pomc∆J3 ) mice . We found that JNK3 deficiency in POMC neurons of HFD-fed mice caused no significant changes in feeding behavior , glucose intolerance , blood glucose concentration , hypertrophy of white and brown adipocytes , and hepatic steatosis compared with control Pomc-cre ( PomcWT ) mice ( Figure 5A , C , E ) . In contrast , JNK3 deficiency in AgRP neurons in HFD-fed mice caused increased feeding , increased glucose intolerance , increased blood glucose concentration , increased hypertrophy of white and brown adipocytes , and increased hepatic steatosis compared with control Agrp-cre ( AgrpWT ) mice ( Figure 5B , D , F ) . Metabolic cage analysis demonstrated that the VO2 , VCO2 , and energy expenditure of HFD-fed Agrp∆J3 mice and Pomc∆J3 mice were similar to control mice ( Figure 5—figure supplement 1 ) . Together , these data demonstrate that JNK3 in AgRP neurons , but not POMC neurons , acts to suppress HFD consumption . 10 . 7554/eLife . 10031 . 012Figure 5 . JNK3 in AgRP neurons , but not POMC neurons , regulates food consumption . ( A , B ) Food consumption by CD-fed and HFD-fed ( 4 wk ) mice was measured ( mean ± SEM; n = 8; p<0 . 05 ) . JNK3 deficiency in POMC neurons was studied by comparing Pomc-cre control mice ( PomcWT mice ) and Pomc-cre Mapk10LoxP/LoxPmice ( Pomc∆J3 mice ) . JNK3 deficiency in AgRP neurons was studied by comparing Agrp-cre control mice ( AgrpWT mice ) and Agrp-cre Mapk10LoxP/LoxPmice ( Agrp∆J3 mice ) . ( C , D ) CD-fed and HFD-fed ( 16 wk ) control mice and mice with JNK3 deficiency in POMC neurons ( C ) and AgRP neurons ( D ) or were tested using glucose tolerance assays and by measurement of fasting blood glucose concentration ( mean ± SEM; n = 8~12; p<0 . 05; p<0 . 01; p<0 . 001 ) . ( E , F ) Representative hematoxylin & eosin-stained sections of liver , epididymal WAT , and interscapular BAT from CD-fed and HFD-fed ( 16 wk ) control mice and mice with JNK3 deficiency in POMC neurons ( E ) and AgRP neurons ( F ) are presented . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 01210 . 7554/eLife . 10031 . 013Figure 5—figure supplement 1 . Effect of JNK3 deficiency in AgRP and POMC neurons on energy expenditure . ( A ) CD-fed and HFD-fed ( 4 wk ) AgrpWT and Agrp∆J3 mice mice were examined using metabolic cages to measure VO2 , VCO2 , and energy expenditure ( mean ± SEM; n=8; p>0 . 05 ) . ( B ) CD-fed and HFD-fed ( 4 wk ) PomcWT and Pomc∆J3 mice were examined using metabolic cages to measure VO2 , VCO2 , and energy expenditure ( mean ± SEM; n=8; p<0 . 05; *p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 013 Leptin and its receptor are known to affect synaptic transmission and modulate AgRP neuron activity ( Pinto et al . , 2004; Baver et al . , 2014 ) . We therefore examined miniature inhibitory postsynaptic currents ( mIPSCs ) and miniature excitatory postsynaptic currents ( mEPSCs ) of AgRP neurons in the ARC of WT and Mapk10-/- mice . This analysis demonstrated that JNK3 deficiency caused no change in mIPSC frequency or amplitude in CD-fed and HFD-fed mice ( Figure 6A–D ) . Similarly , JNK3 deficiency caused no change in mEPSC frequency or amplitude in CD-fed mice ( Figure 6E–H ) . In contrast , HFD-fed JNK3-deficient mice demonstrated increased mEPSC amplitudes in the absence of changes in mEPSC frequency ( Figure 6E–H ) . Studies using the selective antagonist DNQX demonstrated that these mEPSC currents were mediated by AMPA receptors in AgRP neurons ( Figure 6-figure supplement 1 ) . Together , these data indicate that JNK3 deficiency leads to altered excitatory transmission onto AgRP neurons compared to WT mice when fed a HFD . This finding is consistent with the increased expression of AgRP and NPY ( Figure 3C ) and the increased food consumption ( Figure 1F ) observed in HFD-fed JNK3-deficient compared to HFD-fed WT mice . 10 . 7554/eLife . 10031 . 014Figure 6 . JNK3 regulates excitatory transmission onto AgRP neurons . ( A , B ) Mapk10+/+ Npy-GFP and Mapk10-/- Npy-GFP mice were fed a HFD ( 3 wk ) prior to electrophysiological recording of mIPSC from AgRP neurons . ( C , D ) mIPSC frequency ( freq . ) and amplitude ( amp . ) in recordings of CD-fed and HFD-fed mice were quantitated ( mean ± SEM; n=11~12; p<0 . 05 ) . ( E , F ) Mapk10+/+ Npy-GFP and Mapk10-/- Npy-GFP mice were fed a HFD ( 3 wk ) prior to electrophysiological recording of mEPSC from AgRP neurons . ( G , H ) mEPSC frequency and amplitude in recordings of CD-fed and HFD-fed mice were quantitated ( mean ± SEM; n=10; *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 01410 . 7554/eLife . 10031 . 015Figure 6—figure supplement 1 . The AMPA receptor antagonist DNQX blocks mEPSCs in AgRP neurons . mEPSCs of arcuate AgRP neurons from Npy-GFP mice were recorded under baseline conditions ( perfusion with aCSF ) and following addition of 10 µM DNQX . The data presented are representative of three recordings from single neurons . The selective AMPA receptor antagonist ( DNQX ) blocks all mEPSCs recorded at -60 mV . Extracellular Mg2+ in the bath solution blocks NMDA receptor-mediated currents at this holding potential . DOI: http://dx . doi . org/10 . 7554/eLife . 10031 . 015 The JNK signaling pathway is implicated in the metabolic stress response ( Sabio and Davis , 2010 ) . Studies of the ubiquitously expressed isoforms JNK1 and JNK2 demonstrate that the JNK pathway is activated in peripheral tissues by feeding a HFD ( Hirosumi et al . , 2002 ) . Consequences of HFD-stimulated JNK1 and JNK2 activation in peripheral tissues include promotion of insulin resistance in adipose tissue , liver , and muscle ( Sabio et al . , 2008; Sabio et al . , 2010b; Vernia et al . , 2014 ) . In contrast , central actions of JNK1 and JNK2 are mediated by the hypothalamic-pituitary axis by regulation of energy expenditure ( Belgardt et al . , 2010; Sabio et al . , 2010a; Vernia et al . , 2013 ) . Together , these studies indicate that JNK1 and JNK2 play important roles in metabolic stress responses by causing insulin resistance in peripheral tissues and promoting obesity by suppressing energy expenditure ( Sabio and Davis , 2010 ) . JNK3 is expressed in a limited number of tissues , including the brain and testis ( Gupta et al . , 1996 ) . Since JNK1 and JNK2 are expressed ubiquitously , the expression of JNK3 by neurons means that these cells express all three JNK isoforms ( Davis , 2000 ) . To examine the role of JNK in neurons , the effects of ablation of the three genes that encode JNK ( Mapk8 , Mapk9 , and Mapk10 ) in neurons have been examined . This analysis demonstrated that compound JNK-deficiency caused markedly increased survival responses associated with increased autophagy ( Xu et al . , 2011 ) . Roles for individual JNK isoforms in neurons have also been studied ( Coffey , 2014 ) . JNK1 and , to some extent JNK2 , are constitutively activated and are primarily localized to axons and dendrites ( Coffey et al . , 2000; Oliva et al . , 2006 ) where they play a major role in the regulation of the cytoskeleton and axonal/dendritic morphology ( Coffey , 2014 ) . In contrast , JNK3 exhibits low basal activity and is activated in the nucleus when neurons are exposed to stress ( Yang et al . , 1997 ) . Studies of Mapk10-/- mice demonstrate that JNK3 is required for stress-induced cJun phosphorylation and AP-1 activation in neurons ( Yang et al . , 1997 ) . This role of JNK3 in neurons is non-redundant with JNK1 and JNK2 . Here we report that JNK3 in LEPRb+ neurons regulates feeding behavior in mice ( Figure 4 ) . The mechanism of JNK3 function requires metabolic stress ( e . g . feeding a HFD ) to cause JNK3 activation . This distinguishes the JNK3 deficiency phenotype from other negative regulators of leptin signaling . Thus , JNK3 deficiency does not cause hyperphagia when mice are fed a chow diet , but JNK3 deficiency does cause hyperphagia when mice are fed a HFD . In contrast , PTPN1-deficiency causes hypophagia on both CD and HFD ( Bence et al . , 2006 ) . This analysis indicates that JNK3 is not required for fine-tuning leptin receptor signaling , but JNK3 is essential for determining the leptin signaling response during exposure to metabolic stress . JNK3 therefore serves a key role in the establishment of the set-point for the threshold of leptin signaling that controls feeding behavior in response to metabolic stress . Gene ablation studies in sub-populations of LEPRb+ neurons demonstrated that HFD ( but not CD ) hyperphagia was found in mice with JNK3 deficiency in AgRP neurons , but not POMC neurons ( Figure 5 ) . These data demonstrate that JNK3 deficiency in AgRP neurons is sufficient to cause HFD hyperphagia , although possible roles for JNK3 in other LEPRb+ neurons cannot be excluded by this analysis . We conclude that orexigenic signaling by AgRP neurons contributes to the effects of JNK3 deficiency on HFD hyperphagia . Molecular mechanisms that account for JNK3 function include altered excitatory transmission to AgRP neurons in HFD-fed mice . Our recordings measured glutamatergic transmission from all inputs to AgRP neurons and demonstrated an increase in mEPSC amplitude , but not frequency , from HFD-fed JNK3-deficient mice compared with HFD-fed control mice ( Figure 6 ) . This observation is consistent with a possible postsynaptic function of JNK3 in AgRP neurons whereby JNK3 affects AMPA and/or NMDA receptor activity within these neurons . Interestingly , glutamatergic input to AgRP neurons stimulates feeding behavior ( Liu et al . , 2012 ) . Previous studies have established functional connections between the JNK signaling pathway and glutamatergic receptor signaling in neurons . For example , the JNK scaffold proteins JIP1/2 can regulate NMDA receptor signaling ( Kennedy et al . , 2007 ) and AMPA receptor phosphorylation by JNK regulates AMPA receptor function and trafficking ( Thomas et al . , 2008 ) . Further studies are required to identify the complete spectrum of JNK3 targets in AgRP neurons . Nevertheless , since an increased AMPA response was detected in JNK3-deficient AgRP neurons ( Figure 6H and Figure 6-figure supplement 1 ) , we conclude that JNK-mediated AMPA receptor regulation ( Thomas et al . , 2008 ) may contribute to the hyperphagic phenotype of HFD-fed JNK3-deficient mice . The results of the present study indicate that JNK3 plays a major role in the regulation of energy balance . This function of JNK3 to regulate feeding behavior differs from the roles of JNK1/JNK2 to regulate energy expenditure and insulin resistance ( Sabio and Davis , 2010 ) . These conclusions are based on loss-of-function studies . A contrasting conclusion has been reported based on gain-of-function studies using transgenic expression of a MKK7-JNK1 fusion protein ( that mimics constitutively activated JNK1 ) in AgRP neurons that causes a small increase in food consumption by CD-fed mice ( Tsaousidou et al . , 2014 ) . Since JNK1-deficient ( Mapk8-/- ) mice do not exhibit altered feeding behavior ( Sabio et al . , 2008 ) and endogenous JNK1 is constitutively activated in neurons ( Coffey , 2014 ) , it is unclear why transgenic over-expression of an activated Mapk8 allele ( encoding a MKK7-JNK1 fusion protein ) in WT mice would cause a small change in feeding behavior . However , the pro-apoptotic function of this activated Mapk8 allele ( Lei et al . , 2002 ) may cause defects in hypothalamic neuronal circuits that contribute to the reported phenotype . On balance , we favor the conclusion that JNK1 and JNK2 do not influence feeding behavior ( Sabio and Davis , 2010 ) , but JNK3 promotes leptin-mediated suppression of HFD feeding behavior ( when JNK3 is activated ) , but not CD feeding behavior ( when JNK3 is inactive ) . The observation that JNK1 and JNK2 promote obesity ( by inhibiting energy expenditure ) and cause insulin resistance in peripheral tissues indicates that drugs that block JNK signaling may be therapeutically beneficial for the treatment of pre-diabetes ( Sabio and Davis , 2010 ) . However , this study demonstrates that JNK3 inhibition causes HFD-dependent hyperphagia ( Figure 1F ) . This represents a potential problem for drug therapy . While JNK1/2 inhibition may be therapeutically beneficial , hyperphagia may therefore result from JNK3 inhibition . Consequently , the most effective drug strategy for the treatment of pre-diabetes may require a small molecule that inhibits JNK1/2 , but not JNK3 . We have described Mapk10-/- mice previously ( Yang et al . , 1997 ) . We obtained C57BL/6J mice ( stock number 000664 ) , B6 . 129S4-Gt ( ROSA ) 26Sortm1 ( FLP1 ) Dym/RainJ ( Farley et al . , 2000 ) ( stock number 009086 ) , B6 . 129-Leprtm2 ( cre ) Rck/J mice ( DeFalco et al . , 2001 ) ( stock number 008320 ) , B6 . FVB-Tg ( Npy-hrGFP ) 1Lowl/J mice ( van den Pol et al . , 2009 ) ( stock number 006417 ) , Agrptm1 ( cre ) Lowl/J mice ( Tong et al . , 2008 ) ( stock number 012899 ) , and Tg ( Pomc1-cre ) 16Lowl/J mice ( Balthasar et al . , 2004 ) ( stock number 005965 ) from the Jackson Laboratory . These mice were backcrossed to the C57BL/6J genetic background . We established Mapk10LoxP/LoxP mice using homologous recombination in C57BL/6N embryonic stem cells , the generation of chimeric mice , and breeding to obtain germ-line transmission of the floxed Mapk10 allele using standard procedures . The mice used for these studies were backcrossed to the C57BL/6J strain . The Frt-Neo cassette was excised by crossing the mice with FLP transgenic mice . Homologous recombination of 5’ arm of the targeting vector was verified by PCR using the primers 1F: 5’-TGTGACCTTCTAATACAG-3’ and 2R: 5’-CCTAAGACTGTCAGAGAG-3’ ( Mapk10+: 135 bp; Mapk10LoxP: 282 bp ) . Homologous recombination of the 3’ arm of the targeting vector was verified by PCR using the primers ( 3F: 5’-CTGAGTGACGTGTGGAG-3’ and 5R: 5’-TCATTGGGTTGGGATATTC-3’ ) followed by digestion with XhoI ( Mapk10+: 1 , 975 bp; Mapk10LoxP: 1026 bp & 1028 bp ) . Cre-mediated recombination between the LoxP sites was detected by PCR using the primers 1F and 4R: 5’-GATTCTCCCTGTCTGAG-3’ ( Mapk10+: 1008 bp; Mapk10Loxp: 1759 bp; Mapk10∆: 171 bp ) . The Mapk10LoxP/LoxP mice were routinely genotyped by PCR using primers 1F and 2R ( Mapk10+: 135 bp; Mapk10LoxP: 282 bp ) . Male mice ( 8 wks old ) were fed a chow diet ( Iso Pro 3000 , Purina ) or a HFD ( F3282 , Bioserve ) for 4 to 12 wks . Body weight was measured on a weekly basis and whole body fat and lean mass were non-invasively measured using 1H-MRS ( Echo Medical Systems , Houston , TX ) . The mice were housed in a facility accredited by the American Association for Laboratory Animal Care ( AALAC ) . The Institutional Animal Care and Use Committee ( IACUC ) of the University of Massachusetts and the University of Cincinnati approved all studies using animals . The clamp studies were performed at the National Mouse Metabolic Phenotyping Center at the University of Massachusetts Medical School . A 2 hr hyperinsulinemic-euglycemic clamp was conducted using overnight fasted conscious mice with a primed and continuous infusion of human insulin ( 150 mU/kg body weight priming followed by 2 . 5 mU/kg/min; Humulin; Eli Lilly ) , and 20% glucose was infused at variable rates to maintain euglycemia ( Kim et al . , 2004 ) . The analysis was performed by the National Mouse Metabolic Phenotyping Centers at the University of Massachusetts Medical School and the University of Cincinnati . The mice were housed under controlled temperature and lighting with free access to food and water . The food/water intake , energy expenditure , respiratory exchange ratio , and physical activity were measured using metabolic cages ( TSE Systems , Chesterfield , MO ) . Intracerebroventricular treatment with leptin was performed using mice with a cannula stereotaxically implanted into the 3rd ventricle ( coordinates from Bregma: anteroventral , -1 . 8 mm; lateral , 0 . 0 mm; dorsoventral , 5 . 0 mm ) . Mice were monitored daily and allowed to recover for 1 week after surgery . Mice received either solvent ( artificial cerebrospinal fluid; aCSF ) or Leptin ( 5 µg ) in 2 µl delivered over 10 min . Leptin treatment by intraperitoneal ( ip ) injection was performed following 3 consecutive days of sham injection . Tissue isolated from mice starved overnight was used to isolate total RNA using the RNAeasy mini kit ( Qiagen ) . Total RNA ( 500 ng ) was converted into cDNA using the high capacity cDNA reverse transcription kit ( Life Technologies , Carlsbad , CA ) . The diluted cDNA was used for real-time quantitative PCR analysis using a Quantstudio PCR PCR machine ( Life Technologies ) . TaqMan assays ( Life Technologies ) were used to quantify Adipoq ( Mm00456425_m1 ) , Agrp ( Mm00475829_g1 ) , Arg1 ( Mm00475988_m1 ) , Ccl2 ( Mm00441242_m1 ) , Emr1 ( F4/80 ) ( Mm00802530_m1 ) , Il1b ( Mm00434228_m1 ) , Il6 ( Mm00446190_m1 ) , Mapk8 ( Jnk1 ) ( Mm00489514_m1 ) , Mapk9 ( Jnk2 ) ( Mm00444231_m1 ) , Mapk10 ( Jnk3 ) ( Mm00436518_m1 ) , Mgl2 ( Mm00460844_m1 ) , Mrc1 ( Mm00485148_m1 ) , Mrc2 ( Mm00485184_m1 ) , Npy ( Mm03048253_m1 ) , Pomc ( Mm00435874_m1 ) , and Tnf ( Mm00443258_m1 ) . The relative mRNA expression was normalized by measurement of the amount of 18S RNA in each sample using Taqman© assays ( catalog number 4308329; Life Technologies ) . Blood glucose was measured with an Ascensia Breeze 2 glucometer ( Bayer , Pittsburgh , PA ) . Adipokines and insulin in plasma were measured by multiplexed ELISA using a Luminex 200 machine ( Millipore , Billerica , MA ) . Glucose and insulin tolerance tests were performed by intraperitoneal injection of mice with glucose ( 1 g/kg ) or insulin ( 1 . 5 U/kg ) using methods described previously ( Sabio et al . , 2008 ) . Mice ( 8–12 week-old ) were fasted overnight . Hypothalamic extracts were prepared using Triton lysis buffer ( 20 mM Tris-pH 7 . 4 , 1% Triton-X100 , 10% glycerol , 137 mM NaCl , 2 mM EDTA , 25 mM β-glycerophosphate , 1 µM sodium orthovanadate , 1 µM PMSF and 10 µg/mL leupeptin and aprotinin ) . Extracts ( 30–50 µg of protein ) were examined by immunoblot analysis by probing with antibodies to JNK3 ( Cell Signaling Technologies , Danvers , MA ) and GAPDH ( Santa Cruz Biotechnology , Dallas , TX ) . Activated JNK was isolated by immunoprecipitation with the mouse monoclonal p-JNK antibody G9 ( Cell Signaling Technologies ) pre-bound to protein G Sepharose ( GE Healthcare , Pittsburgh , PA ) and detected by immunoblot analysis by probing with an antibody to JNK3 ( Cell Signaling Technologies ) . Immunocomplexes were detected by fluorescence using anti-mouse and anti-rabbit secondary IRDye antibodies ( LI-COR Biosciences , Lincoln , NE ) and quantitated using the Li-COR Imaging system Histology was performed using tissue fixed in 10% formalin for 24 h , dehydrated , and embedded in paraffin . Sections ( 7 µm ) were cut and stained using hematoxylin & eosin ( American Master Tech Scientific , Lodi , CA ) . Paraffin sections were stained with an antibody to F4/80 ( Abcam , Cambridge , MA ) that was detected by incubation with anti-rabbit Ig conjugated to Alexa Fluor 488 ( Life Technologies ) . DNA was detected by staining with DAPI ( Life Technologies ) . Fluorescence was visualized using a Leica TCS SP2 confocal microscope equipped with a 405 nm diode laser ( Leica Microsystems , Buffalo Grove , IL ) . Brain slice preparations were performed using 8–10-weeks-old mice anaesthetized with isoflurane before decapitation and removal of the entire brain . The brains were immediately submerged in ice-cold , carbogen-saturated ( 95% O2 , 5% CO2 ) high sucrose solution ( 238 mM sucrose , 26 mM NaHCO3 , 2 . 5 mM KCl , 1 . 0 mM NaH2PO4 , 5 . 0 mM MgCl2 , 10 . 0 mM CaCl2 , 11 mM glucose ) . Then , 300 µm thick coronal sections were cut with a Leica VT1000S Vibratome and incubated in oxygenated aCSF ( 126 NaCl , 21 . 4 mM NaHCO3 , 2 . 5 mM KCl , 1 . 2 mM NaH2PO4 , 1 . 2 mM MgCl2 , 2 . 4 mM CaCl2 , 10 mM glucose ) at 34°C for 30 min . The slices were maintained and recorded at room temperature ( 20–24°C ) . The intracellular solution for voltage clamp recording contained the following: 140 mM CsCl , 1 mM BAPTA , 10 mM HEPES , 5 mM MgCl2 , 5 mM Mg-ATP , and 0 . 3 mM Na2GTP , pH 7 . 35 and 290 mOsm . To isolate glutamatergic , action potential-independent events , minitature excitatory postsynaptic currents ( mEPSCs ) were recorded in the presence of tetrodotoxin ( 1 µM ) and picrotoxin ( 100 μM ) in whole cell voltage clamp mode . To record miniature inhibitory postsynaptic currents ( mIPSCs ) , the neurons were recorded in the presence of TTX and kynurenic acid ( 1 mM ) . The membrane potential was clamped at −60 mV . All recordings were made using a Multiclamp 700B amplifier , and data were filtered at 1 . 4 kHz and digitized at 20 kHz . Data was analyzed using Clampfit 10 . 2 and Origin Pro 8 . 6 . Differences between groups were examined for statistical significance using the Student’s test or analysis of variance ( ANOVA ) with the Fisher’s test .	Consuming the right amount of food is important for health . Eating too little for a long time causes damage to organs , and overeating can cause harm as well , in the form of conditions such as obesity and type 2 diabetes . Several signaling molecules and brain regions are linked to controlling food consumption and ensuring the body receives the correct amount of nutrients to fuel its activities . Previous studies have found that two proteins called JNK1 and JNK2 , which are found in most tissues of the body , can reduce how much energy cells use . This can trigger insulin resistance and fat accumulation , and so suggests that blocking the activity of these proteins may help to treat type 2 diabetes and obesity . However , the role of another JNK protein – JNK3 , which is mostly found in the brain – was not known . Now , Vernia , Morel et al . have investigated the role of JNK3 in metabolism . It was found that JNK3 reduced the amount of food consumed by mice provided with a cafeteria ( high fat ) diet . Mice that lacked JNK3 ate far more food and gained more weight on a high fat diet than normal mice . However , JNK3 played no role in food consumption when mice were fed a standard chow diet . Treating normal mice with leptin – an appetite-suppressing hormone – caused them to lose weight , but did not affect mice that lacked JNK3 . Examining the brains of the mice revealed that in normal mice , JNK3 in a specific sub-population of neurons decreases the production of proteins that promote eating . However , the proteins continued to be produced in mice that lacked JNK3 , encouraging overeating . Overall , the results suggest that blocking the activity of all the JNK proteins will not help treat obesity and diabetes as shutting down JNK3 could encourage overeating . Therefore , future investigation into treatments for these conditions should focus on drugs that specifically target JNK1 and JNK2 , and not JNK3 .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2016	Excitatory transmission onto AgRP neurons is regulated by cJun NH2-terminal kinase 3 in response to metabolic stress
Cancer cells display aneuploid karyotypes and typically mis-segregate chromosomes at high rates , a phenotype referred to as chromosomal instability ( CIN ) . To test the effects of aneuploidy on chromosome segregation and other mitotic phenotypes we used the colorectal cancer cell line DLD1 ( 2n = 46 ) and two variants with trisomy 7 or 13 ( DLD1+7 and DLD1+13 ) , as well as euploid and trisomy 13 amniocytes ( AF and AF+13 ) . We found that trisomic cells displayed higher rates of chromosome mis-segregation compared to their euploid counterparts . Furthermore , cells with trisomy 13 displayed a distinctive cytokinesis failure phenotype . We showed that up-regulation of SPG20 expression , brought about by trisomy 13 in DLD1+13 and AF+13 cells , is sufficient for the cytokinesis failure phenotype . Overall , our study shows that aneuploidy can induce chromosome mis-segregation . Moreover , we identified a trisomy 13-specific mitotic phenotype that is driven by up-regulation of a gene encoded on the aneuploid chromosome . Aneuploidy , an abnormal number of chromosomes , is a leading cause of mis-carriage and birth defects in humans ( Nagaoka et al . , 2012 ) . In the vast majority of cases , this is due to errors occurring in the oocyte ( Nagaoka et al . , 2012 ) . However , aneuploidy can also arise in somatic cells , and a number of studies have reported age-dependent increases in aneuploidy in human peripheral blood lymphocytes ( Nowinski et al . , 1990; Carere et al . , 1999; Leopardi et al . , 2002 ) . Moreover , aneuploidy was recognized as a common feature of cancer cells already a century ago ( Boveri , 1914 , 2008 ) , and a causal role of aneuploidy in carcinogenesis is currently largely acknowledged ( reviewed in [Pavelka et al . , 2010a; Nicholson and Cimini , 2011] ) . In addition to being aneuploid , cancer cells typically display high rates of chromosome mis-segregation , a phenomenon termed chromosomal instability ( CIN ) ( Lengauer et al . , 1997; Bakhoum et al . , 2014 ) . The observation that even mosaic aneuploidy can cause severe physical and cognitive developmental defects ( Biesecker and Spinner , 2013 ) indicates that aneuploidy has pleiotropic deleterious effects . This idea is further supported by a number of experimental observations: first , knocking down spindle assembly checkpoint genes , which results in high rates chromosome mis-segregation and high levels of aneuploidy , invariably causes embryonic lethality in mouse models ( Foijer et al . , 2008 ) . Second , aneuploid yeast strains were shown to exhibit defects in cell cycle progression and metabolism ( Torres et al . , 2007 ) . Third , MEFs derived from mice carrying specific trisomies were found to display cell proliferation defects and metabolic alterations ( Williams et al . , 2008 ) . Finally , genes involved in stress response were shown to be upregulated in aneuploid yeast and human cells ( Sheltzer et al . , 2012; Stingele et al . , 2012 ) . But in the context of cancer , aneuploidy and CIN strongly correlate with drug resistance ( Lee et al . , 2011 ) and poor patient prognosis ( Bardi et al . , 2004; Carter et al . , 2006; Walther et al . , 2008; Sheffer et al . , 2009 ) , indicating that aneuploidy and CIN may confer a proliferative advantage to cancer cells . In support of this idea , certain aneuploidies were found to confer drug resistance in aneuploid Saccharomyces cerevisiae ( Pavelka et al . , 2010b ) and Candida albicans ( Selmecki et al . , 2006 , 2009 ) . These studies suggest that aneuploidy may confer adaptability by inducing chromosome-specific phenotypic changes , despite general negative effects on cell physiology . However , this problem remains to be investigated in human cells . Recent work in aneuploid budding yeast also showed that aneuploidy is sufficient to cause CIN ( Sheltzer et al . , 2011; Zhu et al . , 2012 ) , but whether this is true in human cells is still a matter of debate ( Duesberg , 2014; Heng , 2014; Valind and Gisselsson , 2014a , 2014b ) . In fact , this question has been difficult to address in cancer cells due to the complexity of cancer karyotypes ( Gisselsson , 2011; Mitelman et al . , 2014 ) , and previous studies in human cancer and non-cancer cells have reached discrepant conclusions ( Lengauer et al . , 1997; Duesberg et al . , 1998; Miyazaki et al . , 1999; Valind et al . , 2013 ) . To determine the effect of aneuploidy on chromosome segregation and cell division in human cells , we utilized a number of diploid human cell types and trisomic counterparts , including: colorectal cancer cell line DLD1 ( 2n = 46 ) and trisomic counterparts carrying extra copies of chromosomes 7 or 13 ( DLD1+7 and DLD1+13 , respectively ) ; diploid amniotic fibroblasts ( AF ) and amniotic fibroblasts with trisomy 13 ( AF+13 ) . These different cell types constitute a good model for our study for two main reasons: first , their karyotypes are aneuploid , but not as complex as typically found in tumors and cancer cell lines; second , they represent different cellular models ( transformed and untransformed ) of aneuploidy . To investigate the effect of aneuploidy on chromosome segregation , we analyzed anaphase lagging chromosomes , a common cause of aneuploidy in normal and cancer cells ( Cimini et al . , 2001; Thompson and Compton , 2008 ) . By analyzing fixed cells with immunostained kinetochores and microtubules , we found that DLD1+7 and DLD1+13 cells displayed significantly higher frequencies of anaphase lagging chromosomes compared to the parental DLD1 cell line ( Figure 2A–B ) . We found no evidence of aneuploidy-dependent increases in other mitotic defects , such as multipolar mitoses and anaphase chromosome bridges ( Figure 2—figure supplement 1 ) . Frequencies of anaphase lagging chromosomes in AF and AF+13 cells could not be analyzed in fixed samples due to low mitotic indices . However , we optimized live-cell imaging of AF and AF+13 cells expressing H2B-GFP/RFP-tubulin ( Figure 2C; Videos 1–2 ) and found higher frequencies of anaphase lagging chromosomes in AF+13 compared to AF cells ( Figure 2D ) . Anaphase chromosome bridges or multipolarity were never observed in AF or AF+13 cells . 10 . 7554/eLife . 05068 . 006Figure 2 . Increased rates of anaphase lagging chromosomes in cells with trisomy 7 or 13 . ( A ) Examples of normal anaphases ( top row ) and anaphase cells with lagging chromosomes ( bottom row ) . Cells were immunostained for microtubules ( red ) and kinetochores ( green ) . DNA is shown in blue . Images represent maximum intensity projections of Z-stacks . Arrowheads point at anaphase lagging chromosomes . Grey scale images at the bottom right corners of the images in the bottom row are single focal planes of DAPI-stained chromosomes shown for easier visualization of the lagging chromosomes . ( B ) Frequencies of anaphase lagging chromosomes were significantly higher ( χ2 test , p < 0 . 0001 ) in both DLD1+7 and DLD1+13 compared to DLD1 cells . Data are reported as mean ± S . E . M and represent the average of three independent experiments in which a total of 613–1115 anaphases were analyzed . ( C ) Time-lapse microscopy of AF and AF+13 cells undergoing mitosis . An example of AF undergoing normal mitosis is shown in the top row and an example of AF+13 displaying an anaphase lagging chromosome is shown in the bottom row . DNA is shown in green ( H2B-GFP ) and microtubules in red ( RFP-tubulin ) . Images are maximum intensity projections of Z-stacks . Insets in the bottom row display enlarged views of the DNA alone ( in grey scale ) in the region around the lagging chromosome . ( D ) None of the 19 AF cells ( from cases #1 and #2 ) imaged displayed anaphase lagging chromosomes , whereas 7 out of 26 AF+13 cells ( 5 out of 18 in case #1 and 3 out of 8 in case #2 ) displayed anaphase lagging chromosomes . Time stamps indicate elapsed time in min:sec . Scale bars , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 00610 . 7554/eLife . 05068 . 007Figure 2—figure supplement 1 . Similar frequencies of multipolar mitoses ( A ) and anaphase chromosome bridges ( B ) in diploid vs aneuploid DLD1 cells . The data are reported as mean ± S . E . M . and represent the average of three experiments in which a total of 931–1179 mitotic cells ( A ) or 616–1190 anaphase cells ( B ) were analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 00710 . 7554/eLife . 05068 . 008Video 1 . Representative video showing normal mitosis in an AF cell . Images were acquired by spinning-disk confocal microscopy at 1 min intervals and they are played back at 5 frames per second . DNA is shown in green ( H2B-GFP ) and microtubules in red ( RFP-tubulin ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 00810 . 7554/eLife . 05068 . 009Video 2 . Representative video showing anaphase lagging chromosome during mitosis in an AF+13 cell . Images were acquired by spinning-disk confocal microscopy at 1 min intervals and they are played back at 5 frames per second . DNA is shown green ( H2B-GFP ) and microtubules in red ( RFP-tubulin ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 009 As an additional method to measure chromosome mis-segregation and to account for events in which two sister chromatids co-segregate to the same spindle pole/daughter cell , we combined the cytokinesis-block assay ( Fenech , 1993 ) with FISH staining using locus-specific probes for chromosomes 3 , 7 , 11 , and 13 in DLD1 cell lines , and probes specific for chromosomes 7 , 11 , 12 , 13 , 18 , and 19 in amniocytes . Using this approach , which allows analysis of the reciprocal distribution of chromosomes between the daughter nuclei of a single mitotic division ( Figure 3A , Figure 3—figure supplement 1 ) , we found a significant increase in chromosome mis-segregation in DLD1+7 , DLD1+13 , and AF+13 compared to the corresponding diploid cell cultures ( Figure 3B–C ) . However , the higher mis-segregation rates were specific to certain chromosomes . Namely , mis-segregation appeared to be increased for chromosome 7 in the DLD1+7 cell line , chromosomes 7 and 13 in DLD1+13 cells , and chromosome 13 in AF+13 cells as compared to their diploid counterparts . We further investigated whether the observed increases in chromosome mis-segregation rates impacted chromosome number variability ( or karyotypic heterogeneity ) in the trisomic cells compared to the diploid counterparts . To this end , we performed chromosome counts in metaphase spreads and found increased karyotypic heterogeneity in both DLD1+7 and DLD1+13 compared to DLD1 cells . We also found higher karyotypic heterogeneity in AF+13 vs AF cells around the modal chromosome number of 47 ( Figure 3D ) . Finally , our chromosome counts revealed the presence of a tetraploid/near-tetraploid sub-population in DLD1+13 cells ( Figure 3D ) , confirming previous findings ( Upender et al . , 2004 ) . Because chromosome counts did not reveal a significant difference in tetraploid cells between AF and AF+13 cell populations ( Figure 3D ) , we decided to further characterize chromosome number variability in these cells by performing FISH analysis with locus-specific probes for chromosomes 7 , 12 , and 18 on interphase nuclei ( Figure 3E ) , which allowed for larger numbers of cells to be examined . This analysis confirmed higher degrees of aneuploidy in AF+13 compared to AF cells ( Figure 3F ) . Furthermore , it revealed the presence of a tetraploid sub-population in AF+13 cells ( Figure 3F ) . The difference between numbers of AF+13 tetraploid interphase cells and tetraploid chromosome spreads ( compare Figure 3D and Figure 3F ) may be due to the inability of tetraploid AF+13 cells to re-enter mitosis ( see also ‘Discussion’ section ) . Taken together , these experiments ( Figures 2–3 ) show that trisomies 7 and 13 cause chromosome mis-segregation , that mis-segregation affects certain chromosomes more than others , and that such increases in mis-segregation rates are associated with karyotypic heterogeneity within the cell population . However , because only trisomies 7 and 13 were examined , and a limited number of chromosomes analyzed in our FISH experiments , it remains elusive whether chromosome mis-segregation is a karyotype-specific or a general effect of aneuploidy in human cells . 10 . 7554/eLife . 05068 . 010Figure 3 . Increased chromosome mis-segregation rates and karyotypic heterogeneity in cells with trisomy 7 or 13 . ( A–C ) Combination of cytokinesis-block assay and FISH staining with chromosome-specific probes shows higher chromosome mis-segregation rates in DLD1+7 and DLD1+13 compared to DLD1 cells and AF+13 compared to AF cells . ( A ) Examples of FISH-stained binucleate ( BN ) DLD1 and DLD1+13 cells . Scale bar , 5 μm . ( B–C ) Frequencies of BN cells displaying mis-segregation events . Two-tailed χ2 test , p < 0 . 005; *two-tailed χ2 test , p < 0 . 0001 , when compared to mis-segregation of the same chromosome in the euploid cell line . Data are presented as mean ± S . E . M . and represent the average of at least three independent experiments/samples in which a total of 460–1229 BN cells were analyzed . ( D ) Beeswarm plot displaying data from chromosome counts in metaphase spreads from the five cell lines . DLD1+7 , DLD1+13 , and AF+13 ( modal chromosome number 47 , shown by the high concentration of sampled points ) displayed increased karyotypic heterogeneity compared to DLD1 and AF cells ( modal chromosome number 46 , shown by the high concentration of sampled points ) , respectively . In addition , DLD1+13 cells displayed a large sub-population of near-tetraploid cells ( modal chromosome number 92 ) . Chromosome counts were performed on 89–303 metaphase spreads . ( E–F ) FISH staining with chromosome-specific probes in interphase nuclei reveals higher rates of aneuploidy and tetraploidy in AF+13 vs AF cells . ( E ) FISH staining in interphase AF and AF+13 cells with probes specific for chromosomes 7 ( blue ) , 12 ( green ) , and 18 ( red ) . Scale bar , 5 μm . ( F ) Quantification of interphase FISH data shown in ( E ) . Cells were classified as having gained or lost 1-few chromosomes or as tetraploid ( 4 signals for each of the three chromosomes analyzed ) . The data show a larger fraction of cells with gain/loss of 1-few chromosomes in the AF+13 population compared to AF cells ( χ2 test , p < 0 . 0001 ) . Additionally , AF+13 cells displayed a larger sub-population of tetraploid cells ( χ2 test , p < 0 . 0001 ) . Data are presented as mean + S . E . M . and represent the average of three different samples in which a total of 2 , 117–2 , 410 cells were analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01010 . 7554/eLife . 05068 . 011Figure 3—figure supplement 1 . Combined cytokinesis-block assay and FISH staining with chromosome-specific probes . The images show examples of binucleate AF , AF+13 , and DLD1+7 cells hybridized with FISH probes specific to chromosomes 11 and 13 . Green arrows in image of mis-segregation in AF+13 cell point at signals for chromosome 13 . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 011 Our chromosome counts ( Figure 3D ) showed that DLD1+13 cells displayed a near-tetraploid sub-population , which was also evident in our FISH experiments in which nuclei with four or more signals per chromosome were more frequent in DLD1+13 compared to DLD1 cells ( 24 . 5% vs 16 . 6% , χ2 , p < 0 . 0001 ) . Similarly , a tetraploid sub-population was evident in AF+13 cells ( 4 . 5% vs 0 . 3% in AF , χ2 , p < 0 . 0001 ) analyzed by interphase FISH , which revealed the presence of nuclei with four signals per chromosome ( Figure 3E–F ) . These observations suggested a possible causal link between trisomy 13 and tetraploidy . Acknowledged mechanisms of tetraploidy induction include mitotic slippage ( Rieder and Maiato , 2004 ) , cytokinesis failure ( Normand and King , 2010 ) , and cell fusion ( Duelli and Lazebnik , 2003 ) . To determine which of these mechanisms cause tetraploidy in DLD1+13 and AF+13 cells , we performed phase contrast time-lapse microscopy and found no evidence of mitotic slippage or cell fusion . Instead , we found that both DLD1+13 and AF+13 cells failed cytokinesis ( Figure 4A–B , Videos 3–6 ) at significantly higher rates than their diploid counterparts ( Figure 4C ) . To identify the molecular mechanism that causes cytokinesis failure in cells with trisomy 13 , we referred to microarray data available for DLD1+13 cells ( Upender et al . , 2004 ) . Interestingly , located on chromosome 13q13 . 3 is the gene SPG20 , which encodes for the protein Spartin , previously suggested to act as a regulator of cytokinesis ( Renvoise et al . , 2010; Lind et al . , 2011 ) , and shown to be overexpressed in DLD1+13 compared to DLD1 cells ( Upender et al . , 2004 ) . None of the other mis-expressed genes ( Upender et al . , 2004 ) was found to have any link with cytokinesis based on published data . We confirmed SPG20 overexpression by western blot in both DLD1+13 and AF+13 cells ( Figure 4D–F ) . Importantly , neither DLD1+7 ( Figure 4E ) nor other trisomic AF cells ( Figure 4—figure supplement 1 ) overexpressed spartin , indicating that high levels of spartin are specifically associated with trisomy 13 . 10 . 7554/eLife . 05068 . 012Figure 4 . DLD1+13 and AF+13 cells overexpress SPG20 and fail cytokinesis at high rates . ( A–B ) Time-lapse microscopy indicates that cells with trisomy 13 frequently fail cytokinesis . Time stamps indicate elapsed time in minutes . Scale bars , 10 μm . ( A ) Still images from time-lapse phase contrast videos of DLD1+13 cells undergoing mitosis and completing ( top row ) or failing ( bottom row ) cytokinesis . ( B ) Still images from time-lapse phase contrast videos of AF ( top row ) and AF+13 ( bottom row ) cells undergoing mitosis and completing ( AF , top row ) or failing ( AF+13 , bottom row ) cytokinesis . ( C ) Quantification of cytokinesis failure from phase-contrast time-lapse videos showing that the rates of cytokinesis failure in DLD1+13 and AF+13 cells are significantly higher than those observed in their diploid counterparts , DLD1 and AF cells ( χ2 test , p < 0 . 01; ** χ2 test , p < 0 . 001 ) . ( D ) Western blot analysis of Spartin across DLD1 cell lines , three samples of AF and three samples of AF+13 cells . β-actin was used as a loading control . ( E ) Quantification of spartin levels ( normalized to β-actin ) in DLD1 , DLD1+7 , and DLD1+13 cells . The data reported are the average of three independent experiments and are displayed as mean + S . E . M . ( F ) Quantification of spartin levels ( normalized to β-actin ) in AF and AF+13 cells . The data reported are the average of the three AF and AF+13 samples shown in ( E ) and are displayed as mean + S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01210 . 7554/eLife . 05068 . 013Figure 4—figure supplement 1 . Trisomies other than 13 are not associated with spartin overexpression . ( A ) Western blot analysis of Spartin levels in AF control and AF+13 , AF+18 , and AF+21 trisomic cells . Spartin levels were normalized to β-actin and are shown as percentage of the amount in the euploid control . ( B ) Western blot with decreasing amounts ( 30 , 20 , and 10 μg ) of total extracts from the trisomies used in ( A ) performed for titration of the western blot conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01310 . 7554/eLife . 05068 . 014Video 3 . Representative video showing normal cytokinesis in a DLD1+13 cell . Images were acquired by phase contrast microscopy at 2 min intervals , and they are played back at 7 frames per second . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01410 . 7554/eLife . 05068 . 015Video 4 . Representative video showing cytokinesis failure in a DLD1+13 cell . Images were acquired by phase contrast microscopy at 2 min intervals , and they are played back at 7 frames per second . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01510 . 7554/eLife . 05068 . 016Video 5 . Representative video showing normal cytokinesis in an AF cell . Images were acquired by phase contrast microscopy at 2 . 5 min intervals , and they are played back at 5 frames per second . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01610 . 7554/eLife . 05068 . 017Video 6 . Representative video showing cytokinesis failure in an AF+13 cell . Images were acquired by phase contrast microscopy at 2 . 5 min intervals , and they are played back at 5 frames per second . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 017 To test whether SPG20 overexpression could explain the cytokinesis-failure phenotype , we transfected the parental cell line DLD1 with YFP-SPG20 ( DLD1-YFP-SPG20; Figure 5A , Videos 7–8 ) , and found that high levels of Spartin ( Figure 5B ) induced high rates of cytokinesis failure ( Figure 5C ) . Moreover , we could rescue the cytokinesis failure phenotype in both DLD1+13 and AF+13 cells by siRNA-mediated Spartin knockdown ( Figure 5D–G ) . Thus , we conclude that the aneuploidy-dependent overexpression of Spartin in DLD1+13 and AF+13 cells induces cytokinesis failure , a karyotype-dependent phenotype . 10 . 7554/eLife . 05068 . 018Figure 5 . Spartin overexpression induces cytokinesis failure . ( A ) Time-lapse microscopy of DLD1 cells transiently transfected with a YFP-N1 vector ( DLD1-YFP , top row ) or a YFP-SPG20-N1 vector ( DLD1-YFP-SPG20 , bottom row ) . Representative still images of time-lapse videos show a DLD1-YFP cell undergoing mitosis and completing cytokinesis ( top row ) and a DLD1-YFP-SPG20 cell undergoing mitosis and failing cytokinesis ( bottom row ) . YFP expression was verified by fluorescence imaging and it is shown in the first panel for each time-lapse series . A copy of the last frame was added at the end of the sequence to highlight cell ( yellow ) and nuclear ( green ) outlines . Scale bar , 10 μm . ( B ) Western blot analysis of Spartin levels in DLD1+13 , DLD1-YFP , and DLD1-YFP-SPG20 cells shows that the levels of YFP-Spartin in DLD1-YFP-SPG20 cells ( center lane ) are much higher than the levels of Spartin in DLD1-YFP cells . WB of Spartin in DLD1+13 cells is shown for comparison . ( C ) Quantification of cytokinesis failure rates in DLD1-YFP and DLD1-YFP-SPG20 showing that overexpression of SPG20 leads to increased rates of cytokinesis failure ( χ2 test , p < 0 . 01 when comparing DLD1-YFP-SPG20 to DLD1-YFP cells ) . ( D ) Western blot analysis of Spartin levels in DLD1+13 cells treated with a SPG20 siRNA or with a control siRNA . ( E ) Reducing the levels of Spartin by SPG20 siRNA significantly reduces the rate of cytokinesis failure in DLD1+13 cells ( χ2 test , p < 0 . 02 when comparing cells treated with a control siRNA to cells treated with SPG20 siRNA ) . ( F ) Western blot analysis of Spartin levels in AF+13 cells treated with a SPG20 siRNA or with a control siRNA . ( G ) Reducing the levels of Spartin by SPG20 siRNA significantly reduces the rates of cytokinesis failure in AF+13 cells ( χ2 test , p < 0 . 02 when comparing cells treated with a control siRNA to cells treated with SPG20 siRNA ) . Average values from three independent experiments; variability between AF+13 samples was not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01810 . 7554/eLife . 05068 . 019Video 7 . Representative video showing normal cytokinesis in a DLD1 cell transiently transfected with a YFP vector ( control ) . Near-simultaneous phase contrast and epifluorescence images were acquired at 4 min intervals at a single focal plane using the Nikon perfect focus function . For clarity , the fluorescent image is displayed only in the first frame , whereas the rest of the video shows only phase contrast images played back at 7 frames per second . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 01910 . 7554/eLife . 05068 . 020Video 8 . Representative video showing cytokinesis failure in a DLD1 cell transiently transfected with a YFP-SPG20 vector ( SPG20 overexpression ) . Near-simultaneous phase contrast and epifluorescence images were acquired at 4 min intervals at a single focal plane using the Nikon perfect focus function . For clarity , the fluorescent image is displayed only in the first frame , whereas the rest of the video shows only phase contrast images played back at 7 frames per second . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 020 To determine how Spartin overexpression may lead to cytokinesis failure , we analyzed the amount and localization of Spartin in fixed DLD1 cells ( Figure 6A–B , Figure 6—figure supplement 1 ) . Spartin localized to the centrosomes throughout mitosis and to some extent along the microtubules of the mitotic spindle ( Figure 6—figure supplement 1 ) , and localized to the midbody during cytokinesis ( Figure 6A ) , as previously described ( Lind et al . , 2011 ) . We quantified the total intracellular amount of Spartin by measuring total Spartin fluorescence intensity in interphase cells , and found that it was significantly higher in DLD1+13 cells compared to DLD1 cells ( Figure 6B; see also Figure 4D ) . However , we did not observe any difference in Spartin localization between the two cell lines during mitosis ( Figure 6A , Figure 6—figure supplement 1 ) , although there was clearly a large amount of Spartin in the cytoplasm , away from the mitotic spindle , in DLD1+13 , but not in the other cell lines ( Figure 6—figure supplement 1 ) . Thus , although Spartin overexpression induces cytokinesis failure ( Figure 5A–C ) , the mechanism by which this happens is not simply mis-localization of Spartin at the midbody ( Figure 6A ) . 10 . 7554/eLife . 05068 . 021Figure 6 . Spartin overexpression impairs Spastin localization to the midbody . ( A–B ) Spartin localization at the midbody is not affected by the high levels of intracellular Spartin in DLD1+13 . ( A ) Images showing Spartin localization ( arrows ) at the midbody of the three DLD1 cell lines . ( B ) The image shows interphase DLD1+13 cells immunostained for Spartin and the yellow and white outline indicate the regions of interest ( ROI ) selected for measurements of total intracellular fluorescence ( yellow ROI ) and background fluorescence ( white ROI ) . The data in the graph report the total intracellular Spartin fluorescence intensity after background subtraction in randomly sampled interphase cells and are represented as mean ± S . E . M ( t-test , p < 0 . 0001 when comparing DLD1+13 to either DLD1 or DLD1+7 ) . ( C ) Superimposition of the MIT domains of Spartin ( orange ) and Spastin ( cyan ) illustrates the considerable degree of structural homology between the two . PDB ID #2DL1 ( Suetake et al . , 2009 ) and #3EAB ( Yang et al . , 2008 ) for Spartin and Spastin , respectively . ( D–E ) Spastin localization at the midbody is impaired in DLD1+13 cells , and is rescued by SPG20 siRNA . ( D ) Images of midbodies of cells immunostained for Spastin ( green ) and microtubules ( red ) . Arrows point at sites of Spastin localization or lack thereof . ( E ) Frequencies of cells lacking Spastin at the midbody . Higher frequencies of cells lacking Spastin at the midbody can be found in DLD1+13 cells compared to either DLD1 or DLD1+7 cells . Spastin localization could be rescued by knocking down Spartin levels in DLD1+13 cells via SPG20 siRNA . Statistical significance was calculated using a χ2 test ( p < 0 . 05; p < 0 . 01; N . S . = not significant ) . Data are presented as mean ± S . E . M . and represent the average of three independent experiments . ( F–G ) Spastin localization at the midbody is impaired in AF+13 cells , and is partially rescued by SPG20 siRNA . ( F ) Images of midbodies of cells immunostained for Spastin ( green ) and microtubules ( red ) . DNA is shown in blue in the merged images . Arrows point at sites of Spastin localization or lack thereof . The insets show 3× magnifications of the midbody region . ( G ) Frequencies of cells lacking Spastin at the midbody are higher in AF+13 compared to AF cells . Spastin localization was partially rescued by SPG20 siRNA-mediated Spartin knock down in AF+13 cells . Statistical significance was calculated using a χ2 test ( p < 0 . 05; *p < 0 . 001; **p < 0 . 0001 ) . Data are presented as mean ± S . E . M . and represent the average of three independent experiments/samples . Scale bars , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 02110 . 7554/eLife . 05068 . 022Figure 6—figure supplement 1 . Spartin localization during mitosis . Cells were immunostained for Spartin ( green ) and microtubules ( red ) and counterstained with DAPI for DNA ( blue ) . For each cell line , the merged images ( top ) and Spartin alone ( bottom ) are shown . Images are maximum intensity projections of Z-stacks . Spartin localizes to the spindle poles during prometaphase , metaphase , and anaphase in DLD1 , DLD1+7 and DLD1+13 . An eccess of cytoplasmic spartin ( away from the mitotic spindle ) can be noticed in DLD1+13 cells , but not in the other two cell lines . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05068 . 022 Spartin is recruited to the midbody by binding hIST1 ( Renvoise et al . , 2010 ) , a component of the ESCRTIII complex , which binds various proteins involved in cytokinesis , including the microtubule severing protein Spastin ( Renvoise et al . , 2010 ) , whose depletion was shown to cause cytokinesis failure ( Bajorek et al . , 2009 ) . Both Spartin and Spastin bind hIST1 through their MIT ( Microtubule Interacting and Trafficking ) domains , which show considerable structural homology ( Figure 6C ) and comparable binding affinities ( Spartin , Kd = 10 . 4 ± 0 . 3 μM , Spastin , Kd = 4 . 6 ± 0 . 1 μM , respectively [Renvoise et al . , 2010] ) . Therefore , we postulated that Spartin overexpression might act in a dominant negative manner by preventing Spastin from binding to hIST1 . To test this , we analyzed Spastin localization at the midbody . Consistent with our hypothesis , DLD1+13 and AF+13 cells frequently lacked Spastin at the midbody ( Figure 6D–G ) . Moreover , by knocking down Spartin we could rescue Spastin mis-localization fully in DLD1+13 ( Figure 6E ) and partially in AF+13 cells ( Figure 6G ) . In summary , we showed that overexpression of SPG20 , a gene on chromosome 13 encoding the protein Spartin , can cause cytokinesis failure in cells with trisomy 13 ( Figures 4–5 ) . Although Spartin overexpression may cause cytokinesis failure by interfering with multiple pathways , here we provide evidence of interference with a pathway responsible for Spastin localization at the midbody ( Figure 6D–G ) . We show here that cells with trisomy 7 or trisomy 13 display rates of anaphase lagging chromosomes that are significantly higher than the rates observed in euploid counterparts . Anaphase lagging chromosomes are a major source of aneuploidy in normal vertebrate cells ( Cimini et al . , 2001 ) and the main type of chromosome segregation defect observed in CIN cancer cells ( Thompson and Compton , 2008; Bakhoum et al . , 2014 ) . Previous studies have shown that anaphase lagging chromosomes can be caused by transient spindle multipolarity in CIN cancer cells ( Ganem et al . , 2009; Silkworth et al . , 2009; Silkworth and Cimini , 2012 ) . However , we can exclude transient multipolarity as a cause of anaphase lagging chromosomes in our experimental systems , given that we did not find differences in the frequencies of multipolar mitoses in DLD1+7 and DLD1+13 compared to DLD1 cells and we did not observe transient spindle multipolarity in live AF+13 . This finding may seem surprising , particularly considering that we found trisomy 13 to be associated with cytokinesis failure , which is believed to result in extra centrosomes and multipolar mitoses ( Storchova and Pellman , 2004; Fujiwara et al . , 2005 ) . However , this finding is in agreement with recent findings showing that experimental inhibition of cytokinesis can produce tetraploid cells with normal centrosome number ( Godinho et al . , 2014 ) . Our finding that aneuploidy is associated with increased rates of anaphase lagging chromosomes , known to arise from errors in mitosis ( Bakhoum et al . , 2014 ) , but not with multipolarity or chromosome bridges , known to arise from errors in centrosome duplication or DNA metabolism ( occurring prior to mitosis ) , suggests that compared to events occurring during other cell cycle stages , mitotic events may be more sensitive to the gene imbalance brought about by aneuploidy . Although abnormal chromosome number was reported by others as not being sufficient to cause CIN ( Lengauer et al . , 1997; Valind et al . , 2013 ) , such difference may be due to the different methods used to evaluate CIN . For example , whereas we examined chromosome mis-segregation in mitosis ( anaphase lagging chromosomes ) or at a post-mitotic interphase ( BN cell analysis ) , Lengauer and colleagues determined the degree of aneuploidy in the overall population after 25 serial passages ( Lengauer et al . , 1997 ) . This kind of analysis may produce a biased result because selective pressure against arising aneuploid cells may mask the ability of aneuploidy to induce chromosome mis-segregation at each cell cycle . Importantly , we were able to perform high-resolution live cell imaging of individual human amniocytes with constitutional trisomy 13 at passage number 2–3 and directly measure how frequently chromosome mis-segregation events occur . The frequency of lagging chromosomes found in mitotic cells was considerably higher compared to the frequency of chromosome mis-segregation measured by FISH analysis both in interphase and post-mitotic fixed cells . This shows how different results are generated from distinct methods and might also explain the divergence between our data and those previously reported by Valind et al . ( 2013 ) using interphase FISH analysis . It should also be noted that , although we do observe an increase in the rates of anaphase lagging chromosomes in trisomic cells , such rates are lower than those reported for most CIN cancer cell lines ( Thompson and Compton , 2008; Nicholson and Cimini , 2013; Bakhoum et al . , 2014 ) , in agreement with findings by Valind et al . ( 2013 ) . This suggests that , although single chromosome gains may result in only modest increases in chromosome mis-segregation rates , the high degrees of aneuploidy typical of most cancer cells may have a cumulative effect and could thus explain the high rates of CIN displayed by many cancer cells ( e . g . , 20–75% anaphase lagging chromosomes in cells with modal chromosome number 66–78; [Lengauer et al . , 1997; Thompson and Compton , 2008; Ganem et al . , 2009; Silkworth et al . , 2009; Nicholson and Cimini , 2011 , 2013] ) . Finally , and perhaps most importantly , the discrepancies in the conclusions reached in different studies may depend on the specific chromosomes analyzed . Indeed , we find that the degree and type of CIN elicited by aneuploidy depend on the aneuploid chromosome ( see Figure 3D ) . This is an agreement with findings in aneuploid budding yeast strains , in which different aneuploidies were found to result in different rates of chromosome loss/CIN ( Sheltzer et al . , 2011; Zhu et al . , 2012 ) . Similarly , Valind and colleagues reported increased rates of CIN for certain chromosomes in specific aneuploid contexts ( e . g . , increased CIN for chromosome 17 in cells with trisomy 18 ) , but not in others ( Valind et al . , 2013 ) . In our study , we specifically found that high rates of mis-segregation for chromosome 7 were observed both in DLD1+7 and DLD1+13 cells , whereas high mis-segregation for chromosome 13 was observed in DLD1+13 and AF+13 cells . The observation that the trisomic chromosome displayed the highest rates of mis-segregation across the trisomies studied ( Figure 3B–C ) suggests that the aneuploid chromosomes may undergo changes that affect their mitotic behavior and segregation , such as delayed replication and/or delayed condensation timing ( DRT and DCT , respectively ) . Indeed , aneuploidy was shown to correlate with DRT and DCT ( Grinberg-Rashi et al . , 2010 ) and previous studies in trisomic cells showed that one of the chromosomes of the trisomic set displays DRT ( Kost-Alimova et al . , 2004 ) . On the other hand , the finding that chromosome 7 mis-segregated in DLD1+13 , but not in AF+13 cells raises the question as to whether types and rates of mis-segregation may vary in a cell type-dependent manner and whether copy number variations in the DLD1-derived cell lines may account for chromosome 7 mis-segregation . Given that the gain of chromosome 7 and chromosome 13 is commonly found in colon cancer ( Ried et al . , 2012 ) , a cell type-specific effect is plausible . Although our cytokinesis-block assay data ( Figure 3B–C ) show the higher mis-segregation rates to be limited to certain chromosomes , the chromosome count data show extensive karyotypic heterogeneity in the trisomic cell populations ( Figure 3D–F ) , thus suggesting that chromosome mis-segregation is more widespread than the cytokinesis-block assay reveals . One explanation is that the limited number of cells and chromosomes analyzed in the cytokinesis-block assay might only reveal differences in mis-segregation rates when such differences are large , but other methods , such as anaphase lagging chromosomes may be a better indicator of general mis-segregation rates . Moreover , some mis-segregation events may result in cell cycle arrest or cell death . Cases of chromosome mis-segregation leading to cell death would only be accounted for when examining cells undergoing mitosis , but not interphase cells . On the other hand , chromosome mis-segregation events leading to cell cycle arrest may only be appreciated when examining interphase nuclei , but not BN or mitotic cells ( Figure 3D–F ) . These considerations argue for the use of multiple assays in studies aimed at dissecting the link between aneuploidy and CIN ( Nicholson and Cimini , 2015 ) . We would also like to point out that chromosome mis-segregation events causing cell death or cell cycle arrest under tissue culture conditions , may not do so in the context of the tumor environment , suggesting that low level aneuploidy could be enough to drive CIN in cancer . Indeed , our data show that the complex aneuploidies observed in cancer cells are not a requirement for increased rates of lagging chromosomes and that at least some low grade and constitutional aneuploidies are sufficient to induce such an effect . This would be in agreement with previous observations showing that haploid budding yeast strains carrying disomies displayed increased genomic instability , and strains with different degrees of aneuploidy displayed variable degrees of chromosomal instability ( Sheltzer et al . , 2011; Zhu et al . , 2012 ) . Similarly , previous reports have shown that lymphocytes of congenitally trisomic individuals displayed aneuploidies for chromosomes other than the congenitally trisomic ones ( Reish et al . , 2006 , 2011 ) . Finally , random aneuploidy and senescent phenotypes were recently reported in aneuploid amniocytes ( Biron-Shental et al . , 2015 ) . Nonetheless , we do not exclude the possibility that certain aneuploidies may not be sufficient to induce chromosome mis-segregation . Our finding that trisomy 13 caused a specific cytokinesis failure phenotype ( Figures 4–5 ) clearly indicates that different karyotypes can be associated with distinct phenotypic changes . It is important to note that , although we observed cytokinesis failure in AF+13 , the frequency of tetraploid metaphase spreads in these cells was very low , as opposed to the DLD1+13 cells ( Figure 3D ) . This may be due to activation of a post-mitotic tetraploidy checkpoint in the AF+13 cells , but not in DLD1+13 . Indeed , previous studies have shown a cell cycle arrest following cleavage failure in untransformed human cells ( Andreassen et al . , 2001; Krzywicka-Racka and Sluder , 2011 ) , whereas transformed cells continue cycling ( Duelli et al . , 2007; Panopoulos et al . , 2014 ) , but the triggering of a p53-dependent arrest in tetraploid cells still remains a matter of debate ( Andreassen et al . , 2001; Stukenberg , 2004; Uetake and Sluder , 2004; Fujiwara et al . , 2005; Wong and Stearns , 2005 ) . The observation that overexpression of a gene mapping on chromosome 13 is specifically linked to the cytokinesis failure phenotype in DLD1+13 and AF+13 cells demonstrates that there is a direct causal relationship between aneuploidy , overexpression of genes on the aneuploid chromosome , and phenotypic changes caused by the consequent proteomic imbalance . These findings support previous studies showing that aneuploidy directly affects transcript and protein levels in various systems in a karyotype-dependent manner ( Pollack et al . , 2002; Upender et al . , 2004; Gao et al . , 2007; Pavelka et al . , 2010b; Ried et al . , 2012; Stingele et al . , 2012; Gemoll et al . , 2013 ) . Previous studies in budding yeast also showed that aneuploid strains displayed karyotype-specific phenotypic variations that conferred resistance to a variety of drugs ( Pavelka et al . , 2010b ) . However , such phenotypic variations in aneuploid yeast strains were only revealed when cells were grown under specific selective conditions ( Pavelka et al . , 2010b ) , whereas we show here that such karyotype-dependent phenotypic changes can be intrinsic to the aneuploid cells . Although some studies have shown that aneuploidy can have overall deleterious effects on cell fitness ( Torres et al . , 2007; Williams et al . , 2008 ) , we provide strong evidence that specific aneuploidies can also induce specific phenotypes , which could , under certain conditions , provide a selective advantage . This particular concept was recently exploited to demonstrate that in fungi , certain drug treatments can lead to the evolution of populations with defined aneuploid karyotypes ( Chen et al . , 2015 ) . In the particular case observed in our study , one could also envision how the increase in tolerance to aneuploidy that tetraploidy was recently shown to confer ( Dewhurst et al . , 2014 ) could enable the evolution of specific aneuploid karyotypes that may allow cells to overcome the detrimental impact of aneuploidy on cellular fitness ( Gordon et al . , 2012 ) . Karyotype-specific phenotypic changes such as those observed in our study can also explain the recurrent aneuploidies that are found in tumors from certain anatomical sites ( e . g . , gain of chromosome 13 and loss of chromosome 18 in colon cancer [Ried et al . , 1996 , 2012; Nicholson and Cimini , 2013] ) . Indeed , aneuploidies for certain chromosomes may result in phenotypes that confer a selective advantage at a certain site ( e . g . , breast ) , but not at a different one ( e . g . , colon ) . And this could explain why , despite the high degrees of aneuploidy and the extensive karyotypic heterogeneity , the distribution of aneuploidies in different cancers is not completely random ( Nicholson and Cimini , 2011; Ried et al . , 2012 ) . The DLD1 cell line was obtained from American Type Culture Collection ( ATCC , BA , USA ) , DLD1+7 and DLD1+13 cell lines were created previously by microcell-mediated chromosome transfer as described in ( Upender et al . , 2004 ) , and DLD1+13 cells were sub-cloned for this study as described in the results section . All cell lines were maintained in RPMI 1640 ( ATCC , BA , USA ) supplemented with 10% FBS ( Gibco , Life Technologies , CA , USA ) , penicillin , streptomycin , and amphotericin B ( antimycotic ) . Passage 1–3 fibroblast cultures were established from surplus amniocentesis samples used in pre-natal diagnosis . Three cases of constitutional trisomy 13 and three diploid controls were used in our study ( Table 1 ) . The study acknowledged the ethics guidelines under national rules and according to the principles of the Declaration of Helsinki , and was approved by the Ethics Committee of Hospital de S . João-Porto ( dispatch 14 November 2012 ) . Informed consent forms with detailed information were provided to all patients . The study did not imply collection of extra material from the healthy female donors ( only surplus cells/tissues were used ) ; the study did not bring any direct benefits to the volunteers; there were no risks or costs for the volunteers; there was no access to patient clinical data ( samples were obtained in anonymous form from the Hospital Genetics Department ) ; participation was volunteer and free to be interrupted at any moment; there are no ethical impacts predicted; there will be no commercial interests . Amniotic fibroblasts were grown in EMEM ( Lonza , Bazel , Switzerland ) supplemented with 15% FBS , 2 . 5 mM glutamine and 1× antibiotic-antimycotic solution ( all from Gibco , Life Technologies , CA , USA ) . All cells were kept in a humidified incubator at 37°C with 5% CO2 . For immunostaining , DLD1 cells were grown on sterilized glass coverslips inside 35 mm Petri dishes , whereas AF cells were grown on sterilized glass coverslips coated with fibronectin ( Sigma Aldrich , MO , USA ) . For analysis of anaphase lagging chromosomes in the DLD1 lines , cells were fixed in freshly prepared 4% paraformaldehyde in PHEM ( 60 mM Pipes , 25 mM HEPES , 10 mM EGTA , 2 mM MgSO4 , pH 7 . 0 ) for 20 min at room temperature and then permeabilized for 10 min at room temperature in PHEM buffer containing 0 . 5% Triton-X 100 . Following fixation and permeabilization , cells were washed with PBS 3 times and then blocked with 10% boiled goat serum ( BGS ) for 1 hr at room temperature . Cells were then incubated at 4°C overnight with primary antibodies diluted in 5% BGS . Cells were washed in PBS-T ( PBS with 0 . 05% Tween 20 ) 3 times , and incubated at room temperature for 45 min with secondary antibodies diluted in 5% BGS . Cells were finally washed , stained with DAPI for 5 min , and coverslips were mounted on microscope slides in an antifade solution containing 90% glycerol and 0 . 5% N-propyl gallate . For analysis of proteins at the midbody , cells were fixed in ice cold methanol for 4 min , washed with PBS 3 times , and blocked in 10% BGS with 0 . 5% Triton-X for 1 hr . The rest of the procedure was the same as described above , except that primary antibodies were diluted in 1% BGS . Primary antibodies were diluted as follows: ACA ( human anti-centromere protein , Antibodies Inc . , CA , USA ) , 1:100; mouse anti-α-tubulin ( DM1A , Sigma Aldrich , MO , USA ) , 1:500; rabbit anti-SPG20 ( Protein Tech Group Inc . , IL , USA ) , 1:300; mouse anti-Spastin ( SP 3G11/1 , Abcam , Cambridge , UK ) , 1:150 . Secondary antibodies were diluted as follows: Rhodamine Red-X goat anti-mouse ( Jackson ImmunoResearch Laboratories , Inc . , PA , USA ) , 1:100; Rhodamine Red-X goat anti-rabbit ( Jackson ImmunoResearch Laboratories , Inc . , PA , USA ) , 1:100; Alexa 488 goat anti-human ( Molecular Probes , Life Technologies , CA , USA ) , 1:200; Alexa 488 goat anti-mouse ( Molecular Probes , Life Technologies , CA , USA ) , 1:200 . Cell cultures were incubated in 50 ng/ml Colcemid ( Karyomax , Invitrogen ) at 37°C for 3–4 hr to enrich in mitotically arrested cells . The cells were then collected by trypsinization and centrifuged at 800 rpm for 8 min . Hypotonic solution ( 0 . 075M KCl ) was added drop-wise to the cell pellet and incubated for 10 min at room temperature . Cells were fixed with an ice-cold 3:1 methanol:acetic acid solution for 5 min and then centrifuged at 800 rpm for 8 min . This last step was repeated two more times and fixed cells were finally dropped on microscope slides . For FISH on binucleate cells , amniocytes were grown in superfrost ultra plus slides ( Thermo Scientific , Life Technologies , CA , USA ) , whereas DLD1 cells were grown on coverslips . For the cytokinesis-block assay , cells were treated with 3 μg/ml dihydrocytochalasin B ( Sigma–Aldrich , MO , USA ) for 24 hr before fixation , and the experiment was repeated at least three independent times . Prior to being processed for FISH staining , cytokinesis-blocked cells were fixed according to previously published protocols: a standard FISH protocol ( Nicholson and Duesberg , 2009 ) was used for all cells; an alternative ( 3-D FISH ) protocol ( Cremer et al . , 2008 ) was used in some experiments with DLD1 , DLD1+7 , and DLD1+13 cells . Bacterial artificial chromosome ( BAC ) contigs using three to six BAC sequences specific to each region were made for the following four probes: CDX2 on chromosome 13q12 , MET on chromosome 7q31 , CHEK1 on chromosome 11q24 , and TERC on chromosome 3q26 . The BAC clone contigs were labeled by nick translation with Spectrum Orange-dUTP ( Abbott Laboratories; IL , USA ) for CDX2 , Dy-505-dUTP ( Dyomics; Jena , Germany ) for MET , Spectrum Orange-dUTP ( Abbott Laboratories; IL , USA ) for CHEK1 , and Dy-505-dUTP ( Dyomics; Jena , Germany ) for TERC . Dual color human whole chromosome paint probes were generated in-house using PCR labeling techniques . Chromosome 7 was labeled with Spectrum Orange ( Abbott Laboratories; Chicago , IL ) and chromosome 13 was labeled with Dy505 ( Dyomics; Jena , Germany ) . Additionally , commercial locus-specific FISH probes for chromosomes 7 , 11 , 19 ( red 5-ROX dUTP ) and 12 , 13 , 18 ( green 5-fluorescein dUTP ) were also used ( Empire Genomics; Buffalo , NY , USA ) . The probe mixtures were co-denatured with the coverslips at 72°C for 5 min before being placed in a moist chamber at 37°C for two nights . After two nights , the coverslips were washed in 2XSSC for 5 min and then mounted on microscope slides with mounting media ( Vectashield; CA , USA ) and DAPI . FISH on DLD1 metaphase spreads was performed with the commercial FISH probes listed above . Additionally , centromeric FISH probe against chromosome 7 ( FITC ) ( Cytocell , Cambridge , UK ) was used in DLD1+7 cells to confirm the presence of centromeric DNA . FISH on metaphase spreads of AF and AF+13 was performed with the XA 13/21 probe ( MetaSystems , Germany ) according to the manufacturer's instructions . Interphase FISH was performed with the Vysis centromeric probes CEP7 Spectrum Aqua , CEP12 Spectrum Green , and CEP18 Spectrum Orange ( Abbott Molecular , IL , USA ) according to the manufacturer's instructions . All the FISH-stained samples were analyzed blindly . DLD1 cells , grown on sterilized coverslips inside 35 mm Petri dishes , were transiently transfected with either P-EYFP-N1 or P-EYFP-N1-SPG20 ( a kind gift from J Bakowska ) using Fugene HD ( Roche , Basel , Switzerland ) according to the manufacturer's protocol . For SPG20 knockdown , cells grown in glass-bottom 35 mm u-dishes ( Ibidi GmbH , Germany ) or on sterilized coverslips inside 35 mm Petri dishes , were transfected with Silencer Select siRNA specific to SPG20 ( S23057 , Ambion , Life Technologies , CA , USA ) using Oligofectamine according to the manufacturer's instructions ( Invitrogen , Life Technologies , CA , USA ) . Silencer Select Negative control siRNA ( Ambion , Life Technologies , CA , USA ) was used as a negative control and was also transfected into cells with Oligofectamine ( Invitrogen , Life Technologies , CA , USA ) . Cells were observed 48–72 hr after transfection . Whole-cell extracts were separated by SDS-PAGE and transferred to PVDF membrane . Membranes were blocked 1 hr at room temperature with 5% milk in tris-buffered saline and then incubated over night with primary antibodies at 4°C . Antibodies were diluted as follows: rabbit anti-Spartin ( Protein Tech Group Inc . , IL , USA ) , 1:1000; rabbit anti-β-actin ( Abcam , Cambridge , UK ) , 1:500 . Blots were detected using goat anti-rabbit secondary antibodies conjugated to horse radish peroxidase and visualized with SuperSignal West Femto ( Thermo Scientific , Life Technologies , CA , USA ) . A GS-800 calibrated densitometer , or alternatively a ChemiDoc XRS system , was used for quantitative analysis of protein levels with the help of ImageLab 4 . 1 software ( BioRad , CA , USA ) . Immunofluorescently labeled DLD1 cells were imaged with a swept field confocal system ( Prairie Technologies , WI , USA ) on a Nikon Eclipse TE2000-U inverted microscope ( Nikon Instruments Inc . , NY , USA ) equipped with a 100×/1 . 4 NA Plan-Apochromatic objective and an automated ProScan stage ( Prior Scientific , Cambridge , UK ) . The confocal head was accessorized with multiband pass filter set for illumination at 405 , 488 , 561 , and 640 nm , and illumination was obtained through an Agilent monolithic laser combiner ( MLC400 ) controlled by a four channel acousto-optic tunable filter . Digital images were acquired with a HQ2 CCD camera ( Photometrics , AZ , USA ) . Acquisition time , Z-axis position , laser line power , and confocal system were all controlled by NIS Elements AR software ( Nikon Instruments Inc . , NY , USA ) on a PC computer ( Dell , TX , USA ) . Both anaphase lagging chromosomes and Spartin localization were analyzed by acquiring Z-sections of cells at 0 . 6 μm steps . The frequencies of anaphase lagging chromosomes were determined from 3 independent experiments performed in duplicate . Spartin localization at the midbody was determined from three independent experiments . Image acquisition of Spastin immunostaining in AFs was carried out on a Zeiss AxioImager Z1 equipped with an Axiocam MR and using a Plan-Apochromat 63×/1 . 4 NA objective . Images of telophase cells from three independent experiments and/or samples were acquired as Z-stacks with 0 . 3 μM steps and scored for the presence or absence of Spastin staining at the midbody . All data were analyzed blindly . FISH samples were viewed and imaged either with a Leica DM-RXA fluorescence microscope ( Leica; Wetzlar , Germany ) or with a swept field confocal system ( Prairie Technologies , WI , USA ) on a Nikon Eclipse TE2000-U inverted microscope ( Nikon Instruments Inc . , NY , USA ) . The Leica DM-RXA fluorescence microscope was equipped with custom optical filters and a 63×/1 . 3 NA objective . The Leica CW 4000 FISH software was used to acquire multifocal images for each fluorescence channel . 15 to 25 images were taken in areas of optimal cell density with minimal cellular clumps and overlapping cells . The Nikon Eclipse TE2000-U inverted microscope was equipped with a 100×/1 . 4 NA Plan-Apochromatic objective and an automated ProScan stage ( Prior Scientific , Cambridge , UK ) . The confocal head was accessorized with a multiband pass filter set for illumination at 405 , 488 , 561 , and 640 nm and illumination was obtained through an Agilent monolithic laser combiner ( MLC400 ) controlled by a four channel acousto-optic tunable filter . Digital images were acquired with a HQ2 CCD camera ( Photometrics , AZ , USA ) . Exposure time , Z-axis position , laser line power , and confocal system were all controlled by NIS Elements AR software ( Nikon Instruments Inc . , NY , USA ) on a PC computer ( Dell , TX , USA ) . FISH-stained samples were analyzed blindly and only cases with a total even number of signals were included in the analysis . DLD1 , DLD1+7 , and DLD1+13 cells were grown on sterilized coverslips inside 35 mm Petri dishes . Coverslips at 60–70% confluency were mounted in Rose chambers filled with L-15 medium supplemented with 4 . 5 g/l glucose . Images were acquired on a Nikon Eclipse Ti inverted microscope ( Nikon Instruments Inc . , NY , USA ) equipped with phase-contrast transillumination , transmitted light shutter , ProScan automated stage ( Prior Scientific , Cambridge , UK ) , and a HQ2 CCD camera ( Photometrics , AZ , USA ) . Cells were maintained at ∼36°C using an air stream stage incubator ( Nevtek , MA , USA ) . For analysis of cytokinesis in untransfected or siRNA-transfected cells , 10–15 different fields of cells were imaged every 2 min for 6–8 hr using a 60×/1 . 4 NA Plan-Apochromatic phase contrast objective controlled by Nikon Perfect Focus ( Nikon Instruments Inc . , NY , USA ) . For P-EYFP-N1 and P-EYFP-N1-SPG20 transfection experiments , 10–15 different fields of cells were imaged by fluorescence and phase contrast using a 20× or 60× objective every 4 min for 8 hr . The time-lapse videos were analyzed using NIS Elements AR ( Nikon Instruments Inc . , NY , USA ) software on a PC computer ( Dell , TX , USA ) . Amniocytes were grown in glass-bottom 35 mm u-dishes ( Ibidi GmbH , Germany ) coated with fibronectin and filled with phenol red-free EMEM complete medium . Images were acquired on a Zeiss Axiovert 200M inverted microscope ( Carl Zeiss , Germany ) equipped with a CoolSnap camera ( Roper , FL , USA ) , XY motorized stage and NanoPiezo Z stage , under controlled temperature , atmosphere and humidity . 20–25 neighbor fields were imaged every 2 . 5 min for 1–2 days using a 20×/0 . 3 NA A-Plan objective . Grids of neighboring fields were generated using the plugin Stitch Grid ( Stephan Preibisch ) from open source Fiji/Image J ( http://rsb . info . nih . gov/ij/ ) . Amniotic cells , glass-bottom 35 mm u-dishes ( Ibidi GmbH , Germany ) coated with fibronectin , were co-transfected with H2B-GFP and pmRFP-tubulin expression plasmids ( Addgene , MA , USA ) using Lipofectamine 3000 transfection reagent and according to the manufacturer's instructions . Live cell imaging was performed 48 hr following transfection under a spinning-disk confocal system Andor Revolution XD ( Andor Technology , UK ) coupled to an Olympus IX81 inverted microscope ( Olympus , UK ) equipped with an electron-multiplying CCD iXonEM Camera and a Yokogawa CSU-22 unit based on an Olympus IX81 inverted microscope . Two laser lines at 488 and 561 nm were used for the excitation of GFP and pmRFP and the system was driven by IQ software ( Andor Technology , UK ) . Z-stacks ( 0 . 8–1 . 0 μm ) covering the entire volume of the mitotic cells were collected every 1 . 5 min with a PLANAPO 60×/1 . 4 NA objective . ImageJ was used to process all the videos .	The DNA in a human cell is divided between forty-six structures called chromosomes . Before a cell divides , it copies every chromosome so that each daughter cell will have the same DNA as the parent cell . These chromosomes align in the center of the cell and then the matching chromosomes are separated and pulled to opposite ends . However , in some cases the separation process does not work properly , which can produce cells that either have too many , or too few , chromosomes . Abnormal numbers of chromosomes within cells—called aneuploidy—is a leading cause of miscarriage and birth defects in humans . Aneuploidy is also a common feature of cancer cells . It is common for the chromosomes in cancer cells to be distributed unequally when the cell divides . This phenomenon is known as chromosomal instability , but the link between aneuploidy and chromosomal instability in cancer cells is not fully understood . Here , Nicholson et al . used live-cell imaging techniques to analyze healthy human cells and cancer cells that had either the normal forty-six chromosomes , or a defined extra chromosome . Nicholson et al . found that when the cells divided , the chromosomes in the cells that had an extra copy of chromosome 7 or 13 were more prone to distributing chromosomes unequally , compared to cells with a normal number of chromosomes . Nicholson et al . also observed that the cells with an extra chromosome 13 were unable to properly divide into two . These cells had increased levels of a protein called Spartin—which is important for the last stage in cell division—and this was responsible for the failure to produce two daughter cells . These findings show that aneuploidy can cause chromosomal instability in human cells . Furthermore , Nicholson et al . have identified a defect in cell division that is specifically caused by the presence of an extra chromosome 13 in human cells . A future challenge will be to determine how , and to what extent , different chromosomes can affect chromosome stability . This could be useful in the development of therapies against cancer cells with aneuploidy .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "cell", "biology" ]	2015	Chromosome mis-segregation and cytokinesis failure in trisomic human cells
Four stress-sensing kinases phosphorylate the alpha subunit of eukaryotic translation initiation factor 2 ( eIF2α ) to activate the integrated stress response ( ISR ) . In animals , the ISR is antagonised by selective eIF2α phosphatases comprising a catalytic protein phosphatase 1 ( PP1 ) subunit in complex with a PPP1R15-type regulatory subunit . An unbiased search for additional conserved components of the PPP1R15-PP1 phosphatase identified monomeric G-actin . Like PP1 , G-actin associated with the functional core of PPP1R15 family members and G-actin depletion , by the marine toxin jasplakinolide , destabilised the endogenous PPP1R15A-PP1 complex . The abundance of the ternary PPP1R15-PP1-G-actin complex was responsive to global changes in the polymeric status of actin , as was its eIF2α-directed phosphatase activity , while localised G-actin depletion at sites enriched for PPP1R15 enhanced eIF2α phosphorylation and the downstream ISR . G-actin's role as a stabilizer of the PPP1R15-containing holophosphatase provides a mechanism for integrating signals regulating actin dynamics with stresses that trigger the ISR . In eukaryotes , regulation of protein biosynthesis defends against proteotoxic stress and balances anabolic growth with nutrient availability ( Jackson et al . , 2010 ) . The eukaryotic translation initiation factor 2 complex recruits the initiator methionyl-tRNA to ribosomes in a GTP-dependent catalytic cycle , but phosphorylation of its alpha subunit ( eIF2α ) by a family of stress-sensing kinases inhibits guanine nucleotide exchange , attenuating translation initiation and with it , global protein synthesis ( Clemens , 1996; Jackson et al . , 2010 ) . During endoplasmic reticulum stress , the eIF2α kinase PERK triggers attenuation of protein synthesis ( Harding et al . , 1999 ) , while other members of the family respond variously to amino acid deprivation , viral infection or heme deficiency ( van 't Wout et al . , 2014 ) . This ability of eIF2α phosphorylation to integrate signals from multiple , apparently unrelated , stresses led this pathway to be named the ‘integrated stress response’ ( ISR ) ( Harding et al . , 2003 ) . Attenuated global protein synthesis is cytoprotective early in the stress response . The ISR also involves the translational induction of the transcription factors ATF4 and CHOP , which activate a transcriptional programme that has pro-survival effects in the short term , whilst prolonged activation leads to a switch to apoptosis ( Zinszner et al . , 1998; Marciniak et al . , 2004; Lu et al . , 2014 ) . In metazoans , inactivation of the ISR is mediated by a catalytic , protein phosphatase 1 ( PP1 ) subunit in complex with a regulatory subunit ( PPP1R15 ) responsible for targeting eIF2α ( He et al . , 1998; Novoa et al . , 2001; Jousse et al . , 2003 ) . In Drosophila , a single PPP1R15 has been described that is required for anabolic larval growth ( Malzer et al . , 2013 ) , while in mammals , two PPP1R15 paralogues exist: a constitutively expressed isoform PPP1R15B ( also known as CReP ) and a stress-inducible isoform PPP1R15A ( also GADD34 ) ( Novoa et al . , 2001; Jousse et al . , 2003 ) . PPP1R15 family members share significant homology in their C-terminal conserved PP1-interacting domain , constituting a core functional domain sufficient to dephosphorylate eIF2α when over expressed in cells ( Novoa et al . , 2001; Malzer et al . , 2013 ) . In contrast , the less well-conserved N-terminal portion of each PPP1R15 determines protein stability ( Brush and Shenolikar , 2008 ) and subcellular localisation ( Zhou et al . , 2011 ) , although the importance of these functions in the regulation of eIF2α phosphatase activity within the cell remains to be worked out . The importance of eIF2α dephosphorylation is highlighted by PPP1R15 loss-of-function phenotypes . In Drosophila , ubiquitous RNAi-mediated depletion of dPPP1R15 leads to embryonic lethality , while failure of blastocyst implantation is seen in Ppp1r15a-Ppp1r15b double knockout mouse embryos ( Harding et al . , 2009; Malzer et al . , 2013 ) . Deficiency of PPP1R15B in isolation permits survival to gestation but leads to defects of haematopoiesis and death in the early neonatal period ( Harding et al . , 2009 ) . In contrast , PPP1R15A-deficient mice are overtly healthy when raised in standard laboratory conditions and show increased resistance to ER stress-induced tissue damage ( Marciniak et al . , 2004 ) . PPP1R15A is regulated transcriptionally ( Novoa et al . , 2001 ) , but relatively little is known about post-transcriptional regulation of its activity or the regulation of the constitutively expressed PPP1R15B or Drosophila dPPP1R15 ( Jousse et al . , 2003; Malzer et al . , 2013 ) . The literature offers numerous examples of proteins that associate with one or other of the PPP1R15 family members ( Hasegawa et al . , 2000a , 2000b; Wu et al . , 2002; Hung et al . , 2003; Shi et al . , 2004 ) , but these are largely single studies with no follow-up or physiological validation . In this study , we set out to characterise conserved elements of the PPP1R15 interactome and in doing so identified a novel mechanism for the regulation of eIF2α phosphatases that links the ISR with cytoskeletal dynamics . Important regulators/components of the PPP1R15-PP1 holoenzyme are likely to be conserved between species and paralogues; therefore , we set out to identify proteins that interact with both mammalian paralogues , PPP1R15A and PPP1R15B , and their non-vertebrate homologue , Drosophila dPPP1R15 . GFP-tagged human PPP1R15A and PPP1R15B were expressed in human embryonic kidney ( HEK ) 293T cells and subjected to GFP-Trap affinity purification followed by mass spectrometry ( Figure 1A , B and Figure 1—figure supplements 1 , 2 ) , whereas V5-His-tagged dPPP1R15 was expressed in Drosophila Schneider 2 ( S2 ) cells and subjected to affinity purification using anti-V5-His resin followed by mass spectrometry ( Figure 1A ) . In addition to the anticipated association of PP1 , we identified a number of other proteins that were bound to each PPP1R15 bait ( as defined by >twofold enrichment over control and the detection of ≥5 identifiable peptides in the mass spectra; Figure 1—figure supplements 1 , 2 ) . 10 . 7554/eLife . 04872 . 003Figure 1 . PPP1R15 associates with actin in mammalian and insect cells . ( A ) Heat map of proteins associated with GFP , GFP-tagged human PPP1R15B ( GFP-hR15B ) and GFP-tagged human PPP1R15A ( GFP-hR15A ) affinity-purified from transiently transfected HEK293T cells ( left panels ) ; heat map of proteins associated with V5 and V5-tagged Drosophila PPP1R15 ( dR15-V5 ) affinity purified from transiently transfected S2 cells ( right panels ) . Samples were analysed by Orbitrap mass spectrometer . Intensity reflects total spectrum count of identified peptides . Proteins identified by at least five spectra and showing at least twofold enrichment over control are shown . ( B ) Coomassie-stained SDS-PAGE of GFP-affinity purified proteins from HEK293T cells expressing indicated proteins . Indicated bands were individually excised and identified by mass spectrometry . ( C ) Coomassie-stained SDS-PAGE of GFP-affinity purified proteins from HEK293T cells expressing indicated proteins . Bands were individually excised and identified by mass spectrometry . ( D ) Coomassie-stained SDS-PAGE of glutathione-affinity purified proteins from HEK293T cells . Indicated bands were individually excised and identified by mass spectrometry . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 00310 . 7554/eLife . 04872 . 004Figure 1—figure supplement 1 . Mass spectrometry results of GFP , GFP-PPP1R15B , and GFP-PPP1R15A expressed in HEK293T cells and purified using GFP-Trap beads . Actin peptides are highlighted in red . Proteins with >twofold enrichment over control and the detection of ≥5 identifiable peptides in the mass spectra are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 00410 . 7554/eLife . 04872 . 005Figure 1—figure supplement 2 . Mass spectrometry results of V5 and dPPP1R15A-V5 expressed in S2 cells and purified using anti-V5 immunoprecipitation . Actin peptides are highlighted in red . Proteins with >twofold enrichment over control and the detection of ≥5 identifiable peptides in the mass spectra are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 005 Actin emerged as the prominent partner conserved across phyla ( Figure 1A , B ) . Confidence in this association was bolstered by finding that Drosophila dPPP1R15 also associated with mammalian actin in stoichiometric amounts ( Figure 1C ) . This association was observed regardless of which terminus of dPPP1R15 was tagged . Actin's presence in complex with PPP1R15 was also observed using other tag combinations: an N-terminal fusion of GST with the catalytic subunit PP1A expressed in HEK293T cells alongside PPP1R15A yielded a complex containing GST-PP1A , PPP1R15A , and actin upon glutathione-affinity chromatography ( Figure 1D ) . GFP-tagged PPP1R15A purified from HEK293T cells failed to associate with filamentous F-actin in a co-sedimentation assay ( Figure 2A ) suggesting selective interaction between PPP1R15 and monomers of soluble G-actin . The distribution of actin between its monomeric G or polymeric F form is influenced by physiological conditions and can be biased pharmacologically by small molecules that stabilise either form ( White et al . , 1983 ) . Jasplakinolide , which stabilises F-actin filaments and depletes the cells of G-actin ( Holzinger , 2009 ) , abolished the interaction between PPP1R15A and actin ( Figure 2B , lane 4 ) . In contrast , latrunculin B , which binds to the nucleotide-binding cleft of actin , thus increasing the cytoplasmic pool of G-actin ( Nair et al . , 2008 ) , potently enhanced the recovery of actin in complex with PPP1R15A ( Figure 2B , lane 3 ) . Cytochalasin D also increases the cellular pool of G-actin , but does so by engaging actin's barbed end , competing with several known G-actin-binding proteins ( Miralles et al . , 2003; Dominguez and Holmes , 2011; Shoji et al . , 2012 ) ; exposure to cytochalasin diminished the recovery of actin in complex with PPP1R15A ( Figure 2B lane 2 ) . 10 . 7554/eLife . 04872 . 006Figure 2 . PPP1R15 selectively associates with monomeric G-actin in cells . ( A ) Immunoblot ( upper panel ) and Coomassie-stained gel ( lower panel ) of affinity-purified GFP-tagged PPP1R15A and purified actin . Samples were incubated and centrifuged to pellet F-actin ( lane 1 ) , leaving G-actin in the supernatant ( lane 2 ) ; pellet P , supernatant S . ( B ) Immunoblot for GFP and actin of GFP-affinity purified proteins ( upper two panels ) from HEK293T cells expressing GFP-tagged PPP1R15A ( hR15A-GFP ) treated with 2 µM of each indicated compound . Immunoblot for actin of 2% of input . ( C ) Fluorescence microscopy image of NIH-3T3 cell F-actin arrangement . NIH-3T3 cells were left untreated ( control ) , cultured in serum-free medium for 24 hr ( serum starved ) , cultured in serum-free medium for 18 hr , followed by addition of medium containing 10% vol/vol FBS for 6 hr ( serum refed ) , then fixed and stained with Alexa-Fluor 568 phalloidin and imaged by confocal microscopy . ( D ) Immunoblot for GFP and actin of NIH-3T3 lysates from cells treated as in ‘C’ then subjected to GFP affinity purification ( upper two panels ) . Immunoblot for actin of 2% of input . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 006 Actin polymerisation is sensitive to physiological growth cues ( Sotiropoulos et al . , 1999 ) . Serum starvation , which resulted in the anticipated conversion of F to G-actin ( Figure 2C ) enhanced recovery of actin in complex with PPP1R15A in NIH-3T3 cell lysates ( Figure 2D ) . On serum re-feeding , cables of F-actin re-formed within the cytoplasm and less actin was recovered in complex with PPP1R15A . The aforementioned confirm that both mammalian and insect PPP1R15 regulatory subunits engage G-actin and that the interaction between them is sensitive to physiological changes in the availability of G-actin . Human PPP1R15A is a 674 amino acid protein , comprising an N-terminal domain required for membrane interaction , a region of proline , glutamate , serine , threonine ( PEST ) rich repeats of uncertain function , and a C-terminal functional core domain that interacts with the PP1 catalytic subunit ( Figure 3A ) and is sufficient for mediating substrate-specific dephosphorylation ( Novoa et al . , 2001; Kojima et al . , 2003; Ma and Hendershot , 2003 ) . Deletion analysis showed that the C-terminus of PPP1R15A ( residues 501–674 ) was also sufficient for the association with actin ( Figure 3B ) . Further deletion revealed that residues C-terminal to amino acid 615 were essential for actin association but not for PP1 binding , which was enfeebled but not abolished ( Figure 3B , C ) . Incorporation of five residues ( W616–R620 of human PPP1R15A ) restored fully the recovery of actin in complex with PPP1R15A ( Figure 3C lane 6 ) , while the W616A and L619A double mutation strongly enfeebled actin recovery in complex with PPP1R15A ( Figure 3D ) . A V556E mutation of the RVxF motif , which all but abolishes PP1 binding and eIF2α dephosphorylation in vivo ( Novoa et al . , 2001 ) , also attenuated recovery of actin in complex with PPP1R15A , but failed to abolish it altogether ( Figure 3C , lane 3 ) . 10 . 7554/eLife . 04872 . 007Figure 3 . Actin associates with the conserved C-terminal portion of PPP1R15 . ( A ) Schematic diagram of human PPP1R15A ( R15A ) constructs used . Green indicates GFP . PEST repeats ( between residues 346 and 494 , orange ) , K555VRF558 ( yellow ) , and W616ARLR620 ( purple ) sequences are identified . ( B ) Immunoblot for GFP , actin , and PP1 of HEK293T lysates from cells expressing indicated constructs and PP1 , and subjected to GFP affinity purification ( upper three panels ) . Immunoblot for actin and PP1 of 2% of input . ( C ) Immunoblot for GFP , actin , and PP1 of HEK293T lysates from cells expressing indicated constructs and PP1 , and subjected to GFP affinity purification ( upper three panels ) . Immunoblot for actin of 2% of input . ( D ) Immunoblot for GFP and actin of HEK293T lysates from cells expressing indicated constructs and subjected to GFP affinity purification ( upper two panels ) . Immunoblot for actin of 5% of input ( lower panel ) . ( E ) Sequence alignment of C-terminal portions of human ( h ) and murine PPP1R15A ( mR15A ) and PPP1R15B ( mR15B ) and Drosophila dPPP1R15 ( dR15 ) with regions of homology boxed . Specific truncations are indicated . ( F ) Immunoblot for GFP and actin of HEK293T lysates from cells expressing indicated constructs and subjected to GFP affinity purification ( upper two panels ) . Immunoblot for actin and PP1 of 2% of input . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 00710 . 7554/eLife . 04872 . 008Figure 3—figure supplement 1 . Immunoblot for GFP , actin , and PP1 of GFP-Trap pull-downs and 2% of input . HEK293T cells were transiently transfected with plasmids encoding the indicated constructs . After 36 hr , cells were lysed in GFP-Trap lysis buffer ( 150 mM NaCl , 10 mM Tris/Cl pH 7 . 5 , 0 . 5 mM EDTA , 1 mM PMSF , and Protease Inhibitor Cocktail [Roche] ) and post-nuclear supernatants were incubated with GFP-Trap beads at 4°C for 2 hr then washed four times in the same buffer . Next , samples were washed thrice in GFP-Trap lysis buffer supplemented with additional NaCl as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 00810 . 7554/eLife . 04872 . 009Figure 3—figure supplement 2 . Immunoblot for GFP , actin , and PP1 of GFP-Trap pull-downs and 2% of input . HEK293T cells were transiently transfected with plasmids encoding the indicated constructs . After 36 hr , cells were lysed in GFP-Trap lysis buffer ( 150 mM NaCl , 10 mM Tris/Cl pH 7 . 5 , 0 . 5 mM EDTA , 1 mM PMSF , and Protease Inhibitor Cocktail [Roche] ) and post-nuclear supernatants were incubated with GFP-Trap beads at 4°C for 2 hr then washed once in the same buffer . Next , samples were washed thrice with no detergent ( GFP-Trap lysis buffer ) , triton buffer ( 150 mM NaCl , 10 nM HEPES pH 7 . 4 , 0 . 5% vol/vol triton X-100 ) , RIPA buffer ( 150 mM NaCl , 50 mM Tris HCl pH 7 . 4 , 1% vol/vol NP40 , 0 . 5% vol/vol sodium deoxycholate , 0 . 1% vol/vol SDS ) or digitonin buffer ( 150 mM NaCl , 50 mM Tris HCl pH 7 . 4 , 0 . 1% vol/vol digitonin ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 009 The quantities of actin and PP1 recovered in complex with PPP1R15A were sensitive to the salt concentration of the buffers used ( Figure 3—figure supplement 1 ) . Actin association with PPP1R15A dropped progressively with increasing salt ( 75% of the actin bound at 150 mM salt was lost at 350 mM ) , as did PP1 association , with no detectable binding at 350 mM . The complex was stable in non-denaturing detergents ( triton X-100 and digitonin ) , but washes in a buffer containing the harsher detergents , sodium deoxycholate ( 0 . 5% vol/vol ) and SDS ( 0 . 1% vol/vol ) , completely abolished interaction between PPP1R15A and both PP1 and actin ( Figure 3—figure supplement 2 ) . Drosophila dPPP1R15 is half the size of the mammalian PPP1R15s . When aligned , mammalian PPP1R15A , PPP1R15B , and dPPP1R15 share significant homology within their C-termini , which drops off at residue 622 of human PPP1R15A ( Figure 3E ) . We therefore truncated the Drosophila protein within and immediately N-terminal to this region of homology ( Y307–H312 ) . Partial truncations reduced the association of dPPP1R15 with actin , while deletion of the entire segment ( at residue 307 ) completely abolished the interaction ( Figure 3F ) . The interaction with actin , thus maps to the conserved portion of PPP1R15 family members and is favoured by a short stretch of hydrophobic residues at the extreme C-terminus of this core . Mutational analysis thus points to a measure of independent association of PP1 or actin with PPP1R15 , but highlights the enhanced recovery of the three proteins in a ternary complex of PPP1R15 , PP1 , and actin . To examine the relevance of G-actin to the endogenous PPP1R15 complex , wild-type Ppp1r15a+/+ and mutant Ppp1r15amut/mut mouse embryonic fibroblasts ( MEFs ) were treated with the ER stress promoting agent tunicamycin to induce the ISR and expression of PPP1R15A . The Ppp1r15amut/mut cells express a C-terminal truncated PPP1R15A that is incapable of binding PP1 ( Novoa et al . , 2003 ) and served as a negative control . As expected , a robust PP1 signal was found associated with endogenous wild-type PPP1R15A in the stressed cells , whilst no signal was detected in PPP1R15A immunoprecipitates from the Ppp1r15amut/mut cells ( Figure 4A , lanes 2 and 5 ) . The poor reactivity of the available antisera to actin and tendency of actin to associate non-specifically with immunoprecipitation reactions frustrated our efforts to detect actin associated with endogenous PPP1R15A in MEFs; however , treatment with jasplakinolide , which depleted the soluble pool of actin led to a marked loss of PP1 association with PPP1R15A in the stressed cells ( compare lanes 2 and 3 , Figure 4A ) . To test the converse interaction , PP1 was affinity purified from MEF lysates using microcystin–agarose beads . Whilst the presence of other known PP1-actin complexes precludes meaningful interpretation of actin purified by microcystin affinity ( Oliver et al . , 2002; Kao et al . , 2007 ) , the PPP1R15A-PP1 interaction detected in stressed wild-type cells was attenuated by jasplakinolide-driven depletion of soluble actin ( Figure 4B ) . Actin's role in the stability of the PPP1R15A-PP1 complex was confirmed in HEK293T cells ( Figure 4C ) . 10 . 7554/eLife . 04872 . 010Figure 4 . G-actin stabilises the PPP1R15A-PP1 complex in vivo . ( A ) Immunoblots of endogenous PPP1R15A ( R15A ) and associated PP1 immunopurified from wild type ( Ppp1r15a+/+ ) or mutant mouse embryonic fibroblasts homozygous for a C-terminal truncation of PPP1R15A that abolishes interaction with PP1 ( Ppp1r15amut/mut ) with an anti-PPP1R15A antiserum ( IP R15A ) . Where indicated , cells were treated with tunicamycin 2 µg/ml ( Tm ) for 8 hr to induce PPP1R15A and jasplakinolide ( 1 µM ) for 1 . 5 hr before harvest . The lower three panels are immunoblots of the input of the immunoprecipitation reactions analysed in the top two panels . Closed and open triangles mark , respectively , the wild type and mutant PPP1R15A lacking the C-terminal functional core . To assess G-actin content of the input , the sample was subjected to ultracentrifugation to remove F-actin . ( B ) PPP1R15A and PP1 immunoblots of PP1-containing complexes purified by microcystin affinity chromatography from cells as in ‘A’ above . The lower three panels report on the content of input material . ( C ) As in ‘B’ , above , but reporting on PP1-continaing complexes purified by microcystin affinity chromatography from HEK293T cells . ( D ) Immunoblots of endogenous or over-expressed GFP-tagged PPP1R15A and associated endogenous PP1 and actin immunopurified with antiserum to PPP1R15A , non-immune rabbit IgG ( as a control ) or antiserum to GFP from lysates of tunicamycin-treated HEK293T cells ( Tm , 2 . 5 µg/ml for 8 hr to induce endogenous PPP1R15A ) or cells transfected with plasmids expressing GFP-PPP1R15A ( GFP-R15A ) or GFP . The protein content of the cell lysate applied to the immunoprecipitations is noted above the immunoblots ( ‘Protein input’ ) . Endogenous PPP1R15A and the larger GFP-PPP1R15A are marked by black and grey arrowheads , respectively . Both heavy and light exposures of the actin and PP1 immunoblots are provided and the relative intensity of the signals is noted . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 010 In order to address the association of actin with endogenous PPP1R15A directly , we used HEK293T cells , which generated less background actin signal in control immunoprecipitation reactions . Purified GFP-tagged PPP1R15 was used as a standard to determine the minimum amount of PPP1R15 that permitted detection of associated actin ( Figure 4D ) . Scaling of input material to immunopurify similar quantities of endogenous and overexpressed PPP1R15A led to recovery of similar amounts of associated endogenous actin ( Figure 4D ) . This supports a role for the interaction in cell physiology . A functional role for actin in PPP1R15 complexes was suggested by the observation that depletion of cellular G-actin by exposure to jasplakinolide promoted a rapid increase in the levels of phosphorylated eIF2α ( Figure 5A , B ) . To extend these observations , cells were treated with the SERCA pump inhibitor thapsigargin , which depletes the ER of calcium and rapidly and transiently activates the ER stress-inducible kinase PERK . As expected , this led to a robust yet transient phosphorylation of eIF2α by PERK ( Figure 5C lanes 1–6 ) . The transient nature of this phosphorylation relates to the rectifying response of PERK on levels of ER stress , but also draws on the combined activities of constitutively expressed PPP1R15B and the induction of PPP1R15A that promote eIF2α dephosphorylation ( Novoa et al . , 2001; Jousse et al . , 2003; Novoa et al . , 2003 ) . In the presence of jasplakinolide , the elevated levels of phosphorylated eIF2α induced by thapsigargin persisted ( Figure 5C , lanes 7–12 ) , while latrunculin B had no effect on the time course of eIF2α phosphorylation ( Figure 5—figure supplement 1 ) . It is noteworthy that peak levels of eIF2α phosphorylation were higher in cells treated with jasplakinolide ( compare lanes 1–2 with 8–9 of Figure 5C ) . This occurred well before the induction of PPP1R15A suggesting that either an endogenous basally expressed phosphatase or the kinase was affected . 10 . 7554/eLife . 04872 . 011Figure 5 . Association of G-actin with PPP1R15 regulates eIF2α phosphatase activity . ( A ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α , and actin . Wild-type ( WT ) mouse embryonic fibroblasts ( MEF ) were treated with jasplakinolide 1 µM for the indicated times . Lysates were subjected to sedimentation assay and immunoblot for G-actin in the supernatant ( G ) or F-actin in the pellet ( F ) . ( B ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α , and actin . WT MEFs were treated with the indicated concentrations of jasplakinolide for 1 hr . Lysates were analysed as in ‘A’ . ( C ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α , and PPP1R15A . WT MEFs were treated with thapsigargin 400 nM for the indicated times , without or with jasplakinolide 1 µM . ( D ) Immunoblot for P-eIF2α and PPP1R15A ( hR15A ) . GFP-hPPP1R15A Tet-On HeLa cells were treated with doxycycline to induce transgene expression and then with thapsigargin 400 nM for the indicated times . Cells were co-treated with jasplakinolide or latrunculin B 1 µM or vehicle as indicated . Accompanying graphs show means ±SEM of n = 3 independent repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 01110 . 7554/eLife . 04872 . 012Figure 5—figure supplement 1 . Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α and PPP1R15A . WT MEFs were treated with thapsigargin 400 nM for the indicated times , without or with latrunculin B 1 µM . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 012 Induction of PPP1R15A was observed in thapsigargin-treated cells from 2 hr onwards but was diminished in cells co-treated with jasplakinolide ( likely , a consequence of profound attenuation of protein synthesis ) . To minimise the confounding effect of diminished PPP1R15A levels in jasplakinolide-treated cells at later time points , the regulatory subunit was conditionally over-expressed under the control of a tetracycline-responsive promoter . As expected , enforced expression of PPP1R15A abrogated the phosphorylation of eIF2α in response to thapsigargin; however , jasplakinolide reversed the inhibitory effect of PPP1R15A over-expression restoring elevated levels of phosphorylated eIF2α to thapsigargin-treated cells ( Figure 5D ) . To examine more closely the effects of the actin cytoskeleton on dephosphorylation of eIF2α within living cells , we sought to gain temporal control over the phosphorylation phase of the cycle . To this end , we made use of a small molecule eIF2α kinase inhibitor , GSK2606414A ( Axten et al . , 2012 ) . At the concentration used , this inhibits the eIF2α kinases PERK and PKR , but not GCN2 . Thus , application of GSK2606414A to Gcn2−/− fibroblasts rapidly abrogates eIF2α phosphorylation . To measure selectively the dephosphorylation phase of the stress response , PERK-mediated phosphorylation of eIF2α was induced by thapsigargin and further phosphorylation was then blocked with the kinase inhibitor GSK2606414A . In the absence of kinase activity , the subsequent decay of the phosphorylated eIF2α signal reflects its de-phosphorylation ( degradation of the protein is not observed over this time scale ) , which was markedly attenuated by jasplakinolide ( Figure 6A , compare lanes 1–5 with lanes 6–10 and Figure 6B ) . 10 . 7554/eLife . 04872 . 013Figure 6 . Jasplakinolide diminishes eIF2α phosphatase activity in vivo . ( A ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α , and ATF4 . Gcn2−/− MEFs were pre-treated with thapsigargin 300 nM for 30 min to induce eIF2α phosphorylation and ATF4 protein levels . GSK2606414A at 2 µM was then added for the indicated times . Protein lysates were analysed by SDS-PAGE and subjected to immunoblot . ( B ) Quantification of ‘A’ using ImageJ software . Mean ± SEM of n = 3 independent repeats . ( C ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) and total eIF2α . MEFs of the indicated genotypes were treated with or without jasplakinolide 1 µM for 1 hr . Protein lysates were analysed by SDS-PAGE and subjected to immunoblot . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 01310 . 7554/eLife . 04872 . 014Figure 6—figure supplement 1 . Immunoblot for P-eIF2α , total eIF2α , and ATF4 ( specific band marked with an asterisk ) in lysates of wild type ( WT ) or eIF2αAA MEFs following treatment with thapsigargin 300 nM for 4 hr and/or jasplakinolide 1 µM for 4 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 014 Jasplakinolide-mediated induction of ATF4 was abrogated in cells in which the serine 51 phosphorylation site of eIF2α had been mutated to alanine ( Figure 6—figure supplement 1 ) , thus validating ATF4 as an indicator of the effects of manipulation of the actin cytoskeleton on ISR activity . ATF4 , whose levels decline rapidly upon GSK2606414A-mediated shutdown of kinase activity , was also stabilised by jasplakinolide ( Figure 6A , lowest panel ) , reflecting the functional significance of the defect in eIF2α dephosphorylation imposed by the depletion of G-actin . Levels of phosphorylated eIF2α induced by jasplakinolide were undiminished in cells lacking any one of the four known eIF2α kinases ( Figure 6C ) , suggesting that the compound's effects on levels of phosphorylated eIF2α reflect its workings on the dephosphorylation phase of the stress cycle and not to off-pathway stress culminating in kinase activation . Actin was recovered in complex with both the inducible and constitutive mammalian PPP1R15 family members ( Figures 1 and 7A ) . To determine if the effects of G-actin were preferentially mediated by complexes containing one or the other PPP1R15 subunit , we compared the effect of jasplakinolide on levels of phosphorylated eIF2α in wild-type MEFs and MEFs deficient in one or the other regulatory subunit . Enhanced levels of phosphorylated eIF2α in jasplakinolide-treated cells and the synergistic effects of depleting G-actin on the response to thapsigargin were observed in wild-type cells and in cells lacking either PPP1R15A or PPP1R15B-directed eIF2α dephosphorylation ( Figure 7B–D ) . These observations indicate that G-actin plays a functional role in holophosphatases constituted with either regulatory subunit . 10 . 7554/eLife . 04872 . 015Figure 7 . Actin associates with PPP1R15B to alter the level of eIF2α phosphorylation . ( A ) Immunoblot for GFP and actin of HEK293T cell lysates expressing either GFP or GFP-PPP1R15B . Upper two panels indicate proteins immunoprecipitated by anti-GFP beads . Lower panel shows 2% of input lysate . ( B ) Immunoblot for P-eIF2α , total eIF2α , and PPP1R15A of lysates from WT or Ppp1r15btm1Dron/tm1Dron MEFs treated for 1 hr with thapsigargin 400 nM , jasplakinolide 1 µM or both . ( C ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α and actin . Ppp1r15atm1Dron/tm1Dron MEFs were treated with jasplakinolide 1 µM for the indicated times . Lysates were subjected to sedimentation assay and immunoblot for G-actin in the supernatant ( G ) or F-actin in the pellet ( F ) . ( D ) Immunoblot for phosphorylated eIF2α ( P-eIF2α ) , total eIF2α , and actin . Ppp1r15atm1Dron/tm1Dron MEFs were treated with the indicated concentrations of jasplakinolide for 1 hr . Lysates were analysed as in ‘C’ . Accompanying graphs show mean ± SEM of n = 3 independent repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 015 To explore in further detail the basis for the correlation between G-actin levels and eIF2α dephosphorylation , we compared in vitro eIF2α-directed phosphatase activity of PPP1R15A-containing complexes recovered from untreated and jasplakinolide-treated cells . PPP1R15A-GFP fusion protein was expressed transiently in HEK293T cells overnight . The following day , cells were treated either with vehicle or with 1 µM jasplakinolide for 1 hr , lysed and then subjected to GFP-affinity purification using GFP-Trap beads . The resulting complexes were divided between four tubes and incubated for the indicated times at 37°C with pre-phosphorylated recombinant eIF2α ( see ‘Materials and methods’ ) . Less actin and PP1 were recovered in complex with tagged PPP1R15A from jasplakinolide-treated cells ( whilst HSP70 binding was unaffected ) ( Figure 8A ) , and the eIF2α-directed phosphatase activity of the purified complexes was likewise diminished ( Figure 8A , B ) . Complex formation with PP1 contributes to dPPP1R15 stability; however , the decline in PPP1R15A levels in cells exposed to the translational inhibitor cycloheximide , was unaffected by the presence of jasplakinolide ( Figure 8C , D ) indicating that stabilisation of a complex between PPP1R15 and PP1 dominates G-actin's role in this experimental system . 10 . 7554/eLife . 04872 . 016Figure 8 . Association of actin with PPP1R15A promotes eIF2α phosphatase activity in vitro . ( A ) Silver-stained SDS-PAGE or GFP-affinity purified PPP1R15A-GFP ( hR15-GFP ) and associated proteins . Asterisks signify identity confirmed by mass spectrometry . Purified complex incubated with phosphorylated recombinant eIF2α N-terminal lobe ( eIF2α ) for incubated time . Note size shift corresponds to dephosphorylation . ( B ) Quantification of ‘A’ using ImageJ software . Mean ± SEM . p value calculated by two-way ANOVA , n = 3 . ( C ) Immunoblot for PPP1R15A and eIF2α of HEK293T cells expressing hPPP1R15A-GFP ( hR15A-GFP ) . Treated with cycloheximide 50 µM for indicated times . ( D ) Quantification of ‘C’ . Mean ± SEM , n = 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 016 In the experiments described thus far , actin polymerisation was manipulated throughout the cell , whereas in vivo the actin cytoskeleton is subject to highly localised changes . In the context of eIF2α dephosphorylation , it seemed particularly relevant to examine the effects of actin polymerisation in the vicinity of the ER membrane , where the majority of PPP1R15 is located ( Brush et al . , 2003; Zhou et al . , 2011; Malzer et al . , 2013 ) . Cells that conditionally expressed a constitutively active mutant of mDia2 , a formin that stimulates localised polymerisation of F-actin ( Pellegrin and Mellor , 2005 ) , were generated . To direct mDia2 to the same membranous compartments as PPP1R15 , we fused the membrane-targeting domain of PPP1R15B ( residues 1–146 , devoid of catalytic activity ) to mDia2 and a GFP tag to facilitate visualisation of the fusion protein . Control cells were generated expressing a PPP1R15B ( 1–146 ) -GFP fusion protein lacking mDia2 . Induction of eGFP-PPP1R15B ( 1–146 ) -mDia2 increased polymerisation of actin adjacent to the ER , as indicated by co-localisation of GFP fluorescence with phalloidin staining ( Figure 9A ) . In control cells , the GFP and phalloidin signals showed little correlation , but as expected , these became strongly correlated ( indicating co-localisation ) on expression of the mDia2 construct ( Figure 9—figure supplement 1 ) . Immunoblot confirmed the expression of each fusion protein on treatment with doxycycline , but only the mDia2 construct led to the induction of ATF4 ( Figure 9B compare lanes 4–6 with 13–15 ) . ISRIB , a small molecule that renders cells unresponsive to eIF2α phosphorylation ( Sidrauski et al . , 2013 ) , blocked formin-mediated induction of ATF4 ( Figure 9B , compare lanes 15 and 16 ) , validating ATF4 as an ISR marker in this assay . 10 . 7554/eLife . 04872 . 017Figure 9 . Localised changes in the polymeric status of actin modulate the sensitivity of the ISR . ( A ) Fluorescence microscopy image of Flp-In T-REx HEK293 cells treated with 1 µg/ml doxycycline for 12 hr to express either ER membrane-localised GFP ( GFP-R15B 1–146 ) or ER membrane-localised GFP-mDia2 fusion ( GFP-R15B 1–146_mDia2 ) then fixed and stained with Alexa-Fluor 568 phalloidin and imaged by confocal microscopy . Bar = 5 µm . ( B ) Immunoblot for GFP and ATF4 in lysates of GFP-R15B 1–146 or GFP-R15B 1–146_mDia2 Flp-In T-REx HEK293 cells following treatment with doxycycline ( Dox ) 0 . 1 µg/ml for indicated times or with ISRIB 100 nM and or thapsigargin 300 nM for 4 hr . Immunoreactivity to ATF4 was quantified using ImageJ software ( ATF4 Intensity ) . Proteins of the expected sizes are marked with a solid triangle GFP-R15B 1–146_mDia2 or an open triangle GFP-R15B 1–146 . ( C ) Immunoblot for P-eIF2α , total eIF2α , and puromycin in lysates of GFP-R15B 1–146 or GFP-R15B 1–146_mDia2 Flp-In T-REx HEK293 cells following pre-treatment—if indicated with doxycycline ( Dox ) 0 . 1 µg/ml for 10 hr followed by treatment with tunicamycin 2 . 5 µg/ml for indicated times . 10 min prior to harvesting , puromycin was added to the culture medium at a final concentration of 10 µg/ml . Immunoreactivity to puromycin within lysates served as a marker of protein translation and was quantified using ImageJ software ( Puromycin intensity ) . Accompanying graphs of mean ± SEM of n = 3 independent repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 01710 . 7554/eLife . 04872 . 018Figure 9—figure supplement 1 . Colocalisation of filamentous actin with ER membrane-localised GFP ( GFP-R15B 1–146 ) or ER membrane-localised GFP-mDia2 fusion ( GFP-R15B 1–146_mDia2 ) . Confocal microscopy images of Flp-In T-REx HEK293 a cells treated with 1 μg/ml doxycycline for 12 hr . Nuclear DAPI staining is shown in blue , Alexa-Fluor 568 phalloidin staining of F-actin is shown in red , and GFP-R15B 1–146 ( A–H ) or GFP-R15B 1–146_mDia2 ( I–P ) is shown in green . The graph presents colocalisation analysis by Costes Pearson's correlation . Ten confocal fields were analysed for each cell type . Higher positive values represent increased colocalisation , whereas negative values represent exclusion of F-actin from GFP containing pixels . DOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 018 Actin polymerisation in the vicinity of the ER also altered the dynamics of the ISR in response to gradually accruing ER stress induced by the glycosylation inhibitor tunicamycin . During ER stress , phosphorylation of eIF2α by PERK attenuates protein translation to offload the ER ( Harding et al . , 1999 ) . The degree of translational attenuation depends upon the intensity and the rapidity of ER stress ( Novoa et al . , 2001 ) . Sudden and intense ER stress caused by depletion of ER calcium stores by thapsigargin induces marked inhibition of translation . In contrast , gradually escalating ER stress by the accumulation of unglycosylated proteins upon treatment with tunicamycin , attenuates translation less dramatically because induction of PPP1R15A limits the degree of eIF2α phosphorylation ( Novoa et al . , 2001 ) . In control cells ( expressing the bland eGFP-PPP1R15B [1–146] targeting fragment ) , tunicamycin induced a transient and minor decrease in translation with a nadir at 2 hr ( Figure 9C , lane 5 ) . By contrast , in cells expressing ER-targeted mDia2 , tunicamycin led to a sustained drop in protein synthesis associated with a sustained increase in eIF2α phosphorylation ( Figure 9C , compare lanes 8 and 9 ) . These experiments are consistent with a pool of G-actin localised in the vicinity of PPP1R15 in promoting eIF2α dephosphorylation . Over the years multiple proteins have been noted to interact with the PPP1R15-PP1 core holoenzyme , but none has proved generalizable across experimental systems or successfully implicated in the genetically well-characterised role of the complex to promote eIF2α dephosphorylation ( Hasegawa et al . , 2000a , 2000b; Wu et al . , 2002; Hung et al . , 2003; Shi et al . , 2004 ) . In this study , an unbiased approach identified actin as a conserved binding partner of PPP1R15 . The affinities of actin for PPP1R15 lay within a physiologically relevant range such that fluctuations of the G:F actin ratio affected the amount of actin recovered in the complex . Alterations to the ratio of G:F actin at the site of PPP1R15 action were seen to modulate cellular sensitivity to ISR stimuli via changes in eIF2α phosphatase activity . Collectively , these findings establish G-actin as an important regulator of PPP1R15-mediated eIF2α dephosphorylation in vivo . Our proteomics analysis also identified other potential binding partners of PPP1R15 . In mammalian cells , tubulin and HSP70 were consistently recovered in complex with overexpressed PPP1R15 and PPP1R15-containing fusion proteins . These interactions are less conserved across phyla than the PPP1R15-actin interaction . Furthermore , in vitro experiments in the accompanying manuscript demonstrate that addition of actin is sufficient to endow the PPP1R15-PP1 complex with selectivity towards eIF2α ( Chen et al . , 2015 ) . Thus , while there is nothing in our observations to argue against tubulin or HSP70 joining the complex and modulating PPP1R15-directed phosphatase activity , the evidence at hand suggesting actin's relevance to the core activity of the eIF2α-directed phosphatase justifies the focus on actin . With polymerisation and depolymerisation , the actin cytoskeleton is highly dynamic and levels of G-actin are subject to large fluctuations . Following polymerisation of actin to the barbed end of a filament , bound ATP is hydrolysed and eventually ADP-actin dissociates from the pointed end ( Dominguez and Holmes , 2011 ) . This dynamic is regulated by proteins that enhance depolymerisation , for example , ADF , or promote the recharging with ATP , which enhances the recycling of monomers , for example , profilin ( Paavilainen et al . , 2004 ) . Capping proteins prevent the consumption of monomers and so increase free G-actin concentrations , while severing proteins can lead to filament disassembly or nucleate more filament formation depending upon the context ( Wear and Cooper , 2004 ) . In contrast , formins like mDia2 remain associated with the barbed end yet promote addition of actin monomers . Other actin-binding proteins have functions unrelated to the cytoskeleton and it is now well recognised that free G-actin can function as a second messenger . For example , MAL , a cofactor of the transcription factor SRF , cycles dynamically between the nucleus and cytoplasm in a manner regulated by its binding to G-actin in quiescent cells ( Miralles et al . , 2003; Vartiainen et al . , 2007 ) . By depleting G-actin , growth signal-driven actin polymerisation releases MAL to enter the nucleus , bind SRF and activate target genes . Other examples include Phactr , a PP1 regulatory subunit whose cytoplasmic localisation depends upon G-actin binding ( Wiezlak et al . , 2012 ) and RNA polymerases II and III for whom actin forms a scaffold for the assembly of enzyme complexes ( Hu et al . , 2004; Kukalev et al . , 2005 ) . Many actin-binding proteins including MAL interact with a hydrophobic target-binding cleft between subdomains I and III of the actin monomer ( Mouilleron et al . , 2008; Dominguez and Holmes , 2011; Shoji et al . , 2012 ) . This site is blocked by cytochalasin D , which inhibits such interactions . Latrunculin B increases the level of actin monomers by binding to a different site on G-actin , the nucleotide-binding cleft , and does not interfere with binding at the hydrophobic cleft . Our observation that cytochalasin D diminishes the recovery of actin in complex with PPP1R15 , is consistent with interaction via the hydrophobic target-binding cleft . While the precise details remain to be worked out , structural and biochemical studies presented in the accompanying manuscript support this idea and further suggest the C-terminal most residues of the functional core of the PPP1R15 family members play an important role in actin engagement ( Chen et al . , 2015 ) . A crystal structure obtained for the binary complex of PPP1R15B and PP1 demonstrated that the N-terminal half of PPP1R15's functional core extensively engages the surface of PP1 following an itinerary previously observed for the regulatory subunit PPP1R9/spinophilin ( Ragusa et al . , 2010; Chen et al . , 2015 ) . Interestingly , the C-terminal portion of PPP1R15's functional core , implicated here in actin binding , was not observed in a high-resolution crystal structure of the PPP1R15B-PP1 binary complex , suggesting that this portion of PPP1R15B remained unstructured in the absence of actin . The crystal structure obtained for the 1:1:1 ternary complex of PPP1R15B-PP1-actin was of too low a resolution to identify these C-terminal residues of PPP1R15's functional core , but unaccounted for density observed in the cleft between lobes I and III of actin suggests a mode of engagement of actin by this portion of PPP1R15B that would be sensitive to disruption by cytochalasin , which binds to the same region of G-actin . Our in vivo findings reported here emphasize the importance of actin binding to the stability of the PPP1R15-PP1 complex and suggest that association of PP1 and actin with PPP1R15 may be cooperative . The accompanying manuscript provides further evidence for the direct binding of PPP1R15 and actin and reveals a role for actin in augmenting the specificity of the holophosphatase for eIF2α ( Chen et al . , 2015 ) . These two mechanisms are likely to work in concert and suggest a crucial role for G-actin in establishing a biologically relevant route to eIF2α dephosphorylation . It would appear that under normal circumstances G-actin is not limiting to eIF2α dephosphorylation in cultured MEFs , as latrunculin B , which enhances the pool of PPP1R15 binding-competent G-actin in some cell types , has no measurable effect on phosphorylated eIF2α ( Figure 5—figure supplement 1 ) . However , regulation of eIF2α phosphatases via the binding of G-actin might plausibly play a role in biological processes that are accompanied by changes in the ratio of G:F actin in other differentiated cell types , for example , in circumstances of cell migration , axonal guidance , or synaptic plasticity . The latter are particularly attractive given the evidence for crosstalk between the ISR and memory formation ( Costa-Mattioli et al . , 2007; Ma et al . , 2013; Sidrauski et al . , 2013 ) . The role of eIF2α phosphorylation in regulating rates of protein synthesis and the coupling of this phosphorylation event to the activation of a gene expression programme are conserved in eukaryotes . However , the mechanism for dephosphorylating eIF2α has diverged considerably . Yeasts rely on direct recruitment of the catalytic phosphatase subunit ( Glc7p ) to the eIF2 substrate , with no PPP1R15 intermediate ( Rojas et al . , 2014 ) , while PPP1R15 family proteins are apparent only in complex animals: insects and vertebrates ( Novoa et al . , 2001; Jousse et al . , 2003; Malzer et al . , 2013 ) . It is tempting to speculate that this more complex mode of regulating eIF2α dephosphorylation co-evolved with mechanisms for regulating the actin cytoskeleton and G-actin availability . Current models suggest that PPP1R15B , which is expressed constitutively , provides a constant background of eIF2α phosphatase activity that is augmented by transcriptional induction of PPP1R15A during later stages of the ISR ( Jousse et al . , 2003 ) . This study reveals that both PPP1R15 isoforms are poised to undergo post-translational regulation through changes in the polymeric status of actin . The focus here has been on the conserved functional core of PPP1R15 , but there remains room for further modulation of both isoforms by their large , poorly characterised N-terminal regions . Our protein discovery effort has identified other interactors that may be unique to each isoform . Thus future studies to explore the possibility of differential regulation of eIF2α phosphatase activity by the different paralogues and their unique interactors seem warranted . Jasplakinolide , thapsigargin , and tunicamycin were from Calbiochem ( Millipore , Hertfordshire , UK ) , cytochalasin D was from Tocris ( Bristol , UK ) , latrunculin B was from Enzo Life Sciences ( Exeter , UK ) , Alexa Fluor 568 Phalliodin was from Life Technologies ( Paisley , UK ) . PPP1R15ApEGFP-C3 and PPP1R15ApEGFP-N1 were kind gifts from S Shenolikar ( Duke-NUS Graduate Medical School Singapore , Singapore ) ( Zhou et al . , 2011 ) . PerkKD-pGEX4T-1 , dPPP1R15pEGFP , 2aOPTx3M ( 1–185 ) pET-30a ( + ) , PPP1R15ApcDNA and dPPP1R15pEGFP have been described previously ( Harding et al . , 1999; Novoa et al . , 2003; Ito et al . , 2004; Malzer et al . , 2013 ) . PP1αpEBG was generated by ligating the human PP1α coding sequence into BamHI and NotI digested pEBG . For inducible HeLa cell lines , GFP-PPP1R15A was excised from PPP1R15ApEGFP-C3 with NheI and XhoI and ligated into NheI and SalI digested pTRE2Hyg ( Clontech Laboratories , USA ) to generate GFP-PPP1R15ApTRE2Hyg . PPP1R15A-GFP was excised from PPP1R15ApEGFP-N1 with BglII and NotI and ligated into BamHI and NotI digested pTRE2Hyg to generated PPP1R15A-GFPpTRE2Hyg . For PPP1R15B-GFP , PPP1R15BpEGFP-C1 was generated by ligating the human PPP1R15B coding sequence into BglII and SalI digested pEGFP-C1 . For Flp-In T-REx HEK293 cell lines expressing GFP-R15B 1–146 and GFP-R15B 1–146_mDia2 , the coding sequence for EGFP and residues 1–146 of human PPP1R15B was mobilized by digestion with NheI ( partially repaired with Klenow-polymerase ) and BamHI , before ligation into pcDNA5_TO_FRT ( Life Technologies , USA ) digested with HindIII ( partially repaired with Klenow polymerase ) and BamHI to generate EGFP_PPP1R15B_1–146pcDNA5_TO_FRT . PCR product encoding residues 532-1171 of mDia2 was ligated into BamHI and XhoI digested EGFP_PPP1R15B_1–146 pcDNA5_TO_FRT to generate EGFP_PPP1R15B_1–146_mDia2_532-1171pcDNA5_TO_FRT . Primers used in this study are listing in Table 1 . 10 . 7554/eLife . 04872 . 019Table 1 . Primers used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 04872 . 019hPPP1R15A truncation or point mutation 1–615ForCCTGCTGCCCGGGCCAGAGCCTGAGCACGCCTCAGGAACCRevGGTTCCTGAGGCGTGCTCAGGCTCTGGCCCGGGCAGCAGG 1–620ForCCTGGGCACGCCTCAGGTAGCCACCTTTAGCCRevGGCTAAAGGTGGCTACCTGAGGCGTGCCCAGG 501–654ForTAAGCTAGCACCATGGAAGCTGAGCCCRevGTCGCGGCCGCTTTACTTGTACAGCTC V556EForCTAAAGGCCAGAAAGGAGCGCTTCTCCGAGAAGGTCACTGRevCAGTGACCTTCTCGGAGAAGCGCTCCTTTCTGGCCTTTAG W616AForCCGGGCCAGAGCCGCGGCACGCCTCAGGAARevTTCCTGAGGCGTGCCGCGGCTCTGGCCCGG R618AForGCCAGAGCCTGGGCAGCCCTCAGGAACCCARevTGGGTTCCTGAGGGCTGCCCAGGCTCTGGC L619AForAGCCTGGGCACGCGCCAGGAACCCACCTTTRevAAAGGTGGGTTCCTGGCGCGTGCCCAGGCT R620AForCCTGGGCACGCCTCGCGAACCCACCTTTAGRevCTAAAGGTGGGTTCGCGAGGCGTGCCCAGG W616A/L619AForCGGGCCAGAGCCGCGGCACGCGCCAGGAACCCRevGGGTTCCTGGCGCGTGCCGCGGCTCTGGCCCG mDia2 fusionForAATCCCGGATCCGTGCCTTGCCACCTGGTACARevAGCTCGCTCGAGTTATAAAGCTCGTAATCTTGCCAGPPP1R15B fusion GFP fusionForAGATTAGATCTGCCACCATGGAGCCGGGGACAGGRevGATCGTCGACACATTGCTTGAGAACATTAAGTCCdPPP1R15 truncation 1–307ForCCAGTTCACCGAGATCGTTAGTACCAAGCTCGATTCTTGCACGRevCGTGCAAGAATCGAGCTTGGTACTAACGATCTCGGTGAACTGG 1–312ForCGTGTCTACCAAGCTCGATAGTTGCACGAGGACGAGCRevGCTCGTCCTCGTGCAACTATCGAGCTTGGTAGACACG All truncations or point mutations in the PPP1R15A coding sequence were made as follows . Fifty nanograms of plasmid template DNA were mixed with 5 µl Pfu turbo DNA polymerase reaction buffer [10×] , 1 µl Pfu turbo DNA polymerase ( Agilent Technologies , Santa Clara , CA ) , 125 ng forward primer , 125 ng reverse primer , 1 µl of 25 mM dNTPs , made up to 50 µl with water . A PCR thermocycler was run using the following program parameters: 95°C for 30 s , 95°C for 30 s , 18 cycles ( 54°C for 1 min , 67°C for 20 min , 94°C for 1 min , 55°C for 1 min , 72°C for 10 min ) . Completed reactions were treated with 1 µl Dpn1 restriction enzyme , incubated at 37°C for 2 hr before using 5 µl of the reaction mix for a standard transformation into One Shot TOP10 chemically competent E . coli ( Life Technologies , Paisley , UK ) . Mammalian cells , HEK293T , MEF ( Ppp1r15btm1Dron/tm1Dron , Ppp1r15atm1Dron/tm1Dron , Pkr−/− , Hri−/− , Perk−/− , Gcn2−/− , eIF2αAA ) , and NIH3T3 , were maintained in DMEM supplemented with 10% vol/vol FBS and antibiotics ( 100U/ml Penicillin G and 100 µg/ml Streptomycin ) and incubated at 37°C with 5% vol/vol CO2 ( Yang et al . , 1995; Harding et al . , 2000; Han et al . , 2001; Novoa et al . , 2003; Scheuner et al . , 2005; Harding et al . , 2009 ) . HeLa Tet-On Advanced cells were purchased from Clontech Laboratories ( Saint-Germain-en-Laye , France ) and maintained in DMEM with 10% vol/vol tetracycline-free FBS and transfected with the expression vectors PPP1R15A-GFPpTRE2Hyg and GFP-PPP1R15ApTRE2Hyg . Stable clones were selected with 600 µM hygromycin . Transgene expression proved optimal when clones were treated with 1 µg/ml doxycycline . Cell lysates were prepared in Harvest lysis buffer ( HEPES pH 7 . 9 , 10 mM; NaCl 50 mM; sucrose 0 . 5M; EDTA 0 . 1 mM; Triton X-100 0 . 5% vol/vol ) supplemented with protease inhibitor cocktail ( Roche , Welwyn Garden City , UK ) and 1 mM PMSF . When analysing phospho-eIF2α , the lysis buffer was supplemented with phosphatase inhibitors ( 10 mM tetrasodium pyrophosphate , 15 . 5 mM β-glycerophosphate , 100 mM NaF ) . Cleared cell extracts were equalized by total cell protein using Bio-Rad protein assay ( Bio-Rad , Hercules , CA , USA ) , boiled in SDS-loading buffer ( 25 mM Tris pH 6 . 8 , 7 . 5% vol/vol glycerol , 1% wt/vol SDS , 25 mM DTT , 0 . 05% wt/vol bromophenol blue ) , subjected to reducing SDS-PAGE , and transferred to nitrocellulose membrane . For GFP-Trap affinity purification , cells were lysed in the manufacturer's recommended buffers ( Chromotek , Planegg-Martinsried , Germany ) and incubated with GFP-Trap A beads according to manufacturer's instructions . Briefly , cells were lysed in GFP-Trap lysis buffer ( 150 mM NaCl , 10 mM Tris/Cl pH 7 . 5 , 0 . 5 mM EDTA , 1 mM PMSF , and Protease Inhibitor Cocktail [Roche] ) and post-nuclear supernatants were incubated with GFP-Trap beads at 4°C for 2 hr then washed four times in the same buffer . Proteins were eluted with SDS-PAGE loading buffer . GST affinity purification was performed using Activated Thiol Sepharose 4B beads ( GE Healthcare , Little Chalfont , UK ) . Briefly , cells were lysed with Harvest buffer , cleared by centrifugation and incubated with rotation with Activated Thiol Sepharose 4B beads for 2 hr at 4°C . Beads were then washed four times with lysis buffer and protein complexes then eluted by boiling with SDS-loading buffer or by addition of 20 mM glutathione . For isolation of endogenous PPP1R15A by immunoprecipitation and PP1 by microcystin-affinity purification , MEF or HEK293T cells were lysed in lysis buffer ( 150 mM KCl , 20 mM HEPES pH 7 . 4 , 2 mM MgCl2 , 1 mM PMSF , and Protease Inhibitor Cocktail ) supplemented with either 0 . 1% ( wt/vol ) digitonin ( Calbiochem , MERK Millipore , Darmstadt , Germany ) or 0 . 5% ( vol/vol ) triton X-100 , respectively . Immunoprecipitation of PPP1R15A was carried out for 16 hr at 4°C prior to three washes with detergent supplemented lysis buffer and elution in SDS-PAGE sample buffer . PP1 isolation by microcystin affinity purification was carried out for 1 hr at 4°C in the presence of 1 mM latrunculin B prior to three washes with detergent-supplemented lysis buffer and elution in SDS-PAGE sample buffer . Lysates were centrifuged at 200 , 000×g for 30 min to remove contaminating F-actin in order that G-actin levels were reflected in input samples . Primary antibodies used were: rabbit anti-PPP1R15A ( 10 , 449-1-AP , 1:1000; Proteintech , Manchester , UK ) mouse anti-GFP antibody ( ab1218 , 1∶1000; Abcam , Cambridge , UK ) , rabbit anti-PP1α antibody ( no . 2582 , 1∶1000; Cell Signaling , Danvers , MA , USA ) , rabbit p-eIF2α ( 3597 , Cell Signaling; 1∶1000 ) , anti-actin ( ab3280 , 1∶1000; Abcam ) , rabbit anti-ATF4 ( C-20 , 1:500; SantaCruz , Santa Cruz , CA , USA ) , mouse anti-puromycin antiserum ( PMY-2A4 , Developmental Studies Hybridoma Bank , University of Iowa , USA ) , anti-total eIF2α mouse monoclonal ( AHO0802 , 1:1000; Invitrogen , Thermo Fisher Scientific , Waltham , MA , USA ) . Gels were stained with InstantBlue Coomassie stain ( Expedeon , San Diego , CA , USA ) as directed by the manufacturer's instructions . For silver staining: gels were fixed ( 10% vol/vol methanol , 7 . 5% vol/vol acetic acid ) for 20 min with agitation followed by two quick rinses with water . Gels were then incubated with 3 . 25 µM DTT in water for 20 min with agitation then 0 . 1% wt/vol AgNO3 in water for 30 min with agitation . Following a 1-min wash with water , gels were developed using 3% wt/vol Na2CO3 , 0 . 02% wt/vol formaldehyde in water until bands became visible and the reaction was stopped with fixative . HEK293T cells were transfected with PP1αpEBG and untagged PPP1R15ApcDNA . After 24 hr , cells were lysed in harvest buffer and subjected to GST affinity purification . Protein complexes were eluted with 20 mM reduced glutathione in 50 mM Tris pH 7 . 5 . The eluate was mixed with 10 µM purified F-actin in actin binding buffer ( 20 mM Tris pH 8 , 100 mM NaCl , 2 mM MgCl2 , 1 mM ATP , 1 mM DTT , 0 . 1 mM CaCl2 ) in a total volume of 200 µl . Samples were centrifuged at 279 , 000×g for 15 min in a TLA120 . 1 rotor . The supernatant was removed and mixed with 50 µl of 4× SDS loading buffer , while the pellet was re-suspended in 250 µl of 1× SDS loading buffer . Samples were then boiled and analysed by SDS-PAGE . Phosphorylated recombinant eIF2α N-terminal domain ( NTD ) was generated as described previously ( Marciniak et al . , 2006 ) . The expression plasmid PerkKD-pGEX4T-1 encoding GST-PERK kinase domain fusion protein of mouse PERK residues 537–1114 wild type has previously been described ( Harding et al . , 1999 ) . eIF2α-NTD encoding residues 1–185 of human eIF2α with three solubilizing mutations was bacterially expressed from codon optimized vector 2aOPTx3M ( 1–185 ) pET-30a ( + ) ( Ito et al . , 2004 ) . Bacterially expressed GST-PERK immobilised on activated thiol sepharose beads was incubated with 10 µl of 1 mM ATP and bacterially expressed eIF2α-NTD at 37°C with shaking in 20 µl kinase buffer ( 5×: 100 mM TRIS pH 7 . 4 , 250 mM KCl , 10 mM Mg ( OAc ) 2 , 10 mM MnCl2 , and 5 mM DTT ) made up to 100 µl total reaction volume . GST-PERK beads were removed by centrifugation and remaining ATP was removed by dialysis against reaction buffer . The resulting phosphorylation eIF2α-NTD was incubated with affinity-purified phosphatase in 20 mM Tris HCL pH 7 . 4 , 50 mM KCl , 2 mM MgCl2 , 0 . 1 mM EDTA , 0 . 8 mM ATP at 30°C for indicates times with shaking . Reactions were terminated by the addition of Laemmli buffer . Cells plated onto glass coverslips were washed twice with PBS and fixed with 4% formaldehyde for 20 min . Following a further two PBS washes , cells were then permeabilised with 0 . 5% vol/vol triton X-100 in PBS for 3 min then blocked with 1% wt/vol BSA in PBS for 1 hr . Cells were then incubated in the dark with Alexa-Fluor 568 phalloidin 1:40 for 1 hr . After three 5-min washes in PBS , the glass coverslips were mounted onto slides using ProLong Gold antifade reagent ( Life Technologies ) ready for visualisation .	For a cell to build a protein , it must first copy the instructions contained within a gene . A complex molecular machine called a ribosome then reads these instructions and translates them into a protein . This translation process involves a number of steps . Proteins called eukaryotic translation initiation factors ( or eIFs for short ) coordinate the first step in the process , which is known as ‘initiation’ . The eIFs also provide the cell with ways to control how quickly it makes proteins . For example , when a cell is stressed , either by starvation or toxins , it adds a phosphate group onto part of an eIF protein , called eIF2α . This modification makes this eIF protein less able to initiate translation , and so the cell builds fewer proteins and conserves more of its resources during times of stress . Once the stressful conditions are over , the phosphate group is removed from eIF2α by an enzyme called a phosphatase . This phosphatase contains two subunits: one that recognizes eIF2α and another that removes the phosphate group . However , experiments that attempted to recreate this phosphatase activity using just these two subunits in a test tube failed to generate a working enzyme that specifically targeted the phosphate group of eIF2α . This suggests that in cells this enzyme contains an additional unknown subunit . Now , Chambers , Dalton et al . ( and Chen et al . ) report the identity of a ‘missing’ third subunit as a protein known as globular-actin or G-actin . Chambers , Dalton et al . engineered human and fruit fly cells to add ‘molecular handles’ on the two known subunits of the phosphatase enzyme . These handles could then be used to essentially pull these proteins out of the mixture of molecules within a cell and see what other proteins came along too . Both of the known subunits ‘pulled’ G-actin along with them; this suggested that it could be the missing part of the phosphatase enzyme . Further experiments confirmed that G-actin works together with the other two subunits to specifically remove the phosphate group from eIF2α in mouse cells that had been stressed using a harmful chemical . Individual G-actin proteins can bind together to form long filaments , and signals that encourage a cell to divide or move also trigger the formation of actin filaments . This reduces the activity of the phosphatase enzyme by depriving it of a crucial component , i . e . , free G-actin proteins . As such , the new mechanism described by Chambers , Dalton et al . suggests how growth and movement signals might also change a cell's sensitivity to stress . These findings may hopefully enable stressed cells to be targeted by drugs to treat disease; but future work is needed to clarify under what circumstances the integration of such signals into the stress response is beneficial to the cell .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "cell", "biology" ]	2015	Actin dynamics tune the integrated stress response by regulating eukaryotic initiation factor 2α dephosphorylation
Mitochondrial DNA copy number ( mtDNA-CN ) is an accessible blood-based measurement believed to capture underlying mitochondrial ( MT ) function . The specific biological processes underpinning its regulation , and whether those processes are causative for disease , is an area of active investigation . We developed a novel method for array-based mtDNA-CN estimation suitable for biobank-scale studies , called ‘automatic mitochondrial copy ( AutoMitoC ) . ’ We applied AutoMitoC to 395 , 781 UKBiobank study participants and performed genome- and exome-wide association studies , identifying novel common and rare genetic determinants . Finally , we performed two-sample Mendelian randomization to assess whether genetically low mtDNA-CN influenced select MT phenotypes . Overall , genetic analyses identified 71 loci for mtDNA-CN , which implicated several genes involved in rare mtDNA depletion disorders , deoxynucleoside triphosphate ( dNTP ) metabolism , and the MT central dogma . Rare variant analysis identified SAMHD1 mutation carriers as having higher mtDNA-CN ( beta = 0 . 23 SDs; 95% CI , 0 . 18–0 . 29; p=2 . 6 × 10-19 ) , a potential therapeutic target for patients with mtDNA depletion disorders , but at increased risk of breast cancer ( OR = 1 . 91; 95% CI , 1 . 52–2 . 40; p=2 . 7 × 10-8 ) . Finally , Mendelian randomization analyses suggest a causal effect of low mtDNA-CN on dementia risk ( OR = 1 . 94 per 1 SD decrease in mtDNA-CN; 95% CI , 1 . 55–2 . 32; p=7 . 5 × 10-4 ) . Altogether , our genetic findings indicate that mtDNA-CN is a complex biomarker reflecting specific MT processes related to mtDNA regulation , and that these processes are causally related to human diseases . No funds supported this specific investigation . Awards and positions supporting authors include: Canadian Institutes of Health Research ( CIHR ) Frederick Banting and Charles Best Canada Graduate Scholarships Doctoral Award ( MC , PM ) ; CIHR Post-Doctoral Fellowship Award ( RM ) ; Wellcome Trust Grant number: 099313/B/12/A; Crasnow Travel Scholarship; Bongani Mayosi UCT-PHRI Scholarship 2019/2020 ( TM ) ; Wellcome Trust Health Research Board Irish Clinical Academic Training ( ICAT ) Programme Grant Number: 203930/B/16/Z ( CJ ) ; European Research Council COSIP Grant Number: 640580 ( MO ) ; E . J . Moran Campbell Internal Career Research Award ( MP ) ; CISCO Professorship in Integrated Health Systems and Canada Research Chair in Genetic and Molecular Epidemiology ( GP ) Mitochondria are semiautonomous organelles present in nearly every human cell that execute fundamental cellular processes including oxidative phosphorylation , calcium storage , and apoptotic signaling . Mitochondrial ( MT ) dysfunction has been implicated as the underlying cause for many human disorders based on mechanistic in vitro and in vivo studies ( Burbulla et al . , 2017; Desdín-Micó et al . , 2020; Sliter et al . , 2018 ) . Complementary evidence comes from recent epidemiological studies that measure mitochondrial DNA copy number ( mtDNA-CN ) , an MT-derived marker that can be conveniently measured from peripheral blood . Since mitochondria contain their own unique set of genomes that are distinct from the nuclear genome , the ratio of mtDNA to nuclear DNA molecules ( mtDNA-CN ) in a sample serves as an accessible marker of MT DNA abundance per cell ( Longchamps et al . , 2020 ) . Indeed , observational studies suggest that individuals with lower mtDNA-CN are at higher risk of age-related complex diseases , such as coronary artery disease , sudden cardiac death , cardiomegaly , stroke , portal hypertension , and chronic kidney disease ( Tin et al . , 2016; Ashar et al . , 2017; Zhang et al . , 2017; Hägg et al . , 2021 ) . Conversely , higher mtDNA-CN levels have been associated with increased cancer incidence ( Kim et al . , 2015; Hu et al . , 2016 ) . While previous studies demonstrate that mtDNA-CN is associated with various diseases , evidence suggests that it may also play a direct and causative role in human health and disease . For example , in cases of mtDNA depletion syndrome , wherein rare defects in nuclear genes responsible for replicating and/or maintaining mtDNA lead to deficient mtDNA-CN ( Gorman et al . , 2016 ) , patients manifest with severe dysfunction of energy-dependent tissues ( heart , brain , liver , and cardiac and skeletal muscles ) . So far , 19 genes have been reported to cause mtDNA depletion ( Oyston , 1998 ) . In addition to these rare monogenic syndromes , the importance of common genetic variation in regulating mtDNA-CN is an active area of research with approximately 50 common loci identified so far ( Cai et al . , 2015; Guyatt et al . , 2019; Longchamps , 2019; Hägg et al . , 2021 ) . In contrast to marked drops in mtDNA-CN by 60–80% as seen in those with rare mtDNA depletion syndromes , the relevance of subtler perturbations in mtDNA-CN in disease risk remains to be determined ( Basel , 2020 ) . Granted , the connection between blood mtDNA-CN and aspects of MT biology remains unclear with many studies showing only moderate correlation between mtDNA-CN and markers of MT function or biogenesis ( Frahm et al . , 2005; Wachsmuth et al . , 2016 ) . Further complicating the interpretation of epidemiological associations between blood mtDNA-CN and disease risk is the fact that ( i ) blood mtDNA-CN is strongly confounded by blood cell composition , particularly , platelets that are devoid of nuclei and that ( ii ) blood mtDNA-CN does not correlate well with mtDNA-CN measured in other tissues in which MT dysfunction may be more relevant ( Picard , 2021 ) . Accordingly , understanding the genetic determinants of blood mtDNA-CN may provide a better understanding of the etiological processes reflected by this poorly understood MT biomarker in the blood . To interrogate mtDNA-CN as a potential determinant of human diseases and to better understand its biological relevance to mitochondria , we performed extensive genetic investigations in up to 395 , 781 participants from the UKBiobank study ( Sudlow et al . , 2015 ) . We first developed and validated a novel method for biobank-scale mtDNA-CN investigations that leverages single nucleotide polymorphism ( SNP ) array intensities , called ‘automatic mitochondrial copy ( AutoMitoC ) . ’ Leveraging AutoMitoC-based mtDNA-CN estimates , we performed large-scale genome-wide association study ( GWAS ) and exome-wide association study ( ExWAS ) to identify common and rare genetic variants contributing to population-level variation in mtDNA-CN . Various analyses were then conducted to build on previous publications regarding the specific genes and pathways underlying mtDNA-CN regulation ( Cai et al . , 2015; Guyatt et al . , 2019; Longchamps , 2019; Hägg et al . , 2021 ) . Finally , we applied Mendelian randomization analyses to assess potential causal relationships between mtDNA-CN and disease susceptibility . We built on an existing framework for processing normalized SNP probe intensities ( log2ratio [L2R] values ) from genetic arrays into mtDNA-CN estimates known as the ‘MitoPipeline’ ( Lane , 2014 ) ( Materials and Methods , Figure 1 ) . The MitoPipeline yields mtDNA-CN estimates that correlate with direct qPCR measurements and has been successfully implemented in several epidemiological investigations ( Ashar et al . , 2017; Zhang et al . , 2017; Figure 1—figure supplements 1–2 ) . We developed a novel method , ‘AutoMitoC , ’ which incorporates three amendments to facilitate large-scale investigations of mtDNA-CN ( Figure 1 ) . First , AutoMitoC replaces autosomal signal normalization of common variants with globally rare variants which negates the need for linkage disequilibrium pruning ( Materials and Methods , Figure 1—figure supplements 3–4 ) . As a result , this simplifies derivation of mtDNA-CN estimates in ethnically diverse cohorts by allowing for use of a single , universal variant set for normalization . Second , to detect potentially cross-hybridizing probes , we empirically assess the association of corrected probe signal intensities with off-target genome intensities ( Materials and Methods , Supplementary file 1 – Tab 1–2 ) , rather than relying on sequence homology of probe sequences , which is not always available . Last , the primary estimate of MT signal is ascertained using principal component analysis ( [PCA]; as opposed to using the median signal intensity of MT probes as per the MitoPipeline ) which improves concordance of array-based mtDNA-CN estimates with those derived from alternative methods ( Materials and Methods ) . A detailed description of the development of the AutoMitoC pipeline is provided in the Materials and Methods . To benchmark performance of AutoMitoC , array-based mtDNA-CN estimates were compared to complementary measures of mtDNA-CN in two independent studies . First , array-based mtDNA-CN estimates were derived in a subset of 34 , 436 UKBiobank participants with available whole exome sequencing ( WES ) data . Reference mtDNA-CN estimates were derived from the proportion of WES reads aligned to the MT genome relative to the autosome ( Longchamps , 2019 ) . AutoMitoC estimates were significantly correlated with WES estimates ( r = 0 . 45; p<2 . 23 × 10–308 ) . Since WES data involves enrichment for nuclear coding genes and therefore could result in biased reference estimates for mtDNA-CN , we also performed an independent validation in an ethnically diverse study of 5791 participants where mtDNA-CN was measured using qPCR , the current gold standard assay ( Fazzini et al . , 2018 ) . Indeed , we observed stronger correlation between AutoMitoC and qPCR-based estimates ( r = 0 . 64; p<2 . 23 × 10–308; Figure 1—figure supplement 5 ) . Furthermore , AutoMitoC demonstrated robust performance ( r > 0 . 53 ) across all ethnic strata in the secondary validation cohort including Europeans ( N = 2431 ) , Latin Americans ( N = 1704 ) , Africans ( N = 542 ) , South East Asians ( N = 471 ) , South Asians ( N = 186 ) , and others ( N = 360; ; Figure 1—figure supplement 5 ) . Bland Altman plots also illustrate the extent of agreement between methods ( Figure 1—figure supplement 6 ) . For every ethnicity , 95% limits of agreement intervals were smaller than expected by chance . Last , while all analyses hitherto followed the MitoPipeline condition of requiring >40 , 000 autosomal variants for normalization , we observed comparable performance using even 1000 random very rare probes ( MAF <0 . 001; r = 0 . 60; p<5 × 10–300 ) for signal normalization which reduced the runtime from several hours to less than 10 min for these 5791 samples running on 10 CPUs . In the larger UKBiobank dataset of 395 , 781 with suitable array-based estimates , the distribution of mtDNA-CN was approximately normal ( Figure 2A ) . To further verify that AutoMitoC-based estimates were indeed capturing mtDNA-CN , we performed association testing between mtDNA-CN values and known predictors including age , sex , ethnicity , and blood cell composition . Consistent with previous reports , every decade increase in age was associated with lower levels ( beta = −0 . 08 SDs; 95% CI , –0 . 07 to –0 . 08; p<2 . 23 × 10–308 ) , and females had higher levels than males ( beta = 0 . 12 SDs; 95% CI , 0 . 11–0 . 12; p=1 . 67 × 10–299 ) . MtDNA-CN levels also differed between ethnicities with South Asians ( beta = −0 . 18; 95% CI , –0 . 16 to –0 . 21; p=2 . 93 × 10–47 ) and Africans ( beta = −0 . 18; 95% CI , –0 . 16 to –0 . 21; p=8 . 39 × 10–48 ) having significantly lower levels than Europeans . Collectively , age , sex , and ethnicity accounted for 0 . 83% of the phenotypic variance in mtDNA-CN levels . All types of cell counts were significantly associated with mtDNA-CN including the standard set of covariates commonly adjusted for in mtDNA-CN investigations , namely , white blood cells ( beta = −0 . 25 SDs in mtDNA-CN levels per 1 SD increase in cell counts; 95% CI , –0 . 25 to –0 . 24; p<2 . 23–308 ) and platelets ( beta = 0 . 07; 95% CI , 0 . 06–0 . 07; p<2 . 23 × 10–308 ) ( Figure 2B ) . White blood cell and platelet counts each accounted for 6 . 3% and 0 . 5% of the variance in mtDNA-CN levels , respectively ( Figure 2B; Figure 2—source data 1 ) . Notably , neutrophil count ( beta = −0 . 31; 95% CI , –0 . 31 to –0 . 30; p<2 . 23–308 ) was the strongest predictor of mtDNA-CN levels and accounted for more variance ( 9 . 9% ) than white blood cell and platelet counts combined ( 8 . 2% ) . Collectively , total white blood cell , platelet , and neutrophil counts explained 12 . 3% variance in mtDNA-CN levels . A GWAS was performed testing the association of 11 , 453 , 766 common genetic variants ( MAF >0 . 005 ) with mtDNA-CN in 383 , 476 UKBiobank participants of European ancestry ( Figure 3 ) . In total , 9602 variants were associated with mtDNA-CN at genome-wide significance ( Figure 3A; Figure 3—figure supplement 1 ) , encompassing 82 independent signals in 72 loci ( Supplementary file 2 – Tab 1; Figure 3—source data 1 ) . The genomic inflation factor was 1 . 16 and the LD score intercept was 1 . 036 , indicating that most inflation in test statistics was attributable to polygenicity . Sensitivity analyses revealed that nuclear mitochondrial DNA ( NUMT ) interference may have played a role in two independent signals ( two loci ) , which were subsequently discarded , leading to a total of 80 independent signals in 70 loci . These 80 independent genetic signals explained 1 . 48% variance in mtDNA-CN levels . Fine-mapping via the FINEMAP algorithm ( Benner et al . , 2016 ) yielded 95% credible sets containing 2363 genome-wide significant variants . Of the 80 independent genetic associations , 17 ( 22% ) mapped to a single candidate causal variant; 32 ( 39% ) mapped to 5 or fewer variants , and 42 ( 51% ) mapped to 10 or fewer variants ( Figure 3B; Supplementary file 2 – Tab 2 ) . Credible sets for 11 genetic signals overlapped with genes responsible for rare mtDNA depletion disorders including DGUOK ( 3 ) , MGME1 ( 2 ) , TFAM ( 2 ) , TWNK ( 2 ) , POLG2 ( 1 ) , and TYMP ( 1 ) ( Supplementary file 2 – Tab 1 & 2 ) . Several associations mapped to coding variants with high posterior probability . DGUOK associations mapped to a synonymous variant ( rs62641680; posterior probability = 1 ) and a nonsynonymous variant ( rs74874677; p=1 ) . TFAM associations mapped to a 5’ untranslated region ( UTR ) variant ( rs12247015; p=1 ) falling within an ENCODE candidate cis-regulatory element with a promotor-like signature and an intronic variant ( rs4397793; p=1 ) with a proximal enhancer-like signature . Last , POLG2 associations mapped to a nonsynonymous variant ( rs17850455; p=1 ) . Beyond the six aforementioned mtDNA depletion genes identified at genome-wide significance , suggestive associations were found for POLG ( rs2307441; p=1 . 0 × 10–7 ) , OPA1 ( rs9872432; p=5 . 2 × 10–7 ) , SLC25A10 ( rs62077224; p=1 . 2 × 10–7 ) , and RRM2B ( rs3907099; p=4 . 7 × 10–6 ) . Given these observations , we hypothesized that mtDNA depletion genes may be generally enriched for common variant associations . Indeed , 10 ( 53% ) of 19 known mtDNA depletion genes ( Oyston , 1998 ) harbored at least suggestive mtDNA-CN associations ( p<5 × 10–6 ) . We also compared our findings to genome-wide significant results from two recent GWAS of mtDNA-CN by Hägg et al . , 2021 and Longchamps et al . , 2022 to better contextualize our findings . Of the 66 independent signals identified by Hägg et al . , 2021 , we found varying levels of evidence to support 43 ( 65% ) genetic associations; 41 ( 62% ) variants were detected at Bonferroni significance; 2 ( 3% ) variants were detected using a suggestive significance threshold ( p<5 × 10–6 ) ( Supplementary file 2 – Tab 3 ) . For the remaining 23 ( 35% ) variants not detected at suggestive significance , we hypothesized that differences in covariate adjustment strategies . Specifically , Hagg et al . did not adjust for platelets which is an important confounder of blood mtDNA-CN levels which might explain the majority of these discrepancies . As a sensitivity analysis , we repeated association testing without adjusting for blood cell traits and were able to recover 19/23 ( 83% ) of the remaining variants at suggestive significance threshold , which suggests that approximately 1 in 3 associations detected by Hägg et al . , 2021 may not be robust to adjustment for platelets ( Supplementary file 2 – Tab 3 ) . While we could not perform the reverse lookup of our top GWAS hits within the Hägg et al . , 2021 GWAS due to a lack of full genome-wide summary statistics , 48 ( 60% ) of our 80 genetic associations are likely novel relative to Hägg et al . , 2021 as these variants were not reported as genome-wide significant nor in linkage disequilibrium ( r2 >0 . 10 ) with their top hits ( Supplementary file 2 – Tab 1 ) . Among the 133 independent signals identified by Longchamps et al . , 2022 , 119 achieved Bonferroni significance , of which 116 were available in our GWAS ( Supplementary file 2 – Tab 4 ) . The remaining three variants ( or their proxies; r2 >0 . 7 ) did not meet the MAF threshold for inclusion ( MAF >0 . 005 ) . Of the 116 detectable variants , we found varying levels of evidence to support 112 ( 97% ) genetic associations; 70 ( 60% ) variants were detected at Bonferroni significance; 15 ( 13% ) variants were detected using a suggestive significance threshold; 27 ( 24% ) variants were detected at nominal significance with concordant directionality ( Supplementary file 2 – Tab 4 ) . While we could not perform the reverse lookup of our top GWAS hits within the Longchamps et al . , 2022 GWAS , 11 ( 14% ) of our 80 genetic associations are likely novel as these variants were not reported as genome-wide significant nor in linkage disequilibrium with their top hits ( Supplementary file 2 – Tab 1 ) . Notably , these 11 associations were also not reported in Hägg et al . , 2021 . Additionally , transethnic meta-analysis inclusive of non-Europeans ( N = 395 , 781 ) was performed but given the small increase in sample size , GWAS findings remained highly similar ( Figure 3—figure supplement 2 ) . However , European effect estimates were significantly and highly correlated with those derived from South Asian ( r = 0 . 97; p=2 . 2 × 10–15 ) and African ( r = 0 . 88; p=9 . 1 × 10–5 ) GWAS analyses ( Figure 3—figure supplement 3 ) . We postulated that differential expression of genes encoded by the MT genome may trigger changes in copy number , and thus a subset of identified loci may affect mtDNA-CN through MT genome transcription . Ali et al . , 2019 recently conducted a GWAS to identify nuclear genetic variants associated with variation in mtDNA-encoded gene expression ( i . e . mitochondrial expression quantitative trait loci [mt-eQTLs] ) ( Ali et al . , 2019 ) . Nonsynonymous variants in LONP1 ( rs11085147 ) and TBRG4 ( rs2304693 ) , as well as an intronic variant in MRPS35 ( rs1127787 ) , were associated with changes in MT gene expression across various tissues ( Supplementary file 2 – Tab 5 ) . Nominally associated mtDNA-CN loci were also observed to influence MT gene expression including intronic variants in both PNPT1 ( rs62165226; mtDNA-CN p=5 . 5 × 10–5 ) and LRPPRC ( rs10205130; mtDNA-CN p=1 . 1 × 10–4 ) . Although differences in MT gene expression may be a consequence rather than a cause of variable mtDNA-CN , the analysis performed by Ali et al . , 2019 was corrected for factors associated with global changes in the MT transcriptome ( Ali et al . , 2019 ) . Moreover , the direction of effect estimates between mtDNA-CN and mt-eQTLs varied depending on gene and tissue context . Altogether , such findings imply that some mtDNA-CN loci may regulate mtDNA-CN by influencing MT gene expression . Heteroplasmy refers to the coexistence of multiple mtDNA alleles within an individual for a particular variant , which is a function of the multicopy nature of the MT genome . A recent GWAS by Nandakumar et al . , 2021 for mean heteroplasmy levels in saliva specimen provided initial evidence supporting a shared genetic basis for heteroplasmy and copy number . To further explore the overlap in genetic determinants of these traits , we searched for the previously reported heteroplasmy loci within our mtDNA-CN GWAS . Of 19 matching variants between the GWAS , 4 heteroplasmy loci were also associated with mtDNA-CN at genome-wide significance including variants nearby or within TINCR/LONP1 ( rs12461806; mtDNA-CN GWAS p=7 . 5 × 10–88 ) , TWNK/MPRL43 ( rs58678340; p=1 . 3 × 10–39 ) , TFAM ( rs1049432; p=1 . 5 × 10–21 ) , and PRKAB1 ( rs11064881; p=2 . 6 × 10–10 ) genes . Consistent with the initial finding from Nandakumar et al . , 2021 that the heteroplasmy-increasing TFAM allele was also associated with higher mtDNA-CN , we also observed concordant directionality for the other three variants . No additional mtDNA heteroplasmy loci were identified to influence mtDNA-CN when using a more liberal suggestive significance threshold . Full genome-wide summary statistics for the heteroplasmy GWAS were not publicly available so we could not perform the reverse lookup to examine whether the 80 mtDNA-CN variants affected heteroplasmy . DEPICT analysis led to the prioritization of 91 out of 18 , 922 genes ( false discovery rate ( FDR ) p<0 . 05; Supplementary file 2 – Tab 6 ) . Among them , 87 of the genes intersected with the GeneMANIA database and were uploaded to the GeneMANIA platform to identify additional functionally related genes ( Warde-Farley et al . , 2010 ) . GeneMANIA analysis discovered an additional 20 related genes ( Supplementary file 2 – Tab 7 ) . Among the 107 total genes prioritized by DEPICT or GeneMANIA ( Figure 3C ) , MT functions were significantly enriched in gene ontology ( GO ) terms including mitochondrion organization ( coverage: 12/225 genes; FDR p=7 . 4 × 10–5 ) , MT nucleoid ( 6/34; FDR p=2 . 2 × 10–4 ) , MT genome maintenance ( 4/10; FDR p=6 . 8 × 10–4 ) , and MT matrix ( 11/257; FDR p=6 . 8 × 10–4 ) . Visual inspection of the links between key genes involved in these functions highlights PPRC1 , a member of the Peroxisome proliferator-activated receptor-gamma coactivator-1alpha ( PGC-1A ) family of MT biogenesis activators ( Scarpulla , 2011 ) , as a potential coordinator of mtDNA-related processes ( Figure 3C ) . MitoCarta3 is a comprehensive and curated inventory of 1136 human proteins ( 1120 nuclear ) known to localize to the mitochondria based on experiments of isolated mitochondria from 14 nonblood tissues ( Rath et al . , 2021 ) . We leveraged this recently updated database , that was absent from GeneMANIA , to conduct a complementary set of targeted analyses focused on MT annotations ( Supplementary file 2 – Tab 7 ) . First , we hypothesized that prioritized genes would be generally enriched for genes encoding the MT proteome . Overall , 27 ( 25% ) of 107 genes had evidence of MT localization corresponding to a 4 . 2-fold enrichment ( null expectation = 5 . 9%; p=1 . 0 × 10–10 ) . Next , given that PPRC1 , an activator of MT biogenesis , was prioritized by DEPICT analyses and then linked to central mtDNA regulators in GeneMANIA , we postulated that prioritized genes may be enriched for downstream targets of PGC-1A . PGC-1A induction resulted in a higher mean fold change among prioritized genes ( beta = 1 . 48; 95% CI , 0 . 60–2 . 37 ) as compared to any MT gene ( beta = 1 . 19; 95% CI , –0 . 76–3 . 13; t-test p=0 . 04 ) . Finally , we categorized the 27 MitoCarta3 genes into their respective pathways . Most ( 16; 57% ) genes were members of the ‘Mitochondrial central dogma’ pathway , which represents a nearly threefold enrichment as compared to the frequency of this pathway in the whole MitoCarta3 database ( null expectation = 20 . 7%; p=1 . 3 × 10–5 ) . Other implicated ( albeit not significantly enriched ) pathways included ‘Metabolism , ’ ‘Mitochondrial dynamics and surveillance , ’ ‘Oxidative phosphorylation , ’ and ‘Protein import , sorting and homeostasis’ ( Figure 3D ) . Four proteins were annotated as part of multiple pathways including TYMP/SCO2 , GTPBP3 , MIEF1 , and OXA1L ( Supplementary file 2 – Tab 7 ) . We performed an ExWAS in 147 , 740 UKBiobank participants with WES data to assess the contribution of rare coding variants . Among 18 , 890 genes tested , SAMHD1 was the only gene reaching exome-wide significance ( Figure 4A; Figure 4—source data 1 ) . The carrier prevalence of rare SAMHD1 mutations was 0 . 75% , and on average , mutation carriers had higher mtDNA-CN than noncarriers ( beta = 0 . 23 SDs; 95% CI , 0 . 18–0 . 29; p=2 . 6 × 10–19; Figure 4—figure supplement 1 ) . Also , while none of the 19 known mtDNA depletion genes reached Bonferroni significance , a suggestive association was found for TFAM ( beta = −0 . 33; 95% CI , –0 . 47 to –0 . 19; p=4 . 2 × 10–6 ) , and this association was independent of the common TFAM variants ( rs12247015; rs4397793 ) previously identified in the GWAS ( beta = −0 . 33; 95% CI , –0 . 47 to –0 . 19; p=8 × 10–6 ) . Rare variants in SAMHD1 and TFAM accounted for 0 . 06% of the variance in mtDNA-CN levels . Collectively , rare and common loci accounted for 1 . 55% . To evaluate whether rare SAMHD1 mutations also influenced disease risk , we conducted phenome-wide association testing of 771 diseases within the UKBiobank . At phenome-wide significance , SAMHD1 mutation carrier status was associated with approximately twofold increased risk of breast cancer ( OR = 1 . 91; 95% CI , 1 . 52–2 . 40; p=2 . 7 × 10–8 ) , as well as greater risk of ‘cancer ( suspected or other ) ’ ( OR = 1 . 52; 95% CI , 1 . 28–1 . 80; p=1 . 1 × 10–6; Figure 4B; Figure 4—source data 2 ) . Exclusion of breast cancer cases was attenuated but did not nullify the association with ‘cancer ( suspected or other ) ’ ( OR = 1 . 36; 95% CI , 1 . 10–1 . 67; p=0 . 004 ) suggesting that SAMHD1 mutations may also increase risk of other cancers , as has been shown for colon cancer ( Rentoft , 2019 ) . To understand whether differences in mtDNA-CN levels between SAMHD1 mutation carriers was a consequence of cancer diagnosis , we repeated association testing with mtDNA-CN excluding cancer patients . In this analysis , the association with mtDNA-CN levels was not attenuated ( beta = 0 . 26; 95% CI , 0 . 19–0 . 32; p=7 . 8 × 10–15 ) suggesting that the effect of rare SAMHD1 variants on mtDNA-CN levels is not driven by its relationship with cancer status . A summary of MT genes and pathways implicated by common and rare loci is provided in Figure 5 . Given that common variant loci overlapped with several mtDNA depletion genes , we postulated that polygenically low mtDNA-CN might cause a milder syndrome with phenotypically similar manifestations . To assess whether mtDNA-CN may represent a putative mediator of mtDNA depletion-related phenotypes , we conducted Mendelian randomization analyses between genetically determined mtDNA-CN and MT disease phenotypes using summary statistics derived from the FinnGen v4 GWAS dataset ( Supplementary file 2 – Tab 8 ) . After accounting for multiple testing of 10 phenotypes , an association between mtDNA-CN and all-cause dementia was found ( OR = 1 . 94 per 1 SD decrease in mtDNA-CN; 95% CI , 1 . 55–2 . 32; p=7 . 5 × 10–4; Figure 6; Figure 6—source data 1 ) . Sensitivity analyses indicated no evidence of global ( MR-PRESSO p=0 . 51; Q-statistic p=0 . 51 ) or directional ( Egger Intercept p=0 . 47 ) pleiotropy . The 27 selected variants accounted for 0 . 70% of the variance in mtDNA-CN and 0 . 13% of the risk for dementia , consistent with a causal effect of mtDNA-CN on dementia risk and not vice versa ( Steiger p=1 . 9 × 10–62 ) . Findings were robust across several different MR methods including the weighted median ( OR = 2 . 47; 95% CI , 1 . 93–3 . 00; p=0 . 001 ) , MR-EGGER ( OR = 2 . 41; 95% CI , 1 . 71–3 . 11; p=0 . 02 ) , and GSMR-HEIDI ( OR = 1 . 95; 95% CI , 1 . 34–2 . 84; p=0 . 001 ) methods . Results also remained statistically significant when using a broader set of genetic instruments including all genome-wide significant loci irrespective of whether genetic variants were located proximally to MitoCarta3 genes ( GSMR-HEIDI OR = 1 . 31; 95% CI , 1 . 02–1 . 68; p=0 . 04 ) . Furthermore , we replicated the dementia association using a second UKBiobank-independent GWAS dataset derived from the International Genomics of Alzheimer’s Disease Consortium ( 2013 ) including 17 , 008 Alzheimer’s disease patients ( OR = 1 . 41; 95% CI , 1 . 0001–1 . 98; p=0 . 04993 ) ( Lambert et al . , 2013 ) . Although commonly viewed as a simple surrogate marker for the number of mitochondria present within a sample , genetic analyses suggest that mtDNA-CN is a highly complex biomarker under substantial nuclear genetic regulation . mtDNA-CN reflects a mixture of MT processes mostly pertaining to mtDNA regulation . Accordingly , the true relationship between mtDNA-CN measured in blood samples with human disease remains to be completely defined though we find evidence for mtDNA-CN as a putative causal risk factor for dementia . Future studies are necessary to decipher if mtDNA-CN is directly involved in the pathogenesis of dementia and other diseases or whether other specific MT processes are truly causative . The Data-driven Expression-Prioritized Integration for Complex Traits ( DEPICT ) v . 1 . 1 tool was used to map mtDNA-CN loci to genes based on shared coregulation of gene expression using default settings ( Pers et al . , 2015 ) . Genome-wide significant variants from the European GWAS meta-analysis were ‘clumped’ into independent loci using PLINK ‘--clump-p1 5e-8 --clump-kb 500 --clump-r2 0 . 05’ with LD correlation matrix derived from 1000 Genomes Europeans ( Purcell et al . , 2007 ) . DEPICT was subsequently run on independent SNPs using default settings . DEPICT identified 91 genes in total at an FDR of 0 . 05 . Of the 91 genes , 4 noncoding genes were excluded from subsequent analyses for lack of a match in the GeneMANIA database ( Warde-Farley et al . , 2010 ) . The excluded genes include a pseudogene ( PTMAP3 ) , an intronic transcript ( ALMS1-IT1 ) , and two long noncoding RNAs ( SNHG15 , RP11-125K10 . 4 ) . The remaining 87 DEPICT-prioritized genes were uploaded to the GeneMANIA web platform ( https://genemania . org/ ) , which mines publicly available biological datasets to identify additional related genes based on functional associations ( genetic interactions , pathways , coexpression , colocalization , and protein domain homology ) . Based on the combined list of DEPICT and GeneMANIA identified genes , a network was formed in GeneMANIA maximizing the connectivity between all input genes , using the default setting , ‘Assigned based on query gene . ’ Functional enrichment analysis was then performed to identify overrepresented GO terms among all network genes ( Ashburner et al . , 2000 ) . All network genes with at least one GO annotation were compared to a background comprising all GeneMANIA genes with GO annotations . To complement the previous pathway analyses , we labeled prioritized genes with MitoCarta3 annotations and performed subsequent statistical enrichment analyses ( Rath et al . , 2021 ) . MitoCarta3 is an exquisite database of MT protein annotations , which draws from mass spectrophotometry and green fluorescent protein colocalization experiments of isolated mitochondria from 14 different tissues to assign all human genes statuses indicating whether the corresponding proteins are expressed in the mitochondria or not . We tested whether prioritized genes were enriched for the MT proteome by using a binomial test in R . The number of ‘trials’ was set to the total number of DEPICT and GeneMANIA-prioritized genes ( 107 ) ; the number of ‘successes’ was set to the aforementioned gene subset that was labeled as MT proteins by MitoCarta3 ( 27 ) ; finally , the expected probability was set to the number of nuclear-encoded MitoCarta3 genes divided by the total number of genes ( 1120/18 , 922 ) . Furthermore , a t-test was used to compare mean PGC-1A induced fold change for the subset of GWAS-prioritized genes expressed in the MT proteome ( 27 ) as compared to the mean PGC-1A induced fold change for all 1120 nuclear MitoCarta3-annotated genes . Also , genes were categorized based on MitoCarta3 ‘MitoPathway’ annotations . MitoCarta3 genes with missing values were excluded from this analysis . Lastly , the 27 genes were labeled based on MitoCarta3 ‘MitoPathways . ’ Only the top-level pathway ( i . e . parent node ) was ascribed to each gene .	Our cells are powered by small internal compartments known as mitochondria , which host several copies of their own ‘mitochondrial’ genome . Defects in these semi-autonomous structures are associated with a range of severe , and sometimes fatal conditions: easily checking the health of mitochondria through cheap , quick and non-invasive methods can therefore help to improve human health . Measuring the concentration of mitochondrial DNA molecules in our blood cells can help to estimate the number of mitochondrial genome copies per cell , which in turn act as a proxy for the health of the compartment . In fact , having lower or higher concentration of mitochondrial DNA molecules is associated with diseases such as cancer , stroke , or cardiac conditions . However , current approaches to assess this biomarker are time and resource-intensive; they also do not work well across people with different ancestries , who have slightly different versions of mitochondrial genomes . In response , Chong et al . developed a new method for estimating mitochondrial DNA concentration in blood samples . Called AutoMitoC , the automated pipeline is fast , easy to use , and can be used across ethnicities . Applying this method to nearly 400 , 000 individuals highlighted 71 genetic regions for which slight sequence differences were associated with changes in mitochondrial DNA concentration . Further investigation revealed that these regions contained genes that help to build , maintain , and organize mitochondrial DNA . In addition , the analyses yield preliminary evidence showing that lower concentration of mitochondrial DNA may be linked to a higher risk of dementia . Overall , the work by Chong et al . demonstrates that AutoMitoC can be used to investigate how mitochondria are linked to health and disease in populations across the world , potentially paving the way for new therapeutic approaches .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "medicine", "genetics", "and", "genomics" ]	2022	GWAS and ExWAS of blood mitochondrial DNA copy number identifies 71 loci and highlights a potential causal role in dementia
Hair greying is a hallmark of aging generally believed to be irreversible and linked to psychological stress . Here , we develop an approach to profile hair pigmentation patterns ( HPPs ) along individual human hair shafts , producing quantifiable physical timescales of rapid greying transitions . Using this method , we show white/grey hairs that naturally regain pigmentation across sex , ethnicities , ages , and body regions , thereby quantitatively defining the reversibility of greying in humans . Molecularly , grey hairs upregulate proteins related to energy metabolism , mitochondria , and antioxidant defenses . Combining HPP profiling and proteomics on single hairs , we also report hair greying and reversal that can occur in parallel with psychological stressors . To generalize these observations , we develop a computational simulation , which suggests a threshold-based mechanism for the temporary reversibility of greying . Overall , this new method to quantitatively map recent life history in HPPs provides an opportunity to longitudinally examine the influence of recent life exposures on human biology . This work was supported by the Wharton Fund and NIH grants GM119793 , MH119336 , and AG066828 ( MP ) . Hair greying is a ubiquitous , visible , and early feature of human biological aging ( O'Sullivan et al . , 2021; Tobin , 2011 ) . The time of onset of hair greying varies between individuals , as well as between individual hair follicles , based on genetic and other biobehavioral factors ( Akin Belli et al . , 2016; Bernard , 2012 ) . But most people experience depigmentation of a progressively large number of hair shafts ( HSs ) from their third decade onward , known as achromotrichia or canities ( Panhard et al . , 2012 ) . The color in pigmented HSs is provided by melanin granules , a mature form of melanosomes continuously supplied to the trichocytes of the growing hair shaft by melanocytes of the hair follicle pigmentary unit ( HFPU ) ( Tobin , 2011 ) . Age-related greying is thought to involve bulb and outer root sheath melanocyte stem cell ( MSC ) exhaustion ( Commo et al . , 2004; Nishimura et al . , 2005 ) , neuroendocrine alterations ( Paus , 2011 ) , and other factors , with oxidative damage to the HFPU likely being the dominant , initial driver ( Arck et al . , 2006; Paus , 2011; Trueb and Tobin , 2010 ) . While loss of pigmentation is the most visible change among greying hairs , depigmented hairs also differ in other ways from their pigmented counterparts ( Tobin , 2015 ) , including in their growth rates ( Nagl , 1995 ) , HF cycle , and other biophysical properties ( Van Neste and Tobin , 2004 ) . Hair growth is an energetically demanding process ( Flores et al . , 2017 ) relying on aerobic metabolism in the HF ( Williams et al . , 1993 ) , and melanosome maturation also involves the central organelle of energy metabolism , mitochondria ( Basrur et al . , 2003; Zhang et al . , 2019 ) . Moreover , mitochondria likely contribute to oxidative stress within the HF ( Lemasters et al . , 2017 ) , providing converging evidence that white hairs may exhibit specific alterations in mitochondrial energy metabolism . Although hair greying is generally considered a progressive and irreversible age-related process , with the exclusion of alopecia areata ( McBride and Bergfeld , 1990 ) , cases of drug- and mineral deficiency-induced depigmentation or repigmentation of hair have been reported ( Kavak et al . , 2005; Kobayashi et al . , 2014; Komagamine et al . , 2013; Reynolds et al . , 1989; Ricci et al . , 2016; Sieve , 1941; Yoon et al . , 2003 ) reflecting the influence of environmental inputs into HFPU function ( Paus et al . , 2014 ) . Because most hairs are continually growing from a living hair follicle , sensitive to changing physiological conditions , into a hardened hair shaft external to the body that retains stable molecular traces of these conditions , the hair shaft represents a bioarchive of recent exposures ( Kalliokoski et al . , 2019 ) . While spontaneous repigmentation can be pharmacologically induced , its natural occurrence in unmedicated individuals is rare and has only been reported in a few single-patient case studies ( Comaish , 1972; Navarini and Trüeb , 2010; O'Sullivan et al . , 2021; Tobin and Cargnello , 1993; Tobin and Paus , 2001 ) . The reversal of hair greying has not been quantitatively examined in a cohort of healthy adults , in parallel with molecular factors and psychosocial exposures . The influence of psychological stress on hair pigmentation is a debated , but poorly documented , aspect of hair greying . In humans , psychological stress accelerates biological aging as measured by telomere length ( Epel et al . , 2004; Puterman et al . , 2016 ) . In mice , psychological stress and the stress mediator norepinephrine acutely causes depigmentation ( Zhang et al . , 2020 ) . However , greying in both mice and humans has been shown to be a relatively irreversible phenomenon driven in part by a depletion of melanocyte stem cells , although some stem cells and transient amplifying cells do remain ( Trueb and Tobin , 2010 ) . In humans , recent evidence suggests hair growth and pigmentation changes in response to stress ( Peters et al . , 2017 ) , but this relationship , along with reversal of greying , remain insufficiently understood . The paucity of quantitative data in humans is mostly due to the lack of sensitive methods to precisely correlate stressful psychobiological processes with hair pigmentation and greying events at the single-follicle level . Here , we describe a digitization approach to map hair pigmentation patterns ( HPPs ) in single hairs undergoing greying and reversal transitions , examine proteomic features of depigmented white hairs , and illustrate the utility of the HPP approach to interrogate the association of life stress and hair greying in humans . Because previous literature suggests that rare repigmentation events are more likely to occur in the early stages of canities ( Van Neste and Tobin , 2004 ) , the current study focuses primarily on pigmentation events in young to middle-aged participants . Finally , we develop a computational model of hair greying to explore the potential mechanistic basis for stress-induced greying and reversibility on the human scalp hair population , which could potentially serve as a resource for the in silico modeling of macroscopic aging events in human tissues . The study was approved by New York State Psychiatric Institute ( NYSPI IRB Protocol #7748 ) . All participants provided written informed consent for their participation in this study and to the publications of data . Dark , white , and bi-color hairs were collected from healthy participants self-identified as ‘having some grey hairs’ or ‘two-colored hairs’ . Exclusion criteria included the use of dye , bleaching , or other chemical treatments on hairs . In addition , participants with hairs shorter than approximately 4 cm were excluded . Participants were recruited via local advertisement and using a snowball recruitment strategy . Some participants were staff of NYSPI and Columbia University Irving Medical Center , but no patients participated in the study . Eligible participants were asked to provide all two-colored hairs present on their scalps or other body regions . A total of 14 individuals ( seven females , seven males ) , mean age 35 ± 13 ( SD , range: 9–65 ) , were recruited . Hairs were plucked , manually or with standard flat tip tweezers , from the scalp or other body regions and archived for future imaging or molecular analyses . Only plucked hairs with follicular tissue attached ( excluding broken hairs ) were used in analyses to enable the interpretation of the direction of pigmentation – transition if the HS tip is dark and the root white , reversal if the HS tip is white and the root dark . Where possible , participants with two-colored hairs also provided fully dark or white hairs for comparison . While the participant age range does not capture the typical range of an aged population , it provides an opportune window to examine the beginning of the aging process as it corresponds to the typical age of onset for greying or canities ( Tobin , 2009 ) . The hair follicles also manifest stochastic hair-to-hair heterogeneity similar to that seen between individual cells in aging organs ( Bernard , 2012; van Deursen , 2014 ) . For this reason , likely as a result of stochastic processes similar to those that drive cellular heterogeneity in gene expression , some HFs reach the end of their pigmented life even in relatively young individuals . Although the course and process of aging in middle age may differ from later aging , rare regimentation events are more likely to occur in the early stages of canities ( O’Sullivan , 2020; Van Neste and Tobin , 2004 ) . This is in accordance with our mathematical model ( Figure 5 ) and with the isolated cases of repigmentation reported in the literature . Thus , to capture this phenomenon without additional confounds that could arise from systemic aging ( comorbidities , systemic inflammation , or other ) , we focused our investigation on this particular age window . As our model predicts , this age window also increases the probability that hairs are near their greying threshold , and as such have the possibility to undergo observable reversal . Whole hairs were first photographed using a Panasonic DC-FZ80 Digital Camera against a white background , with the hair tip and follicle systematically oriented . To facilitate visualization of the images of whole hairs in the figures ( photographic insets of whole hairs ) , the exposure , saturation , sharpness and light/dark tones of the photographs were enhanced . For microscopic imaging of hair follicles and HFPU , individual hair shafts and root-ends were imaged with an Olympus BX61 upright microscope ( Olympus BX61 Upright Wide Field Microscope , RRID:SCR_020343 ) equipped with a digitized stage ( ProScan; Prior Scientific ) , a 2 . 5x/0 . 075 air ( Zeiss , Germany ) , 10x/0 . 4 air ( Zeiss , Germany ) or 40x/1 . 3 oil ( Olympus , MA ) objectives , using DP71 camera ( Olympus , MA ) and MetaMorph software ( MetaMorph Microscopy Automation and Image Analysis Software , RRID:SCR_002368 ) ( Molecular Devices , CA ) version 7 . 7 . 7 . 0 . Images were scaled and analyzed in ImageJ ( ImageJ , RRID:SCR_003070 ) ( version 1 . 42q , NIH , http://rsb . info . nih . gov/ij ) . For microscopic imaging of hair shafts and videos of HPP transitions along the length of hair shafts , hairs were temporarily mounted with water on a glass slide ( 10x magnification , 15 ms exposure , 24-bit , ISO 1600 , 4080 × 3072 digitizer ) . To generate high-resolution HPPs , HSs were digitized as high-resolution 8-bit Greyscale images ( 3200dpi , default adjustments , Epson Perfection V800 Photo Scanner ) , and the scanned images were processed using Image J ( Fiji , RRID:SCR_002285 ) . To capture both the white and dark sections of each hair , hairs were immobilized onto a smooth surface by taping the plucked hair follicle ( proximal to the epidermis ) , straightening the entire length as much as possible without placing too much force on the hair , and immobilizing the tip ( distal to epidermis ) with adhesive tape . HSs were dry and were checked for potential knots or twists caused by handling . Any dust was removed from the hair surface using a kimwipe before being placed on the scanner . Areas of each hair between the immobilized ends were used for analyses . To extract hair darkness at each point along the length , pixel luminosity at each position was estimated as the darkest value across a sliding one-dimensional pixels array perpendicular to the shaft axis , where the hair itself represented the darkest area , and HPP graphs were generated by plotting the intensity in arbitrary units ( A . U . ) by distance ( cm ) . Intensity ranged from 0 to 255 A . U . , with 0 being white and 255 being black . The data was then de-noised using a 100-pixels rolling average , and the resulting HPP was imported into Prism 8 ( GraphPad Prism , RRID:SCR_002798 ) for visualization . To compare intensity profiles across multiple hairs , we transformed numerical intensity values by normalizing to the average intensity of each hair . A total of 100 randomly selected dark hairs were manually plucked from one female and one male individual , including 25 hairs per head region ( left and right temporal , top , and crown ) . Digitized hairs for each individual were graphed as a heatmap , grouped by head region . To examine the hypothesis that hairs exhibit regional variation in HPPs , the intensity of all 25 hairs per region were then averaged to create an 'average' hair from each region . A plot was then made for the four 'average hairs' , one from each head region ( Figure 5—figure supplement 1 ) . Dark and white scalp hairs were plucked from two healthy individuals: a 38-year-old African-American male and a 33-year-old Caucasian male . The African-American hairs were curly and black , while the Caucasian hairs were straight and auburn . Hairs were fixed in a 2% glutaraldehyde solution in 0 . 1 M cacodylate ( TAAB Lab Equipment ) buffer , pH 7 . 4 as described previously ( Picard et al . , 2013 ) . Briefly , plucked hair shafts were cut to 2–3 cm in length , immersed in fixative , and incubated at room temperature for 2 weeks . HS were then post-fixed and dehydrated in ethanol , cut into smaller segments of 0 . 5 cm , and embedded in longitudinal orientation in 100% resin . Orientation and section quality were confirmed with 1 μm-thick sections , and ultrathin sections of 70 nm were cut using a diamond knife on a Leica EM UC7 ultramicrotome ( Leica EM UC7 ultramicrotome , RRID:SCR_016694 ) . Sections mounted on Pioloform filmed copper grids prior to staining with 2% aqueous uranyl acetate and lead citrate ( Leica ) . Ultrathin sections were examined on a Phillips CM 100 Compustage ( FEI ) transmission electron microscope and digital micrographs were captured by an AMT CCD camera . Matched dark and white hairs from the donors were imaged , and three different segments along each hair were analyzed . All images used for analysis were captured at ×7500 magnification , with a pixel size of 0 . 00902 μm/pixel . Images were imported into ImageJ for analysis and all melanin granules contained within a given picture were manually traced ( Intuos tablet ) . In each photograph , the intensity of the melanin granules , cortex , and background ( outside the hair ) were quantified . Cortex and melanin granule intensity were normalized by subtracting the background average intensity ( measured from three different standard regions of interest – ROIs ) to ensure comparability of various micrographs , hair segments , and between dark and white hairs . The intensity of the cortex was also quantified from eight different ROIs devoid of melanin granules . To compute melanin granule size , we obtained the surface area of each manually traced granule . To compute melanin granule density per hair region , the total cortex area in each scaled micrograph was recorded and was divided by the total number of granules that were found in that image , yielding the number of granules/μm2 , which was then multiplied by 100 . The protocol in both experiments 1 and 2 for hair digestion were adapted from a previous protocol establishing that SDS-based protein extraction methods result in higher protein yield than urea-based digestion ( Adav et al . , 2018 ) . This method was adapted with the addition of an initial mechanical homogenization step to extract proteins from minimal amounts of hair tissue ( 1–2 cm ) , which was necessary to analyze multiple hair segments along the same HS . After incubation , the hair was reduced with DTT and then alkylated with iodoacetamide ( IAA ) , as per previous methods ( Goecker et al . , 2020 ) . A subset of participants with noteworthy patterns of single-hair greying and reversal were asked to complete a retrospective stress assessment ( Figure 3—figure supplement 1 ) , completed 1–4 months after hair collection in two individuals ( one male , one female ) . The life event calendar ( LEC ) methodology increases the reliability and validity of recall in retrospective assessments ( Belli , 1998 ) . The use of timeline results in higher accuracy and lowers underreporting as compared to traditional questionnaires ( van der Vaart , 2004 ) . The retrospective psychosocial stress assessment in the present study is an adaptation of LECs , more similar to timelines , which measures one behavioral construct – here ‘stress’ – during a short reference period ( Glasner and van der Vaart , 2009; Sobell et al . , 1988 ) . In our instrument , participants first position landmark events in time ( in this case , the most stressful event or period , and the least stressful ) , and then link other events to these landmark events , a method referred to as sequencing ( Belli , 1998 ) . Additionally , the visual calendar ( see Figure 3—figure supplement 1 ) encourages top-down and parallel retrieval of memories ( Glasner and van der Vaart , 2009 ) , which also contributes to overall accuracy . In the retrospective assessment , participants are first asked to identify the most stressful event or period over the last 12 months and to position it in time along the physical timeline , using their electronic calendar and objective dates , and assign it ‘10’ on the graph . This first positioned event acts as a landmark event from which the other events can then be sequentially linked ( Belli , 1998 ) . Participants then identified the least stressful event or period and assigned it ‘0’ on the physical timeline , acting as another landmark event . Participants then indicated 2–6 additional particularly stressful events or periods , assigned them scores ranging from most stressful to least stressful ( 10 and 0 , respectively ) , marked them on the timeline , and connected these events with a line that best illustrates their stress levels over the past year . The instrument not only asks the participant to mark their stress levels , but also to briefly name/describe each event , which can help with recall of the exact stressor and its intensity , and also allows participants to match up an event with an exact calendar date , enhancing the timing accuracy of events . Stress graphs were then digitized by aligning the retrospective assessment to a grid printed on transparency film with 0 . 25 unit resolution ( number of possible values = 40 units total , from 0 to 10 ) , and the resulting digital values were plotted ( Prism 8 ) . To align stress profiles with HPP , digitized stress profiles were aligned with hairs from the same participant using dates of collection and assuming a hair growth rate of 1 cm/month ( LeBeau et al . , 2011 ) . Each hair segment can then be mapped to specific weeks or month along the stress profile . Dark and white hairs ( n = 10 per person per color ) were collected from the same two individuals whose hairs were analyzed by electron microscopy ( African American male , Caucasian male ) . The follicle and proximal portion ( 2 cm segment ) of the hair shaft were sectioned and separately lysed in 200 μL of lysis buffer containing 500 mM Tris , 1% Tween 20 , 20 μg/μl Proteinase K incubated for 10 hr at 55°C , followed by 10 min at 95°C as described previously ( Picard et al . , 2012 ) . Hair follicles were fully digested whereas the more robust proteinaceous hair shafts were only partially digested , such that the quantified mtDNA abundance is likely an underestimation of the total DNA amount per unit of hair shaft . In addition , nuclear DNA is rapidly degraded by endonucleases and virtually absent in the hair shaft ( Fischer et al . , 2011 ) . We therefore focus our analysis of genomic material in the hair shaft to mtDNA . The number of mtDNA copies per nucleated cell ( mtDNA copy number , mtDNAcn ) was measured by real-time quantitative polymerase chain reaction ( qPCR ) using a duplex Taqman reaction to amplify both mitochondrial ( ND1 ) and nuclear DNA ( B2M , single-copy gene ) amplicons . The primer sequences are: ( ND1-Fwd: GAGCGATGGTGAGAGCTAAGGT , ND1-Rev: CCCTAAAACCCGCCACATCT , Probe: HEX-CCATCACCCTCTACATCACCGCCC-3IABkFQ . B2M-Fwd: CCAGCAGAGAATGGAAAGTCAA , B2M-Rev: TCTCTCTCCATTCTTCAGTAAGTCAACT , Probe:FAMATGTGTCTGGGTTTCATCCATCCGACA-3IABkFQ ) obtained from IDTdna . com . qPCR was performed on QuantStudio 7 Flex Real-Time PCR System ( Life Technologies QuantStudio 7 Real Time PCR System , RRID:SCR_020245 ) ( Applied Biosystems , Foster City , CA ) . Cycling conditions were as follows; 1 cycle of 50°C for 2 min , 95°C for 20 s , followed by 40 cycles of 95°C for 1 s , 60°C 20 s . For plucked hair follicles , all ND1 and B2M Cts were in the dynamic range of the assay and used to compute mtDNAcn from the ∆Ct . All measures were performed in triplicates and the average Ct values taken for each sample . The mean C . V . for ND1 was 0 . 67% in both shafts and follicles , and for B2M 0 . 52% in follicles . mtDNAcn was calculated as 2∆Ct ( ND1 Ct - B2M Ct ) , and multiplied by two to account for the diploid nature of the nuclear genome . To simulate hair greying across the lifespan , a linear mixed effect model with random intercept and slopes to account for the stochastic process of hair greying was implemented in R ( R Project for Statistical Computing , RRID:SCR_001905 ) ( R Development Core Team , 2010 ) . This interactive implementation is available at https://timrain . shinyapps . io/hair ( Shiny , RRID:SCR_001626 ) . We first hypothesized a potential mechanism in which individual hairs are affected by a summation of effects from a random aging factor accumulating over time , random stress factor and random initial greying loading , thus creating variation between hairs within an individual . Once the hair has passed a prespecified threshold , the hair transitions to grey ( Figure 5B ) . This model includes 17 parameters listed in Supplementary file 4 , each of which can be adjusted to simulate various effects on individual hairs in relation to the aging process , including one or two stress exposure periods with customizable intensity and duration . Scaling this model to hair populations with thousands of hairs , the simulation reports trajectories of greying for individual hairs , as well as a graph with the population distribution of white hairs ( shown as frequency distributions ) for a theoretical scalp . First , we simulated the average greying trajectory based on data indicating that the average age of onset for greying is age 35 and that white hairs reach a 40% population frequency at age 65 ( Panhard et al . , 2012 ) . This established a set of default parameters that yielded the greying trajectory shown in Figure 5C . We then simulated two hypothetical scenarios reflecting the total hair population for individuals who accumulate grey hairs at different rates , termed early and late greyers . These variable greying patterns were found to be generated by changing only one parameter , Sigma1 , the standard deviation ( across HFs ) of the rate at which the aging factor increases over time . Additionally , the model also simulated greying reversal , beginning with the parameters of the average greyer and then including also the stress parameters . To show the effect of stress on hair greying , we simulated two stressful periods starting at age 20 and then again at age 50 , with equal intensity and duration . At age 20 the aging factor increases due to the stress but does not induce grey hair as the aging factor is still below the threshold ( Figure 5E ) . On the other hand , at age 50 the same intensity and duration of the stressor will tend to induce additional greying as the aging factor for some hairs increases past the threshold , and then upon the end of the stressor , the aging factor could decrease past the threshold and thus the hair would undergo reversal ( i . e . repigmentation ) to its original color ( Figure 5F ) . An alternative model was considered to explore potential mechanisms for hair transitioning and reversal in response to stress . Specifically , we considered a mechanism in which the rate of accumulation in the aging factor increases during a period of stress ( as opposed to our final model where stress causes a stepwise increase in aging factor ) and then returns to the original rate following the end of the stressor . In this scenario , the threshold remains constant . This mechanism can be rejected because although it adequately simulates hair greying , once a hair has crossed above the threshold , if the stressor only affects the slope , it is not possible for a hair to return below threshold and undergo reversal ( Figure 5—figure supplement 2 ) . To simulate the graying process for a hypothetical person based on our hypothesized mechanism , we posited a linear mixed model for the i th ( i=1 , … , n ) hair with two fixed effects ( β1 aging factor rate and β2 stress sensitive rate ) and three random effects ( bi0 for graying loading at age 0 , bi1 for aging factor rate and bi2 for stress sensitive rate ) . To ensure positivity in age and accumulating stress , the model involves only the absolute value of each random effect . GreyinLoading{i , age}=\|bi0\|+ ( \|bi1\|+β1 ) age+ ( \|bi2\|+β2 ) AccumulatingStress{age}+e{i , age}where AccumulatingStress is defined as:AccumulatingStressage=∑a=age−WindowWidthagestressa The three random effects follow a multivariate normal:bi0 , bi1 , bi2∼N0 , Gwith covariance structure:G=σ02ρ01σ0σ1ρ02σ0σ2ρ01σ1σ0σ12ρ12σ1σ2ρ02σ2σ0ρ12σ2σ1σ22 All the correlations ρ01 , ρ02 , ρ12 in the simulation are constrained to be positive . When the aging factor of hair i reaches a predefined threshold , the i th hair will turn white . The source code is available at https://github . com/junting-ren/hair_simulation ( Ren , 2021; copy archived at swh:1:rev:3a19705969bfca7edc98651c1dd973ca7ae3b23d , RRID:SCR_002630 ) . An ordinary one-way ANOVA with Tukey’s multiple comparison test was used to compare the number of melanin granules per μm2 , granule size , granule intensity , and relative intensity of the cortex in the dark and white hairs , and to compare the pigmentation intensity across head regions . To compare the rate of change in pigmentation per day between greying and reversal hairs , points of transitions visually estimated were used to derive a slope for each greying or reversal segment , which were compared using an unpaired t test . A Mann-Whitney test was used to compare mtDNA levels in dark and white hair shafts and mtDNA copy number in dark and white hair follicles . For univariate and multivariate analyses of proteomic signatures , protein abundance levels were processed in R using the Metaboanalyst 3 . 0 platform ( MetaboAnalyst , RRID:SCR_015539 ) ( Chong et al . , 2019 ) as unpaired data . The data was mean-centered and log transformed prior to statistical analyses and missing ( low abundance ) values were imputed by half of the lowest value for the group ( dark , white ) . Significance was established at an FDR level 0 . 05 and fold changes calculated using ANOVA . Partial least square discriminant analysis ( PLS-DA ) was used to extract meaningful features that distinguish dark and white hairs and visualize groups of hairs or segments along the same hair . Two different strategies were used to generate protein lists subsequently queried for their functional significance: ( i ) For dark vs white comparisons where the whole model is meaningful , the variable importance in projection ( VIP ) scores for each protein were extracted and used to select the top 40 most influential proteins ( Figure 1—figure supplement 4C ) ; ( ii ) For analyses of segments along the hair with greying followed by reversal , the factor loadings for each protein were extracted separately for components 1 and 2 , and the top 20 positive and 20 negative proteins were selected for further analysis . Protein lists derived from both strategies were then used for functional enrichment analysis in ShinyGO and STRING as described above . For the data displayed in Figure 3 , we measured the strength of the temporal relationship between time series on the same scale ( e . g . intensity measures of two hairs ) by calculating the mean squared differences . The overall strength of temporal relationship among multiple time series is measured by the sum of all pairwise mean squared differences . To measure the strength of the temporal relationship between time series on different scales ( specifically , intensity of color for a hair and rated level for stress levels ) we calculated Pearson’s correlation . To provide a reference distribution for comparison , we conducted 1000 random permutations of the data in each instance . For Figure 3A and B , each permutation involved simulating an equivalent number of hairs that transition ( three hairs in 3A; two hairs in 3B ) . Each simulated hair includes a randomly selected transition placed at a random time point , with resampled noise before and after the transition . Two results are reported in each case . For the first , the library of transitions was taken from the transition segments of each observed hair , regardless of which direction ( dark to white or white to dark ) that transition was in . The probability of each direction of transition was determined by the overall rate of each transition direction from the observed data . The second reported result is based on a similar analysis , but the library of transitions included only the hairs that underwent the same directional change ( dark-to-white in 3A; white-to-dark in 3B ) . The noise for uses of resampling was taken from hairs from the same subject after subtracting a smooth function , and resampling was done in segments of length-100 increments to maintain proper temporal correlation patterns . For Figure 3D , under the null hypothesis of no relationship between the hair intensity and stress pattern , each permutation involved choosing a random time point , splitting the stress pattern at that point and rejoining it by concatenating the two segments in the alternative order . To overcome the lack of methodology to map pigmentary states and age-related greying transitions , we developed an approach to digitize HPPs at high resolution across the length of single human HSs . Combined with known hair growth rates on the scalp ( ~1 . 0–1 . 3 cm per month LeBeau et al . , 2011 ) , this approach provides a quantifiable , personalized live bioarchive with the necessary spatio-temporal resolution to map individualized HPPs and greying events along single hairs , and to link HPPs to specific moments in time with unprecedented accuracy . Using this methodology , similar to dendrochronology where tree rings represent elapsed years ( Douglass , 1928 ) , hair length reflects time , and the HS length is viewed as a physical time scale whose proximal region has been most recently produced by the HF , and where the distal hair tip represents weeks to years in the past , depending on the HS length . To examine HPPs in human hairs , we plucked , imaged , digitized , and analyzed hairs ( n = 397 ) from 14 healthy donors ( Figure 1A ) ( see Materials and methods for details ) . Three main pigmentation patterns initially emerged from this analysis: ( i ) Hairs with constant high optical density ( Dark ) , ( ii ) Hairs with constant low optical density ( White ) ; ( iii ) Initially dark hairs that undergo a sharp greying transition from dark to white over the course of a single growing anagen phase of the hair follicle growth cycle ( Transition ) ( Figure 1B–C ) . Dark-to-white transitions demonstrate the existence of rapid depigmentation events within a single anagen hair cycle ( Paus and Cotsarelis , 1999; Slominski et al . , 2005 ) . We confirmed that compared to dark hairs still harboring their ‘young’ pigmentary state , the HFPU of ‘aged’ white HFs from either African American or Caucasian individuals are practically devoid of pigment ( Figure 1D ) , which is consistent with the finding of previous studies ( Cho et al . , 2014 ) . Whereas dark hairs contain melanin granules dispersed throughout the hair cortex when observed by electron microscopy , white hairs from the same individuals show a near complete ( >98% ) absence of melanin , with the few retained melanin granules , when present , being smaller , less dense , and at times vacuolated , a potential response to oxidative stress ( Tobin , 2009; Figure 1E–I , see Figure 1—figure supplement 5 for high-resolution images of mature melanosomes ) . The digitization of HPPs thus reflects the presence of melanosomes within the HS , and rapid greying events are marked by the loss of melanosomes . To gain molecular insight into the greying process , we performed a comprehensive proteomic analysis comparing dark and white HS . Recent work suggests that depigmentation is associated with the upregulation of lipid synthesis enzymes in the HS ( Franklin et al . , 2020 ) . Moreover , in depigmented hairs , the abnormal diameter/caliber of the hair fiber , growth rate , presence/absence of HS medulla as well as the ( dis ) continuity and diameter of the medulla along the hair length ( Van Neste , 2004 ) imply multiple potential proteomic alterations associated with depigmentation . In addition , melanogenesis involves high levels of reactive oxygen species , but dark HFs are equipped with multiple antioxidant mechanisms ( e . g . [Tobin , 2009] ) . Thus , the proteomic features of HSs may provide interpretable information about molecular changes associated HF greying . Protein extraction and LC-MS/MS methods optimized from a previous protocol ( Adav et al . , 2018 ) were used to process the unusually resistant proteinaceous matrix of the hair shaft and to handle the overly abundant keratin proteins over other potential proteins of interest ( see Materials and methods for details ) . Two independent experiments were performed . Experiment 1: matched dark and white hairs collected at the same time from two closely age- and diet-matched individuals ( one female and one male , both 35 years old , each dark and white HS measured twice , total n = 8 ) ; and Experiment 2 ( validation ) : n = 17 hair segments from seven different individuals ( four females and three males ) . In the first experiment , we were able to extract and quantify 323 proteins ( >75% of samples ) from single 2-cm-long HS segments . Compared to dark HS collected at the same time from the same individuals , white hairs contained several differentially enriched ( upregulated ) or depleted ( downregulated ) proteins ( Figure 1J–K see Supplementary file 1 for complete list ) on which we performed GO ( Gene Ontology ) and KEGG ( Kyoto Encyclopedia of Genes and Genomes ) enrichment analysis and explored their protein-protein interaction networks ( Figure 1—figure supplement 2 ) . The protein networks for both downregulated ( <0 . 8 fold , n = 23 ) and upregulated ( >1 . 5 fold , n = 67 ) proteins contain significantly more interactions than expected by chance ( p<0 . 00001 , observed vs expected protein-protein interactions ) . Thus , coherent groups of functionally related proteins are differentially expressed in white hairs , from which two main patterns emerged . The first main pattern relates to protein biosynthesis and energy metabolism . A large fraction ( 34 . 3% ) of upregulated proteins in white hairs was related to ribosome function , protein processing , and associated cytoskeletal proteins . Upregulation of the machinery responsible for protein synthesis and amino acid metabolism included the ribosomal protein RPS15A , which is known to localize to mitochondria . Of all upregulated proteins in white hairs , 26 . 8% were known mitochondrial proteins , based on MitoCarta2 . 0 and others sources ( Calvo et al . , 2016 ) . These proteins are involved in various aspects of energy metabolism , including substrate transport ( carnitine palmitoyltransferase 1A , CPT1A; malonate dehydrogenase 1 , MDH1 ) , respiratory chain function ( Complex III subunit 1 , UQCRC1 ) , and catecholamine homeostasis ( Catechol-O-Methyltransferase , COMT ) . White hairs also contained more proteins involved in glucose ( glucose 6-phosphate dehydrogenase , G6PD; phosphoglycerate kinase 1 , PGK1 ) and lipid metabolism located in either the mitochondria or cytoplasm ( fatty acid synthase , FASN; acyl-CoA thioesterase 7 , ACOT7; mitochondrial trifunctional enzyme subunit beta , HADHB ) or in peroxisomes ( acyl-CoA acyltransferase 1 , ACAA1 ) . The metabolic remodeling in white hairs is consistent with the established role of mitochondria and metabolic regulation of hair growth and maintenance in animal models ( Flores et al . , 2017; Kloepper et al . , 2015; Singh et al . , 2018; Vidali et al . , 2014 ) , and possibly consistent with hair anomalies reported in human patients with mitochondrial disease ( Silengo et al . , 2003 ) . The upregulation of energy metabolism may subserve the likely increased energy demands in depigmented hairs . However , our data and those of others ( Franklin et al . , 2020 ) implicate the upregulation of specific mitochondrial proteins involved , not necessarily in global energy metabolism , but in specific metabolic activities such as amino acid and lipid biosynthesis . A second less robust pattern relates more directly to melanosome biology . In line with the lysosomal origin of melanosomes that are largely absent in depigmented HS ( Slominski et al . , 2005 ) , several lysosomal proteins ( PLD3 , CTSD , HEXB , and LAMP1 ) were downregulated in white hairs , consistent with previous literature ( Franklin et al . , 2020 ) . White hair shafts also showed a depletion of six main keratins ( see Figure 1—figure supplement 2 ) , likely because greying can affect the nature of keratinocytes proliferation ( Van Neste and Tobin , 2004 ) , of proteins associated with exocytosis , such as ITIH4 and APOH ( potentially involved in the secretion of melanosomes from melanocytes to keratinocytes ) , as well as proteins associated with mitochondrial calcium transmembrane transport . Interestingly , calcium exchange between mitochondria and the melanosome is required for melanin pigment production in melanocytes ( Zhang et al . , 2019 ) , and calcium signaling promotes melanin transfer between melanosomes and keratinocytes ( Singh et al . , 2017 ) . Finally , canities-affected white HFs also showed upregulation of antioxidant proteins , specifically those localized to mitochondria and cytoplasm ( superoxide dismutase 1 , SOD1; peroxiredoxin 2 , PRDX2 ) , in line with the role of oxidative stress in HS depigmentation ( Arck et al . , 2006; Wood et al . , 2009 ) . Alterations among these individual metabolic and mitochondrial enzymes were further reflected in KEGG pathways functional enrichment analyses indicating a significant enrichment of metabolic pathways including carbon metabolism and fatty acid synthesis , amino acids , and oxidative phosphorylation ( see below ) . To independently validate these results , we extended this analysis to white and dark HS from six individuals ( three males , three females , range: 24–39 years ) analyzed on a separate proteomic platform and in a different laboratory . In this experiment , a total of 192 proteins ( ≥3 samples ) were quantified from 1cm-long HS segments . This dataset showed a similar trend as the first analysis toward a preferential overexpression of proteins with greying ( 55% upregulated vs 29% downregulated in white HS ) ( Figure 1L–M , see Supplementary file 2 for a complete list ) . The most highly upregulated proteins included mitochondrial components such as the voltage-dependent anion channel 1 ( VDAC1 ) , a subunit of ATP synthase ( ATP5A1 ) , and a regulator of mitochondrial respiratory chain assembly ( Prohibitin-2 , PHB2 ) . Again , the antioxidant enzyme SOD1 was enriched in white relative to dark HSs . To examine the possibility that these relative upregulations are driven by a global downregulation of highly abundant housekeeping proteins , we analyzed the intensity-based absolute quantification ( iBAQ ) data for each sample . This confirmed that the housekeeping proteins , including keratins and keratin-associated proteins , were not downregulated in white hairs , but generally unchanged or slightly upregulated . Moreover , as a whole , upregulated proteins formed a coherent protein-protein interactions cluster ( p<0 . 00001 ) and pathway analysis again showed a significant enrichment of carbon metabolism , glycolysis/glucogenesis , pyruvate metabolism , and amino acid synthesis pathways in white relative to dark HS ( Figure 1—figure supplement 3 , Figure 4E ) . As in experiment 1 , these upregulated pathways all indicate metabolic remodeling in white hair follicles . On the other hand , proteins downregulated in white HSs were related to cholesterol metabolism , complement-coagulation cascades , and secretory processes shared with immune/inflammatory pathways ( Figure 4E ) . The downregulation of secretory pathways is again consistent with reduced transfer of pigmented melanosomes from the melanocytes to the keratinocytes . To verify the robustness of these findings using an alternative analytical approach , we built a simple partial least square discriminant analysis ( PLS-DA ) multivariate model , which provided adequate separation of white vs dark HS ( Figure 1—figure supplement 4 ) . Simple interrogation of this model to extract the features ( i . e . proteins ) that maximally contribute to group separation yielded a set of proteins enriched for estrogen signaling pathways , complement and coagulation cascades , as well as metabolic pathways including NAD+/NADH , cholesterol , pyruvate , and carbon metabolism , similar to results above . Interestingly , we also identified 13 proteins that were undetectable in any of the dark HS ( either not expressed , or below the detection limit ) , but consistently present in white HS ( Supplementary file 3 ) . These proteins are either newly induced or experience a substantial upregulation with greying ( fold change tending toward infinity ) . A separate functional enrichment analysis for these induced proteins also showed significant enrichment for the same aging-related metabolic pathways as for the upregulated protein list: glycolysis/glucogenesis , carbon , pyruvate , and cysteine and methionine metabolism ( all p<0 . 001 ) . These converging proteomics data , which are consistent with previous findings ( Franklin et al . , 2020 ) , support a multifactorial process directly implicating metabolic changes in human hair greying ( Paus , 2011 ) . Moreover , given that metabolic pathways are rapidly and extensively remodeled by environmental and neuroendocrine factors – that is , they naturally exhibit plasticity – these data build upon previous proteomic evidence to show that human hair greying could be , at least temporarily , reversible . Our analysis of HPPs in healthy unmedicated individuals revealed several occasions whereby white hairs naturally reverted to their former dark pigmented state . This phenomenon was previously reported only in a handful of case reports , with only a single two-colored HS in each case ( O’Sullivan , 2020 ) . Here , we document the reversal of greying along the same HS in both female and male individuals , ranging from a prepubescent child to adults ( age range 9–39 years ) , and across individuals of different ethnic backgrounds ( 1 Hispanic , 8 Caucasian , 1 Asian ) . This phenomenon was observed across frontal , temporal , and parietal regions of the scalp ( Figure 2A ) , as well as across other corporeal regions , including pubic ( Figure 2B ) and beard hairs ( Figure 2C ) . The existence of white HS undergoing repigmentation across ages , sexes , ethnicity , and corporeal regions documents the reversibility of hair greying as a general phenomenon not limited to scalp hairs . Nevertheless , we note that this phenomenon is limited to rare , isolated hair follicles . As their occurrence will probably go unnoticed in most cases , it is difficult to assess the true incidence of these repigmentation events . Over an active recruitment period of 2 . 5 years , we were only able to identify 14 participants , indicative of the rarity of this phenomenon . Moreover , more complex HPPs with double transitions and reversions in the same HS were observed in both directions: HS undergoing greying followed by rapid reversal ( Figure 2D ) , and repigmentation rapidly followed by greying ( Figure 2E ) . Importantly , both patterns must have occurred over the course of a single anagen ( growth ) phase in the hair growth cycle , implicating cellular mechanisms within the HFPU . Greatly extending previous case studies of these rare hair repigmentation events , the current study provides the first quantitative account of the natural and transitory reversibility of hair greying in humans . We understand the emergence of a reverted HS – that is , a HS with a white distal segment but with a dark proximal end – as necessarily having undergone repigmentation to its original pigmented state following a period of time as a depigmented ‘old’ white hair ( Figure 2F ) . In double transition HS with three segments , repigmentation must take place within weeks to months after greying has occurred , producing three distinct segments present on the same hair strand ( Figure 2G ) . Microscopic imaging along the length of a single HS undergoing a double transition ( greying followed by rapid reversal ) can be visualized in Video 1 , illustrating the dynamic loss and return of pigmented melanosomes within the same HS . A proposed mechanism for such repigmentation events involve the activation and differentiation of a subpopulation of immature melanocytes located in a reservoir outside of the hair follicle bulb in the upper outer root sheath ( Van Neste and Tobin , 2004 ) , or more likely from transient amplifying melanoblast cells that migrate along the outer root sheath to the interfollicular epidermis ( Birlea et al . , 2017 ) . Our hair digitization approach also provides the first estimates of the rates of change in pigmentation for HS covering a broad range of initial colors and darkness ( Figure 2H ) . Across individuals , assuming a scalp hair growth rate of 1 cm/month ( LeBeau et al . , 2011 ) , the rates of depigmentation in greying hairs ranged widely from 0 . 3 to 23 . 5 units of hair optical density ( scale of 0–255 units ) per day , corresponding to between 0 . 2% and 14 . 4% loss of hair color per day ( Figure 2I ) . The rate at which HS regain pigmentation during reversal was 0 . 1–42 . 5 units per day , which is similar ( ~30% faster on average ) to the rate of greying ( Cohen’s d = 0 . 15 , p=0 . 59 ) ( Figure 2J ) . Given these rates , the fastest measured transitioning hairs grey and undergo full reversal in ~3–7 days ( median: ~3 months ) . Thus , rather than drifting back toward the original color , repigmentation of white human HS occurs within the same time frame and at least as rapidly as the process of greying itself . The spectrum of greying transitions and reversals patterns observed in our cohort , including measured rates of repigmentation along individual hairs , is shown in Figure 2—figure supplement 1 . The HPP results establish the wide range of naturally occurring rates of pigmentary changes in single hairs , which vary by up to an order of magnitude from hair to hair . These data also suggest that reversal/repigmentation may reflect the action of as yet unknown local or systemic factors acting on the HFPU within a time frame of days to weeks . We then asked whether the reversal of greying is governed by a process that is unique to each human scalp HF or if it is likely coordinated by systemic factors that simultaneously affect multiple HFs . Participants’ scalps were visually inspected to identify two-colored hairs , including both greying transitions and reversal . In our combined cohort , three individuals had multiple two-colored hairs collected at either one or two collection times within a one-month interval . In each case , the multiple two-colored hairs originated from independent HFs separated by at least several centimeters ( e . g . left vs right temporal , or frontal vs temporal ) . If the hairs are independent from each other , the null hypothesis is that different HSs will exhibit either greying or reversal changes at different positions along each hair , and will have independent HPPs . If multiple HSs were coordinated by some systemic factor , then we expect HPPs to exhibit similarities . In a first 35-year-old female participant with dark brown hair , three two-colored hairs were identified at a single instance of collection . Notably , all three hairs exhibited dark-to-white greying . Moreover , when the HPPs of the three hairs were quantified and plotted , the HPPs followed strikingly similar greying trajectories ( r = 0 . 76–0 . 82 ) marked by a similar time of onset of greying , similar HPP intermittent fluctuations ( note the rise ~10 cm ) , and a similar time point where all hairs become fully depigmented ( ~15 cm ) ( Figure 3A ) . A permutation test on the similarity of the color trajectories yielded p=0 . 032 , suggesting possible synchrony between different HSs . If our simulation considers only hairs that transition in one direction ( from dark to white ) this gives p=0 . 086 ( see Materials and methods for details ) . In a 37-year-old female participant with brown hair , two transition hairs were identified . The HPPs for both hairs revealed strikingly similar trajectories ( r = 0 . 80 ) , in this case undergoing spontaneous reversal in a near-synchronous manner upon alignment ( p<0 . 001 when considering hairs transitioning in either direction , and similarly , p<0 . 001 considering only hairs transitioning from white to dark , Figure 3B ) . Thus , these findings extend previous reports in single isolated hairs by providing quantitative accounts of coordinated HS ( re ) pigmentation across multiple hairs . Candidate humoral hair pigmentation modulators that could create synchrony in greying or repigmenting hairs include neuropeptides , redox balance , and steroid or catecholamine hormones ( Hardman et al . , 2015; Paus et al . , 2014; Tobin and Kauser , 2005; Zhang et al . , 2020 ) that can directly regulate the human HFPU ( Paus , 2011 ) , impact intrafollicular clock activity ( Hardman et al . , 2015 ) , or regulate the expression of other melanotropic neurohormones in the human HFPU such as α-MSH , ß-endorphin , and TRH ( Gáspár et al . , 2011 ) . These factors must act in parallel with genetic factors that influence inter-individual differences in aging trajectories . In light of these results , we next applied our HPP method to examine the possibility that psychological stress is associated with hair greying/reversal in humans . Anecdotal case reports and a recent pilot study suggest that psychological stress and other behavioral factors accelerate the hair greying process ( Nahm et al . , 2013; Peters et al . , 2017 ) , a notion supported by studies in mice demonstrating that adrenergic stimulation by norepinephrine signaling leads to melanocyte stem cell depletion in mice ( Zhang et al . , 2020 ) . However , contrary to mice where this process appears to be irreversible at the single hair follicle level , our data demonstrates that human hair greying is , at least under some circumstances , reversible . This dichotomy highlights a potential fundamental difference between rodent and human HF biology , calling for a quantitative examination of this process in humans . As evidence that environmental or behavioral factors influence human hair greying , epidemiological data suggests that smoking and greater perceived life stress , among other factors , are associated with premature greying ( Akin Belli et al . , 2016 ) . Chronic psychosocial stress also precipitates telomere shortening , DNA methylation-based epigenetic age , as well as other biological age indicators in humans ( Epel et al . , 2004; Han et al . , 2019 ) , demonstrating that psychological factors can measurably influence human aging biology . In relation to mitochondrial recalibrations , psychosocial factors and induced stress can also influence mitochondrial energetics within days in humans ( Picard et al . , 2018 ) and animals ( Picard and McEwen , 2018 ) . To generate proof-of-concept evidence and test the hypothesis that psychosocial or behavioral factors may influence greying at the single-HF level , we leveraged the fact that HPPs reflect longitudinal records of growth over time – similar to tree rings – which can be aligned with assessments of life stress exposures over the past year . By converting units of hair length into time , perceived stress levels can be quantitatively mapped onto HPPs in both greying and transitional hairs . A systematic survey of two-colored hairs on the scalp of a 35-year-old Caucasian male with auburn hair color over a 2-day period yielded five two-colored HS from the frontal and temporal scalp regions . Again , two-colored hairs could either exhibit depigmentation or reversal . Unexpectedly , all HS exhibited reversal . HPP analysis further showed that all HS underwent reversal of greying around the same time period . We therefore hypothesized that the onset of the reversal would coincide with a decrease in perceived life stress . A retrospective assessment of psychosocial stress levels using a time-anchored visual analog scale ( participants rate and link specific life events with start and end dates , see Materials and methods and Figure 3—figure supplement 1 for details ) was then compared to the HPPs . The reversal of greying for all hairs coincided closely with the decline in stress and a 1-month period of lowest stress over the past year ( 0 on a scale of 0–10 ) following a 2-week vacation ( Figure 3C ) . We were also able to examine a two-colored hair characterized by an unusual pattern of complete HS greying followed by rapid and complete reversal ( same as in Figure 2B ) plucked from the scalp of a 30-year-old Asian female participant with black hair . HPP analysis of this HS showed a white segment representing approximately 2 cm . We therefore hypothesized that this reversible greying event would coincide with a temporary increase in life stress over the corresponding period . Strikingly , the quantitative life stress assessment over the last year revealed a specific 2-month period associated with an objective life stressor ( marital conflict and separation , concluded with relocation ) where the participant rated her perceived stress as the highest ( 9–10 out of 10 ) over the past year . The increase in stress corresponded in time with the complete but reversible hair greying ( Figure 3D ) . This association was highly significant ( p=0 . 007 ) based on our permutation test . Given the low statistical probability that these events are related by chance , life stress is the likely preceding cause of these HS greying and reversal dynamics . These data demonstrate how the HPP-stress mapping approach allows to examine the coordinated behavior of greying and reversal dynamics with psychosocial factors , raising the possibility that systemic biobehavioral factors may influence multiple HFs simultaneously and regulate HPPs among sensitive hairs . To assess whether rapid greying and reversal events among a single hair are molecularly similar or distinct to those described in the two proteomics experiments above , we dissected six segments ( two dark , two white , two reverted ) of the single HS in Figure 3D and quantified their proteomes as part of Experiment 2 . This produced a longitudinal , single-hair , proteomic signature ( Figure 3E ) containing 301 proteins quantified in ≥2 of the six segments . To examine how the proteome as a whole is altered through the greying and reversal transitions associated with psychosocial stress levels , we generated a PLS-DA model with all six segments . Both dark segments clustered together , with similar values on both first and second principal components . The white and reverted segments clustered in separate topological spaces ( Figure 3F ) . Greying was associated with a positive shift largely along the first component ( Component 1 ) , whereas reversal was associated with a negative shift on the second component ( Component 2 ) and a more modest negative shift in Component 1 . We therefore extracted loading weights of each protein on Components 1 and 2 ( reflecting each protein’s contribution to group separation ) and used the top proteins ( n = 20 highest negative , and 20 most positive loadings , total n = 40 per component ) to interrogate KEGG and GO databases . The model’s Component 1 ( associated with greying ) contained proteins that were either ( i ) not expressed in the dark HS but induced selectively in the white HS segment or ( ii ) highly abundant in dark segments but strongly downregulated in white and reverted segments ( Figure 3H , top and bottom , respectively ) . In gene set enrichment analysis of Component 1 ( greying ) , the top three functional categories were carbon metabolism , glycolysis/gluconeogenesis , and Kreb’s cycle ( Figure 4E ) . Component 2 ( reversal ) -associated proteins exhibited distinct trajectories either peaking in the first white segment or upon reversal ( Figure 3I ) and mapped to pathways related to the complement activation cascade , infectious processes , and Parkinson’s and Huntington’s disease ( Figure 4E ) . In contrast , a null set of hair proteins not contributing to either components exhibited enrichment for extracellular exosomes and cell-cell adhesion that reflect hair shaft biology ( Figure 3—figure supplement 2 ) , illustrating the specificity of our findings related to greying and reversal . These data indicate that the reversal of greying at the single-hair level is not associated with a complete reversal in the molecular composition of the HS . Rather , some of the proteomic changes in hair greying are enduring despite successful repigmentation . To systematically examine the overlap among the different proteomic datasets and to derive functional insight into the hair greying process in humans , we then integrated results from the three datasets described above . White HS show consistently more upregulated than downregulated proteins across datasets ( 2 . 91-fold in Experiment 1 , 1 . 89-fold in Experiment 2 ) ( Figure 4A ) . This preferential upregulation suggests that the depigmentation process likely involves active metabolic remodeling rather than a passive loss of some pigmentation-related factor . The overlap in the specific proteins identified across dark-white comparisons and among the 6-segments hair is illustrated in Figure 4B . Five proteins were consistently upregulated between experiments 1 and 2 . These include three well-defined resident mitochondrial proteins involved in lipid metabolism: CPT1A , which imports fatty acids into mitochondria Schlaepfer and Joshi , 2020; ACOT7 , which hydrolyzes long-chain fatty acyl-CoA esters in the mitochondrial matrix and cytoplasm ( Bekeova et al . , 2019 ) ; and SOD1 , which dismutates superoxide anion into hydrogen peroxide ( H2O2 ) in the mitochondrial intermembrane space ( Okado-Matsumoto and Fridovich , 2001 ) . The other two proteins include the actin-depolymerizing protein cofilin-1 ( CFL1 ) and the core glycolysis enzyme phosphoglycerate kinase 1 ( PGK1 ) ( Figure 4C ) . Interestingly , beyond its role in cytoskeleton dynamics , CFL1 promotes mitochondrial apoptotic signaling via cytochrome c release ( Hoffmann et al . , 2019 ) and regulates mitochondrial morphology via its effect on actin polymerization-dependent mitochondrial fission ( Rehklau et al . , 2017 ) . And although PGK1 is a cytoplasmic kinase , it was recently demonstrated to translocate inside mitochondria where it phosphorylates and inhibits pyruvate dehydrogenase and Krebs cycle activity ( Nie et al . , 2020 ) . Thus , all five proteins validated across both experiments are linked to mitochondrial energy metabolism , implicating mitochondrial remodeling as a feature of hair greying . Interestingly , all five proteins have also been linked to the biology of melanocytes ( Bracalente et al . , 2018; Morvan et al . , 2012; Oh et al . , 2014; Sumantran et al . , 2015; Sung et al . , 2016 ) , the source of pigment in the HFPU . The downregulated proteins were keratins , with small effect sizes , and not particularly robust . Analysis of the intensity based on absolute quantification ( iBAQ ) data confirmed the upregulation of these five mitochondrial proteins , and the absence of substantial changes in the keratins . Together , these data suggest that HS proteome profiling may provide a retrospective access to some aspect of melanocyte metabolism , which opens new possibilities to study HF aging biology . Since the observed proteomic signatures are related to specific metabolic pathways rather than the typical high-abundance mitochondrial housekeeping proteins , we reasoned that the upregulation of these mitochondrial components unlikely reflects a bulk increase in total mitochondrial content . To investigate this point using an independent method , we quantified mitochondrial DNA ( mtDNA ) abundance in human HSs by real-time qPCR . Both white and dark HSs contain similarly high levels of mtDNA ( Figure 4D ) . The same was true in the follicles of the same hairs ( Figure 4—figure supplement 1 ) . The similar mtDNA levels between dark and white HSs and HFs increases the likelihood that the reported proteomic changes reflect the induction of specific metabolic pathways associated with hair greying rather than bulk increase in mitochondrial mass . To identify a general proteomic signature of greying hair , we compiled the enrichment scores for KEGG pathways across all datasets ( Figure 4E ) . Consistent with the function of the individual proteins identified in both group comparisons ( Experiments 1 and 2 ) and the multi-segment double-transition hair , white HS showed consistent upregulation of carbon metabolism and amino acid biosynthesis , glycolysis/gluconeogenesis , and general metabolic pathways relevant to aging biology ( Wiley and Campisi , 2016 ) . Comparatively fewer pathways were consistently represented for downregulated proteins across independent experiments . In relation to hair biology in general , our data adds to previous efforts ( Franklin et al . , 2020 ) and provides a quantitative map of the co-expression among keratin and non-keratin HS proteins across dark and white hairs ( Figure 4F ) . Computing the cross-correlations for each protein pair revealed four main clusters among the HS proteome ( Figure 4—figure supplement 2 ) . As expected for hair , keratins were well-represented and constituted the main GO biological processes category for three of the four clusters . The top KEGG categories included glycolysis and estrogen signaling pathways , which also showed strong co-expression with each other , highlighting potential interaction among endocrino-metabolic processes in relation to human hair pigmentation . In general , the identification of several non-keratin metabolism-related proteins in the HS opens new opportunities to investigate greying pathobiology and to non-invasively access past molecular and metabolic changes that have occurred in the aging HFPU of the dynamically growing hair . Finally , to narrow the range of plausible mechanisms for the observed age-related greying and reversal events , we developed a simulation model of HPPs . Greying dynamics of an individual’s hair population ( ~100 , 000 hairs ) across the average 80 year lifespan cannot practically be measured . In the absence of such data , we propose here a mathematical model to simulate hair greying trajectories across the human lifespan ( Figure 5A , available online , see Materials and methods for details ) as has been attempted previously for hair growth cycles ( Halloy et al . , 2000 ) . As basic tenets for this model , it is established that ( i ) the onset of human hair greying is not yet underway and rarely begins in childhood , ( ii ) greying routinely starts between 20 and 50 years of age , ( iii ) greying is progressive in nature ( the total number and percentage of grey hairs increases over time ) , and ( iv ) the proportion of white hairs reaches high levels in old age , although some hairs can retain pigmentation until death , particularly among certain body regions ( Trueb and Tobin , 2010 ) . Additionally , our findings demonstrate that ( v ) age-related greying is naturally reversible in isolated hair follicles , at least temporarily and in individual HS , and may be acutely triggered by stressful life experiences , the removal of which can trigger reversal . Aiming for the simplest model that accounts for these known features of hair greying dynamics , we found a satisfactory model with three components ( Figure 5B ) : ( 1 ) an ‘aging factor’ that progressively accumulates within each hair follicle , based on the fact that biological aging is more accurately modeled with the accumulation of damage , rather than a decline in stem cells or other reserves ( Kinzina et al . , 2019 ) ; ( 2 ) a biological threshold , beyond which hairs undergo depigmentation ( i . e . greying ) , characterizing the transition between the dark and white states in the same HS; and ( 3 ) a ‘stress factor’ that acutely but reversibly increases the aging factor during a stressful event . For modeling purposes , the accumulation of the aging factor is equivalent to the inverse of the decrease in a youth factor ( e . g . loss of telomere length with age ) . Based on the mosaic nature of scalp HFs and our data indicating that not all hairs are in perfect synchrony , the proposed model for an entire population of hairs must also allow a variety of aging rates , as well as differential sensitivity to stress among individual hairs . We find that the model’s predicted hair population behavior ( % of white HSs on a person’s head over time ) across the lifespan is consistent with expected normal human hair greying dynamics ( Figure 5C ) . White hairs are largely absent until the onset of greying past 20 years of age then accumulate before finally reaching a plateau around 70–90% of white hairs , near 100 years . Thus , this model recapitulates the expected between-hair heterogeneity of greying within an individual , producing the common admixture of white and pigmented hairs or ‘salt and pepper’ phenotype in middle-age . However , some individuals also develop hairs with intermediate pigmentation states ( i . e . silver/steel color ) , which our model does not reproduce . This represents a limitation to be addressed in future research . We note that there are natural inter-individual differences in the rate of greying: some individuals begin greying early ( onset in early 20’s ) ; some begin late ( onset in 50’s ) . A higher rate of accumulation of the aging factor ( higher slope for each hair ) or a lower threshold naturally accounts for earlier onset of greying . In addition , our model reveals that within a person , greater hair-to-hair heterogeneity in the rate of aging between HFs , modeled as the standard deviation of slope across hairs , also influences the onset of greying . Greater heterogeneity between HFs allows for earlier onset of greying , whereas decreasing hair-to-hair variation ( i . e . lower heterogeneity ) is associated with a ‘youthful’ later onset of greying ( Figure 5D ) . Interestingly , this unpredicted result aligns with the notion that increased cell-to-cell heterogeneity is a conserved feature of aging ( Bahar et al . , 2006; Enge et al . , 2017; Martinez-Jimenez et al . , 2017 ) and that biological heterogeneity can predict all-cause mortality in humans ( Patel et al . , 2010 ) . Using parameter values that yield the average onset and rate of greying , we then simulated the influence of acute psychosocial stressors , either early in life before the onset of greying , or later once grey HSs have begun to accumulate . Similar to our data , the model also predicts transitory , or temporary reversible events of greying ( see Figure 3D ) . Transitory greying events do not affect all hairs , only those that are close to the threshold at the time of stress exposure undergo greying . Hairs whose cumulative aging factors are substantially lower than threshold do not show stress-induced greying ( a 5-year-old is unlikely to get grey hairs from stress , but a 30-year-old can ) ( Figure 5E–F ) . Similarly , grey hairs far above threshold are not affected by periods of psychosocial stress . Thus , our model accounts for both the overall hair greying dynamics across the lifespan , and how a stressor ( or its removal ) may precipitate ( or cause reversal of ) greying in hairs whose aging factor is close to the greying threshold . We speculate that this simulation opens an attractive possibility whereby HPP data from individuals could be used in models to formulate predictions about future greying trajectories , and to use HPPs and hair population greying behavior to track the effectiveness of behavioral and/or therapeutic interventions aimed at modifying human aging . Extending our high-resolution quantitative digitization approach to hundreds of randomly sampled dark non-transitioning hairs from different scalp regions in the same individuals , we also show that fully dark ( i . e . non-greying ) HSs exhibit mostly unique HPPs , but that hairs among the same scalp regions may exhibit more similar HPPs than hairs sampled from different regions ( Figure 5—figure supplement 1; Stenn and Paus , 2001 ) . This may in part be influenced by the migration of stem cells during embryogenesis to different parts of the scalp , or by other unknown factors . This preliminary extension of the HPP methodology provides a foundation for future studies . Moreover , the regional segregation of HPPs may reflect well-recognized regional differences in the rate of HS formation ( Robbins , 2012 ) . Thus , future models may also be able to leverage information contained within HPPs from non-greying hairs and make specific inference from hairs collected across scalp regions . Similar to how decoding temporal patterns of electroencephalography ( EEG ) provides information about the state of the brain , our data make it imaginable that decoding HPP analysis over time may provide information about the psychobiological state of the individual . Our approach to quantify HPPs demonstrates rapid greying transitions and their natural transitory reversal within individual human hair follicles at a higher frequency and with different kinetics than had previously been appreciated . The literature generally assumes pigment production in the HFPU to be a continuous process for the entire duration of an anagen cycle , but here we document a complete switch-on/off phenomena during a single anagen cycle . The proteomic features of hair greying directly implicate multiple metabolic pathways that are both reversible in nature and sensitive to stress-related neuroendocrine factors . Therefore , this result provides a plausible biological basis for the rapid reversibility of greying and its association with psychological factors , and also supports the possibility that this process could be targeted pharmacologically . Melanogenesis is also known to both involve and respond to oxidative stress , a byproduct of mitochondrial metabolic processes ( Balaban et al . , 2005 ) and driver of senescence ( Vizioli et al . , 2020 ) . Moreover , alterations in energy metabolism are a major contributor to other disease-related aging features ( Kennedy et al . , 2014 ) , including lifespan regulation ( Jang et al . , 2018; Latorre-Pellicer et al . , 2016 ) . The upregulation of specific components related to mitochondrial energy metabolism in white hairs suggests that energy metabolism regulates not only hair growth as previously demonstrated ( Flores et al . , 2017; Mancino et al . , 2020; Vidali et al . , 2014 ) but also HF greying biology . Our findings demonstrating an upregulation of the fatty acid synthesis and metabolism machinery resonate particularly strongly with recent work demonstrating that fatty acid synthesis by FASN ( Fafián-Labora et al . , 2019 ) and transport by CPT1a ( Seok et al . , 2020 ) are sufficient drivers of cell senescence , and that fatty acid metabolism regulates melanocyte aging biology ( Tang et al . , 2019 ) . Approaches combining both high molecular and spatial resolution may be particularly informative ( Vyumvuhore et al . , 2021 ) . Causally linking these putative metabolic changes to canonical aging and/or senescence markers in human hair shafts and among specific HF cell populations will be an important challenge for the field . Although surprising , the reversal of hair greying is not an isolated case of ‘rejuvenation’ in mammals . In vivo , exposing aged mice to young blood in parabiosis experiments ( Rebo et al . , 2016; Villeda et al . , 2014 ) or diluting age-related factors in old animals ( Mehdipour et al . , 2020 ) triggers the reversal of age-related molecular , structural and functional impairments . In human cells , quantitative biological age indicators such as telomere length ( Puterman et al . , 2018 ) and DNA methylation ( Fahy et al . , 2019 ) also exhibit temporary reversal in response to exercise and dietary interventions . Moreover , the reversibility of greying in aging human HFs demonstrated by our data is also consistent with the observed reversibility of human skin aging in vivo when aged human skin is xenotransplanted onto young nude mice ( Gilhar et al . , 1991a ) . Notably this skin ‘rejuvenation’ is associated with a marked increase in the number of melanocytes in human epidermis ( Gilhar et al . , 1991b ) , suggesting plasticity of the melanocyte compartment . Therefore , our HPP data and simulation model adds to a growing body of evidence demonstrating that human aging is not a linear , fixed biological process but may , at least in part , be halted or even temporarily reversed . Our method to map the rapid ( weeks to months ) and natural reversibility of human hair greying may thus provide a powerful model to explore the malleability of human aging biology within time scales substantially smaller than the entire lifespan . A notable finding from both proteomics experiments is the bias toward upregulation rather than the loss of proteins in depigmented grey HS . As noted above , this may reflect the fact that hair greying is an actively regulated process within the HPFU , and that aging is not marked by a loss , but rather an increase in heterogeneity and biological complexity ( Bahar et al . , 2006; Enge et al . , 2017; LaRocca et al . , 2020; Martinez-Jimenez et al . , 2017 ) . Relative to the youthful state , quiescent and senescent cells exhibit upregulation of various secreted factors ( van Deursen , 2014 ) , as well as elevated metabolic activities ( Lemons et al . , 2010 ) , rather than global downregulation of cellular activities . Moreover , similar to the macroscopic appearance of hair greying , age-related senescence markers naturally occur stochastically for DNA methylation changes across the genome ( Franzen et al . , 2017 ) and among cells heterogeneously scattered within tissues in mice ( Baker et al . , 2016; Omori et al . , 2020 ) . Our data reveal that the conserved principle of an age-related increase in molecular and cellular heterogeneity is reflected not only at the tissue level ( mixture of dark and white hairs ) but also in the greying hair proteome . Moreover , our proteomics results are also in line with recent reports of keratin-associated proteins that are downregulated in white vs dark hairs ( Giesen et al . , 2011 ) , and proteins that are upregulated with increasing age of the donor ( Plott et al . , 2020 ) . Specifically , of a previously identified group of 50 potentially age-related , upregulated proteins in the HS ( Plott et al . , 2020 ) , 16 were detected in our second proteomic experiment . Of these 16 , 14 were similarly upregulated in depigmented white hairs relative to dark hairs from the same individuals in our dataset ( Supplementary file 2 ) . Further work will be required to determine if specific molecular aging processes , in specific cell types within the HF , account for the visible macroscopic instability of HFs greying on the human scalp . Finally , in relation to psychobiological processes , the spatio-temporal resolution of the HPP approach provides investigators with an instructive new research tool that allows to link , with an unprecedented level of resolution , hair greying/reversal events with psychosocial exposures . Here , we provided proof-of-concept evidence that biobehavioral factors are linked to human hair greying dynamics . Our optical digitization approach thus extends previous attempts to extract temporal information from human hairs and illustrate the utility of HPP profiling as an instructive and sensitive psychobiology research model . Additional prospective studies with larger sample sizes are needed to confirm the robust reproducibility and generalizability of our findings . Visualizing and retrospectively quantifying the association of life exposures , stress-associated neuroendocrine factors , and HPPs may thus contribute to elucidating the mechanisms responsible for the embedding of stress and other life exposures in human biology .	Hair greying is a visible sign of aging that affects everyone . The loss of hair color is due to the loss of melanin , a pigment found in the skin , eyes and hair . Research in mice suggests stress may accelerate hair greying , but there is no definitive research on this in humans . This is because there are no research tools to precisely map stress and hair color over time . But , just like tree rings hold information about past decades , and rocks hold information about past centuries , hairs hold information about past months and years . Hair growth is an active process that happens under the skin inside hair follicles . It demands lots of energy , supplied by structures inside cells called mitochondria . While hairs are growing , cells receive chemical and electrical signals from inside the body , including stress hormones . It is possible that these exposures change proteins and other molecules laid down in the growing hair shaft . As the hair grows out of the scalp , it hardens , preserving these molecules into a stable form . This preservation is visible as patterns of pigmentation . Examining single-hairs and matching the patterns to life events could allow researchers to look back in time through a person’s biological history . Rosenberg et al . report a new way to digitize and measure small changes in color along single human hairs . This method revealed that some white hairs naturally regain their color , something that had not been reported in a cohort of healthy individuals before . Aligning the hair pigmentation patterns with recent reports of stress in the hair donors’ lives showed striking associations . When one donor reported an increase in stress , a hair lost its pigment . When the donor reported a reduction in stress , the same hair regained its pigment . Rosenberg et al . mapped hundreds of proteins inside the hairs to show that white hairs contained more proteins linked to mitochondria and energy use . This suggests that metabolism and mitochondria may play a role in hair greying . To explore these observations in more detail Rosenberg et al . developed a mathematical model that simulates the greying of a whole head of hair over a lifetime , an experiment impossible to do with living people . The model suggested that there might be a threshold for temporary greying; if hairs are about to go grey anyway , a stressful event might trigger that change earlier . And when the stressful event ends , if a hair is just above the threshold , then it could revert back to dark . The new method for measuring small changes in hair coloring opens up the possibility of using hair pigmentation patterns like tree rings . This could track the influence of past life events on human biology . In the future , monitoring hair pigmentation patterns could provide a way to trace the effectiveness of treatments aimed at reducing stress or slowing the aging process . Understanding how ‘old’ white hairs regain their ‘young’ pigmented state could also reveal new information about the malleability of human aging more generally .	[ "Abstract", "Introduction", "Materials", "and", "methods", "Results", "Discussion" ]	[ "medicine", "cell", "biology" ]	2021	Quantitative mapping of human hair greying and reversal in relation to life stress
Clinical observations indicate that COVID-19 is a systemic disease . An investigation of the viral distribution within the human body and its correlation with tissue damage can aid in understanding the pathophysiology of SARS-CoV-2 infection . We present a detailed mapping of the viral RNA in 61 tissues and organs of 11 deceased patients with COVID-19 . The autopsies were performed within the early postmortem interval ( between 1 . 5 and 15 hr , mean: 5 . 6 hr ) to minimize the bias due to viral RNA and tissue degradation . Very high viral loads ( >104copies/ml ) were detected in most patients' lungs , and the presence of intact viral particles in the lung tissue could be verified by transmission electron microscopy . Interestingly , viral RNA was detected throughout various extrapulmonary tissues and organs without visible tissue damage . The dissemination of SARS-CoV-2-RNA throughout the body supports the hypothesis that there is a maladaptive host response with viremia and multiorgan dysfunction . In December 2019 , several cases of pneumonia caused by a novel Betacoronavirus called SARS-CoV-2 were first described in the city of Wuhan in China ( Zhu et al . , 2020 ) ; the disease was thereafter named ‘coronavirus disease 2019’ ( COVID-19 ) ( Lu et al . , 2020 ) . Within a few months , the initially localized outbreak spread to countries all over the globe and was declared a pandemic ( El Zowalaty and Järhult , 2020 ) . At present , more than 121 million SARS-CoV-2 infections have been reported ( Dong et al . , 2020 ) . The number of deaths attributed to COVID-19 has exceeded 2 . 6 million worldwide ( Dong et al . , 2020 ) . COVID-19 occurs with varying degrees of severity . While approximately 81% of COVID-19 patients experience mild symptoms , 14% suffer from respiratory distress , and 5% have to be hospitalized ( Wu and McGoogan , 2020; Wiersinga et al . , 2020 ) . Of these , 20% enter a critical condition with respiratory failure , endovascular complications , or multiple organ dysfunction . Gastroenterological and neurological symptoms have been reported in 36 . 4% and 18 . 6% of COVID-19 patients , respectively , in case studies ( Wiersinga et al . , 2020; Pan et al . , 2020; Mao et al . , 2020 ) . The clinical observations suggest that COVID-19 is a systemic disease . While little information has been available to date about the molecular regulation of SARS-CoV-2 infections , angiotensin-converting enzyme 2 ( ACE2 ) and transmembrane protease serine 2 ( TMPRSS2 ) , two membrane-bound proteins , have been shown to be crucial for the entry of the virus into cells ( Wang et al . , 2020; Hoffmann et al . , 2020 ) . ACE2 is expressed not only in the epithelia of the lung but also in several other epithelial , endothelial , heart , and renal tissues ( Hamming et al . , 2004 ) . SARS-CoV-2 viral replication and pathogenesis in organisms are currently not well understood due to the lack of appropriate models ( Bar-On et al . , 2020 ) . One crucial step to elucidate viral pathogenesis is the investigation of the distribution of the virus within the entire body . In the present study , we ( 1 ) included full autopsies , ( 2 ) performed autopsies in the early postmortem interval ( 1 . 5–15 hr , mean: 5 . 6 hr ) , ( 3 ) dissected organs and tissues without prior fixation in formalin , ( 4 ) measured SARS-CoV-2 RNA in a high number of samples , ( 5 ) correlated the viral load with tissue damage using comprehensive histopathological investigations , ( 6 ) visualized virus particles in pulmonal tissue samples by means of transmission electron microscopy ( TEM ) , and ( 7 ) determined the postmortem serum levels of inflammatory cytokines and prothrombotic factors . Sampling was performed in the very early postmortem interval to provide reliable viral RNA ( vRNA ) measurements and enabled us to obtain blood serum and well-preserved tissue samples for ultrastructural analysis . The investigation of COVID-19 patients included a full characterization of the clinical characteristics and parameters ( Table 1 ) . In detail , patients 1–7 received intensive care; patients 1–6 were mechanically ventilated , whereas patient 7 was subjected to ECMO ( extracorporeal membrane oxygenation ) . Patients 1–2 and 5–6 were treated with lopinavir/ritonavir . Patients 8–10 were not subjected to intensive care or ventilation according to their patient provisions . Our results show that patients 1–10 died of COVID-19 , whereas patient 11 suffered from a metastasized squamous cell carcinoma of the cervix and died of an ileus following peritoneal carcinosis ( Table 2 provides an overview of the macro- and micromorphological autopsy findings ) . Patient 11 had contracted COVID-19 and received intensive care treatment but was not ventilated . Interestingly , autopsy detected previously undiagnosed malignancies in patients 2 ( chronic lymphatic leukemia , CLL ) and 10 ( endometrial carcinoma ) . In addition , patient 6 had an incidentaloma of the thyroid gland . Amplification of the E-gene of SARS-CoV-2 by using qRT-PCR detected a very high to high mean viral load in the lungs of all patients ( Figure 1 ) . Patients 1–10 showed high viral loads , which were as high as 107 RNA copies/ml ( Figure 1—figure supplement 1 ) . High to moderate to low viral loads were detected in other structures of the respiratory tract , such as the mesopharynx , epiglottis and trachea , in patients 1–8 . In patient 11 , with a non-COVID-19-associated cause of death , vRNA could only be detected in moderate to low amounts in the trachea . Patients 1–10 also showed variable ( very high to very low ) viral loads in at least two samples obtained from the lymphatic system . Lymphatic structures with topological relationships to the respiratory tract were always positive for vRNA . Of the patients subjected to intensive care treatment ( patients 1–7 ) , patients 1 , 3 , 4 , and 5 exhibited moderate to very low viral loads in the cardiac samples ( Figure 1—figure supplement 2 ) Patients 2 , 6 , and 7 exhibited no vRNA in the heart muscle . The vascular samples exhibited higher viral loads overall than the cardiac samples in the majority of patients . Viral RNA was detected in the blood only for patients 3–5 . Patients 1 , 2 , 6 , and 7 , who were treated with lopinavir/ritonavir , were negative for vRNA in blood . Viral RNA was present in the bone marrow of all three patients who tested positive for vRNA in blood and was also found in an additional three patients who tested negative for vRNA in blood . Patients 3–5 had vRNA in variable amounts throughout the small and large intestine . Patients 6 and 7 tested negative for all 12 gastrointestinal samples . Of note , in patient 9 , almost all the gastrointestinal samples exhibited moderate to very high viral loads . Viral RNA could also be detected in endocrine organs , in the urinary tract and in the reproductive organs . Only patients 3–5 tested positive for vRNA in the central nervous system . Skin ( abdominal ) , subcutaneous tissue ( abdominal ) and skeletal muscle ( psoas major ) tested negative in all patients . Due to the very early postmortem interval in which the autopsies were performed , we were able to verify the vRNA findings via TEM of a lung sample from patient 3 by detecting intact SARS-CoV-2 viral particles within lung fibrocytes ( Figure 2 ) . To analyze the proinflammatory responses of the 11 patients , we measured interleukin ( IL ) −6 ( Figure 3a ) and IL-8 ( Figure 3b ) postmortem . Both parameters showed significantly elevated serum levels in all cases compared to the levels in five healthy volunteers , who served as controls . Since abnormalities in the coagulation system were described in COVID-19 patients ( Connors and Levy , 2020 ) , we analyzed the prothrombotic parameters of the blood of the deceased patients . Disseminated intravascular coagulation was described previously , and one study suggested that the role of coagulation is to limit infection dissemination ( Antoniak , 2018 ) . Additionally , coagulation processes cause hyperinflammatory responses in viral infection ( Yang and Tang , 2016 ) . Our results showed significantly higher serum levels of tissue plasminogen activator ( tPa ) ( Figure 3c ) in COVID-19 patients than in healthy volunteers . P-Selectin ( Figure 3d ) , a cell adhesion molecule in platelets necessary for the recruitment of platelets and binding to the endothelium ( Furie and Furie , 2004 ) , was found in all patients in serum at levels significantly higher than those in controls . D dimer serum levels ( Figure 3e ) were slightly elevated in patients . However , no significant elevation could be found . The level of plasminogen activator inhibitor ( PAI , Figure 3f ) , an important inhibitor of tPa ( Cesari et al . , 2010 ) , was significantly elevated in the blood serum of all patients . The level of a biomarker with potential for cardiovascular risk stratification , sCD40L ( Figure 3g ) , was significantly elevated in all patients . Furthermore , coagulation factor IX ( Figure 3h ) was measured at significantly elevated levels in the blood serum of all patients; it has been proposed to function as a mediator between viruses and cells ( Lenman et al . , 2011 ) . The level of P-Selectin Glykoprotein Ligand-1 ( PSGL-1 ) was determined to be not significantly higher in patients than in the healthy volunteer group . Macroscopic signs of severe and extensive lung damage were found in all patients . In patients 1 , 3–5 , 9 , and 10 , the lung tissue was full of blood ( hyperemia ) and fluid ( edema ) and showed weakened consistency ( Figure 4a , b ) . In contrast , the lungs of patients 6–8 displayed a firmer and more consolidated pattern with only a low degree of edema and hyperemia ( Figure 4c–e ) . In the lungs of patient 3 , nodular demarcated damage was found that was correlated with fungal superinfection ( Figure 4f ) . Patients 6 , 9 , and 10 had purulent bronchitis and bronchopneumonia due to bacterial superinfection , and patient nine exhibited severe pharyngitis . Several features of coagulopathies were found , including infarction of the lung ( Figure 4g ) and the spleen ( Figure 4h ) as well as fulminant thromboses of the periprostatic venous plexus ( Figure 4i ) and hemorrhage of the cerebellum ( Figure 4j ) . Vascular stasis and fibrinous thrombi were present in patients 1–2 and 8–10 . Thrombemboli were found in patients 2 , 4 , 5 , and 10 to a variable extent . In patient 2 , pulmonary embolisms were fatal . Microscopically , the lung tissues from patients 1 , 3–5 , 9 , and 10 revealed changes consistent with the early ( exudative ) phase of diffuse alveolar damage ( DAD ) . The consistent acute changes included severe intraalveolar and interstitial hemorrhages ( Figure 5a , Figure 5b ) , architectural injuries with a diffuse alveolar damage pattern ( involving hyaline membranes , fibrinous edema and interstitial proliferation ) , sporadic signs of cellular inflammation ( mostly involving lymphocytes and a few plasma cells ) and severe loss of structured pneumocytes . Frequently , cells with an enlarged cytoplasm and large nuclei were found to be admixed with multinucleated giant cells and to show features of squamous metaplasia and a pattern of bronchiolization ( Figure 5a , b , c ) . Enlarged alveolar cells were detached from the alveolar wall ( Figure 5d ) . The clusters of enlarged cells were strongly positive for AE1/3 ( Figure 5e ) , but only a few cells were colabeled with TTF1 ( Figure 5f ) . The lung histology in patients 6–8 displayed a pattern similar to the latter ( proliferative ) phase of diffuse alveolar damage ( DAD ) . Giant cells and cell aggregates resembling squamous metaplasia were frequently found and sometimes accompanied by fibroblastic proliferation ( Figure 5g , h ) . While the upper lobes of the lung of patient 2 showed only moderate emphysema ( Figure 6a ) , the hemorrhagic tissue damage was restricted to the middle and lower lobes of the right and the lower lobe of the left lung ( Figure 6b ) . Vasculitis-like features were observed in patients 2 , 3 , 4 , 8 , and 9 with sporadic mild lymphoplasmatic cellular infiltrates around the pulmonary artery branches ( Figure 6c , d ) . However , in patient 7 , a strong lymphocyte-predominant intra-alveolar infiltrate was found ( Figure 6e , f ) . In particular , megakaryocytes were sometimes detectable in alveolar capillaries ( Figure 6g , h ) . A common histological feature in all patients was the loss of the follicular architecture of the lymph nodes due to architectural changes ( Figure 7a ) . In the bone marrow of patient 9 , the highest viral loads were found , and significant hemophagocytosis was detectable by microscopy ( Figure 7b ) . Interestingly , a correlation between a high viral load and tissue damage , as seen in the lung , was not found in cardiac or aortic tissues . The cardiac tissues sometimes showed pre-existing changes ( fibrosis and chronic ischemic damage ) , but no severe damage , inflammation or necrosis of cardiomyocytes was found ( Figure 7c ) . Sometimes an increase in cellularity in the otherwise unremarkable cardiac tissue was seen ( e . g . in patient 1 ) , which was suggestive of an activated cardio-mesenchyme ( Figure 7d ) . The large vessels were unremarkable as well . Figure 6e demonstrates a section of the thoracic aorta of patient 5 obtained from the same anatomical location as the sample that tested highly positive for vRNA . Neither the intima ( asterisk ) nor the upper tunica media displayed any inflammatory cells or tissue damage . In the colonic mucosa , strong signs of epithelial damage were not visible by light microscopy ( e . g . patient 3; Figure 7g ) , and the histology of the pancreas was well preserved ( e . g . patient 9; Figure 7h ) . In clinical practice , many critically ill COVID-19 patients show multiple organ involvement in addition to lung failure , in particular vascular dysfunction , including thrombosis and/or impaired microcirculation ( Ji et al . , 2020 ) . A dysregulated immune response was observed , starting with a phase of immunosuppression followed by a proinflammatory phase and then a cytokine storm ( Li et al . , 2020 ) . The cytokine storm may play an important role in COVID-19 , which supports the hypothesis that COVID-19 could have a strong immunopathological component . Some aspects of the viral pathogenesis and the toxicity of the novel virus SARS-CoV-2 are known based on previous studies of SARS-CoV ( Qian et al . , 2013 ) . The virus can infect nasal mucous cells , pneumocytes and alveolar macrophages . ACE2 is the main receptor for the cellular binding process ( Bar-On et al . , 2020 ) . Since the ACE2 receptor is expressed in cells in addition to those in the respiratory system ( Jia et al . , 2005 ) , it is reasonable to assume that other organ systems can also be targeted by SARS-CoV-2 . Quite a few morphological studies have been published so far . Some are single case reports based on necropsies of the lung , liver and heart ( Xu et al . , 2020 ) , partial autopsies of the thoracic cavity ( Karami , 2020 ) or full autopsies ( Aguiar et al . , 2020 ) . Some involve small ( n = 2–3 ) case series based on surgical lung resectates ( Tian et al . , 2020a ) or on full autopsies ( Ding et al . , 2003; Sekulic et al . , 2020; Barton et al . , 2020; Varga et al . , 2020 ) . Larger case studies ( n = 4-28 ) have focused on single organs such as the lungs ( Ackermann et al . , 2020 ) , spleen ( Xu , 2020 ) , or kidney ( Puelles et al . , 2020 ) or on the heart and lungs ( Buja et al . , 2020; Fox et al . , 2020 ) or liver , heart and lungs ( Tian et al . , 2020b ) ; others have reported comprehensive organ findings obtained by minimally invasive sampling ( Duarte-Neto et al . , 2020 ) or full autopsies ( Menter et al . , 2020; Schaller et al . , 2020; Wichmann et al . , 2020; Edler et al . , 2020; Bösmüller et al . , 2020; Bradley et al . , 2020; Remmelink et al . , 2020; Skok et al . , 2020; Hanley et al . , 2020 ) . Several of the aforementioned autopsy studies reported viral loads in selected organs and tissues ( Sekulic et al . , 2020; Puelles et al . , 2020; Tian et al . , 2020b; Menter et al . , 2020; Wichmann et al . , 2020; Bösmüller et al . , 2020; Bradley et al . , 2020; Remmelink et al . , 2020; Hanley et al . , 2020 ) . The postmortem interval between death and autopsy was either not reported ( Tian et al . , 2020b; Bradley et al . , 2020; Skok et al . , 2020 ) or was > 48 hr ( Lax et al . , 2020 ) , 11–84 . 5 hr ( Menter et al . , 2020 ) , 72–96 hr ( Remmelink et al . , 2020 ) , 1–5 days ( Wichmann et al . , 2020 ) , 4 days ( on average with a maximum of 12 days ) ( Edler et al . , 2020 ) , or 6 days ( Hanley et al . , 2020 ) . The small number of patients in these studies is consistent with the sample size of many other studies but is nevertheless a limitation . However , we analyzed a considerably larger number of organs and tissues than the other groups . To our knowledge , the present study is the only one to date that has measured viral loads in a wide variety of organs and tissues by obtaining and processing 61 samples per patient . The present study is the only study so far that has focused on keeping the postmortem interval as short as possible to avoid bias due to the degradation of SARS-CoV-2 virus particles , SARS-CoV-2 RNA and tissue ultrastructure . Regarding vRNA degradation , the values reported by Puelles et al . , 2020 show comparatively low viral loads among their cases , even in the lungs , which is the primary target of SARS-CoV-2 . Wichmann et al . , 2020 reported the highest values in the lungs of 1 . 2 x 104to 9 x 109 copies/ml , while the highest values in our study were ~107 vRNA copies/ml . From Table 2 it becomes clear that the duration of post-mortem intervals in our study did not substantially influence the vRNA copy numbers . The importance of obtaining multiple samples from within one organ is emphasized by the results of patient 10 ( Figure 1—figure supplement 1 ) , in which only two of seven samples from the heart were positive . If only one sample had been obtained , the result of viral detection could have been a false negative . Based on the mapping of SARS-CoV-2 RNA throughout the whole human body , we were able to correlate the viral loads in many organs and tissues with the macro- and micromorphology . TEM investigations of the lung samples revealed the presence of morphologically intact virus particles in the tissue , which was in line with the vRNA mapping , which showed the highest viral loads in the lungs . The morphologically intact virus particles were located in lung fibrocytes . In agreement with data published by Varga et al . , 2020 , viral inclusion bodies were detectable using TEM . In lung tissues , the morphology of the viral particles was clearly observed ( Figure 2 ) , whereas in other tissues such as liver , heart , and intestine , viral inclusions were not visible by TEM . The loss of structural hallmarks could be due to postmortem cell and tissue turnover and the reduced integrity of virus particles during TEM-related preparations . It has to be stressed that the molecular detection of the virus does not depend on the presence of morphologically intact virus particles . We detected the highest viral loads and the most severe tissue damage in the lungs . The lung samples of all patients showed large cells , sometimes multinucleated giant cells , that were similar to the giant cells described in cases of respiratory syncytial virus ( RSV ) infection . The preliminary immunostaining pattern of the enlarged cells indicated that they represented affected pneumocytes . Squamous ( metaplastic ) large cells and clusters of giant cells have been reported by most morphological studies , with one exception ( Ackermann et al . , 2020 ) . The remaining findings from the lung samples agree very well with the findings of other groups , especially the data of Duarte-Neto et al . , 2020 . The strict topological correlation of viral loads and histopathological damage is emphasized by the results in patient 2 . The samples from the upper lung lobes showed normal , unremarkable tissue ( Figure 6a ) corresponding to a negative viral test result , while the samples from the lower lung lobes revealed severe tissue damage corresponding to high and moderate viral loads ( Figure 6b ) . All patients who died due to COVID-19 ( patients 1–10 ) had viral RNA in at least some samples of the lymphatic tissue . Lymphatic tissue with a topological relationship with the respiratory tract ( e . g . tonsils , cervical lymph nodes , and hilar lymph nodes ) was more likely to be positive overall than lymphatic tissue without such a topological relationship ( e . g . mesenteric lymph nodes , spleen , and appendix ) . One remarkable finding in the lymph node samples of all patients was the loss of the follicular structure ( Figure 7a ) . Atrophy of lymphatic tissue has been described in association with SARS-CoV infection by Gu et al . , 2005 and discussed as a crucial determinant of disease outcome by Perlman and Dandekar , 2005 . Lymphocyte depletion has also been reported by Hanley et al . , 2020 . The spleen was positive in patients 1 , 3 , 4 , 5 , and 9 , who presented with the micromorphology of early lung damage , and negative in patients 6 , 7 , 8 , and 10 , who presented with the micromorphology of later lung damage or who did not die of COVID-19 pneumonia ( patient 2 ) . Further interpretation of the viral loads in the appendix is futile since the appendix showed age-related and chronic pathologic changes accompanied by the loss of lymphatic tissue . Viral loads in the cardiac tissue were moderate to very low and systematically ( in all samples ) detected in the patients with early lung damage , while patients with later lung damage displayed high viral loads only sporadically or not at all . The cardiac histology of the left ventricle , anterior wall , and basal portion in patient one with a moderate viral load is presented in Figure 7c , d . Except for the activation of mesenchymal cells , which requires further investigation , the histology was unremarkable . In accordance with Buja et al . , 2020 , in patient 1 , we also observed pericarditis and multifocal acute injury of cardiomyocytes , for example , myocardial contraction band necrosis , which is frequently observed in critically ill patients under catecholamine therapy . The viral loads in the samples from the vascular tissue ( aorta and pulmonary artery ) followed a similar distribution pattern depending on the stage of lung damage but were higher compared to those in the cardiac tissue . The unremarkable histology of the thoracic aorta of patient 5 with a high viral load is presented in Figure 7e , f . The conclusion by Varga et al . , 2020 that SARS-CoV-2 induces endothelitis cannot be comprehended . Interestingly , the presence of detectable vRNA in the gastrointestinal tract was variable . The very high viral loads in patient nine throughout the upper gastrointestinal tract as well as in the small and large bowels were noticeable . The histology of the corresponding tissue samples was unremarkable . According to the clinical documentation , patient 9 did not exhibit any gastrointestinal symptoms . Viral RNA could also be detected in low to very high amounts in the samples from the endocrine organs , urinary tract , nervous system , and reproductive system . Interestingly , the samples of patients 1 , 2 , 6 , and 7 , who were treated with lopinavir/ritonavir , tested negative . Our results support the findings of Remmelink et al . , 2020 . of nonspecific postmortem organ findings despite multiorgan viral spread . The same distribution pattern among the patients was observed regarding viral RNA in blood . Apart from patients 8–11 , who were not subjected to intensive care ( patients 8–10 ) or did not die of COVID-19 ( patient 11 ) , it is noticeable that among the patients who received intensive care prior to death ( patients 1–7 ) , only patients 3 , 4 , and 5 tested positive for vRNA in blood , while patients 1 , 2 , 6 , and 7 tested negative . The latter patients were treated with lopinavir/ritonavir , so the effect of antiviral medication on preventing viremia may be indicated . In our study , 4 of 11 patients were treated with an antiviral medication . None of these four patients showed the presence of vRNA in the blood , but one patient showed the presence of vRNA in the bone marrow ( patient 1 ) . The application of antiviral drugs is currently being investigated in many studies . However , it has not yet been shown that a significant effect can be achieved by their application ( Cao et al . , 2020 ) . However , drugs are often used in severe cases in the ICU based on controversial recommendations ( Meini et al . , 2020 ) . Our data could indicate that lopinavir/ritonavir leads to a reduction in viremia . However , our group size is too small for such a statement , as we could not detect vRNA in the blood , not even in untreated patients in some cases . The patients with vRNA in the blood also showed vRNA in the bone marrow . Patients 1 , 8 , and 9 were negative for vRNA in the blood but positive in the bone marrow . Patient 9 showed the highest viral loads by far in the bone marrow . The histology of the bone marrow , apart from hypercellularity , a left shift and an increased number of megakaryocytes , showed a significant amount of hemophagocytosis ( Figure 6b ) . Hemophagocytosis was also reported by Hanley et al . , 2020 and is a morphological feature of macrophage activation syndrome ( MAS ) or hemophagocytic lymphohistiocytosis ( HLH ) ( Crayne et al . , 2019; Al-Samkari and Berliner , 2018 ) . The clinical characteristics of COVID-19 , including very high ferritin levels and very high levels of proinflammatory interleukins , resemble those of MAS and HLH ( Colafrancesco et al . , 2020 ) and have already led to several therapeutic attempts ( Dimopoulos et al . , 2020 ) . Further studies are needed to clarify this aspect . A positive test result based on qRT-PCR can be determined with the Ct-value . Due to the further spread of the pandemic , clinicians have discussed the release of this value as part of the test result , as the viral load has been identified as an important prognostic indicator ( Magleby et al . , 2020 ) . The authors suggest using the Ct-value to identify patients with a high risk for severe clinical courses . Further studies also indicate that the viral load may serve as a surrogate marker of infectivity associated with mortality rates ( Faico-Filho et al . , 2020 ) . However , the College of American Pathologists reminds us to be cautious in interpreting the results ( Rhoads et al . , 2020; Jaafar et al . , 2020 ) . As a multitude of intra- and interlaboratory factors influence detection , the Ct-value has to be critically evaluated . Due to the relatives' necessary consent before the autopsy , the specimen collection is only possible with a time delay in clinical routine . Our study did this particularly urgently , but there are still differences in the time of sample collection . This is an important limitation of our study and must be noted when considering the results . However , the subsequent processing of the samples was performed identically . We use the same methods for the RNA-extraction as well as the same qRT-PCR method . Another important reason for the detection of Ct-values is the statement about infectiousness . Jaafar et al . showed that the culture of the virus is successful up to a Ct-value of 25 ( 70% ) . At a Ct-value of 30 , this value drops to 20% , and values above 35 as associated with only a low likelihood ( 3% ) of the possibility of culture ( Jaafar et al . , 2020 ) . These values show a correlation to a certain extent , although it should be noted that the culture of viruses from patient material is generally challenging and that no absolute indication can be made based on a Ct-value for this purpose exclusively . To further elucidate the immunological host response , we measured IL-6 and IL-8 in postmortem serum samples . The serum levels of interleukin six were significantly elevated in all patients , including patient 11 , who died of multiple organ failure following ileus . The serum levels of IL-8 were also significantly elevated . In accordance with other autopsy studies ( Wichmann et al . , 2020; Lax et al . , 2020 ) , we observed thromboembolic events in 4 of the 11 patients . Patient 2 died of pulmonary embolism , and patient 4 suffered multiple lung and spleen infarctions due to venous thromboses . Patients 5 and 10 presented with sporadic emboli in the lung histology . In conjunction with the general impairment of the microcirculation , which was histologically visible as homogenous eosinophilic sludge in the small arterioles , capillaries and venules in multiple organ samples from patients 1 , 8 and 9 , 7 ( of 11 ) , several patients suffered hypercoagulation . Some studies hypothesize that SARS-CoV-2 can induce noncoordinated reactions between the coagulation and fibrinolysis systems that result in hypercoagulation and hemorrhage ( Ji et al . , 2020 ) . The levels of the measured pro-thrombotic factors were almost all significantly elevated in all patients . In summary , we presented an autopsy series of 11 patients with COVID-19 . The autopsies were performed in the early postmortem interval to avoid bias due to the degradation of vRNA , virus particles , and tissue structures . SARS-CoV-2 RNA could be detected in very high to high amounts in the lungs and in very high to very low amounts in the lymphatic tissue . TEM visualized SARS-CoV-2 particles in the lung tissue . Viral loads and histological tissue damage were strongly correlated in the lungs even at the organ level ( patient 2 ) . Histological structure changes were also present in the lymph nodes ( atrophy and loss of follicles ) . High viral loads were detected in many other tissue samples from different extrapulmonary organs and tissues without evident tissue damage based on light microscopy . The study was approved by the local ethical board ( registration no . : 2020–1773 ) . Complete autopsies ( inspection of cranial , thoracic and abdominal cavities plus dissection of all internal organs and their surrounding anatomical structures ) of 11 patients with SARS-CoV-2 infection ( proven by nasopharyngeal swab testing during hospitalization ) and a clinical diagnosis of COVID-19 were included in this study . As soon as possible after death , the closest relatives were contacted , who gave their informed consent . The autopsies in this study were performed 1 . 5–15 hr ( mean 5 . 6 hr ) postmortem , and the organs were dissected directly without prior fixation . The same team including two experienced forensic pathologists conducted all autopsies . The lungs , with expected high viral loads , were removed first and dissected last to avoid the transfer of vRNA to other organs/tissues . At each autopsy , a total of up to 61 native and nonfixed samples ( five locations in the nervous system , 14 in the respiratory tract with double sampling of the lungs , 10 in the cardiovascular system , 12 in the gastrointestinal tract , 3 in the urinary tract , 4 in the reproductive system , 2 in the endocrine system , 6 in the lymphatic system , 2 in hematological tissues , and 3 in abdominal skin , abdominal subcutaneous tissue and the musculus psoas major ) were collected after rinsing them in clean tap water . Tissue samples were transferred for virological processing immediately after autopsy; blood samples were centrifuged to obtain serum . Samples from the same anatomical locations were fixed in 5% buffered formalin solution for comparative histopathological analysis , and selected samples were placed in a special fixative ( see below ) for TEM . Since the correct electron microscopic identification of virus particles in tissue obtained by autopsy can be challenging ( Goldsmith et al . , 2020 ) , all possible precautions were taken . That included maintaining the shortest possible post-mortem time , the very meticulous handling of the small pieces of tissues , the exclusive use of TEM-grade , fresh chemicals , the strict adherence to the schedule and the execution of tasks by trained and experienced staff . In addition , a reference sample was prepared from a cell culture ( Vero-76 cells ) infected with SARS-CoV-2 to be able to clearly distinguish the virus findings from normal cytoplasmic structures such as parts of the endoplasmic reticulum , Golgi apparatus , coated vesicles , or artefacts caused by degradation . All tissues were homogenized in RPMI medium by using the FastPrep-24 5G Instrument ( MP Biomedicals , Schwerte , Germany ) , and disposable homogenizer beads ( Zymo Research Bashing Bead Lysis Tubes , Freiburg , Germany ) were used to avoid contamination . We placed 200 mg of each tissue/organ in 1000 µl RPMI-1640 ( Roswell Park Memorial Institute , Thermo Fisher Scientific GmbH , Dreieich , Germany ) . After a centrifugation step ( 2 min , 12 , 000 rpm ) , the supernatants were collected for the determination of the viral load . RNA extraction was performed by using the QIAcube RNeasy Viral Mini Kit ( Qiagen , Hilden , Germany ) according to the manufacturer's guide . qRT-PCR was performed using RIDAgene ( r-biopharm , Darmstadt , Germany ) with the Rotor-Gene Q ( Qiagen , Hilden , Germany ) to detect the E-gene of SARS-CoV-2 by determining the cycle threshold ( Ct ) value . The RNA standard curve , prepared by amplification of the positive control with the RIDAgene ( r-biopharm , Darmstadt , Germany ) kit , was applied for quantification . SARS-CoV-2 RNA is represented as the decadic logarithm of the number of copies/ml . The following scale was applied: very high ( >104 copies/ml ) , high ( 103–104 copies/ml ) , moderate ( 102–103 copies/ml ) , low ( 101–102 copies/ml ) , and below the detection limit ( bdl ) . For the measurement of proinflammatory cytokines and coagulation parameters , a Legendplex Human Thrombosis Panel ( 13-plex ) ( BioLegend , San Diego , CA , USA ) was used . Twenty-five microlitres of each serum sample was transferred in duplicate into a 96-well filter plate , and the Legendplex panel was performed by following the manufacturer’s instructions . The samples were measured on the same day on a flow cytometer ( BD , Accuri ) , and the protein amount was calculated by comparison to a standard curve . Serum samples from five healthy volunteers without any signs of infection were age-correlated and analysed as a control . After fixation for at least 24 hr in 10% neutral buffered formalin , the tissue samples were dehydrated in a graded series of ethanol and xylene , mounted in paraffin and cut into 3-μm-thick sections . In addition to hematoxylin and eosin ( HE ) , Elastica-van-Gieson ( EvG ) , Berlin Blue ( Fe ) , periodic acid Schiff stain ( PAS ) , Alcian Blue-periodic acid Schiff stain ( abPAS ) , Giemsa , Gomori Trichrome , and Kongo Red stain were used by following routine protocols . For immunohistochemistry , the following antibodies were used: AE1/3 ( Dako/IR053 ) , TTF-1 ( Dako/IR056 ) , CK7 ( Dako/IR619 ) , CK5/6 ( Dako/IR780 ) , p40 ( Zytomed/MSK097 ) , Ki67 ( Dako/IR626 ) , CD68 ( Dako/M0876 ) , CD61 ( Dako/M0753 ) , CD31 ( Dako/IR610 ) , CD34 ( Dako/IR623 ) , ASMA ( Dako/IR611 ) , CD3 ( Dako/IR503 ) , CD20 ( Dako/IR604 ) , MUM1 ( Dako/IR644 ) , collagen IV ( Dako/M0785 ) , and tenascin ( Chemicon/MAB19101 ) . All immunostaining were performed with the Dako Omnis immunostainer ( Agilent ) by following routine procedures . The sections were examined microscopically ( Axio Imager . M2 , Carl Zeiss Microscopy GmbH ) , and representative photographs were obtained ( Axiocam 506 color , Carl Zeiss Microscopy GmbH; ZEN 2 . 6 ( blue edition ) , Carl Zeiss Microscopy GmbH ) . During each autopsy , several small pieces of lung tissue ( 2 mm cubes ) were immediately fixed with freshly prepared modified Karnovsky fixative ( 4% w/v paraformaldehyde and 2 . 5% v/v glutaraldehyde in 0 . 1 M sodium cacodylate buffer , pH 7 . 4 ) for 24 hr at room temperature . After washing three times for 15 min each with 0 . 1 M sodium cacodylate buffer ( pH 7 . 4 ) , the tissue was further cut into 1 mm cubes and postfixed with 2% w/v osmium tetroxide for 1 hr at room temperature . During the subsequent dehydration in an ascending ethanol series , poststaining with 1% w/v uranyl acetate was performed . Afterwards , the samples were embedded in epoxy resin ( Araldite ) and sectioned using a Leica Ultracut S ( Leica , Wetzlar , Germany ) . Based on the examination of semi-thin sections , regions of interest of approximately 500 µm x 500 µm in size were selected and trimmed . Finally , the ultrathin sections were mounted on filmed Cu grids , post-stained with lead citrate , and studied in a transmission electron microscope ( EM 900 , Zeiss , Oberkochen , Germany ) at 80 kV and 3000–50 , 000x magnification . For image recording , a 2K slow scan CCD camera ( TRS , Moorenweis , Germany ) was used .	Since the discovery of the new coronavirus that causes COVID-19 , scientists have been scrambling to understand the different features of the virus . While a lot more is now known about SARS-CoV-2 , several key questions have proved more difficult to answer . For example , it remained unclear where the virus travels to in the body and causes the most harm . To help answer this question , Deinhardt-Emmer , Wittschieber et al . performed postmortem examinations on 11 patients who had recently died of COVID-19 . After sampling 61 different organs and tissues from each patient , several tests were used to detect traces of SARS-CoV-2 . The experiments showed that the largest pool of SARS-CoV-2 was present in the lungs , where it had caused severe damage to the alveolae , the delicate air sacs at the end of the lungs’ main air tubes . Small amounts of the virus were also detected in other organs and tissues , but no severe tissue damage was seen . In addition , Deinhardt-Emmer , Wittschieber et al . found that each patient had increased levels of some of the proteins involved in inflammation and blood clotting circulating their bloodstream . This suggests that the inflammation caused by SARS-CoV-2 leads to an excessive immune reaction throughout the entire body . This research provides important new insights into which areas of the body are most impacted by SARS-CoV-2 . These findings may help to design more effective drug treatments that target the places SARS-CoV-2 is most likely to accumulate and help patients fight off the infection at these regions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "immunology", "and", "inflammation" ]	2021	Early postmortem mapping of SARS-CoV-2 RNA in patients with COVID-19 and the correlation with tissue damage
Insulin gene mutations are a leading cause of neonatal diabetes . They can lead to proinsulin misfolding and its retention in endoplasmic reticulum ( ER ) . This results in increased ER-stress suggested to trigger beta-cell apoptosis . In humans , the mechanisms underlying beta-cell failure remain unclear . Here we show that misfolded proinsulin impairs developing beta-cell proliferation without increasing apoptosis . We generated induced pluripotent stem cells ( iPSCs ) from people carrying insulin ( INS ) mutations , engineered isogenic CRISPR-Cas9 mutation-corrected lines and differentiated them to beta-like cells . Single-cell RNA-sequencing analysis showed increased ER-stress and reduced proliferation in INS-mutant beta-like cells compared with corrected controls . Upon transplantation into mice , INS-mutant grafts presented reduced insulin secretion and aggravated ER-stress . Cell size , mTORC1 signaling , and respiratory chain subunits expression were all reduced in INS-mutant beta-like cells , yet apoptosis was not increased at any stage . Our results demonstrate that neonatal diabetes-associated INS-mutations lead to defective beta-cell mass expansion , contributing to diabetes development . Pancreatic beta-cells maintain glucose homeostasis via the regulated secretion of insulin . Although the etiologies of type 1 , type 2 and monogenic diabetes are different , they share similarities in the molecular pathways that become dysregulated in beta-cells during disease progression . Among these , endoplasmic reticulum ( ER ) stress and unfolded protein response ( UPR ) seem to be critical for the proper function and resilience of the beta-cell , and their role has been studied in different diabetes models ( Brozzi and Eizirik , 2016; Cnop et al . , 2017; Herbert and Laybutt , 2016 ) . High quantities of insulin are transcribed , translated and ultimately secreted by beta-cells . This requires the establishment of appropriate mechanisms for proinsulin translation , folding , processing , storage and eventual secretion of mature insulin ( Steiner et al . , 2009 ) . To cope with both the constant basal insulin secretion and the dynamic demand in response to elevated circulating glucose , the UPR is highly efficient in beta-cells , and adapts the ER loading and protein folding capacity to the insulin biosynthesis rate ( Back and Kaufman , 2012; Vander Mierde et al . , 2007 ) . High levels of insulin biosynthesis generate a chronic sub-threshold ER-stress that suppresses beta-cell proliferation ( Szabat et al . , 2016 ) , while induction of mild ER-stress in the context of hyperglycemia has been shown to induce beta-cell proliferation ( Sharma et al . , 2015 ) . These findings highlight the important link between insulin expression , UPR levels and beta-cell proliferation . Permanent neonatal diabetes mellitus ( PNDM ) is caused by mutations in genes controlling beta-cell development or functionality , and is usually diagnosed before 6 months of age ( Greeley et al . , 2011; Murphy et al . , 2008 ) . The development of efficient differentiation protocols has enabled the generation of beta-like cells in vitro from human pluripotent stem cells ( hPSC ) ( Pagliuca et al . , 2014; Rezania et al . , 2014; Russ et al . , 2015 ) . Combined with genome editing technologies , they make possible the establishment of in vitro models for detailed studies of pathogenic mechanisms of PNDM ( Balboa and Otonkoski , 2015; Saarimäki-Vire et al . , 2017; Shang et al . , 2014; Zhu et al . , 2016 ) . Insulin gene mutations are among the most common causes for PNDM globally ( Huopio et al . , 2016; Støy et al . , 2010 ) . Dominant negative heterozygous mutations that disrupt cysteine bridges within proinsulin lead to its misfolding , aggregation and accumulation in the ER ( Herbach et al . , 2007; Liu et al . , 2010a; Park et al . , 2010; Rajan et al . , 2010 ) . Accordingly , these high molecular weight proinsulin aggregates increase ER-stress and activate the UPR . Sustained UPR activation results in beta-cell dysfunction and the eventual onset of diabetes ( Colombo et al . , 2008; Liu et al . , 2010b ) . This phenomenon has been studied extensively in the Akita mouse model of diabetes , which carries a proinsulin cysteine disruption mutation ( C96Y ) that leads to mutant proinsulin accumulation in the ER , enlarged ER , reduction of secretory granules and mitochondrial swelling ( Izumi et al . , 2003; Kayo and Koizumi , 1998; Wang et al . , 1999; Yoshioka et al . , 1997; Zuber et al . , 2004 ) . Similar findings have been reported from the Munich mouse model carrying Ins2 C95S mutation ( Herbach et al . , 2007 ) . Although further studies suggested that unresolved UPR resulted in beta-cell apoptosis via Chop ( Ddit3 ) induction ( Oyadomari et al . , 2002 ) , significant differences in the number of apoptotic beta-cells were not observed in either model ( Herbach et al . , 2007 ) . To study the role of proinsulin cysteine disrupting mutations in human beta-cells , we derived human induced pluripotent cell lines ( iPSC ) from Finnish people carrying C96R ( the same cysteine as the Akita C96Y mutation ) and C109Y insulin mutations ( Huopio et al . , 2016 ) and differentiated them in vitro to beta-like cells . To circumvent the challenges associated with variable iPSC differentiation efficiency , we employed isogenic correction of the cell lines using CRISPR-SpCas9 ( Balboa and Otonkoski , 2015; Hsu et al . , 2014 ) . Single-cell RNA sequencing of the in vitro differentiated cells showed increased expression of ER stress-associated transcripts in the INS mutant cells , in concert with reduced proliferation . To further study the properties of these beta-like cells in vivo , we performed transplantation experiments into immunocompromised mice . Transplanted INS mutant beta-like cells exhibited lower insulin secretion and increased levels of ER-stress markers , together with reduced mTORC1 signaling and beta-cell size , without any apparent increase in apoptosis . Our findings suggest that PNDM-associated insulin mutations lead to inadequate development of a functional beta-cell mass . We studied two Finnish families with hereditary neonatal diabetes due to heterozygous insulin gene mutations ( Figure 1—figure supplement 1A ) . Both missense mutations affect cysteine residues , resulting in the disruption of the proinsulin inter-chain disulphide bonds A7-B7 ( mutation C96R ) and A20-B19 ( mutation C109Y ) ( Figure 1B ) which are essential for the proper folding and biological activity of the insulin molecule ( Chang et al . , 2003 ) . Affected individuals of these families become hyperglycemic 3–4 months after birth . Prior knowledge about the insulin gene mutations in the families enabled the neonatal genetic diagnosis in newborn siblings . Two affected newborns were monitored for the development of the disease . They were born at normal gestational age , presenting normal birth weight and fasting plasma c-peptide levels ( Figure 1—figure supplement 1A ) . Continuous subcutaneous glucose monitoring demonstrated the gradual deterioration of glycemia during the first months of life ( Figure 1A , Figure 1—figure supplement 1B ) . We derived induced pluripotent stem cells ( iPSC ) from the affected parents using retroviral and Sendai virus-mediated delivery of reprogramming factors OCT4 , SOX2 , KLF4 and MYC to dermal fibroblasts obtained from skin biopsies ( Figure 1—figure supplement 2A ) . Established iPSC were cultured for at least ten passages and then characterized . They expressed hallmark pluripotency markers , presented normal karyotypes and were able to spontaneously differentiate to the three germ layers in embryoid bodies ( Figure 1—figure supplement 2C–E ) . Sequencing of the insulin locus confirmed the presence in heterozygosis of the T to C change causing C96R mutation ( iPSC line HEL71 . 4 ) ( Figure 1C ) and G to A change causing C109Y mutation ( iPSC line HEL107 . 2 ) ( Figure 1—figure supplement 2B ) . Differentiation protocols devised to obtain beta-cells from human pluripotent stem cells ( PSC ) are not equally efficient across cell lines ( Nostro et al . , 2015; Sui et al . , 2018 ) . For disease modeling purposes , this variation in the differentiation from iPSC with different genetic backgrounds might obscure the phenotype caused by the mutation under study . Thus , we generated mutation-free isogenic iPSC lines from the patient-derived HEL71 . 4 line by correcting the C96R mutation using CRISPR/SpCas9 genome editing ( Figure 1C ) . Different guide RNAs were designed to target as close as possible to the C96R mutated codon in the insulin locus and tested in HEK293 cells . A guide RNA ( Ins8 ) cutting 9 base pairs away from the point mutation showed high cutting efficiency by T7 assay ( Figure 1—figure supplement 2F–G ) . We tested a mutation correction strategy based on homology directed repair ( HDR ) stimulated by the Ins8 guide RNA cutting activity using a single stranded DNA oligo ( ssODN ) of 70 bases as a donor template ( Figure 1C , Figure 1—figure supplement 2F ) . The correction donor ssODN contained the wild-type nucleotide in the mutation position ( wild-type Cys TGT instead of mutant Arg CGT ) and a synonymous point mutation in the next codon that created a restriction site for BsrGI ( Thr ACC instead of Thr ACA ) enabling rapid screening of the recombinant clones . The recombination of the ssODN was efficient in HEK293 cells ( Figure 1—figure supplement 2F–G ) . Next , we delivered the SpCas9-T2A-EGFP expressing plasmid , Ins8 gRNA and ssODN to patient iPSC HEL71 . 4 and sorted them based on EGFP+ expression . Sorted cells were pooled together , expanded and single-cell sorted to 96-well plates . We recovered 27 iPSC colonies out of 384 sorted single cells ( Cloning efficiency = 7 . 03% ) . We screened the clones by PCR followed by BsrGI restriction and found 17 recombinant clones ( 17/22 = 77% recombinant ) ( Figure 1—figure supplement 2H ) . Recombinant clones were further examined by Sanger sequencing ( Figure 1C ) . Clones F2 , F10 , G6 and A2 presented correction of the mutation with no chromosomal abnormalities ( Figure 1—figure supplement 2I ) . No mutations in putative off-targets were found in any clone ( Supplementary file 1 – Table 1 ) . For the differentiation and transplantation experiments in this study we have used the C96R mutant iPSC line HEL71 . 4 ( INS C96R ) , its corrected clones ( INS corrected ) and the C109Y mutant iPSC line HEL107 . 2 ( INS C109Y ) . We utilized a previously described protocol to differentiate beta-cells from iPSC ( Saarimäki-Vire et al . , 2017 ) ( Figure 1—figure supplement 3A ) . All INS mutant and INS corrected iPSC lines differentiated efficiently into definitive endoderm on day 3 ( Figure 1D ) and progressed to the pancreatic progenitor stage on day 12 , presenting abundant PDX1+ , NKX6 . 1+ and SOX9+ cells , with few NEUROG3+ and endocrine CHGA+ cells ( Figure 1—figure supplement 3B ) . At the pancreatic progenitor stage ( Stage 4 , 12 days of differentiation ) cells were dissociated and plated in suspension in a rotational platform , forming 3D islet-like aggregates that differentiated further to the endocrine lineage . After 30 days of differentiation ( final Stage 7 , S7 ) , we characterized the islet-like aggregates by cytometry . Differentiations from INS mutant and INS corrected iPSC yielded S7 aggregates that were composed of cells expressing PDX1 ( an average across all iPSC lines of 73 ± 12% SD PDX1+ cells , n = 10 ) and NKX6 . 1 ( an average across all iPSC lines of 37 ± 15% SD , NKX6 . 1+ , n = 23 ) ( Figure 1E ) ( Figure 1—figure supplement 3B–C ) . On average across all iPSC lines , 41% ( ±14% SD , n = 25 ) of the S7 cells were INS+ , with 18% ( ±6% SD , n = 23 ) of the cells expressing both INS and NKX6 . 1 , a sign of bona fide beta-cells ( Nostro et al . , 2015 ) ( Results pooled by genotype presented in Figure 1E–F , results of individual iPSC lines presented in Figure 1—figure supplement 3C ) . The number of INS+ cells at S7 was significantly different between INS mutant ( 31 ± 10% SD; n = 9 ) and corrected cells ( 49 ± 13% SD; n = 16 ) ( Figure 1E , Figure 1—figure supplement 3C ) . Subsequent analyses focused particularly on the INS+ cells at the Stage 7 of in vitro differentiation or 1 , 3 or 6 months after transplantation under the kidney capsule of immunodeficient NSG mice . Bulk RNA isolation has been traditionally used to study gene expression in cell samples using RT-qPCR or RNA-seq analysis . Using these methods , the identity of the cell source of a given RNA transcript is lost . Moreover , cell sample heterogeneity might introduce a lot of variation when studying a particular cell-specific transcript . Of note , independent iPSC to beta-like cell differentiation experiments vary in the yield of INS+ cells ( Figure 1—figure supplement 3C ) . The use of bulk RNA analysis to study the effects of the INS mutation in the beta-like cells might therefore obscure subtle transcriptional differences between the INS mutant and corrected cells . To overcome this problem we performed InDrop droplet-based single-cell RNA sequencing ( scRNAseq ) on differentiated islet-like cells ( Baron et al . , 2016; Klein et al . , 2015 ) . A total of 2 287 single cells from INS C96R and INS corrected Stage 7 islet-like aggregates were sequenced with a mean depth of 38 329 aligned reads/cell . Out of these cells , 2 171 ( 94 . 9% ) passed quality control . An average of 3 321 unique transcripts ( UMI ) and 1391 genes were detected per cell ( Figure 2—figure supplement 1A , Supplementary file 1 – Table 9 , Figure 2—source data 1 ) . Clustering analysis of all cells from both genotypes distinguished four different cell populations , expressing markers of beta-cells ( beta-like cells ) , endocrine progenitor cells ( progenitor cells ) , alpha cells ( alpha-like cells ) , and proliferating alpha cells ( proliferating alpha-like cells ) ( Figure 2A–B , Figure 2—figure supplement 1 , Figure 2—figure supplement 4A–B ) ( Supplementary file 1 – Table 2 ) ( Segerstolpe et al . , 2016 ) . The highest levels of INS expression were observed in the beta-like cluster ( Figure 2B–C ) . A resampling procedure confirmed the robustness of the clustering results ( Figure 2—figure supplement 4C–D ) . To confirm and to strengthen the identity of the sequenced cells , we compared our scRNA-seq data with a previous published human adult islet scRNAseq dataset generated with the InDrop single-cell platform ( Baron et al . , 2016 ) . Mapping of the individual cells to the Baron et al . ( 2016 ) dataset confirmed the identity of the beta-cells and was utilized to further refine the clustering of the beta-like and progenitor clusters , filtering out cells that mapped to the alpha cell cluster ( Figure 2—figure supplement 1B , Figure 2—figure supplement 4A–B ) . We performed differential gene expression analysis on both refined beta-like and progenitor clusters between INS C96R and INS corrected cells to determine the transcriptional changes caused by the INS C96R mutation at the single-cell level ( Figure 2D , Figure 2—figure supplement 1C ) ( Supplementary file 1 – Tables 3 and 4 ) . Mutant beta-like cells presented significant upregulation of chaperone genes HSPA5 , HSPA8 and HSP90B1 , disulfide isomerase PDIA6 , ubiquitin gene UBB and components of the ER associated degradation ( ERAD ) pathway HM13 , HERPUD1 , SEC61B and SDF2L1 . SDF2L1 expression is induced upon overexpression of the Akita mouse model C96Y mutant INS and it has been shown to interact with misfolded proinsulin and mediate its degradation ( Figure 2—figure supplement 1C ) ( Hartley et al . , 2010; Tiwari et al . , 2013 ) . These transcriptional changes indicate that the INS C96R mutation causes ER-stress and the subsequent activation of the UPR . Expression of the CDKN1C gene was also increased . The gene product , p57/Kip2 , is an important inhibitor of human beta-cell proliferation ( Avrahami et al . , 2014 ) , suggesting reduced proliferation of the INS C96R cells . Interestingly , the INS gene was upregulated in INS C96R cells , together with other insulin secretion related genes ( CPE , SCGN , DLK1 ) , a phenomenon previously described in young Akita mice ( Oyadomari et al . , 2002 ) ( Figure 2D , Figure 2—figure supplement 1C , Supplementary file 1 – Tables 3 and 4 ) . Conversely , genes encoding mitochondrial respiratory chain subunits ( MT-CO1 , MT-CO2 ) , immediate early gene IER2 , transcription factors PAX6 and RFX6 , and the mTOR regulator LAMTOR5 were significantly downregulated in INS C96R beta-like cells ( Figure 2D , Figure 2—figure supplement 1C ) . Similar to the beta-like cluster cells , INS C96R progenitor cells presented increased INS transcription and downregulated immediate early genes ( FOS , EGR1 , IER2 , JUN ) , mitochondrial genes ( MT-CO1 , MT-CO2 , MT-CYB ) , pancreatic transcription factors ( FOXA2 , GATA6 , RFX6 , PAX6 ) and progenitor proliferation-associated ID genes ( ID2 , ID3 ) ( Figure 2D , Figure 2—figure supplement 1C ) . These results indicate that the ER-stress caused by the INS C96R proinsulin results in downregulation of genes associated with beta-cell proliferation and function . Gene set enrichment analysis ( GSEA ) was performed on the differentially expressed genes in the beta-cell cluster . Ribosomal translation , insulin synthesis and processing , diabetes pathways and ATF6-controlled UPR chaperones gene sets were overrepresented among the genes upregulated in INS C96R cells . On the contrary , the genes downregulated in INS C96R were enriched in gene sets related to EGF signaling and respiratory electron transport , processes important for beta-cell development and function . In the progenitor cluster , GSEA showed overrepresentation of GLIS3 targets and peptide chain elongation in the INS C96R upregulated genes , while SRC , TGF-beta and EGF signaling gene sets were enriched in the downregulated genes ( Supplementary file 1 – Table 5 , Figure 2—source data 2 ) . Mitochondrial oxidative phosphorylation , regulation of macromolecule metabolic processes and mitotic cell cycle Gene Onthology ( GO ) biological processes were enriched among the genes downregulated in the beta-like cluster INS C96R cells . Intracellular transport , protein targeting to ER and protein folding GO terms were enriched among the upregulated genes . Similar GO terms were enriched in the down and upregulated genes of the progenitor cluster ( Supplementary file 1 – Table 6 , Figure 2—source data 2 ) . We hypothesized that the endocrine progenitor population identified by the clustering is in the process of differentiating towards endocrine ( mainly alpha-like and beta-like ) cells . To investigate the hierarchy of differentiation events , we performed pseudotime analysis to infer the possible trajectories of these populations ( Qiu et al . , 2017 ) . The analysis algorithm imposed a branched trajectory , in which the progenitor cluster cells were split in two distinct progenitor branches that merge to give rise to the beta-like cell branch ( Figure 2E , Figure 2—figure supplement 2A ) . The differentially expressed genes between these two progenitor populations suggest that Progenitor one may represent early endocrine progenitors , expressing higher levels of HES1 , PAX6 , PROX1 and other genes related to cytoskeleton regulation , cell adhesion , migration and TGF-beta signaling modulation ( Figure 2E , Figure 2—figure supplement 2A–C , Supplementary file 1 – Table 7 ) . The Progenitor two population presents higher levels of NEUROD1 , CHGA and FEV , suggesting a more advanced , already granulated , late endocrine progenitor identity . Analysis of differentially regulated genes along the pseudotime trajectory showed that the expression dynamics of INS and other proliferation ( ID2 , CDKN1C ) and UPR-related genes ( HSPA5 , PDIA6 , SDF2L1 ) were significantly different between INS C96R and INS corrected cells ( Figure 2F , Figure 2—figure supplement 2D , Supplementary file 1 – Table 8 ) . This illustrates the significantly higher upregulation of ER-stress markers , INS and CDKN1C cell cycle inhibitor earlier in the differentiation trajectory of INS C96R progenitors to beta-like cells , indicating the immediate negative effect of misfolded proinsulin in the recently committed beta-like cells upon INS expression . To determine the expression of different ER-stress markers at the protein level , we performed immunohistochemistry of stem cell-derived Stage 7 beta-like cells cultured in vitro . Immunoreactivity for ER-stress associated proteins BIP ( HSPA5 ) , GRP170 ( HYOU1 ) and MANF was significantly increased in the INS C96R beta-like cells as compared with the corrected cells ( Figure 3A–B , Figure 5—figure supplement 1C ) . Conversely , the number of KI67+ and PCNA+ proliferative insulin-positive cells was significantly reduced in the INS C96R cells ( Figures 3A–B and 6E–F ) . We quantified single cell levels of INS immunostaining intensity using flow cytometry and from immunohistochemistry preparations ( Figure 1—figure supplement 3D–E ) . We did not find significant differences in the signal intensity between INS mutant and INS corrected cells , indicating that impaired proliferation of INS mutant beta-like cells is likely the major contributor to the reduced percentage of INS+ cells observed by cytometry . qRT-PCR analysis confirmed the significantly increased expression of ER-stress markers BIP , sXBP1 , MANF and GRP170 , in line with the findings of immunostainings and scRNAseq results ( Figure 3C ) . Other ER-stress associated genes , including CHOP ( DDIT3 ) , which has been reported to be upregulated in Akita mice ( Oyadomari et al . , 2002 ) , were not differentially expressed at the mRNA level between INS C96R and corrected cells ( Figure 3C ) . However , the expression of CHOP and ATF transcription factors might be also regulated post-transcriptionally ( Cnop et al . , 2017 ) . To determine if the INS C96R beta-like cells under ER-stress are more sensitive to apoptosis , TUNEL assays were performed on S7 aggregates . The ratio of TUNEL+/INS+ cells was similar in INS C96R and INS corrected cells in the basal conditions ( about 1% ) ( Figure 3D ) . Induction of additional ER-stress by treatment with brefeldin A ( BFA ) , thapsigargin ( TGA ) or tunicamycin ( TM ) resulted in increased apoptosis for both genotypes . BFA , but not the other stressors , induced a significantly higher level of apoptosis in the INS C96R cells compared to mutation corrected cells ( Figure 3D ) . Thus , INS C96R beta-like cells present higher sensitivity to apoptotic cell death induced by further increasing ER-stress with BFA treatment . The insulin secretory responses of S7 beta-like cells were assessed by sequential static incubations in the presence of 1 µM forskolin to increase cAMP levels . While the response to high glucose alone was minimal ( 1 . 15- and 1 . 05-fold for INS C96R and corrected cells , respectively ) , tolbutamide and KCl triggered robust insulin secretion ( 2 . 5- and 3 . 2-fold for both INS C96R and corrected cells ) . The fractional release of insulin ( as % of content ) was not significantly different between INS C96R and corrected cells ( Figure 3E ) . However , the insulin content of the INS C96R S7 islet-like aggregates was significantly reduced ( 5 . 2-fold lower than in corrected ) ( Figure 3F ) . The ratio of human proinsulin content to human insulin content was not significantly different between INS C96R and corrected cells at this stage ( Figure 3—figure supplement 1A ) , but the total proinsulin content was significantly reduced in INS C96R ( Figure 3—figure supplement 1B ) as well as the ratio of secreted proinsulin in maximal stimulation with KCl ( Figure 3—figure supplement 1C ) . Taken together , these results show that a ) Stage 7 beta-like cells are not functionally mature enough to respond to glucose stimulation alone , and b ) the INS C96R beta-like cells maintain their responsiveness to pharmacological stimulation despite markedly decreased proinsulin and insulin content . Islet-like cell clusters from INS C96R , INS C109Y and corrected cell lines were transplanted under the kidney capsule of NSG mice to study the effect of the insulin mutation on beta-like cells in vivo ( Figure 4A ) . Graft functionality was tracked from 1 to 6 months after transplantation by measuring circulating human C-peptide in plasma samples from randomly fed mice . Mice carrying INS C96R and INS C109Y grafts presented significantly lower levels of human C-peptide than mice carrying INS corrected grafts ( Figure 4B ) . The levels of C-peptide increased from the 2 month time point onwards for the INS corrected grafts ( p < 0 . 0001 , One-way ANOVA ) , while no change was observed in the INS C96R and INS C109Y grafts ( Figure 4B ) . The increase in circulating human C-peptide in INS corrected grafts can be attributed to further expansion and maturation of the transplanted beta-like cells . Additionally , differentiation of co-transplanted pancreatic progenitors to the beta-cell lineage is likely to contribute as well . Intraperitoneal glucose tolerance test performed at 3 or 6 months after grafting showed that INS corrected grafts presented higher insulin secretion levels than INS C96R and INS C109Y grafts , both upon fasting and after glucose injection ( Figure 4C–D ) . Mice transplanted with INS C96R grafts presented a trend towards elevated ratio of circulating proinsulin to circulating C-peptide ( Figure 4E ) , a phenomenon that has been previously described in humans with mutant insulin diabetes ( Liu et al . , 2010b; Liu et al . , 2015; Rajan et al . , 2010 ) . Grafts were retrieved at 1 , 3 and 6 months and examined by immunohistochemistry . Almost all INS C96R and INS C109Y INS+ cells in 3-month-old grafts presented higher levels of immunoreactivity for proinsulin in comparison to INS corrected grafts ( Figure 4F–H ) . A similar pattern was observed in the S7 cells , as well as 1 month and 6 month grafts ( Figure 5—figure supplement 1D ) . Proinsulin immunoreactivity occupied most of the cytoplasm in INS C96R cells , while it was punctate in INS corrected cells . This indicates defective proinsulin transport , resulting in its accumulation ( Figure 4G ) . Immunohistochemistry for ER-stress markers revealed profound differences in the expression of BIP , MANF and GRP170 at 1 , 3 and 6 months , with remarkably increased levels in the INS C96R and C109Y cells compared with INS corrected cells ( Figure 5A–E and Figure 3—figure supplement 1E for 3 months grafts ) ( Figure 5—figure supplement 1 for 1 and 6 month grafts , Figure 5—figure supplement 1C shows the quantification of BIP intensity on a per cell basis ) . The percentage of INS C96R beta-like cells expressing high levels of BIP increased dramatically after transplantation ( Figure 5F , Figure 5—figure supplement 1C ) , indicating a progressive aggravation of ER-stress in vivo . Mild UPR increase has been shown to induce beta-cell proliferation in hyperglycemic conditions ( Sharma et al . , 2015 ) . We examined the relationship between the proliferation marker PCNA and the ER-stress marker BIP , but we did not find a consistent correlation ( Figure 5—figure supplement 2A ) . We performed a similar analysis for proliferation marker KI67 and the intensity of INS and PROINS immunostaining , since decreased levels of INS expression have been associated with increased beta-cell proliferation ( Szabat et al . , 2016; Xin et al . , 2018 ) . KI67+ cells presented reduced INS immunostaining intensity across genotypes and time points , suggesting a possible correlation between reduced INS expression and increased proliferation ( Figure 5—figure supplement 2B–C ) . Similar to proinsulin , MANF immunoreactivity occupied most of the cytoplasm in INS C96R cells , indicating accumulation in a distended ER ( Figure 5D ) ( Lindahl et al . , 2014 ) . To detect if MANF was released from the cells under ER-stress , we measured it in vitro and in transplanted animals . Overnight MANF secretion tended to be elevated in the INS C96R in vitro cells , but this difference was not significant ( Figure 3—figure supplement 1D ) . Circulating human MANF levels were below detection limits in the blood of transplanted animals . Apoptotic INS+ cells assayed by TUNEL or CASP3 staining were very rare at all time points ( <0 . 5% at 3 months , not significantly different ) ( Figure 3—figure supplement 1F ) , suggesting that aggravated ER-stress does not lead to increased apoptosis of the INS C96R or INS C109Y cells in vivo . The proportion of INS+ cells was significantly reduced in INS C96R grafts at 3 months . This finding was further confirmed by immunohistochemistry for C-peptide ( CPEP ) ( Figure 6A–B ) . On the contrary , the glucagon ( GCG ) positive cell compartment was increased ( Figure 6A–B ) . We examined other endocrine hormones and found that there were no significant differences in the ratios of somatostatin ( SST ) positive or pancreatic polypeptide ( PP ) positive cells . However , the percentage of ghrelin ( GHRL ) positive cells was significantly increased in INS C96R grafts ( Figure 6A–B ) . Double hormone positive INS+GCG+ , CPEP+GCG+ , INS+SST+ , INS+GHRL+ or CPEP+PP+ cells were very rare ( <2% ) and not significantly different between INS C96R and corrected grafts ( Figure 6C ) . The proportion of cells co-expressing PDX1 and C-peptide was significantly reduced in 3-month-old grafted INS C96R beta-cells ( Figure 6A , D ) . Proliferation of INS C96R beta-like cells was significantly impaired at S7 in vitro ( Figures 3B and 6E–F ) . Following transplantation , this difference gradually disappeared as the proliferation of INS corrected cells decreased ( Figure 6D–F ) . This may recapitulate human postnatal beta-cell development , with a postnatal peak of proliferation that declines rapidly during the first two years of life ( Gregg et al . , 2012; Meier et al . , 2008 ) . mTORC1 signaling is required for the proper postnatal growth and maturation of beta-cells ( Ni et al . , 2017; Sinagoga et al . , 2017 ) . Single-cell RNA-seq revealed downregulation of genes involved in proliferation , oxidative phosphorylation and mTORC1 regulation ( LAMTOR5 ) in INS C96R beta-like cells . These processes are in part regulated by mTORC1 signaling , suggesting that it could be dysregulated in INS C96R cells . We found that immunoreactivity for phosphorylated S6 , a central downstream signaling target of mTORC1 , was significantly reduced in INS+ 3 month-old INS C96R grafts ( Figure 7A–C ) . Consistent with reduced mTORC1 signaling , INS+ cell size was also significantly reduced in the INS C96R grafts ( Figure 7B–C ) . Based on the findings of the sc-RNAseq analysis on the in vitro beta-like cells , we examined the expression of mitochondria respiratory chain subunits in 3 month old grafts ( Figures 2D and 7D ) . Immunostaining for MT-CO1 , MT-CO2 and TOM20 revealed an altered , more condensed and globular mitochondrial morphology as well as decreased immunofluorescence intensity in INS+ mutant cells ( Figure 7D–E ) . Interestingly , INS staining intensity was also decreased at this point , suggesting a reduced INS protein content in the grafted beta-like cells ( Figure 7E ) . We interrogated the impact of insulin gene mutations on beta-cell development by generating a model based on genome edited patient-derived iPSC ( Balboa and Otonkoski , 2015; Saarimäki-Vire et al . , 2017; Shang et al . , 2014; Zhu et al . , 2016 ) . iPSC provide a novel possibility to study mechanisms of beta-cell dysfunction using patient cells . However , this approach is still being developed and presents important caveats that need to be taken into account . The in vitro-differentiated beta-like cells are functionally immature , which limits their usefulness particularly to model metabolically controlled insulin secretion . Also , the variability of differentiation efficiency across cell lines has been a confounding factor . Genome editing technologies have partially solved this problem by enabling the generation of isogenic cell lines . We combined this approach with single-cell transcriptomics , enabling the detection of ER-stress early during the development of beta-cells as the consequence of accumulation of misfolded mutant proinsulin . Impaired beta-cell proliferation was the most striking phenotype of the mutated cells . Following transplantation , the INS mutant cells presented increased proinsulin accumulation and further increased signs of ER-stress , associated with reduced PDX1 expression and reduced beta-cell size , as well as mitochondrial alterations . Many of these features are attributable to the observed decreased mTORC1 signaling . Since populations of endocrine cells generated with the available differentiation protocols ( Pagliuca et al . , 2014; Rezania et al . , 2014; Russ et al . , 2015 ) are heterogeneous , single-cell RNA-seq methods provide a robust approach to identify and characterize specifically bona-fide beta-like cells ( Carrano et al . , 2017 ) . Single-cell RNA-seq has been recently used to study both mouse and human islets in different stages of development and disease ( Baron et al . , 2016; Segerstolpe et al . , 2016; Xin et al . , 2018; Zeng et al . , 2017 ) . These datasets can now be compared with in vitro hPSC-derived beta-cells to verify the identity of the endocrine cells , as we present here ( Figure 2—figure supplement 1B ) . Moreover , single-cell transcriptomic data enables the use of pseudotime analysis to infer differentiation trajectories . This is particularly useful to elucidate hierarchical and temporal relationships between cell types in developing tissues and in vitro differentiation experiments ( Qiu et al . , 2017 ) . Using this approach , we identified a differentiation trajectory with two distinct endocrine progenitor stages , an earlier progenitor population marked by PROX1 , HES1 and ID2 , and a more differentiated progenitor population marked by CHGA , NKX2 . 2 , FEV , NEUROD1 and MNX1 . These two types of progenitors likely represent sequential stages in the differentiation to beta-like cells , which are depicted as independent populations due to the branching trajectory imposed by the analysis algorithm . Interestingly , two putative endocrine progenitor populations have been also observed in a recent study where the stages of hPSC differentiation towards beta-cells were examined by single-cell qRT-PCR ( Petersen et al . , 2017 ) . Further research will be required to determine if these different endocrine progenitor populations have in vivo counterparts in human embryonic pancreas or rather represent an in vitro artifact . Diabetes caused by misfolded proinsulin has been studied extensively in the Akita mouse model carrying the INS C96Y mutation ( Izumi et al . , 2003; Oyadomari et al . , 2002; Wang et al . , 1999 ) and in the Munich mouse model carrying the INS C95S mutation ( Herbach et al . , 2007 ) . These mice become progressively diabetic 4 to 8 weeks after birth , presenting a reduced beta-cell mass which has commonly been attributed to increased apoptosis ( Oyadomari et al . , 2002 ) . However , significantly increased beta-cell apoptosis was not detected in some of the studies published with these models ( Herbach et al . , 2007; Izumi et al . , 2003 ) . A closer examination of the Akita mouse postnatal development has revealed reduced proliferation and impaired function of neonatal beta-cells , in the absence of increased apoptosis ( Riahi et al , accompanying paper ) . These findings are well in line with the results of our study , since we could not detect increased apoptosis in the INS mutant cells at any point , despite the elevated ER-stress manifested by the high expression of BIP , MANF and GRP170 . A potential caveat of our model is that grafted mice employed in this study were normoglycemic . It remains to be elucidated if additional ER-stress imposed by increased insulin demand in a diabetic environment would eventually lead to increased apoptosis of human INS mutant beta-cells in vivo . Disruption of mTORC1 signaling is a crucial link between aggravated ER-stress and defective beta-cell expansion in the neonatal period ( Sinagoga et al . , 2017 ) . Inhibition of mTORC1 signaling in Raptor KO mouse beta-cells leads to impaired postnatal beta-cell growth , function and mitochondrial function ( Ni et al . , 2017 ) . Restoration of the mTORC1 signaling was sufficient to rescue the beta-cell proliferation defect in Akita mice ( Riahi et al . , accompanying paper ) . All these results are in agreement with our findings , where the INS mutant beta-cells show reduced mTORC1 signaling ( decreased LAMTOR5 expression , reduced S6 phosphorylation ) and would explain their diminished cell size , altered mitochondria and INS protein content . ER-stress and mitochondrial function are closely associated in the etiology of diverse diseases . For example , ER-stress signaling pathway component PERK has been shown to regulate mitochondrial morphology ( Lebeau et al . , 2018 ) . In the context of ER-stress induced by misfolded proinsulin , Akita mouse beta-cells present increased mitochondrial dysfunction , with mitochondrial fragmentation and reduced respiration ( Mitchell et al . , 2013 ) . We observed similar mitochondrial defects by scRNA-seq and immunostaining INS mutant beta-cells in vitro and in vivo . In both rodents and humans , beta-cell proliferation peaks in the neonatal period ( Finegood et al . , 1995; Gregg et al . , 2012; Zeng et al . , 2017 ) . A recent scRNA-seq study shows that neonatal proliferative mouse beta-cells are characterized by high mitochondrial membrane potential , expression of immediate early genes Fos , Egr1 , Jun and Srf and increased PI3K-mTOR signaling ( Zeng et al . , 2017 ) . Similar scRNA-seq analysis of human adult beta-cells has shown that beta-cells with high expression of UPR and lower expression of insulin are more prone to proliferate ( Xin et al . , 2018 ) . These cells are more metabolically active , as reflected by their higher expression of glycolytic pathway , tricarboxylic acid cycle and electron transport chain genes . Interestingly , our scRNA-seq analysis shows downregulation of similar gene sets , including mitochondrial respiratory chain subunits ( MT-CO1 , MT-CO2 ) , immediate early genes ( FOS , EGR1 and IER ) and LAMTOR5 in INS C96R beta-like cells ( Figure 2D , Figure 2—figure supplement 1C ) . Furthermore , basic helix-loop-helix ( bHLH ) transcription factors ID1 , ID2 and ID3 were downregulated in the INS C96R cells . These genes , regulated by BMP signaling , are important for the proliferation and differentiation of beta-cells ( Hua and Sarvetnick , 2007 ) . Taken together , our single-cell transcriptomic results provided a basis for the reduced proliferation and altered mitochondria of INS mutant beta-like cells , that was later confirmed in vitro and in vivo . MANF is an ER-stress induced prosurvival factor whose role in reducing UPR activation is critical for mouse beta-cell development and postnatal beta-cell mass expansion and maintenance ( Danilova et al . , 2018; Lindahl et al . , 2014 ) . Manf KO mice presented reduced beta-cell mass and beta-cell proliferation at birth , without increased apoptosis , highlighting the importance of maintaining a physiological level of ER-stress for perinatal beta-cell proliferation . Thus , it is possible that the lack of ER-stress-induced beta-cell apoptosis is in part explained by the high MANF expression triggered by the misfolded proinsulin . Interestingly , we detected increased apoptosis in the INS C96R beta-like cells after inducing additional ER-stress with brefeldin A , but not with other stressors ( Figure 3D ) . This difference could be the result of the particular brefeldin A mechanism of action , the inhibition of protein transport from ER to Golgi apparatus , which results in additional ER overloading ( Helms and Rothman , 1992 ) . Impaired beta-cell proliferation resulted in skewed endocrine cell proportions with less insulin-positive and more glucagon- and ghrelin-positive cells . Ghrelin-positive cells represent a transient fetal endocrine cell population ( Arnes et al . , 2012 ) , rare in adult human islets ( Segerstolpe et al . , 2016 ) . Mouse adult beta-cells have been shown to lose their identity and misexpress ghrelin upon Pax6 deletion , together with an expansion of the islet alpha-cell population ( Swisa et al . , 2017 ) . Interestingly , we detected reduced expression of PAX6 in vitro and lower PDX1 expression in vivo in the INS mutant cells . Therefore , a small percentage of the cells could have differentiated into alternative cell identities although our results indicate that impaired proliferation and growth is likely to be the main cause for the demise of the INS mutant cells . A potential limitation in the analysis of grafted cells is the extensive ischemia induced cell death upon transplantation ( Faleo et al . , 2017 ) , which could skew cell type proportions and increase graft to graft variability . Overall , our results indicate that misfolded proinsulin triggers ER-stress concomitantly with INS expression , affecting the development of the INS mutant beta-cells by impairing their proliferation without increased apoptosis . Decreased proliferation results in the reduced percentage of INS+ cells observed in vitro and in vivo in the INS mutant cells . Elevated ER-stress leads to reduced mTORC1 signaling and altered mitochondria , which are critical for beta-cell proliferation and function . Importantly , our findings demonstrate that INS mutations leading to neonatal diabetes are already pathogenic during pancreatic development due to failure of neonatal beta-cell expansion . This could theoretically open up new possibilities for the treatment of mutant insulin-associated diabetes through transient stimulation of mTORC1 , but this treatment would have to be applied within the neonatal period . Our study extends the observations from the diabetic Akita model into human diabetes and further emphasizes the role of ER-stress in controlling beta-cell proliferation . These findings may be of relevance for the risk of developing type two diabetes later in life , since the functional beta-cell reserve is established in the perinatal period ( Gregg et al . , 2012; Meier et al . , 2008 ) . Dermal fibroblasts obtained from a skin biopsy were reprogrammed using retroviral delivery of the OCT4 , SOX2 , MYC and KLF4 transcription factors , as described elsewhere ( Toivonen et al . , 2013 ) . hiPS cells were cultured on Matrigel ( BD Biosciences ) -coated plates with E8 medium ( Life Technologies , A1517001 ) and passaged using 5 mM EDTA ( Life Technologies , 15575–038 ) as a dissociation agent . For pluripotency characterization , cells were spontaneously differentiated using embryoid-body assay ( Balboa et al . , 2017 ) . Karyotype analyses based on chromosomal G-banding were performed at Yhtyneet Medix Laboratories , Helsinki , Finland . All hiPSC lines were authenticated using Sanger sequencing for the insulin gene mutations and were negative for mycoplasma contamination test . Guide RNAs ( gRNAs ) targeting the insulin locus were designed using web-based tool http://crispr . mit . edu ( Hsu et al . , 2013 ) , selecting for guide RNAs with high quality scores to avoid possible off-targets . Transcriptional units for gRNA expression were prepared by PCR ( Balboa et al . , 2015 ) and transfected to HEK293 cells together with WT SpCas9 expressing plasmid CAG-Cas9-T2A-EGFP-ires-puro ( Addgene plasmid # 78311 ) . Cutting efficiency was determined using T7 endonuclease I ( New England Biolabs ) assay ( PCR primers: hIns_1229_Fw: GGGTGACCCTCCCTCTAACC , 3’Ins-Rv: TCAGCGGCCGCTCCACAGGGACTCCATCAGA ) . gRNA Ins8 ( CTGGTAGAGGGAGCAGATGC-TGG ) was found to cut with high efficiency 9 bp away from the INS C96R mutation . A correction strategy was devised , based on the recombination with a 70 bases single stranded DNA oligo ( ssODN ) , complementary to the Ins8 gRNA . This ssODN corrects the C96R mutation and introduces a synonymous coding nucleotide change , disrupting the protospace adjacent motif ( PAM ) and creating a novel BsrGI restriction site that facilitates the screening of recombinant clones ( Figure 1—figure supplement 2 ) ( ssODN_Ins8_BsrGI:GCAGAAGCGTGGCATTGTGGAACAATGCTGTACAAGCATCTGCTCCCTCTACCAGCTCGAGAACTACTGC ) . For correction of the mutation in the patient-derived iPSC , two million HEL71 . 4 cells were electroporated with 6 μg of CAG-Cas9-T2A-EGFP-ires-puro endotoxin-free plasmid , 500 ng of gRNA-PCR Ins8 product and 6 μg ssODN ( Neon Transfection System , 1100 V , 20 ms , two pulses , ThermoFisher ) . Cells were immediately plated onto Matrigel-coated plates containing E8-medium with 5 μM ROCK inhibitor ( Y-27632 2HCl , Selleckchem ) . Cells positive for GFP fluorescence were pool-sorted 48 hr later and expanded . Single-cell sorting was performed as previously described ( Saarimäki-Vire et al . , 2017 ) . Plasmids and detailed protocols have been deposited on Addgene ( http://www . addgene . org/78311/ ) . For differentiation of iPSC to beta-cells a modification of previously published protocols was used ( Pagliuca et al . , 2014; Rezania et al . , 2014 ) , as described previously ( Saarimäki-Vire et al . , 2017 ) ( Figure 1—figure supplement 3A ) . Cells were dissociated with 5 mM EDTA treatment for 10 min and seeded at 1 . 5–2 million cells/3 . 5 cm well on Matrigel-coated plates with E8 medium containing 5 μM ROCK inhibitor ( Y-27632 2HCl , Selleckchem ) . Differentiation was started 24 hr later and proceeded through seven stages differentiation protocol ( Stages 1 to 4 in adherent culture and stages 5 to 7 in suspension culture ) : To improve reproducibility and standardize the differentiation , small molecule compounds were prepared in batches of stage-specific supplements , enabling rapid differentiation media preparation and consistency between experiments: Cytometry for definitive endoderm marker CXCR4+ was performed as previously described ( Saarimäki-Vire et al . , 2017 ) . For intracellular antigen pancreatic marker cytometry of Stage 4 or Stage 7 , cells were dissociated with TrypLE for 5–10 min at 37°C and resuspended in 5% FBS-containing PBS . Cells were fixed and permeabilized using Cytofix/Cytoperm ( 554714 , BD Biosciences ) as recommended by manufacturer . Primary or conjugated antibodies were incubated with the cells overnight at 4°C in Perm/Wash buffer ( 554714 , BD Biosciences ) containing 4% FBS . Cells were washed 2x with Perm/Wash buffer and analysed using FACSCalibur cytometer ( BD Biosciences ) and FlowJo software ( Tree Star Inc . ) . Total RNA was isolated using NucleoSpin Plus RNA kit ( Macherey-Nagel ) . SimpliNano ( General Electric ) spectrophotometer was used to measure RNA quality and concentration . A total of 1 . 5 μg RNA was denatured at 65° C for 1 min and reverse transcribed ( RT ) with 0 . 5 μL Moloney murine leukemia virus ( MMLV ) reverse transcriptase ( M1701 , Promega ) , 0 . 2 μL Random Primers ( C1181 , Promega ) , 1 μL Oligo ( dT ) 18 Primer ( SO131 , ThermoFisher ) and 0 . 5 μL Ribolock RNAse inhibitor ( EO0382 , ThermoFisher ) for 90 min at 37° C . qRT-PCR reactions were prepared with 50 ng of retrotranscribed RNA were amplified with 5 μL of forward and reverse primer mix at 2 μM each using 5x HOT FIREPol EvaGreen qPCR Mix Plus ( no ROX ) in a final volume of 20 μL . QIAgility ( Quiagen ) liquid handling system was used for pipetting the reactions into 100 well disc that were subsequently sealed and run in Rotor-Gene Q ( Qiagen ) with a thermal cycle of 95° C for 15 min , followed by 40 cycles of 95° C , 25 s; 57° C , 25 s; 72° C , 25 s , followed by a melting step . Relative quantification of gene expression was analysed using ΔΔCt method , with cyclophilin G ( PPIG ) as endogenous housekeeping control gene . RT-reaction without template was used as negative control and exogenous positive control was used as a calibrator . Expression levels were normalized in each sample by the percentage of INS+ cells determined by cytometry and presented as relative to INS corrected cells . See Key Resource Table for primer list . One hundred manually picked Stage 7 islet-like aggregates were incubated in full Stage 7 media with the corresponding concentration ER-stress inducers . Brefeldin A ( B5936 , Sigma-Aldrich ) was used at 1 μg/mL in DMSO for 24 hr . Thapsigargin ( T9033 , Sigma-Aldrich ) and tunicamycin ( T7765 , Sigma-Aldrich ) were used at 1 μM and 5 μg/mL respectively for 48 hr . DMSO was used as a vehicle control at 5 μL/mL . Aggregates were collected and PFA-fixed for immunohistochemistry after treatment . Stage 7 islet-like aggregates were sampled in groups of 100 to 1 . 5 mL tubes . They were washed twice with Krebs buffer containing no glucose and then transferred to 12-well plate placed in a rotating platform for incubation in 3 . 3 mM glucose-containing Krebs buffer for 1 hr ( low glucose ) . This was performed twice . Then aggregates were incubated sequentially in 3 . 3 mM glucose , 20 mM glucose , 20 mM glucose + 100 μM tolbutamide and 3 . 3 mM + 30 mM KCl , for a period of 30 min , with two washes with 1 mL Krebs buffer containing no glucose between treatments . 500 μL of supernatant from each treatment incubation were collected , centrifuged to remove possible cells in suspension and stored at −80°C for ELISA-based determination of human insulin concentration . After the last treatment incubation , samples were retrieved and lysed in acid ethanol for determination of total insulin content and DNA content . Stimulated insulin secretion results are presented as fractional release of total human insulin content after cell mass normalization using total DNA content . NOD-SCID-gamma ( NSG ) ( Jackson Laboratories; 005557 ) mice were housed at Biomedicum Helsinki animal facility , on a 12 hr light/dark cycle and food ad libitum . Transplantations were performed on 3- to 12-month- old mice as described previously ( Saarimäki-Vire et al . , 2017 ) . Briefly , aggregates equivalent to approximately 5 million cells were loaded on a PE-50 tubing and transplanted under the kidney capsule . Mice were anesthetized with isoflurane . Carprofen ( Rimadyl , 5 mg/kg , subcutaneously , Prizer , Helsinki , Finland ) and Buprenorphine ( Temgesic , 0 , 05–0 , 1 mg/kg , subcutaneously , RB pharmaceuticals Lmt , Berkshire , UK ) were used as analgesics during the operation and in the following day . Mouse blood samples were collected monthly from the saphenous vein using heparinized capillary tubes . Blood plasma was separated by centrifugation ( 5000 rcf , 5 min , RT ) . IPGTT was performed after 6–8 hr fast . 2 g glucose/kg of body weight was injected intraperitoneally in the form of a 30% glucose solution in water . Blood glucose levels were measured with glucometer ( OneTouch Ultra , Lifescan , Milpitas; USA ) at 0 , 20 , 40 and 60 min after glucose injection . Blood samples for measuring human c-peptide levels were taken before , and 40 min after glucose injection . Human c-peptide and proinsulin levels were measured from plasma samples and cell supernatants with Ultrasensitive C-PEPTIDE ELISA ( Mercodia , Sweden ) and PRO-INSULIN ( Mercodia , Sweden ) according to manufacturer's instructions . Human MANF levels were measured using in-lab ELISA ( Galli et al . 2016 ) . Cells on adherent cultures were fixed in 4% PFA for 15–20 min , permeabilized with 0 . 5% triton-X100 in 1x PBS , blocked with UltraV block ( ThermoFisher ) for 10 min and incubated with primary antibodies diluted in 0 . 1% Tween in 1 x PBS at 4°C overnight . Cells were washed with 1 x PBS , incubated with secondary antibodies diluted in 0 . 1% Tween in PBS . Same procedure was used for whole-mount staining of Stage 7 cell aggregates . For paraffin embedding , Stage 7 cell aggregates were fixed with 4% PFA at RT overnight and briefly stained with Eosin . After this , they were embedded in low-melting Agarose ( Sigma-Aldrich ) and transferred to paraffin blocks . Grafts were retrieved after IPGTT , dissected and fixed with 4% PFA in RT overnight and placed into cassettes and processed for tissue transfer and paraffin embedding . Paraffin blocks were cut into 5 μm sections . For immunohistochemistry , slides were deparaffinised and antigen retrieval was performed by boiling slides either in 1 mM EDTA or 0 . 1 M citrate buffer . Blocking and incubation with primary and secondary antibodies were done as described for fixed cells above . For TUNEL analysis , paraffin sections were processed with In Situ Cell death Detection Fluorescein kit ( Roche , #11684795910 ) according to manufacturer's instructions . See Key Resource Table for list of antibodies ( ICC: immunocytochemistry on fixed cells; IHC: immunohistochemistry on paraffin sections; FC: flow cytometry ) Immunofluorescence stainings of adherent cells were imaged with EVOS inverted microscope ( LifeTechnologies ) . Paraffin sections and whole-mount stainings were imaged with Zeiss Axio Observer Z1 with Apotome two and processed with ZEN2 software blue edition . To ensure reliable quantification of the immunostainings , all paraffin sections were stained simultaneously and imaged on the same session with the same microscope parameters . Image quantifications were performed blindly using Fiji software ( Schindelin et al . , 2012 ) . Quantification of individual cell immunostaining intensity was performed manually using Fiji ROI Manager and Multi Measure tools . Fiji pixel intensity and Cell counter tools were used to score percentages of cells positive , negative , low or high across the different immunostainings . Statistical analyses were performed with GraphPad Prism ( version 7 . 0 c , GraphPad Software ) . Data were tested for normal distribution using Shapiro-Wilk normality test . Normally distributed data were analyzed to compare the means of two samples using unpaired two-tailed Student’s t test , with Welch’s correction in the case of samples with unequal variance as determined by F test . One-way ANOVA with multiple comparison Tukey test was used to compare the means of more than two samples . When the data groups were not normally distributed or the sample size was too small , the non-parametric Mann-Whitney U test and Kruskal-Wallis test with the Dunn multiple comparisons test were used to compare the sum of ranks . Details on the statistical analyses performed are described in the figure legends . Data are presented as individual value points and/or the mean as summary statistic with error bars representing the Standard Error of the Mean ( SEM ) . P-values under 0 . 05 were considered statistically significant ( p < 0 . 05 , p < 0 . 01 , **p < 0 . 001 ) . The Coordinating Ethics Committee of the Helsinki and Uusimaa Hospital District ( no . 423/13/03/00/08 ) approved the patient informed consent for the derivation of the hiPSC lines used in this study: HEL71 . 4 and HEL107 . 2 . Animal care and experiments were approved by National Animal Experiment Board in Finland ( ESAVI/9978/04 . 10 . 07/2014 ) .	Insulin is a hormone that is crucial for maintaining normal blood sugar levels and is produced by so called beta cells in the pancreas . If the beta cells in the body stop making insulin , blood sugar levels start to rise , which can lead to diabetes . A form of diabetes known as neonatal diabetes , where the body stops making insulin , usually appears during the first six months of life . Infants affected by this early onset of diabetes often have mutations in one copy of the gene that encodes insulin . This means that they can still produce half of the amount of insulin , but it is not enough to keep blood sugar stable . Instead , insulin production stops completely after a few months . Scientists believe that this is because the mutant insulin has a toxic effect on beta cells . Mutations in the insulin gene can affect the structure of insulin . As a result , insulin accumulates inside the beta cells , which stresses them and eventually makes them fail . The mechanisms behind this process are still unclear . Now , Balboa et al . used stem cells ( which can turn into other cell types ) taken from patients with this rare type of insulin mutation to find out more . They corrected the mutant insulin gene in these stem cells with a technique called CRISPR and then induced the mutant and corrected stem cells to turn into beta cells . The results showed that the mutant beta cells slowed down their rate of cell division but did not die more frequently . When the cells were implanted into mice their growth and development changed . The mutant cells were more stressed and smaller than the cells with the repaired genes . They also had fewer signalling molecules that help cells grow . As a consequence , the cells were struggling to grow and mature . Although this type of diabetes is rare , beta cells come under stress in other forms of the disease . In a separate study , Riahi et al . found that boosting molecular signals for cell growth could protect beta cells in mice with mutant insulin . If this could also work in humans , it may lead to new ways to prevent diabetes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine" ]	2018	Insulin mutations impair beta-cell development in a patient-derived iPSC model of neonatal diabetes
The nucleolus is a membrane-less organelle formed through liquid-liquid phase separation of its components from the surrounding nucleoplasm . Here , we show that nucleophosmin ( NPM1 ) integrates within the nucleolus via a multi-modal mechanism involving multivalent interactions with proteins containing arginine-rich linear motifs ( R-motifs ) and ribosomal RNA ( rRNA ) . Importantly , these R-motifs are found in canonical nucleolar localization signals . Based on a novel combination of biophysical approaches , we propose a model for the molecular organization within liquid-like droplets formed by the N-terminal domain of NPM1 and R-motif peptides , thus providing insights into the structural organization of the nucleolus . We identify multivalency of acidic tracts and folded nucleic acid binding domains , mediated by N-terminal domain oligomerization , as structural features required for phase separation of NPM1 with other nucleolar components in vitro and for localization within mammalian nucleoli . We propose that one mechanism of nucleolar localization involves phase separation of proteins within the nucleolus . The nucleolus , a membrane-less organelle , is the site of ribosome biogenesis and a cellular stress sensor ( Boisvert et al . , 2007 ) . Nucleoli contain three substructures: the fibrillar centers ( FCs ) and dense fibrillar component ( DFC ) are engulfed in the granular component ( GC ) ( Boisvert et al . , 2007 ) , which exhibits ATP-dependent liquid-like features ( Brangwynne et al . , 2011 ) . Ribosomal RNA ( rRNA ) genes are transcribed between the FC and DFC , and rRNAs are processed while migrating into the GC , wherein they assemble with ribosomal proteins to form pre-ribosomal particles ( Boisvert et al . , 2007 ) . Nucleophosmin ( NPM1 , also known as B23 ) , a highly abundant marker of the GC , functions as a nucleolar chaperone and plays a role in cellular stress responses ( Colombo et al . , 2011 ) . While appreciated ( Brangwynne et al . , 2011; Chen and Huang , 2001; Negi and Olson , 2006; Weber and Brangwynne , 2015 ) , the molecular basis of the GC’s fluidity is unknown . While Npm1 loss is embryonic lethal in mice , mouse fibroblasts derived from Npm1-/-/Trp53-/- embryos readily proliferate in culture ( Colombo et al . , 2005 ) , indicating that NPM1 is dispensible for ribosome biogenesis . However , NPM1 is known to influence ribosome biogenesis , genome stability and tumor suppression ( Lindstrom , 2011 ) and to participate in responses to cellular stresses , including DNA damage ( Lee et al . , 2005 ) , chemotoxicity ( Chan , 1992; Yao et al . , 2010b ) , and oxidative stress ( Paron et al . , 2004 ) . Furthermore , NPM1 depletion is associated with disruption of nucleolar structure ( Holmberg Olausson et al . , 2014 ) . We propose that NPM1 participates in the organization of the liquid-like structure of the GC and consequently may actively participate in stress signal integration and transmission , thereby explaining its known roles in ribosome biogenesis , tumor suppression and other processes ( Lindstrom , 2011 ) . Accordingly , NPM1 interacts with a vast array of partners ( http://thebiogrid . org/110929/summary/homo-sapiens/npm1 . html ) , many involved in ribosome biogenesis , and controls the nucleolar localization of ribosomal ( Lindstrom , 2012; Rosorius et al . , 2000 ) , viral ( Duan et al . , 2014; Fankhauser et al . , 1991 ) , and certain tumor suppressor proteins ( Bertwistle et al . , 2004 ) , many of which engage the N-terminal oligomerization domain ( OD ) of NPM1 ( Figure 1a ) via arginine-rich short linear motifs ( R-motifs ) ( Mitrea et al . , 2014 ) . Furthermore , NPM1 binds nucleic acids ( e . g . , rRNA and DNA ) through its C-terminal domain ( nucleic acid binding domain , NBD ) ( Wang et al . , 1994 ) . We seek to understand the molecular mechanisms of NPM1’s multifarious functional interactions . Here , we show that NPM1 undergoes liquid-liquid phase separation in the presence of two classes of nucleolar macromolecules: proteins and RNA . Using a multidisciplinary strategy , we identify the structural features that mediate NPM1 phase separation and its nucleoar localization . These results provide a novel perspective on the mechanisms involved in the formation of the liquid-like structure of the nucleolus and in the nucleolar localization of biological macromolecules . 10 . 7554/eLife . 13571 . 003Figure 1 . Multivalency of acidic tracts within NPM1 and R-motifs within nucleolar substrates mediates liquid-liquid phase separation . ( a ) Composite model of NPM1 structure; the oligomerization domain ( OD , green , PDB ID 4N8M ) , containing the A1 acidic tract ( red ) , is connected via a disordered region ( IDR , grey ) , containing two additional acidic tracts ( A2 & A3 , red ) , to the C-terminal nucleic acid binding domain ( NBD , blue , PDB ID 2VXD ) ; ( b ) Phase separation diagrams for mixtures of N130 and four R-motif containing peptides ( bottom dot graphs ) ; phase separation was assessed by the formation of liquid-like droplets detected using light microscopy ( grey dot , clear solution observed; green dot , liquid-like droplets observed ) . Representative examples of liquid-like droplets formed between 50 μM N130 and the lowest peptide concentration associated with phase separation , visualized by DIC ( top panel ) and Alexa Fluor488 emission of labeled N130 ( bottom panel ) are illustrated . The composition of the R-motif peptides is given at the top of each pair of images; R is arginine; and X is any other amino acid . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 00310 . 7554/eLife . 13571 . 004Figure 1—figure supplement 1 . Representative ITC curves for titrations of multivalent R-motif containing peptides into N122 and N130 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 00410 . 7554/eLife . 13571 . 005Figure 1—figure supplement 2 . Multivalent R-motifs reside in disordered regions of proteins . IUPRED predictions of protein disorder . The location of R-motif peptides is highlighted . rpL5 and rpL23 structures have been resolved in the context of the assembled ribosome , where they adopt helical structures in complex with other ribosomal proteins and rRNA . In isolation , however , these structural features may not be highly populated . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 005 We first investigated interactions of R-motif-containing proteins with NPM1 by analyzing the results of a whole cell NPM1 pull-down experiment ( see Materials and Methods ) . Of 132 NPM1-binding proteins , 97% exhibited at least one R-motif ( RXn1R , where X is any amino acid and n1 ≤ 2; Supplementary file 1 ) and 78 . 8% were annotated with GO terms indicating association with membrane-less organelles ( Supplementary file 2 ) . Amongst all 132 NPM1-binding proteins , 73% exhibited multiple R-motifs ( Supplementary file 1 ) ; in contrast , only 44% of all human proteins exhibited multiple R-motifs ( p<0 . 0001; Supplementary file 3 ) . Thus , multivalent R-motifs are enriched in proteins that bind to NPM1 . R-motifs bind to a region of NPM1 which includes the OD and a short disordered region ( residues 1–130; termed N130; Figure 2a ) . The interactions engage two highly conserved acidic tracts , termed A1 ( residues 34–39 ) and A2 ( residues 120–130 ) , at the interface between monomer subunits and within the disordered region ( Mitrea et al . , 2014 ) , respectively ( Figure 2a ) . These A tracts and the pentameric nature create multivalency within N130 . Multivalent interactions involving low complexity sequences cause assembly and phase separation of biopolymers within membrane-less organelles ( Fromm et al . , 2014; Li et al . , 2012 ) ; therefore , we tested the ability of multivalent N130 and peptides containing multiple R-motifs derived from NPM1-binding proteins to undergo phase separation . Titration of four R-motif-containing peptides ( Table 1 ) caused phase separation into liquid-like droplets ( Figure 1b and Videos 1–4 ) at critical concentrations that varied with R-motif composition and affinity for N130 ( Table 2 and Figure 1—figure supplement 1 ) . At 200 µM N130 , upon titration of the divalent rpL5 peptide , phase separation was observed when the rpL5:N130 ratio reached ~3:1 ( Figure 2b ) . At the same N130 concentration , phase separation was not observed upon titration of a monovalent R-motif peptide ( rpL5-RA; Figure 2b ) , even though it bound , albeit with lower affinity ( Table 2 and Figure 1—figure supplement 1 ) , confirming that R-motif multivalency is required for phase separation with N130 . The inability to phase separate was not due to reduced binding affinity ( rather than loss of multivalency ) ; a poly-R peptide , containing a single but longer R-motif , with affinity similar to that of rpL5 , also failed to phase separate ( Figure 2—figure supplement 1 ) . Additionally , phase separation was not observed when an NPM1 construct containing only the OD ( N122 , residues 1–122; Figure 2a ) was titrated with the rpL5 peptide ( Figure 2c and Table 2 ) . We thus conclude that the minimal multivalency requirements for phase separation are the acidic A1 and A2 tracts within NPM1 and at least two complementarily charged R-motifs within a polypeptide binding partner . 10 . 7554/eLife . 13571 . 006Figure 2 . Multivalency within both the rpL5 peptide ( of R-motifs ) and N130 ( or A tracts ) is required for phase separation . ( a ) Schematic representation of the NPM1 constructs used in this study; ( b ) Titrations of rpL5 ( blue ) or rpL5-RA peptide lacking the second R-motif ( red ) , into 200 μM N130 , monitored by light scattering at 340 nm; ( c ) Titrations of rpL5 peptide into N130 ( blue ) and N122 ( orange ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 00610 . 7554/eLife . 13571 . 007Figure 2—figure supplement 1 . A monovalent peptide with similar affinity to rpL5 for binding N122 does not phase separate with N130 . ( a ) ITC curve of 6R peptide titrated into N122; ( b ) Light scattering assays of divalent rpL5 ( blue ) , monovalent rpL5- RA ( red ) and monovalent , but tighter binding 6R ( aqua ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 00710 . 7554/eLife . 13571 . 008Table 1 . Amino acid sequences of the synthetic multivalent R-motif containing peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 008Peptide Name#Peptide amino acid sequencerpL521RRRREGKTDY10YARKRLV37GNL2682RRRAVRQQRP10KKVGVRYYET20HNVKNRNR709SURF6299RRAQRQRRWE10KRTAGVVEKM20QQRQDRRR326rpL23a47RRPKTLRLRR10QPKYPRKSAP20RR68rpL5-RA21RRRREGKTDY10YAAKALV37rpL5-2xLinkerRRRREGKTDY10YAEGKTDYYA20RKRLV#The peptides are referred to by the same name as the protein they originate from and the residue numbers of their N- and C-termini are indicated . 10 . 7554/eLife . 13571 . 009Video 1 . In vitro droplets formed between 100 µM N130 and 500 µM rpL5 . 1 µM N130 was labeled with AlexaFluor488 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 00910 . 7554/eLife . 13571 . 010Video 2 . In vitro droplets formed between 25 µM N130 and 70 µM GNL2 . 1 µM N130 was labeled with AlexaFluor488 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01010 . 7554/eLife . 13571 . 011Video 3 . In vitro droplets formed between 100 µM N130 and 200 µM SURF6 . 1 µM N130 was labeled with AlexaFluor488 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01110 . 7554/eLife . 13571 . 012Video 4 . In vitro droplets formed between 50 µM N130 and 80 µM rpL23 . 1 µM N130 was labeled with AlexaFluor488 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01210 . 7554/eLife . 13571 . 013Table 2 . Binding affinities for interactions between N122 ( NPM1 residues 1–122 , displaying only A1 ) and N130 ( displaying both A1 and A2 ) and the multivalent R-motif-containing peptides , determined using isothermal titration calorimetry ( ITC ) , at concentrations below the critical phase separation threshold . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 013PeptideN122N130N ( sites ) KD ( μM ) N ( sites ) KD ( μM ) rpL50 . 63 ± 0 . 0919 . 0 ± 2 . 72 . 18 ± 0 . 1920 . 5 ± 5 . 0GNL20 . 66 ± 0 . 0410 . 2 ± 1 . 3n . d . n . d . SURF60 . 55 ± 0 . 069 . 2 ± 0 . 51 . 4 ± 0 . 0219 . 0 ± 0 . 8rpL23an . d . #n . d . #n . d . n . d . rpL5-RA1 . 0 ± 0 . 0650 . 9 ± 3 . 180 . 84 ± 0 . 07140 . 1 ± 9 . 29* Not determined due to multiple , unresolved binding events# Heat change was too weak for accurate data analysisAverage values from a minimum of three independent experiments are reported ± SD Next , to understand the mechanism of phase separation , we characterized the structural features of complexes of rpL5 with N130 before and after droplet formation . Fluorescence anisotropy ( FA ) of a N130S125C mutant , labeled with Alexa Fluor594 within A2 , increased in a biphasic manner upon titration of rpL5 ( Figure 3a ) . Results from small-angle neutron scattering ( SANS ) and analytical ultracentrifugation ( AUC ) experiments showed that monodisperse , soluble complexes formed at the first FA transition , which occurred at ~1:1 rpL5:N130 stoichiometry ( Figure 3a , inset; Figure 3a–c; Figure 3—figure supplement 1; Table 3 ) . Nuclear magnetic resonance ( NMR ) spectroscopy showed that rpL5 bound to both the A1 and A2 tracts within 15N-N130 ( Mitrea et al . , 2014 ) with a global KD of 57 ± 14 µM ( see Analysis methods ) in agreement with isothermal titration calorimetry measurements ( Table 2 ) . Site-specific KD values calculated from chemical shift perturbations of individual N130 peaks upon titration with rpL5 ( Figure 4a ) are presented in Table 4 . We investigated NMR transverse dipole-dipole/CSA cross relaxation and longitudinal relaxation for backbone amide moieties in 2H/15N-N130 titrated with rpL5 to determine their dynamic parameters before phase separation ( Figure 4b , c ) . Residues within the A2 tract of the apo state , analyzed separately from the folded pentameric core ( see Analysis methods ) , experienced fast , local motions [average local correlation time ( τc , local ) , 2 . 10 ± 0 . 05 ns; average local order parameter ( Sf2 ) , 0 . 44 ± 0 . 01]; these motions were slowed when N130 was ~93% saturated with rpL5 ( τc , local , 4 . 56 ± 0 . 06 ns; Table 5 ) and were reduced in amplitude ( Sf2 , 0 . 76 ± 0 . 01; Table 5 ) . Furthermore , a comparison of the overall tumbling time for core residues of N130 in the free and 93% rpL5 saturated state indicated 2:1 binding stoichiometry ( see Analysis methods ) . 10 . 7554/eLife . 13571 . 014Figure 3 . The liquid-like phase formed by rpL5 and N130 is characterized by molecular ordering and is accompanied by soluble , oligomeric intermediates . ( a ) Fluorescence anisotropy of Alexa Fluor594-labeled N130 ( at S125C within acidic tract A2; 1 μM ) upon titration of the rpL5 peptide; the total [N130] was 200 μM . Insets: light microscopy images of the 1:1 ( cyan box ) , 2:1 ( green box ) and 3:1 ( red box ) rpL5:N130 solutions . The same stoichiometry color coding is used in all panels . ( b ) SANS curves , I ( q ) versus q , for rpL5:N130 ( 200 μM ) solutions at 0:1 , 1:1 , 2:1 , and 3:1 stoichiometry . The curve for the 0:1 solution is on the absolute I ( q ) scale ( cm-1 ) with the others shifted in 1 , 2 and 4 decade increments for clarity . Fits ( solid lines ) of the curves to obtain Rg values ( 0:1 and 1:1 solutions ) and correlation distances ( 2:1 and 3:1 solutions ) are shown ( See Analysis methods for details on curve fitting ) . ( c ) Pair-distribution , P ( r ) , curves for 0:1 and 1:1 rpL5:N130 were calculated from the corresponding SANS curves ( fits shown in Figure 3b ) . For apo N130 , the Dmax ~ 78 . 65 Å with a resulting Rg = 23 . 04 ± 0 . 09 Å and I ( 0 ) = 0 . 2344 ± 0 . 0006 cm-1 . From I ( 0 ) and using Eq . S1 , the estimated molecular mass , M = 76 kDa , was determined; this mass is consistent with the expected five subunits within the pentamer ( subunit M = 14 . 6 kDa ) . For 1: 1 rpL5:N130 , the Dmax ~ 80 . 32 Å with a resulting Rg = 24 . 3 ± 0 . 1 Å and I ( 0 ) = 0 . 2744 ± 0 . 0009 cm-1 . Here , I ( 0 ) yields M = 89 kDa , indicative of ~5 rpL5 ( M = 2 . 2 kDa ) molecules bound to N130 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01410 . 7554/eLife . 13571 . 015Figure 3—figure supplement 1 . Sedimentation velocity analytical ultracentrifugation ( SV-AUC ) profiles for N130 titrated with rpL5 . SV-AUC profiles for apo N130 ( purple ) , or N130 with rpL5 at 1:1 ( light blue ) , 2:1 ( green ) and 3:1 ( red ) stoichiometric ratio ( rpL5:N130 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01510 . 7554/eLife . 13571 . 016Table 3 . Results of sedimentation velocity analytical centrifugation analysis ( SV-AUC ) of N130 in the absence or presence of increasing concentrations of the rpL5 peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 016rpL5:N130mg/mlas20 ( Svedberg ) bs20 , w ( Svedberg ) cMW ( kDa ) df/f0 e0:12 . 764 . 63 ( 98 . 7% ) 4 . 8288 . 91 . 537 . 90 ( 1 . 20% ) 8 . 22198 . 01 . 531:13 . 095 . 23 ( 96 . 3% ) 5 . 4499 . 31 . 467 . 63 ( 3 . 7% ) 7 . 94175 . 01 . 462:1 ( 1-300 min ) 3 . 631 . 07 ( 2 . 6% ) 1 . 116 . 81 . 205 . 80 ( 48% ) 6 . 0386 . 31 . 208 . 11 ( 23% ) 8 . 44143 . 01 . 2012 . 33 ( 25% ) 12 . 82268 . 01 . 2014-19 ( 7% ) 14-19300-5001 . 202:1 ( 100-300 min ) f3 . 315 . 75 ( 56% ) 5 . 9881 . 11 . 165 . 80 ( 25% ) 6 . 03128 . 01 . 168 . 11 ( 17% ) 8 . 44182 . 01 . 163:1 ( 1-300 min ) 2 . 010 . 95 ( 17% ) 0 . 987 . 41 . 435 . 80 ( 31% ) 6 . 00112 . 01 . 4310 . 32 ( 32% ) 10 . 73266 . 01 . 4314-27 ( 21% ) 14-27500-8501 . 433:1 ( 100-300 min ) f1 . 291 . 18 ( 4% ) 1 . 221 . 11 . 495 . 73 ( 41% ) 5 . 96117 . 01 . 497 . 41 ( 20% ) 7 . 71173 . 01 . 499 . 17 ( 15% ) 9 . 54238 . 01 . 4911 . 33 ( 15% ) 11 . 79266 . 01 . 49a Total concentration in mg/ml . b Sedimentation coefficient taken from the ordinate maximum of each peak in the best-fit c ( s ) distribution at 20 °C with percentage protein amount in parenthesis . Sedimentation coefficient ( s-value ) is a measure of the size and shape of a protein in a solution with a specific density and viscosity at a specific temperature . c Standard sedimentation coefficient ( s20 , w-value ) in water at 20 °C . d Molar mass values ( MW ) taken from the c ( s ) distribution that was transformed to the c ( M ) distribution . The theoretical molar mass is in parenthesis . e Best-fit weight-average frictional ratio values ( f/f0 ) w taken from the c ( s ) distribution . f Fit of the data in the 100-300 min sedimentation time frame , for the analysis of the lower MW rpL5:N130 complex intermediates . The MW cutoff of ~ 300 kDa corresponds to a tetramer of N130 pentamers . 10 . 7554/eLife . 13571 . 017Figure 4 . NMR chemical shift perturbations and amide backbone relaxation analysis of apo N130 and soluble complexes at 2 . 5:1 rpL5:N130 stoichiometry . ( a ) 1H-15N TROSY-HSQC spectra of apo N130 ( magenta ) , in a soluble complex with rpL5 ( 2 . 1:1 rpL5:N130; green ) and in the phase separated state ( 4:1 rpL5:N130; red ) . ( b ) Examples of R2 , β and R2 , α peak intensity decay curves measured at 800 MHz ( i ) and 1000 MHz ( ii ) , and R1 peak intensity decay curves measured at 800 MHz ( iii ) for free N130 . In ( i ) and ( ii ) , points and curves in black and red correspond to the R2 , β and R2 , α experiments , respectively . In ( i ) – ( iii ) , circles and squares indicate data points for residues Tyr29 ( within the folded pentamer core ) and Glu121 ( within the disordered A2 tract ) , respectively . Solid lines correspond to fits using a simple exponential decay model from which relaxation rates and intensities at zero time were extracted . ( c ) ( i ) Comparison of S2 values for the N-terminus for which data could be collected for apo N130 ( magenta bars ) and 2 . 5:1 rpL5:N130 ( green bars ) . Comparison of τc , local ( ii ) and S2 ( iii ) values for C-terminal residues in apo N130 ( magenta bars ) and 2 . 5:1 rpL5:N130 ( green bars ) . S125 is overlapped in the apo N130 spectrum; therefore , under this condition , these analyses could not be performed . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 01710 . 7554/eLife . 13571 . 018Table 4 . NMR-derived dissociation constant ( KD ) values determined by monitoring chemical shift perturbations of individual N130 nitrogen backbone resonances while titrating the rpL5 peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 018ResidueStructural contextKD ( μM ) Met5N-terminus385 ± 108Asp6N-terminus116 ± 25Met7N-terminus142 ± 30Ser10N-terminus56 ± 13Leu12N-terminus10 ± 5Arg13N-terminus58 ± 13Gln15Core36 ± 10Tyr17Core19 ± 9Leu18Core205 ± 51Val33Core1041 ± 649Asp36A1 tract329 ± 93Glu37A1 tract1549 ± 1168Glu39A1 tract115 ± 37His40Core123 ± 23Leu42Core24 ± 9Ser43Core166 ± 38Ala64Core1227 ± 562Asn66Core1530 ± 958Tyr67Core245 ± 58Glu68Core147 ± 29Val74Core53 ± 14Phe92Core14 ± 6Glu93Core329 ± 78Ile94Core73 ± 17Thr95Core34 ± 9Leu116Core101 ± 24Val117Core61 ± 17Ala118Core35 ± 9Glu120A2 tract68 ± 16Glu121A2 tract101 ± 18Asp122A2 tract279 ± 76Ala123A2 tract182 ± 50Glu124A2 tract614 ± 410Glu126A2 tract576 ± 471Asp127A2 tract2162 ± 1612Glu130A2 tract4258 ± 198610 . 7554/eLife . 13571 . 019Table 5 . NMR-derived dynamic parameters for monodisperse apo N130 and a 2 . 5:1 rpL5:N130 soluble complex . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 019Apo N130 ( N-terminus/A2 tract ) 2 . 5:1 rpL5:N130 ( N-terminus/A2 tract ) τc , local ( ns ) --a / 2 . 10 ± 0 . 05--a / 4 . 56 ± 0 . 06Sf20 . 16 ± 0 . 01a / 0 . 44 ± 0 . 010 . 26 ± 0 . 02a / 0 . 76 ± 0 . 01aDynamic parameters for the N-terminus were determined using a global τc value; thus , the only residue-specific motional parameter for residues within the N-terminus was Sf2 ( see Materials and Methods ) . The second transition in the FA curve ( starting at ~2 . 5:1 rpL5:N130 ) corresponded to phase separation ( Figure 3a ) . To gain insight into conformational changes associated with phase separation , we performed single-molecule Förster resonance energy transfer ( smFRET ) experiments with a mutant of N130 , N130Q15C/S125C , that could be dually labeled with Alexa Fluor594 & 680 at sites within the pentamer core ( Q15C ) and the A2 acidic tract ( S125C ) , respectively . The FRET efficiency ( EFRET ) for this dye pair within the droplet phase ( EFRET ~0 . 15 ) was dramatically reduced in comparison with that observed in the absence of rpL5 ( EFRET ~0 . 85; Figure 5 ) , consistent with the extension of the A2 tract from the N130 pentamer core due to interactions with rpL5 molecules upon phase separation ( Figure 6a ) . In contrast to the monodisperse character of the SANS curves for samples of rpL5 and N130 at 0:1 and 1:1 ( rpL5:N130 ) stoichiometry ( Figure 3b , c ) , the curve for the 1:3 sample indicated periodic structural organization within the liquid-like droplets ( Figure 3b , red curve ) . We interpreted these features in terms of inter-molecular correlation distances ( 55 Å , 77 Å , and 119 Å; see Analysis methods ) due to organization of N130 and rpL5 induced by phase separation . Notably , nascent , higher order structural organization was evident in the SANS curve for a solution of rpL5 and N130 with 2:1 stoichiometry ( Figure 3b , green curve ) . Assembly of rpL5 and N130 into higher order , soluble intermediates at and above the 2:1 stoichiometric ratio was also demonstrated using sedimentation velocity analytical ultracentrifugation ( SV-AUC; Figure 3—figure supplement 1 , Table 3 ) . The appearance of high molecular weight species within the detection range ( <1 MDa ) was accompanied by a progressive decrease in the total mass detected , likely due to sedimentation of rpL5:N130 droplets in the sample cell during the dead-time of the experiment . Above the phase separation threshold ( >3:1 rpL5:N130 ) , resonances for residues within the N130 core broadened beyond detection in 2D 1H-15N TROSY spectra but not those for residues within the A2 tract ( Figure 4a ) . Chemical shift values indicated that these residues remained disordered . 10 . 7554/eLife . 13571 . 020Figure 5 . The A2 tract extends away from the folded core of N130 on the pathway to phase separation . smFRET histograms for Alexa Fluor594/680-labeled N130Q15C/S125C ( ~100 pM; total [N130] , 200 μM ) at 0:1 , 0 . 5:1 , 2 . 7:1 and 3:1 ( phase separated state ) rpL5:N130 stoichiometry; the FRET events corresponding to conformational states with similar EFRET values are indicated by the shaded regions . The solid lines are Gaussian fits of the data , from which average EFRET values were determined . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 02010 . 7554/eLife . 13571 . 021Figure 6 . Schematic model of the structural rearrangements in N130 associated with liquid-liquid phase separation in the presence of the multivalent rpL5 peptide . ( a ) Structural representation of results from NMR and smFRET that revealed dramatic changes in dynamics and spatial orientation of the A2 track of N130 upon rpL5 peptide binding: apo N130 ( top ) and N130 in complex with rpL5 ( bottom ) . Residues within the A2 tract of N130 are shown as colored spheres with diameters proportional to the τc , local value for the 15N nucleus of amide group . The sites of fluorescent labeling , Q15C and S125C , are indicated as purple and yellow spheres , respectively ( the spheres for Q15C do not encode dynamic information ) . In the apo state , A2 randomly samples relatively compact conformations , while in liquid-like droplets , these residues extend away from the N130 core . ( b ) Schematic representation of apo N130 ( i ) and the proposed structural model of phase separation . Upon saturation of the two principal binding sites within the A1 and A2 tracts on N130 ( ii ) , bound rpL5 peptides extend toward and engage in weak interactions with neighboring N130 pentamers ( iii ) , thus creating 3D , expandable cross-links . In ( iii ) , only a subset of the possible inter-N130 pentamer crosslinks are shown for clarity . We suggest that one of the correlation distances observed using SANS ( 77 Å ) corresponds to the inter-N130 pentamer spacing . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 021 Integrating our structural results , we propose a model of rpL5/N130-dependent phase separation , as follows ( Figure 6b ) . As rpL5 is titrated into N130 up to 2:1 stoichiometry ( rpL5:N130 ) , the R1 motifs , comprised of four Arg residues each , bind to acidic residues within the A1 binding groove and disordered A2 tract of N130 , with the R2 motif available for interactions . Upon further titration of rpL5 , at the critical phase separation concentration , when the higher affinity sites reach a critical saturation threshold , R2 motifs within rpL5 molecules already bound to N130 pentamers transiently engage A tracts of other pentamers , possibly within the longer and disordered A2 tract , establishing inter-N130 pentamer cross-links . Together , our data support the hypothesis that the molecular basis of phase separation is the formation of non-covalent , inter-N130 pentamer interactions via the two R-motifs within the same rpL5 peptide molecule . We propose that these rpL5-mediated interactions establish the inter-pentamer spacing within the droplet phase that was detected by SANS ( Figure 3b ) . In order to validate this model , we synthesized a variant rpL5 peptide in which the eight residue-long linker connecting the two R-motifs ( see Table 1 ) was duplicated . The phase separated sample formed between the peptide with the longer linker ( rpL5-2xLinker ) exhibited altered inter-molecular spacing , as indicated by altered correlation distances derived from the SANS curve ( Figure 7 ) . These shifted to positions corresponding to larger correlation distances , in agreement with the hypothesis that the R-motif peptides establish inter-N130 pentamer spacing in the liquid-like phase . We envision that such cross-links are dynamically formed and broken but that , once phase separation occurs , they dominate the structural organization detected by SANS . 10 . 7554/eLife . 13571 . 022Figure 7 . The length of the linker between R-motifs in the rpL5 peptide influences the molecular organization within rpL5:N130 liquid-like droplets . ( a ) SANS curves , I ( q ) versus q , for 3:1 rpL5:N130 ( red ) and rpL5-2xLinker:N130 ( orange ) solutions; the N130 concentration for both was 200 μM . The curve for the droplets containing rpL5 2xLinker is displayed on absolute I ( q ) scale ( cm-1 ) with the curve corresponding to rpL5 containing droplets shifted in 4 decade increments for clarity . Fits ( solid lines ) of the curves to obtain correlation distances are shown ( See Analysis methods for details on curve fitting ) . Peak positions for the droplets with the rpL5-2xLinker correspond to correlation distances ( d ) of 63 Å & 100 Å whereas those observed with the wild-type rpL5 peptide correspond to d values of 56 Å , 79 Å & 119 Å . Inset: wide-field fluorescence microscopy ( left ) and DIC ( right ) images of droplets formed from the 3:1 rpL5-2xLinker:N130 solution . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 022 Our data showed that the minimal construct , N130 , supported droplet formation . However , both the OD ( Enomoto et al . , 2006; Jian et al . , 2009 ) and NBD ( Hisaoka et al . , 2014; Negi and Olson , 2006 ) are required for nucleolar localization of transfected NPM1 , suggesting that interactions with ribosomal RNA ( rRNA ) are also involved in integration of NPM1 within the GC matrix . Full length NPM1 also exhibits a central intrinsically disordered region ( IDR ) that , in addition to the A2 tract , contains a longer acidic tract termed A3 ( Figure 1a ) . We tested and confirmed the ability of full length NPM1 ( N294 ) to phase separate with either wheat germ rRNA or rpL5 using light scattering ( Figure 8a & b , respectively ) and fluorescence microscopy ( Figure 9a , Rows 1 & 2 , respectively ) . Binary mixtures of N294 with either rpL5 or rRNA formed droplets that increased in size over time ( Figure 9b ) ; those with rpL5 fused more rapidly and were larger in size than those with rRNA . Furthermore , droplets of NPM1 with rpL5 formed at ~5-fold lower concentrations of both components ( Figure 8b ) than were required for phase separation with N130 and rpL5 ( Figure 1b ) due to the higher valency of acidic tracts within the full-length protein . Interestingly , rpL5 ( and other R-motif peptides ) and rRNA phase separated in the absence of N294 ( Figure 8c , d ) , forming very small puncta ( Figure 9a , Row 3 ) . We cannot explain the physical basis for the different dynamics of liquid-like droplets formed by N294 with either rRNA or rpL5 ( Figure 9a , Rows 1 and 2 , respectively ) or the morphology of the punctate structures formed by rRNA and rpL5 ( Figure 9a , Rows 3 ) due to the lack of data on the interaction between NBD of NPM1 or rpL5 , and rRNA ( i . e . , binding affinities , number and location of binding sites , etc . ) . The size difference between the two types of droplets with N294 ( Figure 9a , Rows 1 and 2 ) may , however , arise from differences in binding affinity between rRNA and NBD versus rpL5 and A tracts within OD/IDR . 10 . 7554/eLife . 13571 . 023Figure 8 . Phase separation by pairwise mixtures of NPM1 ( N294 ) and molecules representing two fundamental components of the nucleolus , R-motif containing proteins ( represented by R-motif peptides ) and rRNA , as determined by light scattering assays . Titration of ( a ) N294 into 50 μg/mL rRNA and ( b ) rpL5 into 40 μM N294; ( c ) Titrations of rRNA into 50 μM R-motif containing peptides ( rpL5 , GLN2 , SURF6 , and rpL23a ) . ( d ) Titrations of R-motif containing peptides into 50 μg/mL rRNA ( as in c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 02310 . 7554/eLife . 13571 . 024Figure 9 . Multi-modal binding of NPM1 mediates formation of liquid-like droplets with both rRNA and rpL5 in vitro . ( a ) Confocal microscopy images of droplets after 15 min incubation formed between 40 μM N294 and 50 μg/mL rRNA ( Row 1 ) , 40 μM N294 and 200 μM rpL5 ( Row 2 ) , 200 μM rpL5 and 50 μg/mL rRNA ( Row 3 ) and 40 μM N294 and 50 μg/mL rRNA droplets treated with 40 μM rpL5 , a concentration below the critical phase separation threshold for N294/rpL5 binary system ( Row 4 ) . rpL5 and N294 were labeled with Alexa Fluor594 and Alexa Fluor488 , respectively . Scale bar = 10 μm; ( b ) Quantitation of droplet size over time for the samples presented in panel a; values plotted are the mean ± SD , n = 4 fields of view . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 024 Intrigued by the physical differences between the structures formed by pairwise combinations of N294 , rpL5 and rRNA , we next sought to understand the behavior of these three species in ternary mixtures by examining interactions between pre-formed droplets comprised of N294 and rRNA to which freely diffusing rpL5 was added , at a concentration below that which caused phase separation with N294 alone ( Figure 9a , Row 4 ) . The peptide accumulated into the rRNA/N294 droplets , highlighting NPM1’s capacity to bind the two fundamental classes of macromolecules present in the nucleolus; these droplets also slowly grew over time ( Figure 9b ) . Importantly , this multi-modal binding mediates the co-assembly of rRNA and rpL5 within a dense , multi-component liquid-like phase . We hypothesize that a similar molecular mechanism is responsible for the phase separation and co-localization of nucleolar components within the GC . We next examined the roles of the different domains of NPM1 in phase separation to form multi-component liquid-like droplets using two truncation mutants , one lacking the NBD ( N240 , residues 1–240; Figure 2a ) and another lacking the OD ( ΔN , residues 120–294; Figure 2a ) . In agreement with a mechanistic model wherein multivalent interactions between pentameric A1 , A2 , and A3 tracts within NPM1 and multivalent R-motifs within its nucleolar protein partners mediate phase separation , the N240 , but not the ΔN construct phase separated with rpL5 ( Figure 10a ) . Furthermore , neither of these truncated constructs experienced phase separation in the presence of rRNA , confirming that multivalent display of the NBD is required for the co-localization of rRNA with NPM1 within liquid-like droplets ( Figure 10b ) . 10 . 7554/eLife . 13571 . 025Figure 10 . The OD and A tracts of NPM1 are required for phase separation with rpL5 , while phase separation in the presence of rRNA requires both folded domains ( OD & NBD ) . Light scattering assays of ( a ) titrations of rpL5 peptide into 40 μM NPM1 and ( b ) titrations of NPM1 constructs into 50 μg/mL rRNA . Values plotted are the mean ± SD , n = 3 experiments . The dashed line at 0 . 1 AU indicates the threshold , above which visible turbidity and microscopic droplets can be detected . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 025 Multi-modal binding to two classes of macromolecules , R-motif-containing nucleolar proteins ( binding mode 1 ) and rRNA ( binding mode 2 ) , is likely critical for NPM1-dependent formation of multi-component liquid-like droplets . In the absence of NPM1 , rRNA and rpL5 , representing these two classes , phase separated into small puncta ( Figure 8c , d , Figure 9 ) . Given this result , we next asked whether addition of NPM1 constructs to these puncta would cause reorganization and co-localization of the constituent macromolecules within larger , liquid-like droplets . In order to differentiate between the effects of interactions between the OD/A tracts and rpL5 ( mode 1 ) and OD/NBD and rRNA mode ( mode 2 ) on phase separation , we first formed rRNA/rpL5 puncta at two concentrations of rpL5 and then added the NPM1 constructs and monitored phase separation using confocal-microscopy . At the high rpL5 concentration ( rpL5 , 200 μM; NPM1 constructs , 30 μM; termed the 'excess' condition ) , rpL5 and NPM1 could independently phase separate . However , at the low concentration ( rpL5 , 50 μM; NPM1 constructs , 30 μM; termed the 'limiting' condition ) , rpL5 and NPM1 could not independently phase separate . The addition of N294 to pre-formed , rpL5/rRNA puncta caused spontaneous formation of large droplets under conditions of both limiting and excess rpL5 , whose sizes increased with increasing N294 concentration ( Figure 11 ) . In contrast , addition of the N240 deletion construct , lacking the NBD , caused phase separation under conditions of excess rpL5 , but not with a limiting amount of rpL5 ( Figure 11 ) . Finally , addition of the ΔN construct , lacking the OD and therefore displaying dramatically reduced multivalency , failed to cause phase separation under both excess and limiting rpL5 conditions ( Figure 11 ) . Together , these results support a mechanism wherein the high valency associated with the NPM1 OD , together with the multiple acidic tracts and NBD , are required for the dissolution of rpL5/rRNA puncta and co-localization of these molecules within large liquid-like droplets that readily grow in size . We thus propose that , through multi-modal interactions with two major classes of nucleolar macromolecules , NPM1 localizes to the nucleolus and also mediates the co-localization of other protein-binding partners and rRNA within the GC of the nucleolus . 10 . 7554/eLife . 13571 . 026Figure 11 . Both the OD and NBD of NPM1 are required for co-localization of rRNA and rpL5 within liquid-like droplets in vitro . Confocal microscopic images 15 min after the addition of the specified NPM1 construct ( 30 μM ) to preformed rpL5/rRNA puncta formed from 50 μg/mL rRNA and 200 μM rpL5 ( a ) and from 50 μg/mL rRNA and 50 μM rpL5 ( b ) . Scale bar = 10 μm . Quantification of the growth of droplets formed through disassembly of rpL5/rRNA puncta upon addition of individual NPM1 constructs . ( c ) NPM1 constructs were titrated into puncta formed between 200 μM rpL5 and 50 μg/mL rRNA , where rpL5 is above the threshold for independent phase separation with NPM1 . The star at 40 μM N294 indicates that the droplets experienced wetting on the slide surface and expanded above the maximum threshold of 100 μm2 set in the analysis; ( d ) NPM1 constructs titrated into puncta formed between 50 μM rpL5 and 50 μg/mL rRNA , where rpL5 is below the threshold for independent phase separation with NPM1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 026 We next explored the relevance of our in vitro findings to nucleolar physiology . First , we monitored the accumulation of a series of recombinant eGFP-fused NPM1 constructs within isolated nucleoli ( Figure 12a ) using confocal fluorescence microscopy . Consistent with in vitro results showing that all domains of NPM1 were required for the integration of both an R-motif peptide and rRNA into large , liquid-like droplets , the eGFP-N294 fusion protein accumulated within purified nucleoli , even at sub-micromolar concentrations ( Figure 12b , c ) , while constructs lacking either the rRNA binding or oligomerization domains ( eGFP-N240 or eGFP-△N , respectively ) did not ( Figure 12c ) . These results suggested that the nucleolar localization of NPM1 requires multi-modal interactions with both R-motif-containing nucleolar proteins and rRNA . 10 . 7554/eLife . 13571 . 027Figure 12 . Synergistic activity of the OD and NBD are required for NPM1 incorporation in mammalian nucleoli . ( a ) DIC image of isolated nucleoli; ( b ) Titrations of recombinant eGFP-N294 into purified nucleoli . Fluorescence intensity over background is quantified . ( c ) Fluorescence profile through sections of nucleoli treated with 100 μM recombinant eGFP-NPM1 constructs . Values in panels ( b ) and ( c ) represent mean ± SD , n = 20 nucleoli; ( d ) Fluorescence confocal microscopy images of MEF cell lines; DAPI was imaged to identify nuclei , mCherry to identify the NPM1 constructs , and a fluorescently-labeled antibody to NOPP140 to identify the GC of nucleoli . Scale bar = 5μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 02710 . 7554/eLife . 13571 . 028Figure 12—figure supplement 1 . The multi-modal binding properties of the OD and NBD of NPM1 are required for localization within mammalian nucleoli . Confocal fluorescence microscopy images of Trp53-/- and Npm1-/-/p53-/- ( DKO ) MEF cell lines , and DKO cell lines stably expressing mCherry-fused NPM1 constructs , stained with DAPI to visualize nuclei and with a fluorescent antibody for fibrillarin to visualize nucleoli . NPM1-expressing cells were also imaged for mCherry fluorescence . Scale bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 02810 . 7554/eLife . 13571 . 029Figure 12—figure supplement 2 . Flow cytometric analysis of transduced MEFs . ( a ) Surface expression of Thy1 . 1 was analyzed in parental ( filled histogram ) and DKO MEFs transduced with empty vectors ( mock; black curve ) or those containing mCherry NPM1 variants: N294 ( red ) , N240 ( green ) and ΔN ( blue ) . ( b ) The expression of mCherry was determined in Thy1 . 1-expressing cells ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13571 . 029 To extend our studies to the cellular setting , we next investigated the domain requirements for nucleolar localization by creating Trp53-/-/Npm1-/- mouse embryonic fibroblasts cell lines ( DKO MEFs ) that stably expressed a related series of mCherry-fused NPM1 constructs . The parental DKO cell line was created by replacing exons 2–7 of Npm1 with eGFP , and therefore constitutively expresses the fluorescent protein ( Grisendi et al . , 2005 ) . The stable DKO MEF cell lines were created through infection with retroviruses carrying the Thy1 . 1 cell surface marker in frame with an internal ribosomal entry site ( IRES ) and an mCherry-NPM1 variant gene . The two genes are translated from the same bicistronic mRNA transcript at similar levels ( Gurtu et al . , 1996 ) . Therefore , the Thy1 . 1 cell surface marker is an unbiased reporter of mCherry-NPM1 variant expression . mCherry is primarily localized to the cytoplasm and is stable in cells ( Shaner et al . , 2004 ) ; therefore , with the constructs under study here , any alteration of its localization and stability can be attributed to its NPM1 variant fusion partner . In agreement with the in vitro and ex cellulo results , full length mCherry-NPM1 accumulated in nucleoli and co-localized with the nucleolar markers NOPP140 ( Figure 12d ) and fibrillarin ( Figure 12—figure supplement 1 ) . Despite the fact that all cell lines expressed comparable levels of the Thy1 . 1 marker ( Figure 12—figure supplement 2 ) , and all NPM1 constructs encode the native bi-partite nuclear localization signal ( residues 152–157 and 191–197 [Hingorani et al . , 2000] ) , mCherry fluorescence was undetected in the mCherry-N240 cell lines , while the mCherry-ΔN cell lines exhibited diffuse mCherry fluorescence throughout the nucleus ( Figure 12d and Figure 12—figure supplement 1 ) . Notably , Enomoto , et al . , previously showed that NPM1 mutants which lacked the ability to accumulate within nucleoli exhibited dramatically reduced half lives in cells ( Enomoto et al . , 2006 ) , a potential explanation for the undetectable levels of mCherry-N240 in our study . We note that our observations through use of DKO cell lines contrast with those of others ( Enomoto et al . , 2006; Negi and Olson , 2006 ) wherein NPM1 constructs lacking the C-terminal domain accumulated within nucleoli . These previous studies are confounded by the possiblility for the formation of heteromeric oligomers comprised of endogenous wild-type and mutant NPM1 proteins . Collectively , these data suggest that nucleolar localization of NPM1 is achieved only when interactions with both R-motif-containing nucleolar proteins and rRNA are possible . Deletion of either the NBD or OD abrogated the ability of NPM1 to simultaneously interact with an R-motif-containing peptide and rRNA within droplets in vitro ( Figure 11 ) , suggesting that co-localization of NPM1 with the two types of nucleolar components in dense liquid-like droplets and localization within nucleoli arise through similar mechanisms involving multi-modal , multivalent interactions that promote phase separation . The nucleolus is a membrane-less organelle comprised of a vast array of macromolecules that mediate ribosome biogenesis and participate in stress signaling . However , the molecular mechanism that underlies the localization of macromolecules within the nucleolus is poorly understood . In contrast to the nuclear localization process , for example , which relies on nuclear transport receptors that recognize their cargo via a specific short linear motif ( Christie et al . , 2015 ) and transport them across the nuclear membrane through the nuclear pore complex , there are no known molecular transporters that mediate localization of proteins to the nucleolous . It has been hypothesized that enrichment of proteins within the nucleolus occurs though selective retention within the nucleolar matrix . Short basic sequences , termed nucleolar localization signals ( NoLS ) , often rich in R and lysine ( K ) residues , have been identified experimentally to be necessary and , in some cases , sufficient for localization of certain proteins within the nucleolus ( Scott et al . , 2010 ) . However , the molecular basis for how these canonical signals cause nucleolar localization is presently unknown . Here , we identified multivalent , low complexity R-motifs within proteins that are associated with binding to the abundant nucleolar protein , NPM1 . Furthermore , we showed that these multivalent R-motifs underwent phase separation with NPM1 to form dense , liquid-like droplets , and became co-localized within these types of droplets that also contained rRNA . Strikingly , the multivalent R-motifs that we identified to mediate interactions with NPM1 have features in common with the canonical NoLSs: they are enriched in basic amino acids and are often located within disordered and solvent accessible regions of proteins ( Scott et al . , 2010 ) ( Figure 1—figure supplement 2 ) . This observation , however , raises several questions . Do the links between canonical NoLSs and nucleolar localization , and multivalent R-motifs and NPM1 binding , have a common mechanistic basis ? And , does nucleolar localization rely on the ability of proteins to experience phase separation within a multi-component , liquid-like phase ? To address these questions , we compared our list of NPM1 interacting proteins that contained multivalent R-motifs with the database of all predicted NoLSs within the human proteome ( [Scott et al . , 2010]; http://www . compbio . dundee . ac . uk/www-nod/downloads/AllPredictedHumanNoLSs . txt ) . Interestingly , 83 ( 63% ) of the proteins from our curated list of NPM1 interactors corresponded to those with predicted NoLSs . Of these 83 , 69 ( 83% ) contained 2 or more R-motifs , and there was extensive overlap between the sequence locations of the multivalent R-motifs that we identified and those of the predicted NoLSs ( Supplementary file 1 ) . Based on these observations , we propose that canonical NoLSs , which closely match our definition of multivalent R-motifs , enable proteins that are freely diffusing through the nuclear compartment to phase separate within the nucleolar matrix , thereby mediating their nucleolar localization . Furthermore , while the NoLSs described by Scott , et al . , account for the nucleolar localization of hundreds of proteins , more than 4 , 500 proteins are known to reside within the nucleolus ( Ahmad et al . , 2009 ) , highlighting the need to identify additional localization mechanisms . Interestingly , along these lines , NPM1 , although well established to be highly enriched in the nucleolus , does not exhibit a canonical NoLS . Rather , here we describe a multi-modal interaction model that enables the multifunctional protein NPM1 to interact with R-motif-containing proteins and rRNA , representative of two major classes of macromolecules resident of the nucleolus . In support of this , we found that disruption of its multi-modal interactions , through abrogation of NBD oligomerization by the deletion of the OD or removal of the NBD , hindered NPM1’s ability to phase separate in vitro and to accumulate in nucleoli within mammalian cells ( Figures 10–12 ) . Consistent with our model , a previously described putative NoLS involves two buried Trp residues which are essential for proper folding of the NBD ( Grummitt et al . , 2008 ) . Mutations of these residues destabilized the NBD , thereby impeding interactions with nucleic acids and nucleolar localization ( Banuelos et al . , 2013; Chiarella et al . , 2013; Falini et al . , 2006; Grummitt et al . , 2008 ) . Thus , we propose that multi-modal binding by NPM1 may play an important role in the structural organization of the nucleolus and the recruitment of proteins to the nucleolus . In our structural studies of the minimal phase separating system , N130/rpL5 , we showed that within the liquid-like matrix of the droplets , with diameters on the microns to tens of microns length scale , we observed local ordering on the length scale of 5 to 12 nanometers ( Figures 3 and 7 ) . We propose that the intermolecular spacing indicated by SANS is dictated , at least in part , by the length of the linker connecting the two R-motifs in the divalent rpL5 peptide . Another factor in the spatial organization within these droplets is the 5-fold symmetry of the A1 binding grooves within the N130 pentamer that bind R-motifs within the rpL5 peptide ( Figure 1 and [Mitrea et al . , 2014] ) . We suggest that the SANS data indicate the 'persistence length' of local structural order that , despite constant formation and dissociation of R-motif peptide-mediated cross-links , is sufficient to provide 'structure' to the liquid-like droplets . However , for several reasons , this is likely a highly simplified view of NPM1 structural organization within the GC of the nucleolus . First , NPM1’s binding partners display diverse patterns of multivalent R-motifs ( Supplementary file 1 ) , providing many ways of forming cross-links with NPM1 . Second , in comparison with N130 , full length NPM1 exhibits the A3 acidic tract within the IDR , enabling additional interactions with R-motifs within proteins , and the NBD , that can bind to rRNA . Therefore , due to these factors , NPM1’s structural organization within the nucleolus is likely highly heterogeneous , in contrast to the orderly organization of N130 within droplets with rpL5 . This model for the the structurally heterogenous integration of NPM1 within the nucleolar matrix provides a mechanistic explanation for its central role in many nucleolar functions , such as ribosome biogenesis and nucleolar stress responses ( Lindstrom , 2011 ) . Through studies of an in vitro model system based upon formation of multi-component liquid-like droplets , representing a simplified form of the GC of the nucleolus , we identified three fundamental types of interactions that mediated phase separation: ( 1 ) R-motifs ( from nucleolar proteins ) binding to oligomeric , multivalent A-tracts ( within NPM1 ) , ( 2 ) low complexity multivalent R-motifs binding to rRNA and ( 3 ) oligomeric , folded nucleic acid binding domains ( within NPM1 ) binding to rRNA . Disrupting either type ( 1 ) or ( 3 ) interactions abrogated accumulation of NPM1 within nucleoli ( Figure 12 ) , demonstrating the relevance of our in vitro findings to the physiological setting . Complete loss of NPM1 from nucleoli ( e . g . , in the Trp53-/-/Npm1-/- mouse embryonic fibroblasts cells ) did not , however , preclude other nucleolar proteins ( NOPP140 and fibrillarin; Figure 12 and Figure 12—figure supplement 1 , respectively ) from accumulating within punctate nuclear structures , in support of a model wherein collective interactions between many different nucleolar components drive their assembly into dense liquid-like structures . While the R-motif-rich sequences are recognized to be associated with nucleolar localization ( Scott et al . , 2010 ) , the other two protein features identified in our study to be important for interactions with nucleolar components , namely multivalent acidic tracts and multivalent , folded nucleic acid binding domains , are not . Notably , the association of multivalency mediated by homo-oligomerization and acidic low complexity regions with phase separation is , to the best of our knowledge , unique to NPM1 . We expect , however , that similar features may be utilized by other proteins for incorporation into the nucleolar matrix . For example , NOPP140 , known to be localized within the GC , exhibits 11 acidic , serine-rich clusters ( [UniProt , 2014]; http://www . uniprot . org/uniprot/Q14978 ) . Furthermore , nucleolin , which also lacks a canonical NoLS , exhibits 3 acidic tracts ( with lengths between 15 and 29 amino acids ) , 4 tandem folded RNA recognition motifs in addition to multivalent R-motifs ( [UniProt , 2014]; http://www . uniprot . org/uniprot/P19338 ) , which may mediate multi-modal binding to protein and rRNA components within the dense liquid-like phase of the nucleolus . While the established NoLS is consistent with our mechanistic model for nucleolar localization , other proteins , which we suggest function by contributing to the structural organization of the nucleolus , such as NPM1 , are not encompassed by this definition . We propose that the three fundamental types of interactions noted above mediate the localization of a wide veriety of proteins , along with rRNA , within the liquid-like structure of the nucleolus , providing new perspective on the term 'nucleolar localization signal' . A list of 132 NPM1 binding partners obtained from BioGRID ( deposited by Dr . Steven Gygi , Harvard Medical School; available at http://thebiogrid . org/166968/publication/high-throughput-proteomic-mapping-of-human-interaction-networks-via-affinity-purification-mass-spectrometry . html ) was analyzed using the DAVID Bioinformatics Resources ( http://david . abcc . ncifcrf . gov ) ( Huang et al . , 2009; 2008 ) to identify proteins with known involvement in nucleolar structure and/or function . GO terms were available for 125 of the 132 NPM1-binding proteins ( Supplementary file 2 ) . To gain insight into the molecular basis for interactions with NPM1 , we determined the occurrence of multivalent R-motifs within the sequences of the 132 NPM1-binding proteins . Minimal R-motifs were defined as follows: a minimal , single R-motif as the sequence pattern , RXn1R , where n1≤2 , R is arginine and X is any amino acid; a minimal , multivalent R-motif as the sequence pattern , RXn1RXn2RXn3R , where n1 , n3≤2 , and n2≤20 ) ( Supplementary file 1 ) . In Supplementary file 1 , the identified R-motifs were extended if followed by another Arg residue within two or fewer residues . A Python algorithm , available for download at https://github . com/dlaszlo88/eLIFE-NPM_NMRrelaxation was developed to identify proteins exhibiting multivalent R-motifs and was applied to the list of NPM1 binding proteins as well as a list of 20 , 193 non-redundant human proteins obtained from the UniProtKB/Swiss-Prot database ( http://www . uniprot . org/uniprot/ ) ( UniProt , 2014 ) . Of the 132 NPM1 binding partners , 73% exhibited at least one multivalent R-motif; in comparison , only 44% of all human proteins exhibited at least one multivalent R-motif ( Supplementary file 3 ) . These data show NPM1 binding partners are enriched in multivalent R motifs , when compared to the majority of the human proteome ( p<0 . 0001 ) . The N130 protein was expressed and purified as described ( Mitrea et al . , 2014 ) . The eGFP-NPM1 constructs were subcloned in the pET28 vector ( Novagen , Darmstadt , Germany ) , with an N-terminal poly-His tag . eGFP was amplified from pEGFP-C1 , a gift from Dr . Douglas Green , and inserted between NdeI and BamHI restriction sites . NPM1 construct genes , derived from human NPM1 , were cloned between BamHI and XhoI restriction sites . The original Thrombin cleavage site following the affinity tag was replaced with a PreScission cleavage site . Proteins were expressed in E . Coli strain BL21 ( DE3 ) in Luria Broth . Protein expression was induced at OD600 ~0 . 6 with 100 mg/L IPTG ( Goldbio , St . Louis , MO ) and the cultures were incubated overnight at 20 °C . Bacterial pellets were harvested by centrifugation and lysed by sonication in 20 mM Tris , 150 mM NaCl , 5 mM β-mercapto-ethanol ( BME ) , pH 7 . 5 , supplemented with protease inhibitors ( SigmaFAST , Sigma-Aldrich , St . Louis , MO ) . eGFP-NPM1 proteins were further purified using Ni-NTA affinity chromatography , using 0 . 5 M NaCl in the buffers . The eluted proteins were dialysed overnight against 10 mM Tris , 50 mM NaCl , 2 mM DTT pH 7 . 5 in the presence of HRV3C protease ( BioVision , Milpitas , CA ) to remove the poly-His tag and further purified on a Mono Q HR 5/5 ( GE Healthcare , Pittsburgh , PA ) ion exchange column , followed by size exclusion chromatography on HiLoad 26/60 Superdex 200 ( GE Healthcare , Pittsburgh , PA ) in 10 mM Tris , 150 mM NaCl , 2 mM DTT , pH 7 . 5 . The N122 construct was a gift from Dr . Yuh Min Chook ( UT Southwestern , Dallas ) . The protein was expressed with an N-terminal GST tag as described above . The N122 protein was affinity purified on a reduced glutathione agarose column ( Qiagen , Hilden , Germany ) , followed by GST tag proteolysis with thrombin and finally purified to homogeneity using size exclusion chromatography , as described above . Peptides were synthesized in house , by the Macromolecular Synthesis resource within the Hartwell Center for Bioinformatics and Biotechnology at St . Jude Children’s Research Hospital , using standard solid phase peptide synthesis chemistry . The samples were incubated at room temperature for 5 min and UV-Vis absorbance spectra were recorded in triplicate on a NanoDrop 2000c spectrophotometer ( Thermo Scientific , Waltham , MA ) . N130 was labeled with Alexa Fluor488 C5 maleimide ( Life Technologies , Carlsbad , CA ) , according to the manufacturer’s protocol . The labeling was performed in 10 Tris , 150 mM NaCl , pH 7 . 5 buffer to maintain N130 in its folded pentameric state ( Mitrea et al . , 2014 ) and ensure selective labeling at solvent exposed Cys104 , thereby avoiding labeling the buried Cys21 . The rpL5 peptide was N-terminally labeled using Alexa Fluor594 Succinimidyl ester ( Life Technologies , Carlsbad , CA ) following the manufacturer’s protocol . Titrations were performed using a GE Auto-iTC200 instrument ( Malvern , Malvern , UK ) , at 25 °C , with the NPM1 construct in the cell . Proteins and peptides were dialyzed overnight against the reaction buffer , consisting of 10 mM Na phosphate , 150 mM NaCl , 2 mM DTT , pH 7 . 0 . The concentrations were selected to be below the phase separation threshold and were determined from A280nm of the protein or peptide diluted in 6 M guanidine hydrochloride buffer . N130 constructs with Q15C and S125C single and double mutations in the background of a C104T mutation were used . For ensemble fluorescence anisotropy measurements , N130 labeled at C125 with Alexa Fluor594 ( C5 maleimide derivative; Life Technologies , Carlsbad , CA ) was used . The protein was labeled in 10 mM Na phosphate , 500 mM NaCl , pH 7 . 5 with 2-fold molar excess of the dye overnight at 4 °C in the dark . For dual labeling of N130 at C15 and C125 , the following protocol was used: a 2-fold molar excess of donor dye ( Alexa Fluor594 C5-maleimide derivative , Life Technologies , Carlsbad , CA ) , and a 8-fold molar excess of the acceptor dye ( Alexa Fluor680 C2-maleimide derivative , Life Technologies , Carlsbad , CA ) were incubated together in 10 mM Na phosphate , 500 mM NaCl , pH 7 . 5 overnight at 4 °C for in the dark . The excess dye in all labeling reactions was removed by multiple rounds of washing with the labeling buffers using a 3K MWCO centrifugal filter device ( Millipore , Darmstadt , Germany ) . The purity of all the samples was confirmed by ESI-mass spectroscopy ( Scripps Center for Mass Spectrometry ) . Ensemble fluorescence measurements were carried out using an automated temperature controlled PC1 spectrofluorometer ( ISS , Champaign , IL ) in 10 mM Tris , 150 mM NaCl , pH 7 . 5 . The single-molecule fluorescence measurements were performed on freely diffusing molecules with a 532 nm excitation line ( CrystaLaser , Reno , NV ) ( operating on a 500 µW power ) using a home-built instrument . The details of smFRET instrumentation , data collection and data analysis have been described elsewhere ( Ferreon et al . , 2009; 2013 ) . All experiments were performed at room temperature in 10 mM Tris , 150 mM NaCl , 2 mM DTT , pH 7 . 5 using a dual-labeled protein concentration of ~100 pM in the presence of 200 µM unlabeled N130 . To minimize photobleaching , 0 . 5% N-propyl gallate ( final concentration ) was used from a 200x concentrated stock in acetonitrile . For all NMR experiments , the N130 construct was used . Samples of perdeuterated , 15N uniformly labeled , N130 were dissolved in 90% H2O/10% D2O containing 10 mM sodium phosphate buffer at pH 7 . 0 with 150 mM NaCl and 5 mM DTT . All experiments were performed at 298 K . TROSY-HSQC based titration experiments were collected with a Bruker Avance I spectrometer operating at a Larmor frequency of 800 MHz ( Bruker , Billerica , MA ) . 15N backbone relaxation experiments which measured the transverse cross-relaxation and longitudinal relaxation rates were performed on the free and bound N130 at static magnetic fields of 600 , 800 and 1000 MHz . For N130 in the droplet state the same experiments were performed , but only at 600 and 800 MHz . Data were processed and visualized using NMRPipe and CARA , respectively . Fits to all NMR data for the extraction of parameters describing the observed motion were performed using in-house written scripts in Python and Mathematica ( Wolfram Research , Champaign , IL ) . All scripts are available for download at available for download at https://github . com/dlaszlo88/eLIFE-NPM_NMRrelaxation . See below for details on all NMR experiments and analysis . Sedimentation velocity experiments were conducted in a ProteomeLab XL-I analytical ultracentrifuge ( Beckman Coulter , Indianapolis , IN ) following standard protocols unless mentioned otherwise ( Schuck , 2000; Zhao et al . , 2013 ) . The samples in a buffer containing 10 mM NaP pH 7 . 0 , 150 mM NaCl , and 2 mM DTT were loaded into a cell assembly comprised of a double sector charcoal-filled centerpiece with a 12 mm path length and sapphire windows . The cell assembly , containing identical sample and reference buffer volumes of 380 µL , was placed in a rotor and temperature equilibrated at rest at 20 °C for 2 hr before it was accelerated from 0 to 40000 rpm . Rayleigh interference optical data were collected continuously for 12 hr , 400 min for 1:0 and 1:1 and 300 min for 1:2 and 1:3 and analyzed . The velocity data were modeled with diffusion-deconvoluted sedimentation coefficient distributions c ( s ) in SEDFIT ( https://sedfitsedphat . nibib . nih . gov/software/default . aspx ) , using algebraic noise decomposition and with signal-average frictional ratio and meniscus position refined with non-linear regression . The s-value was corrected for time , temperature and radial position and finite acceleration of the rotor was accounted for in the evaluation of Lamm equation solutions ( Zhao et al . , 2015 ) . Maximum entropy regularization was applied at a confidence level of P-0 . 68 . SANS experiments were performed on the extended Q-range small-angle neutron scattering ( EQ-SANS , BL-6 ) beam line at the Spallation Neutron Source ( SNS ) located at Oak Ridge National Laboratory ( ORNL ) . In 30 Hz operation mode , a 4 m sample-to-detector distance with 2 . 5–6 . 1 and 9 . 4–13 . 4 Å wavelength bands was used ( Zhao et al . , 2010 ) to obtain the relevant wavevector transfer , q = 4π sin ( θ ) /λ , where 2θ is the scattering angle . rpL5:N130 samples at 0:1 , 1:1 , 2:1 , and 3:1 mol ratios were prepared in 10 mM Tris , 150 mM NaCl , 2 mM DTT D2O ( pH measured = 7 . 5 ) . The samples were loaded into 2 mm pathlength circular-shaped quartz cuvettes ( Hellma USA , Plainville , NY ) and SANS measurements were performed at 25 °C . Data reduction followed standard procedures using MantidPlot ( Arnold et al . , 2014 ) . The measured scattering intensity was corrected for the detector sensitivity and scattering contribution from the solvent and empty cells , and then placed on absolute scale using a calibrated standard ( Wignall and Bates , 1987 ) . Wide field DIC and 488 nm fluorescence images used for phase separation assays with the N130 construct were collected on a Nikon C1Si microscope ( Nikon Instruments , Melville , NY ) using a 60x 1 . 45 NA magnification oil objective . 10 µl samples were incubated at room temperature for 5 min prior to analysis . Droplets were defined as having an area greater than 9 squared pixels ( 0 . 2 μm/pixel ) and circularity 0 . 5–1 . Particles smaller than 9 squared pixels were visible by fluorescence but not DIC at low protein and peptide concentrations; however , these objects were below the threshold to confidently measure circularity . Therefore , droplets formation was not recorded until particles could be observed by DIC . For ternary mixture assays , the various AlexaFluor488-tagged NPM1 constructs were titrated into the mixtures of rRNA and rpL5 peptide . Confocal images of phase separation droplets were collected on Nikon C1Si ( Nikon Instruments , Melville , NY ) or Zeiss Axio Observer ( Carl Zeiss Microscopy , Jena , Germany ) microscopes , using a 60X 1 . 45 NA or 63X 1 . 40 NA oil magnification objective , respectively , in μ-slide VI0 . 4 6 channel flow cells ( ibidi , Madison , WI ) or CultureWell 16-well chambered coverglass ( Grace Biolabs , Bend , OR ) , coated with PlusOne Repel-Silane ES ( GE Healthcare , Pittsburgh , PA ) . Images of MEFs and isolated nucleoli were recorded on a Zeiss Axio ObserverZ . 1 microscope equipped with a CSU-22 spinning disk ( Yokagawa , Tokyo , Japan ) , Delta Evolve EMCCD camera and 100X 1 . 45 NA oil objective . Nucleoli were prepared from Cos7 cells ( ATCC number CRL 1651 ) using an established protocol ( Rosner et al . , 2013 ) . Briefly , 1 x 107 cells were harvested by trypsin digestion followed by serum inactivation and washing in two exchanges of PBS . Cells were incubated in nucleolar isolation buffer ( NIB; 10 mM Tris , 2 mM MgCl2 , 0 . 5 mM EDTA , pH 7 . 5 ) for 2 min at room temperature followed by 10 min on ice . The cells were disrupted by the addition of NP-40 to a final concentration of 1% . The mixture was centrifuged for 3 min at 500 x g to collect the nuclear fraction , followed by washing in NIB containing NP-40 prior to resuspension in buffer without additional NP-40 . The nuclear suspension was sonicated on ice at 20% power for 12 cycles of 1 s ON / 5 s OFF , using a 1/8’’ tip on a Fisher Scientific Model 505 Sonic Dismembrator ( Thermo Fisher Scientific , Waltham , MA ) . Nucleoli were collected as the pellet by centrifugation at 4 °C , 500 x g , for 3 min and resuspended in NIB . The integrity of purified nucleoli was verified by ( i ) morphology and ( ii ) distribution of established nucleolar proteins via immunostaining and confocal microscopy . We note that incubation in the presence of 150 mM NaCl to match the physiological millieu caused isolated nucleoli to dissolve . Trp53-/- and NPM1-/- / Trp53-/- ( DKO ) MEF cell lines were a kind gift from Drs . C . Sherr ( St . Jude Children’s Research Hospital ) and P . Pandolfi ( Beth Israel Deaconess Medical Center ) . Retroviruses were produced by transfecting the Thy1 . 1-IRES-mCherry-NPM1 variant plasmids or the empty vector into Phoenix cells , using Xfect transfection reagent ( Clontech Laboratories , Mountain VIew , CA ) , in 6 well plates . We note that all Phoenix cells harboring plasmids that encoded for mCherry exhibited red fluorescence under the light microscope . Transfected cells were incubated overnight at 37 °C in a 5% CO2 atmosphere . The supernatant containing viruses was collected , filtered through a 0 . 45 µm seringe filter , and transferred immediately onto 6 well plates seeded with 8 x 104 DKO MEFs , in the presence of 5 µg/mL polybrene . Infections were carried twice a day , for a total of five infections . Virally transduced cells were sorted based on Thy1 . 1 expression using FACS; sorted cells were expanded in culture . MEFs were trypsinized using 0 . 05% trypsin + EDTA prior to resuspension in PBS containing 1% BSA and 1 mM EDTA . Cells were surface labeled with APC-conjugated anti-Thy1 . 1 ( clone OX-7; Biolegend , San Diego , CA ) at [0 . 4 ng/ml] for 30 min on ice . Following incubation , cells were washed three times prior to resuspension in PBS containing 1 mM EDTA and analysis using a Fortessa cytometer ( BD Biosciences , San Jose , CA ) . Enrichment of eGFP-tagged NPM1 constructs in purified nucleoli was determined using Slidebook 6 . 0 ( Intelligent Imaging Innovations , Gottingen , Germany ) . Briefly , each nucleolus was bisected with a line of 80 pixels in length , based on the DIC image . The resulting fluorescence intensity of each pixel along the line was plotted according to pixel position . For the quantification of droplet size , the lower and upper limits of fluorescence intensity were matched for all images in Slidebook 6 . 0 and the particle count was subsequently performed in ImageJ , as follows: 24-bit RGB images were converted into 8-bit images , background subtraction and auto threshold were applied , images were converted to mask and merged objects were separated using the binary watershed operation . Particles were analyzed using Analyze Particles , imposing the size restriction of 5–2500 pixel ( ~100 µm2 ) and the circularity restriction 0 . 2–1 . >90% of all observed droplets had areas within the selected window and were included in the analysis , unless otherwise noted; >50 particles were counted per image . Purified nucleoli were allowed to settle onto poly-D-Lysine-coated chambered coverslips prior to fixation for 10 min with 4% paraformaldehyde in PBS . Nucleoli were subsequently treated for 10 min with cold acetone at -20 °C prior to incubation with blocking buffer ( 20 mM Tris pH 8 . 0 , 100 mM NaCl , 2% bovine serum albumin , 0 . 05% Tween-20 ) for 30 min . The samples were stained overnight at 4 °C in blocking buffer containing anti-Fibrillarin ( Novus Biologicals , Littleton , CO; NB300-269 , 2 . 5 μg/mL ) , anti-rpL5 ( Santa Cruz Biotech , Dallas , TX; sc-103865 , 2 μg/mL ) and anti-NPM ( Abgent , San Diego , CA; AP2834b , 1:200 dilution ) antibodies . The samples were washed in 20 mM Tris pH8 . 0 , 100 mM NaCl and 0 . 05% Tween-20 prior to detection with fluorescently labeled secondary antibodies ( Life Technologies , Carlsbad , CA ) and imaged using a Zeiss Axio Observer Z . 1 equipped with a CSU-22 spinning disk , Delta evolve EMCCD camera and 100X 1 . 45NA oil objective , and Slidebook 6 . 0 ( Intelligent Imaging Innovations , Gottingen , Germany ) . MEFs were seeded onto 4-well chambers at 40000 cells/well and incubated overnight at 37 °C , in 5% CO2 incubator . Cultured MEFs were fixed in 4% methanol-free paraformaldehyde for 10 min at room temperature , followed by incubation in TBS ( 20 mM Tris , 100 mM NaCl ) containing 0 . 3 M glycine for 10 min . The cells were permeabilized for 3 min in TBS containing 0 . 1% Triton-100 , followed by 30 min incubation in TBS containing 2% bovine serum albumin . The slides were incubated overnight at 4 ° with anti-Fibrillarin ( clone 38F3; Genetex , Irvine , CA ) or anti-Nopp140 ( sc-28672 , Santa Cruz Biotech , Dallas , TX ) antibodies each at [1 ng/ml] diluted in TBS + 2% BSA . Primary antibodies were detected with either AF647-labeled goat anti-Mouse ( ThermoFisher ) or AF647-labeled goat anti-Rabbit ( ThermoFisher ) secondary antibodies at [1 ug/ml] for 1 hr at room temperature . Samples were post-fixed in 1% PFA after washing , and prior to imaging in TBS containing 1 ug/ml Hoescht ( Thermo Fisher Scientific , Walthman , MA ) . Images were acquired using a Marianas spinning-disk laser scanning confocal ( Intelligent Imaging Innovations ) comprising a Zeiss AxioObserverZ . 1 equipped with a 63x 1 . 4NA objective and Evolve EMCCD camera ( Photometrics , Tucson , AZ ) .	Inside cells , machines called ribosomes assemble proteins from building blocks known as amino acids . Cells can alter the numbers of ribosomes they produce to match the cell’s demand for new proteins . For instance , when cells grow they require a lot of new proteins and therefore more ribosomes are produced . However , when cells face harsh conditions that cause stress ( e . g . exposure to UV radiation or a harmful chemical ) they generally stop growing and therefore need fewer ribosomes . In human and other eukaryotic cells , ribosomes are assembled in a structure called the nucleolus . However , because the nucleolus is not separated from the rest of the cell by a membrane , it was not clear how it is able to accumulate large quantities of the proteins and other molecules needed to make ribosomes . Recent work suggests that the nucleolus is formed through a process referred to as “phase separation” in which the liquid in a particular region of the cell has different physical properties to the liquid surrounding it . This is like how oil and water form separate layers when mixed . A protein called nucleophosmin is found at high levels in the nucleolus where it interacts with many other proteins , including those involved in making ribosomes . Nucleophosmin binds to motifs within these proteins that contain multiple copies of an amino acid called arginine ( referred to as R-motifs ) . Now , Mitrea et al . investigate how nucleophosmin binds to R-motif proteins and whether this is important for assembling the nucleolus . A search for R-motifs in a list of over a hundred proteins known to bind to nucleophosmin showed that the majority of these proteins contained multiple R-motifs . Furthermore , when high levels of nucleophosmin and the R-motif proteins were present , they underwent phase separation . Next , Mitrea et al . examine the changes in how nucleophosmin and a ribosomal protein interact before and after phase separation . The experiments show that many molecules of nucleophosmin bind to each other and that multiple regions in nucleophosmin are able to interact with the R-motifs . Together , these interactions produce large assemblies of proteins that result in the creation of separate liquid layers . Furthermore , the experiments show that R-motif proteins and other molecules needed to make ribosomes can be brought together within the same liquid phase by nucleophosmin . Mitrea et al . ’s findings provide the first insights into the role of nucleophosmin in the molecular organisation of the nucleolus . The next challenge is to understand how this organisation promotes the production of ribosomes and helps the cell to respond to stressful situations .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2016	Nucleophosmin integrates within the nucleolus via multi-modal interactions with proteins displaying R-rich linear motifs and rRNA
Engineering of GPCR constructs with improved thermostability is a key for successful structural and biochemical studies of this transmembrane protein family , targeted by 40% of all therapeutic drugs . Here we introduce a comprehensive computational approach to effective prediction of stabilizing mutations in GPCRs , named CompoMug , which employs sequence-based analysis , structural information , and a derived machine learning predictor . Tested experimentally on the serotonin 5-HT2C receptor target , CompoMug predictions resulted in 10 new stabilizing mutations , with an apparent thermostability gain ~8 . 8°C for the best single mutation and ~13°C for a triple mutant . Binding of antagonists confers further stabilization for the triple mutant receptor , with total gains of ~21°C as compared to wild type apo 5-HT2C . The predicted mutations enabled crystallization and structure determination for the 5-HT2C receptor complexes in inactive and active-like states . While CompoMug already shows high 25% hit rate and utility in GPCR structural studies , further improvements are expected with accumulation of structural and mutation data . G-protein coupled receptors ( GPCRs ) represent the largest family of transmembrane proteins , which is involved in regulation of all major physiological functions and comprises more than 25% of established therapeutic targets ( Lagerström and Schiöth , 2008; Rask-Andersen et al . , 2014 ) . However , high conformational flexibility and low thermostability of these receptors have always presented major challenges for their structural , biophysical , and biochemical characterization . With exception of visual rhodopsin , structural characterization of all other 50 GPCRs so far required substantial efforts in protein engineering to design GPCR constructs suitable for crystallization ( Cherezov et al . , 2007; Warne et al . , 2009; Chun et al . , 2012; Katritch et al . , 2013; Stevens et al . , 2013; Pándy-Szekeres et al . , 2018 ) . The design typically involves truncations of N- and C- termini , replacements of flexible loops and/or termini with soluble fusion domains ( Chun et al . , 2012 ) , stabilizing co-crystallization partners ( Zhang et al . , 2015 ) , and in many cases introduction of one or more point mutations ( reviewed in ( Heydenreich et al . , 2015 ) ) . Point mutations have shown to be especially important for thermostabilizing GPCR and making them amenable for structure-based drug design applications , which involve receptor co-crystallization with typically low-affinity hit or lead compounds . For example , point mutations were used to thermostabilize β1-adrenergic ( ADRB1 ) and A2A adenosine ( A2AAR ) receptors in both agonist and antagonist bound states , resulting in more than a dozen structures for each receptor including co-crystals with ligands in a micromolar affinity range ( Moukhametzianov et al . , 2011; Christopher et al . , 2013; Warne et al . , 2008; Warne et al . , 2011 ) . In the case of thermostabilized A2AAR , structural and biophysical characterization of initial hits led to structure-based discovery and optimization of preclinical candidates for Parkinson disease ( Langmead et al . , 2012 ) . Moreover , thermostabilized GPCR constructs can streamline biochemical characterization of ligand binding in sensor-based high-throughput screening ( HTS ) ( reviewed in ( Kumari et al . , 2015 ) ) and measurements of ligand-binding kinetics by surface plasmon resonance ( SPR ) ( Christopher et al . , 2013; Congreve et al . , 2011; Rich et al . , 2009 ) . However , currently employed experimental identification of stabilizing mutations by alanine scanning ( Errey et al . , 2015 ) or directed evolution approaches ( Egloff et al . , 2014; Schlinkmann et al . , 2012 ) is a very resource consuming process , and only a few GPCRs have been successfully stabilized so far ( reviewed in ( Vaidehi et al . , 2016 ) ) . Furthermore , stabilizing mutations obtained by these methods have shown very limited transferability between different GPCRs ( Heydenreich et al . , 2015; Serrano-Vega and Tate , 2009 ) , requiring extensive stabilization campaigns to be performed for each individual receptor . Computational approaches could provide a cost- and time-effective alternative for GPCR stabilization . The already existing in silico prediction tools for soluble proteins ( Kumar et al . , 2006; Khan and Vihinen , 2010 ) , however , are not effective for GPCRs because they do not take into account peculiarities of the 7-transmembrane ( 7TM ) nature of the receptors . At the same time , although some of the recently developed GPCR-specific methodologies can be successful in explaining known experimentally-derived mutations ( Vaidehi et al . , 2016; Bhattacharya et al . , 2014 ) , their success in prediction of new stabilizing mutations has been limited so far , and has not resulted yet in successfully solved crystal structures of new GPCRs . In this study , we present a set of complementary approaches for predicting stabilizing mutations in GPCRs combined into a CompoMug tool ( COMputational Predictions Of MUtations in GPCRs ) . CompoMug consists of four modules: knowledge-based , sequence-based , structure-based , and machine-learning-based , taking maximum advantage of accumulated structural and biophysical data . We applied CompoMug to identify stabilizing point mutations for the 5-HT2C receptor , which is an important pharmacological target for the treatment of obesity and neuropsychiatric disorders . Experimental assessment showed that 10 out of the 39 predicted mutations improved stability of the receptor by more than 1 . 5°C , and one mutation resulted in increase of the apparent melting temperature by up to ~8 . 8 ± 1 . 3°C , which is among the highest reported improvements in thermostability by a single point mutation in GPCRs . Moreover , combinations of two or three mutations led to even higher thermostability gains , some of which were compatible with both agonist and antagonist binding . Finally , the mutants predicted by CompoMug allowed for the determination of two 5-HT2C crystal structures in both agonist-bound and antagonist-bound complexes . The CompoMug provides a computational platform for thermostabilization of other GPCRs and can be further evolved with an accumulation of experimental mutation data . The knowledge-based module employs a short list of established point mutations that have been already shown to improve stability and helped to solve structures for multiple GPCRs . Although in general stabilizing point mutations are not transferable across different GPCRs ( Serrano-Vega and Tate , 2009 ) , several specific mutations located in structurally or functionally conserved sites have shown increased chances to be beneficial for multiple receptors . Such known point mutations , listed in Table 1 , could be good candidates for new GPCR targets , even for those with relatively low homology to solved GPCRs . For example , the mutation of a residue in position 3 . 41 to Trp ( X3 . 41W , where X stands for any residue , and superscript shows GPCRdb numbering ( Isberg et al . , 2015 ) ) , first identified in the β2 adrenergic receptor ( Roth et al . , 2008 ) , has been now tested in more than 20 receptors by the GPCR Network ( Stevens et al . , 2013 ) and has shown to increase stability for several of them , helping to solve crystal structures for at least six receptors so far . The list also includes mutations that target residues in the sodium binding pocket , e . g . D2 . 50N , S3 . 39A , and D7 . 49N mutations ( Kruse et al . , 2012; Fenalti et al . , 2014; Katritch et al . , 2014 ) . Sodium ions play an important role in class A GPCR signaling ( Katritch et al . , 2014 ) , and , therefore , modifications in the sodium-binding site , e . g . by D7 . 49N mutation , can decouple ligand binding from conformational changes in the intracellular side of the receptor ( Katritch et al . , 2014; Massink et al . , 2015 ) . Such decoupling apparently reduces conformational heterogeneity of the receptors , resulting in thermostabilization of some receptors , like A2AAR ( White et al . , 2018 ) , and facilitating their structure determination , especially in complexes with agonists ( see Table 1 ) . Note , that while currently only a few mutations in class A can be classified as transferrable ‘knowledge-based’ , the list may continue to grow with an accumulation of additional knowledge on mutations , and also expand to include specific transferrable mutations in other GPCR classes . Algorithmically , we implemented the knowledge-based module as a simple procedure , which checks mutations from Table 1 , and assigns score 1 . 0 if the mutation is potentially applicable ( i . e . the wild type residue in the target GPCR corresponds to a residue in Table 1 ) , and 0 . 0 otherwise . The sequence-based module looks for residues of the target receptor that deviate from a standard conservation pattern in an evolutionarily related group of GPCRs , e . g . receptor orthologs , a subfamily or a branch of the GPCR tree . We hypothesized that such residues in GPCRs are more likely to be destabilizing , and restoring conserved amino acids in such positions might result in receptor stabilization . In CompoMug , the ‘deviation score’ for an amino acid residue is estimated based on multiple sequence alignment ( MSA ) of evolutionary related homologous sequences: ( 1 ) Scorekaa=Ckmax−CkaaNMSA−CkaaCkmax , where NMSA is the total number of sequences in the MSA , Ckmax is the number of sequences with the most conserved amino acid residue at the position k , and Ckaa is the number of sequences that have the same residue aa as the target sequence in this position . As one can see from Equation 1 , the first term is the highest when the target sequence has the most infrequent amino acid in the position k , that is , it approaches 1 , when Ckaa=1 and Ckmax≈NMSA . The second term penalizes the position k if it lacks a dominating conserved amino acid at the position , that is , the penalty is increased as Ckmax is decreased . The total score varies from -1 . 0 to 1 . 0 , where maximum score 1 . 0 is ascribed to a deviating amino acid in a super-conserved position ( e . g . x . 50 in GPCRs ) . In other words , the preference is given to mutations of rarely observed amino acids in the otherwise highly conserved positions . Figure 2A and B schematically show the score computation given an MSA . Apparently , any conservation-related score depends on the set of sequences used to construct the MSA . For example , orthologs share very high sequence similarity with respect to the target GPCR resulting in a few , but usually very clear deviation patterns . On the other hand , comparison with GPCR sequences from different branches has a much more complex conservation pattern that may result in many false positive candidates . To capture the sequence deviations at different levels of GPCR hierarchy , we composed several sets of sequences to construct various MSAs . Specifically , we used five MSAs: ( 1 ) ortholog sequences corresponding to the species variations of the target receptor , ( 2 ) sequences corresponding to the common sub-family ( sequence identity for the TM regions >40% ) , ( 3 ) sequences corresponding to the common GPCR branch ( sequence identity for the TM regions >30% ) , ( 4 ) sequences corresponding to the whole non-olfactory class A GPCR ( Rios et al . , 2015 ) , and ( 5 ) sequences corresponding to the crystallized receptors . MSAs were generated with the structure-based alignment tool of the GPCRdb ( Isberg et al . , 2015 ) , and in case of the whole class A alignment updated using MAFFT software ( Katoh and Frith , 2012 ) . Although the last MSA is not directly related to the evolutionary variation , it may contain information relevant for the GPCR stability and propensity for crystallization . At the same time , the MSA for whole class A GPCR would capture rare variations in the most conserved residue positions of class A , including N1 . 50 , D ( E ) R3 . 50Y , FxxxCWxP6 . 50 and NP7 . 50xxY . Given all five MSAs , we computed positional scores for each MSA , as well as the global score as the average of the individual MSA scores . In a special case of non-conserved Gly residues , we multiplied the ‘deviation score’ by factor of 2 , to account for Gly usually destabilizing effect on α-helical secondary structure in the transmembrane helices of the receptor . Figure 2 schematically shows the workflow of the sequence-based module applied to the 5-HT2C receptor . The structure-based module is focused on identifying pairs of residues , which could form a salt bridge ( also called ionic lock ) when replaced with charged amino acids , or disulfide bonds when replaced with cysteines . Such ionic locks and covalent bonds can help to restrict the conformational flexibility of the receptor and improve stability . A successful use of the structure-based approach requires an accurate 3D structural model , which can be derived based on the close homology with a known crystallographic structure . In this study , structural models were obtained using the template-based homology modeling implemented in ICM-Pro v . 3 . 8 molecular modeling suite ( molsoft . com ) , followed by the backbone regularization and exhaustive Monte-Carlo side-chain refinement in internal coordinates . To predict potential ionic locks in the structural model , the search is performed for pairs of residues that satisfy the following criteria: i ) residues are separated in sequence by at least five residues to exclude pairs of residues belonging to the same α-helix , ii ) side chains point toward each other and do not point to the lipid membrane , iii ) residue’s Cβ-Cβ distance lies in the range from 7 . 0 Å to 10 . 0 Å , and iv ) mutations of residues to at least one of four charged pairs ( E-K , E-R , D-K , D-R ) improve predicted free energy of the receptor after thorough local conformational optimization of the mutants ( Equation 2 ) ( 2 ) Efoldedmut−Eunfoldedmut<Efoldedwt−Eunfoldedwt We used energy calculation implemented in the Molsoft ICM-Pro v . 3 . 8 . software ( molsoft . com ) . The structural model of the mutant type was obtained by mutation of a given residue followed by Monte Carlo sampling of the flexible side chains for the mutated residue and the neighboring residues . Then the free energy of the unfolded and folded states for the wild and mutant types was approximated by a sum of the empirically derived residue-specific energies . In order to predict stabilizing disulfide bonds in the receptor , we first employed the DbD software ( Craig and Dombkowski , 2013 ) to obtain the initial list of candidates . DbD scans all pairs of residues in a protein and selects those that satisfy geometrical parameters of the disulfide bond , when replaced with cysteines . The geometrical parameters , e . g . χ3 angle and Cβ-Cβ distance , were obtained from analysis of protein structures in PDB . Given the DbD predictions , the final list of candidates was derived using the energy criterion implemented in ICM-Pro ( see Equation 2 ) . Figure 3 schematically represents the structure-based module . With the accumulation of experimental data on the stability of GPCR mutants , it becomes feasible to derive powerful prediction models using machine learning techniques . Our prediction model is derived using ( i ) a training benchmark , composed from site-specific mutations performed on GPCRs with known structure , ( ii ) a feature vector , consisting of structure-based and energy-based descriptors , which reflect important changes in the protein upon a point mutation , and ( iii ) a support vector machine method as implemented in the libsvm package ( Chang and Lin , 2011 ) . Each of these steps is described below in details . Given the output predictions from each module , we then filtered out point mutations that may affect ligand binding . For this purpose , we analyzed GPCR-ligand interactions in solved GPCR structures ( Munk et al . , 2016 ) and excluded residue positions that appear in the binding pocket in more than five different class A GPCR structures . We also did not consider predictions in the less conserved regions that lack secondary structure , e . g . loops and N/C – termini , since the modeling accuracy for these regions is much lower , compared to the transmembrane alpha-helical core . The initial training set benchmarking of the CompoMug prediction algorithms was performed with the alanine scanning data available for neurotensin receptor NTS1 ( Shibata et al . , 2009 ) , adenosine receptor AA2AR ( Magnani et al . , 2008 ) , and β1 adrenergic receptor ADRB1 ( Serrano-Vega et al . , 2008; Heydenreich et al . , 2015 ) . Due to the nature of the experimental data , such comparison is limited to only X to A ( where X is any residue ) and A to L point mutations , and the benchmark employed only sequence-based and machine learning modules . For each receptor , we kept top 40 predicted single point mutations and compared the results with the experimental alanine data for the three receptors ( see Supplementary file 1 ) . For the human AA2AR , turkey ADRB1 , and rat NTS1 receptors CompoMug successfully predicts 20 , 11 , and 9 stabilizing mutations out of 39 , 18 , and 20 reported mutations in the transmembrane region , suggesting about 50% recall rate in this initial benchmark . To test the algorithms in a real case of a blind predictions for a new target prospective screening , we applied CompoMug to predict stabilizing point mutations for the serotonin 5-HT2C receptor . The 5-HT2C receptor is widely expressed within the central and the peripheral nervous systems and appears to play a prominent role in psychiatric disorders . Thus , obtaining the structure of this receptor could help for better understanding and treatment of the pathophysiology of obesity and psychiatric disorders including schizophrenia , anxiety , and depression ( Wacker et al . , 2013 ) ( Peng et al . , 2018 ) . To select candidates for point mutations we used the knowledge-based , sequence-based , structure-based and machine-learning modules of CompoMug as described in Computational Methods . In the sequence-based module , we composed five different MSAs ( see Supplementary file 2 ) : orthologs of 5-HT2C receptor , orthologs of all 5-Hydroxytryptamine GPCRs , aminergic receptors ( human only ) , crystallized receptors ( class A only ) , and class A alignment ( non-olfactory ) ( Rios et al . , 2015 ) . For the structure-based module , we first constructed the 5-HT2C homology model based on the structure of the 5-HT2B receptor ( PDB ID 4IB4 ) ( Wacker et al . , 2013 ) . These two serotonin receptor subtypes share 62% of identical residues in the 7TM region ( 49% for the full sequence ) . This structural model was also used to generate 239*19 = 4541 models ( considering 239 residues in the TM regions and 19 possible amino acid substitutions ) with conformationally optimized point mutations as the input for the machine-learning-based module , followed by the score assignment with the derived prediction models . After the post-processing procedure , a list of 39 mutations from different modules was selected for experimental testing , as presented in Table 2 Note , that several mutations were predicted by more than one module . A total of 39 mutations predicted by CompoMug ( see Table 2 ) were tested on the apo 5-HT2C receptor , using the base construct with N- and C- termini truncations and BRIL fusion as described in Experimental Assays section . The optimal insertion position for BRIL , as well as C- and N-terminal truncations were determined experimentally starting from the WT construct ( without mutations ) , as described in the structural paper ( Peng et al . , 2018 ) . For each point mutation , the receptor was expressed in a modified pFastBac1 vector in sf9 insect cells , and the aSEC and CPMs profiles were measured for the unliganded receptor ( apo ) to quantify its thermostability . Point mutations that decreased the receptor expression yield or stability , or for which we could not accurately measure the apparent melting temperature , or did not affect the stability of the protein were disregarded from further experiments . The Tm measurements were repeated for the 10 stabilizing mutations that improved expression and increased apparent melting temperature by at least 1 . 5°C ( bold rows in Table 2 ) . The most remarkable effect was observed for the C3607 . 45N point mutation predicted with the sequence-based module , which increased the thermostability of the receptor by 8 . 8 ± 1 . 3°C in the initial CPM assays . Other mutations showed a moderate effect on thermostability , increasing the apparent melting temperature by 1 . 5–3 . 9°C . Six out of ten mutations are substitutions to the hydrophobic residues ( A , L , or V ) , three point mutations are substitutions to the polar or charged residues ( T , N , or D ) , and one double mutation corresponds to an engineered disulfide bridge ( see Table 2 ) . We also observed that improvements in aSEC and thermostability were well correlated , meaning that point mutations augmented both aSEC quality and apparent melting temperature . After testing single mutations , we devised a list of potentially additive double and triple combinations of point mutations , all of them including the C3607 . 45N mutation . Specifically , we first tested the C3607 . 45N mutation in combinations with all other mutations , as well as double mutation C3607 . 45N-G3627 . 47A in combination with other mutants . These double and triple combinations were tested for the apo receptor and the receptor in complex with different 5-HT2C binding ligands , including an agonist ergotamine and five different antagonists . As Figure 6 shows ( see Figure 6—source data 1 for raw data ) , the tested combinations further improve thermostability of apo receptor , with the maximal observed increase in Tm reaching ~13°C for the triple mutation C3607 . 45N , G3627 . 47A , A1714 . 42L . Moreover , binding of antagonist mesulergine improved thermostability of this triple mutant by additional ~8°C , resulting in a total of 21°C increase in Tm , as compared to the apo base receptor construct . Interestingly , this same triple mutation was destabilized by binding of agonist ergotamine as compared to the apo mutant . In general , while the C3607 . 45N point mutation makes most substantial contribution to the stability of the apo and agonist-bound receptor , the addition of most other point mutations ( except for V2405 . 64A ) predominantly stabilizes the antagonist-bound receptor conformation , which was previously less amenable to crystallization . The biggest contrast between agonist and antagonist bound state thermostability ( ~16°C ) was observed for the quadruple mutant construct with an engineered disulfide bond ( C3607 . 45N , G3627 . 47A , A982 . 49C/A1403 . 38C ) , suggesting that the introduction of the rigid covalent link between the TM2 and TM3 fixes receptor in the inactive conformational state . The predicted stabilizing point mutations made it possible to obtain first crystals of the 5-HT2C receptor in complex with an antagonist , as well as to improve the diffraction of the agonist-bound crystals from >4 Å to <3 . 0 Å , as described in our recent paper ( Peng et al . , 2018 ) . The predicted mutations were introduced in the context of an available 5-HT2C construct that included optimized fusion partner and N- , C- termini truncations . In this context , multiple combinations of CompoMug-derived mutants resulted in diffracting crystals of the 5-HT2C receptor . At the same time , the single C3607 . 45N mutation was found as sufficient to solve structures in complex with agonist ergotamine ( at 3 . 0 Å resolution ) , as well as antagonist ritanserin ( at 2 . 7 Å ) , which is the first antagonist-bound structure of a serotonin receptor ( Peng et al . , 2018 ) . Determination of the crystallographic structure of the 5-HT2C receptor ( Peng et al . , 2018 ) now allows more detailed analysis of the stabilizing nature of the discovered by CompoMug mutations . The mutations were modeled based on the atomic structure of the 5-HT2C receptor as shown in Figure 7 . For example , the A1714 . 42 residue , located at the intracellular side of TM4 , is surrounded by hydrophobic side chains of Y902 . 41 , F912 . 42 , I1754 . 46 , and its replacement with a longer Leu side chain could form more favorable hydrophobic contacts . The G1844 . 55 in the middle of TM4 is exposed to the lipid membrane and does not form any contacts with the side chains , and its replacement with Ala could have a stabilizing effect on the α-helix conformation and more favorable hydrophobic contacts with the lipid environment . The V2405 . 64 residue does not form any specific contacts and it is located close to the membrane intracellular boundary , so the V2405 . 64A mutation may reduce unfavorable contacts with predominantly charged and polar lipid headgroups in this environment . The L3336 . 57 residue points to the membrane and does not form any specific contacts with the neighboring side chains , and the L3336 . 57V might improve stability by forming more favorable hydrophobic contacts with lipids . The C3607 . 45 amino acid is rarely observed at the 7 . 45 position , and it is known that N7 . 45 plays important role in the sodium coordination as a part of the sodium binding pocket ( Katritch et al . , 2014; Liu et al . , 2012 ) . Thus , the C3607 . 45N point mutation restores the conserved residue in the sodium binding pocket and improves the stability of the receptor . Given that this point mutation was necessary to obtain the crystallographic structures of the 5-HT2C receptor in both agonist-bound and antagonist-bound conformations , while D992 . 50N was detrimental , the integrity of the sodium binding pocket in 5-HT2C receptor apparently plays an important role for the overall receptor stability . The G3627 . 47 residue is partially exposed to the lipid environment , thus both the G3627 . 47L/A point mutations improve the stability of the receptor by stabilizing the secondary structure of TM7 and ameliorating hydrophobic interactions with the membrane environment . The I3748 . 49 residue is surrounded by positively charged K83ICL1 , K3738 . 48 , R3768 . 51 , and R3778 . 52 side chains , so the I3748 . 49D/T point mutations may form salt bridges or polar interactions resulting in improved stability of the receptor . Finally , the double mutant A982 . 49C/A1403 . 38C can form a disulfide bridge between TM2 and TM3 , apparently fixing the inactive conformation of the receptor . The latter observation is corroborated by the highest differential in thermostability between antagonist and agonist bound states measured for the combination construct containing the A982 . 49C/A1403 . 38C mutant ( Figure 6 ) . In this study , we present CompoMug - a computational tool to predict stabilizing point mutations in GPCRs . The four modules of CompoMug synergistically use different types of information on known transferable mutations , natural sequence variations , structural interactions , and machine learning of a large dataset of GPCR mutations , respectively , to maximize success rate of predictions . Applied to the 5-HT2C receptor , CompoMug helped us to identify as many as 10 stabilizing mutations ( 25% hit rate ) , supporting the importance of all four modules . One of the predicted mutations , C3607 . 45N , improved the apparent melting temperature of the apo 5-HT2C receptor by 8 . 8 ± 1 . 3°C . Moreover , a triple mutant C3607 . 45N , G3627 . 47A , A1714 . 42L had its thermostability improved by as much as ~13°C , as compared to the base construct apo receptor . Moreover , this C3607 . 45N mutation in the optimal fusion construct yielded crystal structures of the 5-HT2C receptor in two distinct conformations , agonist-bound active like and antagonist-bound inactive . CompoMug is being applied to other receptors of the GPCR family , and performance of its modules can be further improved via the feedback loop with newly generated experimental data .	The trillions of cells in the human body rely on receptors that sit in their cell membranes to communicate with each other . Hundreds of different receptors belong to the G protein-coupled receptor superfamily ( called GPCRs for short ) and play vital roles in the all organs and bodily systems . Indeed , GPCRs are the targets for almost 40% of therapeutic drugs . As such , deciphering the shape and activity of GPCRs is key to understanding the normal workings of the human biology and could help scientists discover new treatments for various diseases , from depression to high blood pressure to cancer . These receptors , however , are notoriously flimsy and unstable , making them difficult to work with in the laboratory . Different approaches have been developed to make GPCRs more stable , usually by swapping one or a few of the amino acid building blocks in the protein for other amino acids . Currently , this requires a costly and slow trial-and-error approach in which each amino acid out of 300-400 in the protein is mutated and tested experimentally . To speed up and reduce the cost of the process , Popov et al . asked if a computer could predict which mutations in the protein would stabilize it , meaning that fewer proteins would actually need to be tested . Four computer algorithms based on four different principles were developed and verified . The first one compares the target GPCR to other closely related receptors , trying to detect variations that cause the instability . The second tries to build in specific stabilizing interactions , or “bridges” , between different parts of the receptor . The third algorithm searches the known structures of other GPCRs for useful mutations . Finally , the fourth one uses accumulated data on the stability of hundreds of mutations in different GPCRs to train a machine learning predictor to recognize stabilizing mutations . All four algorithms produced useful predictions in a real-life project . Indeed , when combined in one computational tool , named CompoMug , the algorithms made it possible to detect optimal mutations in a human GPCR called 5-HT2C . This made the protein much easier to work with in the laboratory , and ultimately helped to solve its three-dimensional structure ( which was reported in a separate study , published earlier in 2018 ) The 5-HT2C receptor is involved in regulating , among other things , mood and appetite . Details of its structure might therefore help researchers to design new antidepressants and obesity treatments . Moreover , CompoMug is already helping structural biologists to solve the structures of other GPCRs , which will further facilitate many aspects of GPCR drug discovery .	[ "Abstract", "Introduction", "Computational", "methods", "Results", "Discussion" ]	[ "structural", "biology", "and", "molecular", "biophysics", "computational", "and", "systems", "biology" ]	2018	Computational design of thermostabilizing point mutations for G protein-coupled receptors
Imprinted X-inactivation silences genes exclusively on the paternally-inherited X-chromosome and is a paradigm of transgenerational epigenetic inheritance in mammals . Here , we test the role of maternal vs . zygotic Polycomb repressive complex 2 ( PRC2 ) protein EED in orchestrating imprinted X-inactivation in mouse embryos . In maternal-null ( Eedm-/- ) but not zygotic-null ( Eed-/- ) early embryos , the maternal X-chromosome ectopically induced Xist and underwent inactivation . Eedm-/- females subsequently stochastically silenced Xist from one of the two X-chromosomes and displayed random X-inactivation . This effect was exacerbated in embryos lacking both maternal and zygotic EED ( Eedmz-/- ) , suggesting that zygotic EED can also contribute to the onset of imprinted X-inactivation . Xist expression dynamics in Eedm-/- embryos resemble that of early human embryos , which lack oocyte-derived maternal PRC2 and only undergo random X-inactivation . Thus , expression of PRC2 in the oocyte and transmission of the gene products to the embryo may dictate the occurrence of imprinted X-inactivation in mammals . X-chromosome inactivation results in the mitotically-stable transcriptional inactivation of one of the two X-chromosomes in female mammals in order to equalize X-linked gene expression between males and females ( Morey and Avner , 2011; Plath et al . , 2002 ) . Two different forms of X-inactivation characterize the mouse embryo , imprinted and random . Imprinted X-inactivation results in the silencing of genes exclusively on the paternal X-chromosome and initiates during preimplantation embryogenesis ( Huynh and Lee , 2003; Mak et al . , 2004; Monk and Kathuria , 1977; Okamoto et al . , 2004; Takagi and Sasaki , 1975 ) . In later stage embryos , imprinted X-inactivation of the paternal-X is stably maintained in the extraembryonic lineage but reversed in the embryonic lineage ( Harper et al . , 1982; Mak et al . , 2004; Okamoto et al . , 2004; Takagi and Sasaki , 1975; West et al . , 1977 ) , which subsequently undergoes random inactivation of either the maternal or the paternal X-chromosome ( Lyon , 1961 ) . Notably , imprinted X-inactivation is a paradigm for both mitotic as well as meiotic , or transgenerational , epigenetic regulation , due to its stable parent-of-origin-specific inactivation pattern . X-inactivation is characterized by a well-defined series of epigenetic events ( Kalantry , 2011 ) . Both imprinted and random X-inactivation are prefaced by the expression of X-linked non-protein coding Xist RNA from the prospective inactive-X ( Kay et al . , 1994; Penny et al . , 1996 ) . During imprinted X-inactivation in the mouse embryo , Xist is expressed at the two-cell stage and the RNA visibly begins to coat the paternal-X at the four-cell stage ( Kalantry et al . , 2009; Namekawa et al . , 2010; Patrat et al . , 2009 ) . The progressive accumulation of Xist RNA coincides with the gradual and stereotyped silencing of paternal X-linked genes that is only completed after the blastocyst stage of embryogenesis ( Kalantry et al . , 2009; Namekawa et al . , 2010; Patrat et al . , 2009 ) . Coincident with Xist RNA coating , Polycomb repressive complex 2 ( PRC2 ) proteins and the PRC2-catalyzed chromatin mark histone H3K27me3 accumulate on the inactive-X , correlating with the silencing of X-linked genes ( Mak et al . , 2004; Okamoto et al . , 2004; Plath et al . , 2003; Silva et al . , 2003 ) . Moreover , the mis-expression of Xist results in the concomitant accumulation of PRC2 proteins and H3K27me3 ( de la Cruz et al . , 2005; Kohlmaier et al . , 2004; Plath et al . , 2003; Silva et al . , 2003 ) , suggesting that Xist RNA directly or indirectly recruits PRC2 to the inactive-X . PRC2 has thus been suggested to contribute to the establishment of X-inactivation ( Plath et al . , 2003; Silva et al . , 2003 ) . Consistent with a role for PRC2 in X-inactivation , we and others previously showed that post-implantation female mouse embryos mutant for the Polycomb gene Eed fail to maintain silencing of paternal X-linked genes during imprinted X-inactivation ( Kalantry and Magnuson , 2006; Kalantry et al . , 2006; Wang et al . , 2001 ) . EED is a non-catalytic component of the PRC2 complex , but EED binding to the PRC2 enzyme EZH2 is required for the full methyltransferase activity of EZH2 ( Cao et al . , 2002; Czermin et al . , 2002; Kuzmichev et al . , 2002; Müller et al . , 2002 ) . When EED is mutated other core PRC2 proteins are degraded and the histone H3K27me3 mark is lost ( Montgomery et al . , 2005 ) . Thus , EED is an essential component of PRC2 and EED function is canonically equated with H3K27me3 catalysis ( Margueron and Reinberg , 2011; Montgomery et al . , 2005 ) . Although Eed-/- embryos fail to maintain imprinted X-inactivation , the mutant embryos initiate imprinted X-inactivation properly ( Kalantry and Magnuson , 2006; Kalantry et al . , 2006 ) . A potential answer for this difference is that Eed-/- embryos inherit maternal EED protein that is present in the oocyte ( Kalantry and Magnuson , 2006; Plath et al . , 2003; Shumacher et al . , 1996 ) . The presence of maternally-derived EED protein could explain the absence of a defect in establishing imprinted X-inactivation in Eed-/- embryos . Such maternal control of imprinted X-inactivation would also be consistent with a transgenerational epigenetic effect that underlies genomic imprinting ( Barlow , 2011; Ferguson-Smith and Bourc'his , 2018; Lee and Bartolomei , 2013; van Otterdijk and Michels , 2016 ) . Here , we test the hypothesis that oocyte-derived PRC2 orchestrates imprinted X-inactivation in the early embryo . PRC2 proteins and H3K27me3 are first enriched on the prospective inactive paternal X-chromosome in the early mouse embryo at the 8–16 cell morula stage ( Okamoto et al . , 2004 ) . We assessed the accumulation of EED , H3K27me3 , and Xist RNA by immunofluorescence ( IF ) combined with RNA fluorescent in situ hybridization ( FISH ) in wild-type ( WT ) embryonic day ( E ) 3 . 5 blastocyst embryos ( Cloutier et al . , 2018; Hinten et al . , 2016 ) , which are in the process of silencing paternal X-linked genes and establishing imprinted X-inactivation ( Borensztein et al . , 2017; Namekawa et al . , 2010; Patrat et al . , 2009; Wang et al . , 2016 ) . As expected , females displayed coincident accumulation of EED , H3K27me3 , and Xist RNA in a vast majority of the nuclei ( 72–100% ) . Males , by contrast , lacked such enrichment ( Figure 1A ) . Our previous work suggested that zygotically-null preimplantation embryos harbor WT maternal EED protein ( Kalantry and Magnuson , 2006; Kalantry et al . , 2006 ) . To test for the presence of maternally-derived EED protein in Eed-/- embryos , we employed our previously generated conditional Eed mutation ( Figure 1—figure supplement 1A ) ( Maclary et al . , 2017 ) . We generated E3 . 0-E3 . 5 blastocyst-stage embryos zygotically-null and heterozygous for Eed ( Eed-/- and Eed+/- , respectively ) from a cross of Eed+/- females with Eedfl/-;Prm-Cre males . Prm-Cre is active during spermatogenesis and catalyzes the deletion of the loxp flanked ( floxed ) Eed allele in the mature sperm ( Figure 1—figure supplement 1B ) ( O'Gorman et al . , 1997 ) . As a result , about half of the embryos generated from the above cross are expected to be genotypically Eed-/- and the other half Eed+/- . In the derived embryos , we assayed inactive-X enrichment of EED , H3K27me3 , and Xist RNA by combined IF/FISH ( Figure 1B ) . Of the 41 female embryos examined , nine showed coincident accumulation of EED and/or H3K27me3 with Xist RNA in over 70% of the nuclei and were not significantly different from WT embryos in Figure 1A ( p>0 . 1 ) . An additional nine embryos were devoid of EED or H3K27me3 enrichment overlapping with the Xist RNA coat . We presumed the former to be Eed+/- embryos and the latter to be Eed-/- embryos . The remaining 23 embryos displayed 2–70% of nuclei with EED and/or H3K27me3 enrichment . This intermediate class likely represents Eed+/- or Eed-/- embryos that had not yet fully depleted maternally-inherited EED protein or Eed+/- embryos which had not yet robustly expressed zygotic EED . Male embryos from the cross , distinguished by a lack of Xist RNA coating , did not show enrichment of EED or H3K27me3 in the nucleus , as in the WT male embryos in Figure 1A . To confirm that there is no bias in the sex ratio or genotype of the embryos , we performed PCR genotyping of embryos derived from the above cross ( Figure 1C ) . Embryos from 12 litters showed no statistical difference in the distribution of Eed+/- and Eed-/- male or female embryos ( p>0 . 05 ) , suggesting that the intermediate class of 23 embryos in Figure 1A are likely a mixture of Eed+/- or Eed-/- embryos . Together , the results in Figure 1 suggest that genotypically null Eed-/- embryos inherit oocyte-derived maternal EED protein and that expression of EED transitions from maternal to zygotic at or slightly before the blastocyst stage . To define the kinetics of depletion of maternal EED and induction of zygotic EED prior to the blastocyst stage , we quantified EED and H3K27me3 nuclear IF signals in 2- , 4- , 8- , and 16 cell embryos from the following series of crosses . The first was Eedfl/fl females crossed to Eedfl/fl males , which yielded control Eedfl/fl embryos . The second was a cross of Eedfl/- females to Eedfl/fl;Prm-Cre males to generate Eedfl/- and Eed-/- embryos ( Eedfl/- / Eed-/- ) . Whereas both Eedfl/- and Eed-/- embryos are expected to harbor maternal EED protein , Eedfl/- but not Eed-/- embryos would express zygotic EED . The third cross was of Eedfl/fl;Zp3-Cre females to WT males to yield embryos that are devoid of maternal EED ( Eedm-/- ) but which are capable of expressing zygotic EED . Zp3-Cre is active in the growing oocyte , where it efficiently deletes the Eedfl allele and generates embryos devoid of maternal EED ( Figure 5C and Figure 2—figure supplement 1A ) ( Lewandoski et al . , 1997 ) . The final cross was a cross of Eedfl/fl;Zp3-Cre females with Eedfl/fl;Prm-Cre males to generate embryos devoid of both maternal and zygotic EED ( Eedmz-/- ) . Eedfl/fl and Eedfl/- / Eed-/- 2-cell embryos exhibited similar levels of EED and H3K27me3 , whereas Eedm-/- and Eedmz-/- embryos were devoid of both EED and H3K27me3 ( Figures 2A , C and D; Supplementary file 1 ) . These data are consistent with the 2-cell embryo harboring only maternally-derived EED and H3K27me3 . Four-cell embryos displayed a similar pattern to 2-cell embryos , although a subset of Eedfl/- / Eed-/- ~4-cell embryos displayed reduced EED and H3K27me3 levels , consistent with expression of zygotic EED beginning at or slightly before this stage and its failure in Eed-/- embryos ( Figure 2C and Figure 2—figure supplement 1B; Supplementary file 1 ) . At the ~8-cell stage , Eedfl/- / Eed-/- embryos showed highly variable EED and H3K27me3 levels , suggesting further differentiation of the two genotypes . In agreement with increasing zygotic Eed expression , Eedm-/- ~8-cell embryos displayed higher levels of EED and H3K27me3 than the corresponding Eedmz-/- embryos ( Figure 2C and Figure 2—figure supplement 1B; Supplementary file 1 ) . By the ~16-cell stage , Eedfl/- / Eed-/- embryos were clearly separated into two categories . One group had statistically lower levels of EED , while the other group was statistically indistinguishable from Eedfl/fl embryos ( Figures 2B , C and D; Supplementary file 1 ) . Therefore , the likely genotypes of the two groups are Eed-/- and Eedfl/- , respectively . Eedm-/- 16-cell embryos continued to display higher levels of EED and H3K27me3 than the Eedmz-/- embryos , but nevertheless harbored significantly lower EED and H3K27me3 levels than Eedfl/fl embryos ( Figures 2B , C and D; Supplementary file 1 ) . In order to visualize how EED levels are changing across early embryogenesis , we plotted the mean fluorescence intensity values of EED for each genotype by embryonic stage ( Figure 2E ) . Maternally-derived EED starts declining at ~4-cell stage but is still present at the 16-cell stage . Conversely , while zygotic Eed transcription initiates at ~4-cell stage , zygotic EED levels are still low in ~16-cell embryos , suggesting that EED in WT Eedfl/fl 16-cell embryos is a combination of maternally-derived and zygotically generated protein ( Figure 2F ) . To test if zygotic Eed-/- embryos initiate and establish imprinted X-inactivation of the paternal X-chromosome , we compared X-linked gene expression in an allele-specific manner in individual hybrid Eedfl/fl , Eedfl/- , and Eed-/- E3 . 5 blastocysts by RNA sequencing ( RNA-Seq ) ( Figure 3—figure supplement 1A ) . In these embryos , the maternal X chromosome was derived from the Mus musculus 129/S1 mouse strain and the paternal-X from the divergent Mus molossinus JF1/Ms strain ( Materials and methods ) . We exploited single nucleotide polymorphisms ( SNPs ) to assign RNA-Seq reads to either the maternal or paternal X-chromosome in the hybrid embryos ( Cloutier et al . , 2018; Maclary et al . , 2017 ) . A subset of X-linked genes was expressed more robustly from the paternal allele relative to the maternal allele in Eedfl/- and Eed-/- female embryos compared to Eedfl/fl embryos ( Figure 3A; Supplementary file 2 ) . However , when the allelic expression ratio of all X-linked genes in Figure 3A was averaged , paternal X-linked gene expression was not significantly higher in Eed-/- blastocysts compared to Eedfl/- ( p = 0 . 72 ) or Eedfl/fl ( p = 0 . 76 ) female embryos ( Figure 3B and Figure 3—figure supplement 1B; Supplementary file 2 and Supplementary file 3 ) . X-linked genes were expressed predominantly from the maternal allele in all three genotypes . Thus , the ratio of maternal:paternal X-linked gene expression in Eed-/- female blastocysts was broadly similar to that in Eedfl/fl and Eedfl/- embryos . We next sought to validate the RNA-Seq data via Pyrosequencing . Pyrosequencing is a low-throughput technique that can accurately capture allelic expression ratios of individual genes ( Cloutier et al . , 2018; Gayen et al . , 2015 ) . We analyzed the expression of Xist and three X-linked genes subject to X-inactivation , Rnf12 , Atrx , and Pgk1 . Xist expression analysis by Pyrosequencing was especially important , as there was variability in Xist SNP-overlapping read coverage in the RNA-Seq data due potentially to the highly repetitive sequence of Xist RNA . We did not detect any significant changes in maternal:paternal allelic expression in hybrid Eed-/- vs . Eedfl/fl and Eedfl/- blastocysts ( Figure 3C and Figure 3—figure supplement 1C; Supplementary file 4 ) . Whereas Xist was expressed predominantly from the paternal allele , Rnf12 , Atrx , and Pgk1 were preferentially expressed from the maternal allele in all three genotypes . As an independent validation of the RNA-Seq and Pyrosequencing results , we also performed RNA FISH to test Xist RNA coating and nascent RNA expression of Rnf12 in Eed-/- and Eedfl/fl female ( Figure 3D ) and male ( Figure 3—figure supplement 1D ) blastocysts . RNA FISH has the added benefit of providing single cell expression resolution in embryos ( Cloutier et al . , 2018; Hinten et al . , 2016 ) . We distinguished Eedfl/fl from Eed-/- female embryos by assaying H3K27me3 enrichment by IF on the Xist RNA-coated X-chromosome ( Figure 3D and E ) . We classified embryos displaying fewer than 5% of the nuclei with this H3K27me3 enrichment as Eed-/- ( Figure 3E ) . Xist RNA coating and Rnf12 expression in female Eed-/- embryos did not differ significantly from Eedfl/fl blastocysts ( Figure 3D and F ) . Both sets of embryos displayed Xist RNA coating of one X-chromosome and Rnf12 expression from the other X-chromosome in a majority of the cells . Male Eed-/- or Eed+/- embryos also did not differ significantly from Eedfl/fl embryos in their Rnf12 expression patterns ( Figure 3—figure supplement 1D ) . Thus , by three independent assays – allele-specific RNA-Seq , Pyrosequencing , and RNA FISH – zygotic Eed expression appears to be largely dispensable for the initiation and establishment of imprinted X-inactivation . Since early Eed-/- embryos harbor WT oocyte-derived EED protein , we next examined the role of maternal EED in initiating imprinted X-inactivation in Eedm-/- and Eedmz-/- blastocysts , which are devoid of oocyte-derived EED . Eedm-/- blastocysts exhibited a small percentage of nuclei with H3K27me3 enrichment coinciding with the Xist RNA coat ( Figure 4A ) . Eedmz-/- blastocysts , on the other hand , lacked all such overlapping accumulation ( Figure 4A ) . H3K27me3 enrichment on the Xist RNA-coated X-chromosome in Eedm-/- but not Eedmz-/- blastocysts is likely due to the expression of zygotic Eed in Eedm-/- but not Eedmz-/- embryos ( Figure 2 ) . To test if maternal EED regulates imprinted X-inactivation , we conducted allele-specific RNA-Seq on individual hybrid Eedm-/- and Eedmz-/- E3 . 5 blastocysts ( Figure 4—figure supplement 1A ) . Strikingly , the RNA-Seq data revealed a relative increase in paternal X-linked gene expression in Eedm-/- and Eedmz-/- embryos compared to Eedfl/fl , Eedfl/- , and Eed-/- embryos ( Figure 4B and C , and Figure 4—figure supplement 1B; Supplementary file 2 and Supplementary file 3 ) . Furthermore , Eedmz-/- embryos appeared to express paternal X-linked genes to a greater degree compared to Eedm-/- embryos ( Figure 4B ) . When allelic expression ratios of all X-linked genes in Figure 4B were averaged , however , the difference between Eedm-/- and Eedmz-/- embryos did not reach statistical significance ( p=0 . 14 ) ( Figure 4C; Supplementary file 3 ) . The shift in the ratio of X-linked gene expression towards the paternal allele in Eedm-/- and Eedmz-/- embryos could be due to increased paternal X-linked gene expression or to decreased maternal X-linked gene expression . To determine the source of the expression change , we calculated the normalized expression of genes on the maternal and paternal X-chromosomes for all genotypes ( Figure 4D and Figure 4—figure supplement 1C ) . Whereas paternal X-linked genes significantly increased in expression , maternal X-linked gene expression decreased in Eedm-/- and Eedmz-/- embryos compared to Eedfl/fl , Eedfl/- , and Eed-/- embryos . The increase in paternal X-linked gene expression in Eedm-/- and Eedmz-/- embryos was significant when compared to the three other genotypes . The decrease in maternal X-linked gene expression in Eedm-/- and Eedmz-/- embryos reached significance only vs . Eedfl/fl embryos and not vs . Eedfl/- and Eed-/- embryos . The lack of a significant decrease between Eedm-/- and Eedmz-/- embryos compared to Eedfl/- and Eed-/- embryos is likely due to the greater variation in maternal X-linked gene expression in Eedfl/- and Eed-/- embryos ( Supplementary file 3 ) . Finally , Eedmz-/- embryos displayed a significant increase in paternal X-linked gene expression compared to Eedm-/- embryos ( p=0 . 02; Supplementary file 2 and Supplementary file 3 ) , suggesting that zygotic EED can contribute to the silencing of a subset of X-linked genes in blastocysts . To validate the Eedm-/- and Eedmz-/- blastocyst RNA-Seq data , we again analyzed allele-specific expression of Xist , Rnf12 , Atrx , and Pgk1 in E3 . 5 blastocysts by Pyrosequencing . Pyrosequencing also showed a significant defect in the initiation and establishment of imprinted X-inactivation in Eedm-/- and Eedmz-/- embryos ( Figure 4E and Figure 4—figure supplement 1D; Supplementary file 4 ) . In Eedm-/- and Eedmz-/- embryos , Xist expression unexpectedly increased from the maternal-X relative to the paternal-X . Conversely , the expression of Rnf12 and Atrx increased from the paternal-X relative to the maternal-X in Eedm-/- embryos . In Eedmz-/- embryos , in addition to Rnf12 and Atrx , Pgk1 also displayed nearly equal levels of expression from the maternal and paternal alleles . The Pyrosequencing results thus recapitulate the defects in imprinted X-inactivation observed by RNA-Seq . Together , the RNA-Seq and Pyrosequencing data lead to several suggestions . The first is that maternal EED depletion in the oocyte induces Xist from the maternal X-chromosome in the early embryo . This derepression is consistent with maternal PRC2 repressing the maternal Xist locus , which is marked by H3K27me3 in the oocyte [Figure 4—figure supplement 1E; ( Zheng et al . , 2016 ) . Ectopic Xist induction from the maternal-X then results in the silencing of genes on that X-chromosome . The second major suggestion is that loss of maternal EED induces paternal X-linked genes . Finally , the data implicate zygotic EED expression in the silencing of a subset of paternal X-linked genes at the onset of imprinted X-inactivation . To validate the RNA-Seq and Pyrosequencing data from the maternal Eed mutants , we performed RNA FISH in Eedm-/- and Eedmz-/- blastocysts for Xist and Rnf12 ( Figure 5A ) . Whereas most nuclei in Eedm-/- and Eedmz-/- females displayed a single Xist RNA coat and monoallelic expression of Rnf12 , a subset displayed Xist RNA coating of both X-chromosomes . The majority of these nuclei also lacked Rnf12 expression , suggesting silencing of Rnf12 on both X-chromosomes . We similarly examined Eedmz-/- male blastocysts ( Figure 5B ) . A subset of nuclei in Eedmz-/- male mutant embryos also exhibited ectopic Xist RNA coating of their sole , maternally-inherited X-chromosome . Interestingly , Eedmz-/- male embryos were present in two distinct morphological classes . The first category was comprised of large , well-developed embryos , which displayed few or no nuclei with Xist RNA coating . The second category consisted of underdeveloped embryos , which displayed Xist RNA-coating in much higher proportions ( 20–60% of nuclei ) . In both sets of embryos , Xist RNA coating was often accompanied by a loss of Rnf12 expression from the Xist RNA-coated X-chromosome . These data suggest that Xist RNA coating hinders developmental progression by silencing genes on the ectopically Xist RNA-coated X-chromosome . Eedmz-/- embryos that adaptively repress Xist may overcome this developmental deficiency . The correlation between reduced frequency of ectopic Xist RNA-coated nuclei and development of Eedmz-/- embryos led us to test the developmental competency of maternal-null Eed embryos . We assessed if Eedm-/- embryos could yield live born animals . To our surprise , a small number of Eedm-/- female as well as male embryos could live to term ( Figure 5C ) , suggesting that the ectopic Xist RNA expression and coating could be resolved in maternal-null embryos of both sexes . Interestingly , significantly more females were born compared to males ( p=0 . 02 , Two-tailed Student’s T-test ) , suggesting that females can more robustly extinguish ectopic Xist RNA expression compared to males . These data further suggest that zygotic EED expression is sufficient to compensate for the absence of maternal EED in a subset of the early embryos . Eedmz-/- embryos are expected to be inviable , since loss of zygotic Eed expression results in lethality of both female and male embryos ( Faust et al . , 1995; Shumacher et al . , 1996; Wang et al . , 2001 ) . The relative paucity of ectopic Xist RNA-coated nuclei in female Eedm-/- and Eedmz-/- blastocysts observed by RNA FISH in Figure 5A–B is inconsistent with the robust ectopic Xist RNA expression from and silencing of maternal X-linked genes and the increased expression of paternal X-linked genes that were detected via Pyrosequencing and RNA-Seq ( Figure 4B–D ) . We thus postulated that instead of undergoing imprinted inactivation of the paternal X-chromosome , Eedm-/- and Eedmz-/- blastocysts switch to random X-inactivation of either the maternal- or the paternal-X in individual cells . Such mosaicism would explain the silencing of maternal X-linked genes and the induction of paternal X-linked gene expression in Eedm-/- and Eedmz-/- female embryos detected by RNA-Seq and Pyrosequencing . To test the above model of X-inactivation mosaicism , we developed and applied an allele-specific Xist RNA FISH strategy on hybrid control Eedfl/+ and test Eedm-/- female E3 . 5 blastocysts ( Materials and methods; Figure 6—figure supplement 1 ) . Allele-specific Xist RNA FISH allowed us to discriminate Xist RNA expression from the maternal vs . the paternal X-chromosome in individual cells . Allele-specific Xist RNA FISH displayed Xist RNA expression from only the paternal-X in Eedfl/+ female blastocysts ( Figure 6A ) , as would be expected from embryos stably undergoing imprinted X-inactivation of the paternal-X . In Eedm-/- female blastocysts , however , we saw a mosaic distribution of Xist RNA expression and coating . Whereas some Eedm-/- blastocyst nuclei displayed Xist RNA expression from and coating of the maternal-X , others exhibited Xist RNA expression from and coating of the paternal-X . A subset of nuclei in Eedm-/- blastocysts exhibited Xist RNA expression from both the maternal and paternal X-chromosomes ( Figure 6A ) , consistent with the non-allele specific Xist RNA FISH data from Eedm-/- and Eedmz-/- blastocysts in Figure 5A . Male Eedm-/- embryos similarly displayed ectopic Xist RNA expression from and coating of their sole maternally-inherited X-chromosome in approximately 50% of nuclei ( Figure 6B ) . From the blastocyst data , we extrapolated that earlier Eedm-/- embryos may harbor a higher proportion of cells with ectopic Xist RNA coating of the maternal-X . This pattern would be later resolved into the mosaic Xist RNA coating pattern observed at the blastocyst stage in females and loss of the Xist RNA coat in males . We therefore performed allele-specific Xist RNA FISH on 3–16 cell control Eedfl/+ and test Eedm-/- hybrid embryos . In the Eedfl/+ female embryos , Xist RNA was expressed from and coated only the paternal X-chromosome ( Figure 7A ) . Most Eedm-/- female embryos , by contrast , displayed a high percentage of nuclei with Xist RNA expression and coating of both X-chromosomes ( Figure 7A ) . In male 3–17 cell embryos , Eedfl/+ embryos did not show any nuclei with Xist RNA coating ( Figure 7B ) . In Eedm-/- male embryos , by contrast , almost every nucleus exhibited ectopic Xist expression from and coating of the maternally-inherited X-chromosome ( Figure 7B ) . Thus , in the absence of maternal EED most cells express Xist from both X-chromosomes in early female embryos and from the sole X in early male embryos . By the blastocyst stage , however , one of the two Xist alleles is stochastically silenced in most female cells and the sole Xist allele is silenced in most male cells . Intriguingly , the Xist RNA coating of both X-chromosomes in female and of the single X in male early preimplantation Eedm-/- and Eedmz-/-mouse embryos resemble the pattern observed in preimplantation human female and male embryos ( Okamoto et al . , 2011; Petropoulos et al . , 2016 ) . In early preimplantation human embryos , females display Xist RNA coating of both Xs and males of their sole maternally-inherited X-chromosome . We therefore hypothesized that the Xist RNA expression profile in early human embryos may reflect the absence of maternally-derived EED and other core PRC2 proteins in human oocytes . To test this hypothesis , we analyzed RNA-Seq data from mouse and human oocytes to determine the expression levels of core PRC2 genes Eed , Ezh2 , Ezh1 , and Suz12 ( Kobayashi et al . , 2012; Macfarlan et al . , 2012; Reich et al . , 2011 ) . Compared to mouse oocytes , human oocytes expressed all four genes at negligible levels ( Figure 8A ) . This difference in the expression of PRC2 components in oocytes may underlie why early mouse but not human embryos undergo imprinted X-inactivation . Genomic imprinting is a paradigm of transgenerational epigenetic inheritance , since the two parental alleles undergo diametrically divergent transcriptional fates in a parent-of-origin-specific manner in the embryo . Imprinted X-inactivation is an extreme example of genomic imprinting in that most genes on the paternally-inherited X-chromosome undergo silencing . The maternal X-chromosome , by contrast , remains active . Here , we test the role of core PRC2 protein EED in the initiation of imprinted X-inactivation during early mouse embryogenesis . We defined the transition of maternal to zygotic EED expression in the early embryo and found the presence of maternal and a relative absence of zygotic EED when imprinted X-inactivation begins . Upon ablation of Eed in the oocyte and the absence of maternally-derived EED in the embryo , the initiation of imprinted X-inactivation is compromised ( Figure 8B ) . Maternal-null ( Eedm-/- and Eedmz-/- ) but not zygotic-null ( Eed-/- ) early preimplantation female and male embryos ectopically induced Xist RNA from the maternal X-chromosome . Early Eedm-/- female embryos therefore display Xist RNA-coating of both X-chromosomes and Eedm-/- mutant males of the sole maternally-inherited X-chromosome . PRC2-catalyzed H3K27me3 marks the Xist locus on the maternal X-chromosome during oogenesis ( Zheng et al . , 2016 ) . In agreement , the injection of the H3K27me3 demethylase Kdm6b in the zygote resulted in the derepression of the Xist locus on the maternal X-chromosome in 8–16 cell embryos ( Inoue et al . , 2017 ) . Female morulas derived from Kdm6b-injected zygotes displayed Xist RNA coating of both the maternal and the paternal X-chromosome in most blastomeres , suggesting inactivation of both Xs in the embryo . Nullizygosity of X-linked gene expression due to inactivation of both Xs in females or of the single-X in males is expected to result in cell and embryo lethality ( Gayen et al . , 2015 ) . The conditional deletion of Eed in the oocyte , however , yielded live born mice , implying that ectopic Xist expression due to H3K27me3 loss and the ensuing inactivation of the maternal-X in the early embryo is resolved later [this study; ( Prokopuk et al . , 2018 ) ] . In agreement , our study shows that by the blastocyst stage most nuclei in Eedm-/- and Eedmz-/- female embryos exhibit only one Xist RNA coat . However , instead of Xist RNA coating exclusively of the paternal X-chromosome as in WT embryos , the maternal Eed mutants express Xist RNA from and coat either the maternal or the paternal X-chromosome , a hallmark of random X-inactivation . This randomization persists later in development in extraembryonic tissues ( data not shown ) , which normally maintain imprinted inactivation of the paternal-X . Like Eedm-/- and Eedmz-/- females , Eedm-/- and Eedmz-/- male blastocysts also extinguish ectopic Xist induction . In addition to maternal EED , our data argue that zygotically generated EED contributes to imprinted X-inactivation in the early embryo . In comparison to Eedm-/- embryos , Eedmz-/- female blastocysts displayed a further increase in paternal X-linked gene expression . One interpretation of these data is that the onset of zygotic EED expression results in the preferential installation of H3K27me3 at the Xist locus on the maternal-X in some cells of early Eedm-/- embryos . These cells thus forestall or extinguish Xist expression from the maternal X-chromosome and inactivate the paternal-X , ultimately resulting in more cells in the embryo in which the paternal-X is inactive compared to the maternal-X . Loss of both maternal and zygotic EED would annul such biased inactivation of the paternal-X and thereby cause a greater increase in paternal X-linked gene expression in Eedmz-/-vs . Eedm-/- embryos . An alternative possibility is that zygotic EED functions to maintain silencing preferentially of paternal X-linked genes in the early embryo . The differential sensitivity of genes on the maternal vs . paternal X-chromosomes to zygotic EED in Eedm-/- embryos may reflect the different kinetics of inactivation of the two X-chromosomes . The ectopic induction of Xist and X-linked gene silencing on the maternal-X may occur more slowly compared to that on the paternal-X . Due to this delay , genes on the maternal-X would still be in the process of undergoing silencing in Eedm-/- blastocysts . A subset of paternal X-linked genes , on the other hand , may have established silencing and are now in the maintenance phase of X-inactivation in the blastocysts . In the absence of both maternal and zygotic EED , then , Eedmz-/- blastocysts fail to maintain silencing of these paternal X-linked genes . Our previous work has shown that zygotic EED is in fact required to maintain silencing of a discrete set of paternal X-linked genes during imprinted X-inactivation ( Kalantry and Magnuson , 2006; Kalantry et al . , 2006; Maclary et al . , 2017 ) . The ability of the cells of early Eedm-/- and Eedmz-/- embryos to resolve Xist RNA coating of both Xs in females or of the single X in males implies that the early mouse embryo has an X-chromosome counting mechanism that ensures that a single X-chromosome remain active in females as well as in males , irrespective of its parent of origin . Such a counting mechanism has previously been proposed by Takagi and colleagues to explain the kinetics of Xist RNA induction in XX and XY androgenetic embryos , which harbor only paternal X-chromosomes ( Okamoto et al . , 2000 ) . Like in Eedm-/- embryos , androgenetic 4 and 8–16 cell embryos also initially induce Xist RNA from all Xs , which is resolved at the blastocyst stage and results in females displaying a single Xist RNA coat in most nuclei and males exhibiting few or no nuclei with Xist RNA coating ( Okamoto et al . , 2000 ) . Molecular sensing of the X-chromosomal complement in imprinted X-inactivation is also suggested by studies of diploid XX parthenogenetic or gynogenetic embryos , which harbor two maternal X-chromosomes . In these preimplantation bi-maternal XX embryos , Xist expression is delayed and appears to occur stochastically from one or the other X-chromosome ( Kay et al . , 1994 ) . In agreement , the extraembryonic tissues of post-implantation XX parthenogenotes display hallmarks of random X-inactivation instead of the imprinted form observed in WT extraembryonic cells ( Rastan et al . , 1980 ) . Randomization of X-inactivation in extraembryonic cells of mouse embryos with two paternal or maternal X-chromosomes led Takagi and colleagues to suggest that imprinted X-inactivation in placental mammals may have arisen from random X-inactivation ( Matsui et al . , 2001 ) , a notion that our data from Eedm-/- and Eedmz-/- embryos agree with . Evidence suggests that the X-linked Rnf12 gene may be a key component of the X-chromosome counting mechanism during imprinted X-inactivation . The maternal-X allele of Rnf12 is required to induce Xist from the paternal-X in preimplantation mouse embryos ( Shin et al . , 2010 ) . Upon Xist RNA coating , Rnf12 is rapidly silenced on the paternal X-chromosome ( Kalantry et al . , 2009; Namekawa et al . , 2010; Patrat et al . , 2009 ) . In Eedm-/- and Eedmz-/- embryos , in addition to the paternal Rnf12 allele , the maternal Rnf12 allele is also stringently silenced due to ectopic Xist RNA coating of the maternal-X . Since Rnf12 is required for Xist RNA induction in the preimplantation embryo , the silencing of all Rnf12 alleles in Eedm-/- and Eedmz-/- female and male embryos may paradoxically lead to the loss of Xist RNA expression from both Xs in females or from the sole X-chromosome in males . In females , this transient state of two active-Xs may then be followed by random X-inactivation , analogously to how differentiating pluripotent epiblast cells undergo random X-inactivation ( Gayen et al . , 2015; Maclary et al . , 2014; Mak et al . , 2004 ) . The X-chromosome counting process and randomization of X-inactivation in the early embryo may explain how Eedm-/- embryos can yield live born animals [this study and ( Prokopuk et al . , 2018 ) . In the course of preparing this manuscript , a publication reported that extraembryonic tissues of Eed maternal-null female post-implantation embryos exhibit random X-inactivation ( Inoue et al . , 2018 ) . The primary piece of data in the study supporting this conclusion is the expression of maternal and paternal X-linked genes , including Xist , in post-implantation E6 . 5 female Eedm-/- extraembryonic tissues by allele-specific RNA-Seq . Although in agreement with our conclusions , the study did not directly demonstrate when imprinted X-inactivation switches to random X-inactivation and whether loss of zygotic Eed would result in a similar outcome . Our study , by contrast , genetically dissects the relative contributions of maternal vs . zygotic EED in the initiation and establishment of imprinted X-inactivation . We are thus able to pinpoint when and how the loss of maternal EED converts imprinted X-inactivation to random X-inactivation in preimplantation embryos . Genetically testing the requirement of maternal vs . zygotic EED is necessary to determine that the establishment of imprinted X-inactivation in the preimplantation embryo is maternally but not zygotically controlled . Xist RNA expression in Eedm-/- mouse embryos mimics the pattern observed in human embryos , which do not undergo imprinted X-inactivation and ultimately display only random X-inactivation ( Okamoto et al . , 2011; Petropoulos et al . , 2016 ) . In agreement , like the Eedm-/- and Eedmz-/-oocytes , human oocytes are devoid of expression of Eed , as well as expression of the other core PRC2 genes , suggesting that the presence or absence of maternal PRC2 or related chromatin modifying proteins may dictate whether placental mammals undergo imprinted X-inactivation . This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All animals were handled according to protocols approved by the University Committee on Use and Care of Animals ( UCUCA ) at the University of Michigan ( protocol #s PRO6455 and PRO8425 ) . Mice harboring a conditional mutation in Eed were described in our prior publication ( Maclary et al . , 2017 ) . A Mus molossinus JF1 X-chromosome was introgressed to generate Eedfl/fl;XJF1Y males . Mus musculus Eedfl/fl females were backcrossed onto the 129/S1 background . The X-linked Gfp transgenic ( X-Gfp ) and JF1 strains have been described previously ( Hadjantonakis et al . , 1998; Kalantry and Magnuson , 2006; Kalantry et al . , 2006; Kalantry et al . , 2009; Maclary et al . , 2017 ) . Embryos generated for the purpose of allele-specific RNA sequencing ( RNA-Seq ) , Pyrosequencing , or allele-specific RNA fluorescence in situ hybridization ( FISH ) were sired by males harboring the XJF1 X-chromosome . Embryos generated for immunofluorescence ( IF ) and non-allele specific RNA FISH were sired by males harboring the X-Gfp transgene . The paternal X-Gfp is only transmitted to daughters . Thus , GFP fluorescence conferred by the paternally-transmitted X-Gfp transgene was used to sex the embryos . For derivation of embryos lacking zygotic Eed , the Protamine-Cre ( Prm-Cre ) transgene was bred into an Eedfl/fl or Eedfl/- background . Prm-Cre is expressed only during spermatogenesis ( O'Gorman et al . , 1997 ) , thus resulting in the deletion of the Eed floxed allele in the male germline . For derivation of embryos lacking maternal EED , a Cre transgene controlled by the Zona pellucida three gene promoter ( Zp3-Cre ) ( Lewandoski et al . , 1997 ) , was used to delete the floxed Eed alleles in growing oocytes . Embryonic day ( E ) 3 . 5 embryos were isolated essentially as described ( Maclary et al . , 2014 ) . Embryos were flushed from the uterine limbs in 1X PBS ( Invitrogen , #14200 ) containing 6 mg/ml BSA ( Invitrogen , #15260037 ) . Two to sixteen cell embryos were flushed from oviducts of superovulated females with 1X PBS ( Invitrogen , #14200 ) containing 6 mg/ml BSA ( Invitrogen , #15260037 ) or M2 medium ( Sigma , #M7167 ) . For superovulation , 4–5 week-old , or 9–12 week-old females were treated with 5 IU of pregnant mare’s serum gonadotropin ( PMSG , Sigma , # G-4877 ) and 46 hr later with 5 IU of human chorionic gonadotropin ( hCG , Sigma , #CG-5 ) . Embryos were harvested 48–74 hr post hCG . The zona pellucida surrounding embryos was removed through incubation in cold acidic Tyrode’s solution ( Sigma , #T1788 ) , followed by neutralization through several transfers of cold M2 medium ( Sigma , #M7167 ) . Isolated E3 . 5 embryos were either lysed for RNA isolation or plated onto 0 . 2% gelatin- ( Sigma , #G2500 ) and/or 0 . 01% Poly-L-Lysine ( PLL , Sigma , # P4707 ) -coated glass coverslips ( 22mm X 22mm , Thermo Fisher Scientific , #12548B ) in 0 . 25X PBS for immunofluorescence ( IF ) coupled with RNA in situ hybridization ( FISH ) . 2–16 cell embryos were plated on coverslips coated in 0 . 01% Poly-L-Lysine for IF . E3 . 5 or 4–16 cell embryos were plated on coverslips coated with 1X Denhardt’s ( Sigma , #D9905 ) solution for allele-specific RNA FISH . For plated embryos , excess solution was aspirated , and coverslips were air-dried for approximately 15–30 min . After drying , embryos were permeabilized and fixed in 50 µL solution of either 0 . 05% or 0 . 1% Tergitol ( Sigma , #NP407 ) with 1% paraformaldehyde ( Electron Microscopy Sciences , #15710 ) in 1X PBS for 5 min , followed by 1% paraformaldehyde in 1X PBS for an additional 5 min . Excess solution was gently tapped off onto paper towels , and coverslips were rinsed 3X with 70% ethanol and stored in 70% ethanol at −20°C prior to IF and/or RNA FISH . For embryo DNA isolation , embryos were isolated as described above . Individual blastocysts were lysed in 15 µL buffer composed of 50 mM KCl , 10 mM Tris-Cl ( pH 8 . 3 ) , 2 . 5 mM MgCl2 , 0 . 1 mg/mL gelatin , 0 . 45% NP-40 , 0 . 45% Tween-20 , and 0 . 4 mg/mL Proteinase K ( Fisher , #BP1700 ) . Embryos in lysis buffer were incubated at 50°C overnight , then stored at 4°C until use . Genomic PCR used 1–3 µL lysate per sample . Reactions for Eed were carried out in ChromaTaq buffer ( Denville Scientific ) with 2 . 5 mM MgCl2 added . XX vs . XY sexing PCR reactions were carried out in Klentherm buffer ( 670 mM Tris pH 9 . 1 , 160 mM ( NH4 ) SO4 , 35 mM MgCl2 , 15mg/ml BSA ) . Both used RadiantTaq DNA polymerase ( Alkali Scientific , #C109 ) . Primer sequences are described in Supplementary file 5 . Liveborn animals from the cross of Eedfl/fl;Zp3-Cre female by WT male were genotyped for Eed to confirm deletion of the floxed allele . Ear punches were taken after weaning and lysed in 50 µL of lysis buffer ( above ) . Ear punches were incubated at 50°C overnight , then stored at 4°C until use . 1 µL of DNA lysate was used per reaction . Eed PCRs were carried out as above . Allele-specific expression was quantified using the Qiagen PyroMark sequencing platform , as previously described ( Gayen et al . , 2015 ) . Briefly , the amplicons containing SNPs were designed using the PyroMark Assay Design software . cDNAs were synthesized using Invitrogen SuperScript III One-Step RT-PCR System ( Invitrogen , #12574–026 ) . Following the PCR reaction , 5 µL of the 25 µL reaction was run on a 3% agarose gel to assess the efficacy of amplification . The samples were then prepared for Pyrosequencing according to the standard recommendations for use with the PyroMark Q96 ID sequencer . All amplicons spanned intron ( s ) , thus permitting discrimination of RNA vs . any contaminating genomic DNA amplification due to size differences . Control reactions lacking reverse transcriptase for each sample were also performed to rule out genomic DNA contamination . E3 . 5 embryos of similar sizes for all genotypes were used in the Pyrosequencing assays . Pyrosequencing primer sequences are described in Supplementary file 5 . Embryos mounted on gelatin- , PLL- , and/or PLL/gelatin-coated glass coverslips were washed 3 times in 1X PBS for 3 min each while shaking . Coverslips were then incubated in blocking buffer consisting of 0 . 5 mg/mL BSA ( New England Biolabs , #B9001S ) , 50 µg/mL yeast tRNA ( Invitrogen , #15401–029 ) , 80 units/mL RNAseOUT ( Invitrogen , #10777–019 ) , and 0 . 2% Tween 20 ( Fisher , #BP337-100 ) in 1X PBS in a humid chamber for 30 min at 37°C . The samples were next incubated with primary antibody diluted in blocking buffer for 45 min −2 hr in the humid chamber at 37°C . The samples were then washed 3 times in 1X PBS/0 . 2% Tween 20 for 3 min each while shaking . After a 5 min incubation in blocking buffer at 37°C in the humid chamber , the samples were incubated in blocking buffer containing fluorescently-conjugated secondary antibody for 30 min in the humid chamber at 37°C , followed by three washes in PBS/0 . 2% Tween 20 while shaking for 3 min each . For samples undergoing only IF staining , DAPI was added to the third wash at a 1:250 , 000 dilution . Coverslips were then mounted on slides in Vectashield ( Vector Labs , #H-1000 ) . For samples undergoing IF combined with RNA FISH , the samples were processed for RNA FISH following the third wash . Antibody information is described in Supplementary file 5 . RNA FISH with double-stranded and strand-specific probes was performed as previously described ( Gayen et al . , 2015; Hinten et al . , 2016; Kalantry et al . , 2009 ) . The Rnf12 dsRNA FISH probe was made by random-priming using BioPrime DNA Labeling System ( Invitrogen , #18094011 ) and labeled with Cy3-dCTP ( GE Healthcare , #PA53021 ) using a previously described fosmid template ( Kalantry et al . , 2009 ) . Strand-specific Xist probes were generated from templates as described ( Maclary et al . , 2014; Sarkar et al . , 2015 ) . Probes were labeled with Fluorescein-12-UTP ( Roche , #11427857910 ) or Cy5-CTP ( GE Healthcare , #25801087 ) . Labeled probes from multiple templates were precipitated in a 0 . 5M ammonium acetate solution ( Sigma , #09691 ) along with 300 µg of yeast tRNA ( Invitrogen , #15401–029 ) and 150 µg of sheared , boiled salmon sperm DNA ( Invitrogen , #15632–011 ) . The solution was then spun at 15 , 000 rpm for 20 min at 4°C . The pellet was washed consecutively with 70% ethanol and 100% ethanol while spinning at 15 , 000 rpm at room temperature . The pellet was dried and resuspended in deionized formamide ( VWR , #97062–010 ) . The probe was denatured by incubating at 90°C for 10 min followed by an immediate 5 min incubation on ice . A 2X hybridization solution consisting of 4X SSC and 20% Dextran sulfate ( Millipore , #S4030 ) was added to the denatured solution . All probes were stored in the dark at −20°C until use . Following IF , embryos mounted on coverslips were dehydrated through 2 min incubations in 70% , 85% , 95% , and 100% ethanol solutions and subsequently air-dried . The coverslips were then hybridized to the probe overnight in a humid chamber at 37°C . The samples were then washed 3 times for 7 min each at 37°C with 2X SSC/50% formamide , 2X SSC , and 1X SSC . A 1:250 , 000 dilution of DAPI ( Invitrogen , #D21490 ) was added to the third 2X SSC wash . Coverslips were then mounted on slides in Vectashield ( Vector Labs , #H-1000 ) . Allele-specific Xist RNA FISH probes were generated as described ( Levesque et al . , 2013 ) . Briefly , a panel of short oligonucleotide probes were designed to uniquely detect either the M . musculus or the M . molossinus alleles of Xist ( Supplementary file 5 ) . Five probes were designed for each Xist allele . Each probe overlapped a single nucleotide polymorphism ( SNP ) that differs between the two strains , with the SNP located at the fifth base pair position from the 5’ end . The same panel of five SNPs was used for both sets of allele-specific probes . The 3’ end of each oligonucleotide probe is fluorescently tagged using Quasar dyes ( Biosearch technologies ) . M . musculus-specific oligos were labeled with Quasar 570 and M . molossinus-specific oligos labeled with Quasar 670 . In addition to labeled SNP-overlapping oligonucleotides , a panel of 5 ‘mask’ oligonucleotides were also synthesized . These ‘mask’ probes are complimentary to the 3’ end of the labeled allele-specific probes and will hybridize to the allele-specific oligonucleotides , leaving only 9–10 base pairs of sequence surrounding the polymorphic site available to initially hybridize to the target Xist RNA . Since this region of complementarity is short , the presence of a single nucleotide polymorphism is sufficient to destabilize the hybridization with the alternate allele . Sequences of detection and mask probes are listed in Supplementary file 5 . Allele-specific Xist RNA FISH probes were combined with a strand-specific Xist RNA probe , labeled with Fluorescein-12-UTP ( Roche , #11427857910 ) , which served as a guide probe that hybridizes to Xist RNA generated from both Xist alleles and ensured the fidelity of the allele-specific probes in detecting the cognate Xist RNA molecules . The guide Xist RNA probe was first ethanol precipitated as previously described , then resuspended in hybridization buffer containing 10% dextran sulfate , 2X saline-sodium citrate ( SSC ) and 10% formamide . The precipitated guide RNA probe was then mixed with the M . musculus and M . molossinus detection probes , to a final concentration of 5 nM per allele-specific oligo , and 10 nM mask probe , yielding a 1:1 mask:detection oligonucleotide ratio . Coverslips were hybridized to the combined probe overnight in a humid chamber at 37°C . After overnight hybridization , samples were washed twice in 2X SSC with 10% formamide at 37°C for 30 min , followed by one wash in 2X SSC for five min at room temperature . A 1:250 , 000 dilution of DAPI ( Invitrogen , #D21490 ) was added to the second 2X SSC with 10% formamide wash . Coverslips were then mounted on slides in Vectashield ( Vector Labs , #H-1000 ) . Stained samples were imaged using a Nikon Eclipse TiE inverted microscope with a Photometrics CCD camera . The images were deconvolved and uniformly processed using NIS-Elements software . For four color images ( blue , green , red , and white ) , the far-red spectrum was employed for the fourth color ( AlexaFluor 647 secondary antibody and Cy5-UTP labelled riboprobes for RNA FISH ) . Additional antibody information is outlined in Supplementary file 5 . EED and H3K27me3 IF intensity quantifications were performed using the ‘3D Measurement; 3D thresholding , 3D viewing and voxel based measurements’ Nikon Elements software package ( Nikon Instruments , 77010582 ) . Individual nuclei were marked by creating a binary image , using the ‘Threshold’ function , over the DAPI stain of the nuclei . Each nucleus was designated as a Region of Interest ( ROI ) by converting the binary image to an ROI . An additional polygonal ROI was manually created over a non-nuclear region , which was thensubtracted from the nuclear fluorescence intensity . For each channel , average intensity of each nucleus was taken as the intensity measurements from individual ROIs . These intensity values of individual nuclei of an embryo were then averaged to get the average intensity per embryo . Embryos with 2–3 cells were categorized as being at the ~2-cell stage in development . The ~4-cell stage encompassed embryos with 4–5 cells . Embryos with 6–10 cells were classified as being at the ~8-cell stage in development , and the ~16-cell stage encompassed embryos with 14–19 cells . To preserve IF intensities , the images of embryos were not deconvolved . Intensity data for individual nuclei is presented in Figure 2—source data 1 . The Threshold function of the software cannot always distinguish between two nuclei that are overlapping . Similarly , if a single nucleus is an odd shape , it may be counted as multiple nuclei by the software . Some embryos therefore had different numbers of nuclei measured than the number of cells in the embryo . If the number of cells in an embryo differs from the number of nuclei listed , the actual number of cells is indicated in parenthesis next to the embryo label in Figure 2—source data 1 . mRNA was isolated from whole embryos using the Dynabeads mRNA DIRECT Kit ( Thermo Fisher , # 610 . 11 ) according to the manufacturer’s instructions . E3 . 5 embryos of similar sizes of all genotypes were used for RNA-Seq . Eedfl/- and Eed-/- embryos were genotyped by Eed RT-PCR and all embryo genotypes were confirmed by quantifying the relative expression of the floxed Eed exon seven to the sample’s number of mapped reads ( Figure 3—figure supplement 1 and Figure 4—figure supplement 1 ) . Samples were submitted to the University of Michigan DNA Sequencing Core for strand-specific library preparation using the Takara SMARTer Seq V4 stranded low input kit ( Takara , #634889 ) . All libraries were sequenced on the Illumina HiSeq2000 or HiSeq4000 platforms to generate 50 bp paired-end reads . Quality control analysis of the RNA-Seq data was conducted using FastQC . SNP data from whole-genome sequencing of the 129/S1 ( M . musculus ) and JF1/Ms ( M . molossinus ) mouse strains were substituted into the mm9 mouse reference genome build ( C57BL/6 J ) using VCFtools to generate in silico 129/S1 and JF1/Ms reference genomes ( Keane et al . , 2011; Maclary et al . , 2017; Takada et al . , 2013; Yalcin et al . , 2011 ) . Sequencing reads were separately mapped to each of the two in silico genomes using STAR ( Dobin et al . , 2013 ) , allowing 0 mismatches in mapped reads to ensure allele-specific mapping of SNP-containing reads to only one strain-specific genome . STAR was selected for read mapping , in part due to the improved ability to handle structural variability and indels , with the goal of reducing mapping bias to the genome most similar to the reference genome . STAR is a spliced aligner capable of detecting structural variations and is able to handle small insertions and deletions during read mapping . STAR additionally permits soft-clipping of reads during mapping , trimming the ends of long reads that cannot be perfectly mapped . This function would permit clipping of reads that end near indels , thus preserving mappability at SNPs near indels . Prior work showed that the variability due to mapping bias between the 129/S1 and JF1/Ms genomes is minimal in our RNA-Seq analysis pipeline ( Maclary et al . , 2017 ) . Although small biases may affect allelic mapping at a subset of SNP sites within a gene , the effect is mitigated since most genes contain multiple SNPs ( Figure 3—figure supplement 1 ) . For allelic expression analysis , only RNA-Seq reads overlapping known SNP sites that differ between the 129/S1 and JF1/Ms genomes were retained . All multi-mapping reads were excluded from the allele-specific analysis . For each SNP site , reads mapping to the 129/S1 and JF1/Ms X chromosomes were counted and the proportion of reads from each X chromosome identified . Allelic expression was calculated individually for each SNP site; for genes containing multiple SNPs , the paternal-X percentage for all SNPs was averaged to calculate gene-level allelic expression . All SNP sites with at least 10 SNP-overlapping reads were retained . Genes containing at least one SNP site with at least 10 SNP-overlapping reads were retained for further analysis and are referred to in the text as informative . In X-linked genes , the SNP frequency is ~1 SNP/250 bp in transcribed RNAs ( Keane et al . , 2011; Maclary et al . , 2017; Takada et al . , 2013; Yalcin et al . , 2011 ) . To calculate expression from the maternal vs . paternal X-chromosomes , all reads were first merged into a single alignment file and the number of reads per RefSeq annotated gene was counted using HTSeq . To calculate the percentage of expression arising from the paternal X-chromosome , the total read counts from HTSeq were normalized by number of mapped reads . Then , the normalized number of mapped reads for each gene was multiplied by the proportion of SNP-containing reads mapping to the paternal X-chromosome . This analysis was done in R using the following formula:{totalreads× ( paternalreadsmaternalreads+paternalreads ) } For analysis of publicly available oocyte RNA-Seq data , raw Fastq files were obtained from the NCBI Sequence Read Archive . Quality control analysis was conducted using FastQC . Reads were aligned to the mm9 ( mouse ) or hg19 ( human ) reference genome using STAR ( Dobin et al . , 2013 ) and counted using FeatureCounts ( Liao et al . , 2014 ) . BioProject and Run numbers for samples analyzed are listed here . Human oocyte RNA-SeqMouse oocyte RNA-SeqBioProject IDRun numberBioProject IDRun numberPRJNA146903SRR351336PRJDB21DRR001701PRJNA146903SRR351337PRJDB21DRR001702PRJEB8994ERR841204PRJNA154207SRR385627 Welch’s two-sample T-tests were used to test for significant differences between the means of Pyrosequencing and RNA-Seq allelic expression data . This test was chosen due to the unequal variance and sample sizes between different genotype groups . In the RNA-Seq allelic expression significance tests , the average percent paternal expression of all informative X-linked genes was calculated for each sample . The total paternal expression value for each genotype group was obtained by calculating the mean of the informative percent paternal values for all samples in that genotype group . A two-tailed Student’s T-test was used to determine the significance of RNA FISH and IF data . All barplots and heatmaps were made using the ggplot and Pheatmaps R packages , respectively . Dotplots were made using Python’s Seaborn package . Only genes that were informative in all samples were included in the heatmaps .	Almost every one of our cells , with a few exceptions , contains the complete set of genes needed to build and maintain the human body . Yet , not all of these genes are active in every cell . Instead , some genes are tagged for activation , while others are silenced . These changes do not alter the genetic code , only how it is read by the cell , and are collectively referred to as epigenetics . Female mammals have two X-chromosomes compared to males' one . As such , females will silence one of those chromosomes to avoid getting a double-dose from those genes located on the X-chromosome . This epigenetic process is called X-chromosome inactivation , and it lasts for the life of the animal . Epigenetic information can also be passed on to future generations . In early female embryos of mice , for example , it is always the X-chromosome inherited from the father that is suppressed , which suggests that the instructions for which X-chromosome to inactivate must have come from the parents . Harris , Cloutier et al . set out to dissect the mechanics of the specialised form of X-chromosome inactivation seen in female embryos of mice , which is known as imprinted X-inactivation . A protein called EED was suspected to play a key role . Embryos inherit EED protein from the mother's egg , so it was reasoned that this protein may be the epigenetic link between the generations . The cascade of epigenetic events leading to imprinted X-inactivation in the early embryo has been well-defined , but the role of maternal EED was yet to be tested . The experiments showed that the mother's EED protein was needed to silence the father's X-chromosome in female mouse embryos . Without EED from the mother's egg , early embryos failed to initiate imprinted X-inactivation and reverted instead to random X-inactivation , where either X-chromosome is chosen for silencing in female cells . This pattern resembles what happens in early human embryos , which are unable to undergo imprinted X-inactivation because a woman's eggs lack the EED protein . Together these new findings trace the passage of epigenetic information from parent to offspring at the molecular level . With evidence like this , scientists can better understand mechanisms of non-genetic inheritance more broadly , including from parent to offspring .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2019	Conversion of random X-inactivation to imprinted X-inactivation by maternal PRC2
Within neuroscience , psychology , and neuroimaging , the most frequently used statistical approach is null hypothesis significance testing ( NHST ) of the population mean . An alternative approach is to perform NHST within individual participants and then infer , from the proportion of participants showing an effect , the prevalence of that effect in the population . We propose a novel Bayesian method to estimate such population prevalence that offers several advantages over population mean NHST . This method provides a population-level inference that is currently missing from study designs with small participant numbers , such as in traditional psychophysics and in precision imaging . Bayesian prevalence delivers a quantitative population estimate with associated uncertainty instead of reducing an experiment to a binary inference . Bayesian prevalence is widely applicable to a broad range of studies in neuroscience , psychology , and neuroimaging . Its emphasis on detecting effects within individual participants can also help address replicability issues in these fields . Within neuroscience , psychology , and neuroimaging , the common experimental paradigm is to run an experiment on a sample of participants and then infer and quantify any effect of the experimental manipulation in the population from which the sample was drawn . For example , in a psychology experiment , a particular set of stimuli ( e . g . visual or auditory stimuli ) might be presented to a sample of human participants , who are asked to categorise the stimuli or perform some other task . Each participant repeats the procedure several times with different stimuli ( experimental trials ) , and their responses and reaction times are recorded . In a neuroimaging experiment , the same procedure is employed with neuroimaging signals recorded in addition to behavioural responses . The researcher analyses these responses to infer something about brain function in the population from which the participants were drawn . In this standard experimental paradigm , the implicit goal is usually to determine the presence of a causal relationship between the experimental manipulation and the response of interest . For example , between a stimulus property and the neural activity in a particular brain region , or between neural activity and a behavioural measure ( e . g . accuracy , reaction time ) . A properly controlled experimental design in which other extraneous factors are held constant ( i . e . a randomised control trial ) enables a causal interpretation of a correlational relationship ( Angrist and Pischke , 2014; Pearl , 2009 ) . We use statistical tools to evaluate the measured effect and ensure that we are not being fooled by randomness—that is , it should be unlikely our observed result was produced by random fluctuations with no effect present . This is often formalised as null hypothesis significance testing ( NHST ) . We reject the null hypothesis ( often that the population mean is zero ) when the probability of observing a given effect size ( or larger ) is less than some prespecified value ( often p=0 . 05 ) if the null hypothesis was true ( i . e . if the population mean really was zero ) . Simply stated , we would be unlikely to obtain the observed effect size if the null hypothesis was true . Researchers performing such studies usually wish to infer something about the population from which the experimental participants are selected ( Holmes and Friston , 1998 ) , rather than about the specific sample of participants that were examined ( as in a case study ) . Importantly , any statistical inference from a sample to the population requires a model of the population itself . The ubiquitous approach used in psychology and neuroimaging is to model the effect in the population with a normal distribution and perform inference on the mean of this model: the population mean ( see Materials and methods ) . For example , the null hypothesis is often that the population mean is zero , and the probability of the observed sample data is computed under this assumption , taking into account the variance between individuals in the sample of participants . However , an alternative and equally valid question is to ask how typical is an effect in the population ( Friston et al . , 1999b ) . In this approach , we infer an effect in each individual of the sample , and from that infer the prevalence of the effect in the population—that is , the proportion of the population that would show the effect , if tested ( Allefeld et al . , 2016; Donhauser et al . , 2018; Friston et al . , 1999a; Rosenblatt et al . , 2014 ) . The results obtained using the population mean ( vs . population prevalence ) can differ , particularly when effects are heterogenous across participants . Here , we argue that in many experimental applications in psychology and neuroscience , the individual participant is the natural replication unit of interest ( Little and Smith , 2018; Nachev et al . , 2018; Smith and Little , 2018; Thiebaut de Schotten and Shallice , 2017 ) . This is because many aspects of cognition , and the neural mechanisms underlying them , are likely to vary across individuals . Therefore , we should seek to quantify effects and ensure that our results can be reliably distinguished from chance within individual participants . We argue that with such a shift in perspective towards experimental assessment within individual participants , we should also shift our statistical focus at the population level from NHST of the population mean to estimating the population prevalence: the proportion of individuals in the population who would show a true positive above-chance effect in a specific experiment . This can be thought of as the expected within-participant replication probability . Although we focus here on a broad class of experiments in psychology and neuroimaging that feature human participants and non-invasive recording modalities , the arguments we present are general and apply equally well to other experimental model organisms or sampling units . To support this shift in perspective , we present a simple Bayesian method to estimate population prevalence based on the results of within-participant NHST , including prevalence differences between groups of participants or between tests on the same participants . This approach can also be applied without explicit within-participant NHST to estimate prevalence of different effect sizes , giving a new view on what can be learned about a population from an experimental sample . We suggest that applying Bayesian population prevalence in studies that are sufficiently powered within individual participants could address many of the recent issues raised about replicability in psychology and neuroimaging research ( Benjamin et al . , 2018; Ioannidis , 2005 ) . Bayesian prevalence provides a population prevalence estimate with associated uncertainty and therefore avoids reducing an entire experiment to a binary NHST inference on a population mean ( McShane et al . , 2019 ) . To illustrate , we simulate data from the standard hierarchical population model that underlies inference of a population mean effect based on the normal distribution ( Friston , 2007; Holmes and Friston , 1998; Penny and Holmes , 2007 ) ( see Materials and methods ) . We emphasise that this standard hierarchical population simulation is the simplest possible illustrative example , intended to demonstrate the different perspective of within-participant vs . population mean inference . The model simulated here is used implicitly in almost every random-effects inference employed in psychology and neuroimaging studies , in which the tested null hypothesis is that the population distribution is normal with zero mean and non-zero variance . We also emphasise that the simulated data values within individuals are considered to represent the effect of a controlled experimental manipulation—that is , a causal relationship . However , this could be any causal relationship that we can quantify in psychology , neuroscience , or neuroimaging . For example , we could consider a within-participant difference in reaction time between two classes of stimuli , a coefficient or contrast from a linear model ( e . g . a General Linear Model of fMRI data ) , a cross-validated out-of-sample predictive correlation from a high-dimensional stimulus encoding model ( e . g . a model predicting auditory cortex MEG responses to continuous speech stimuli ) , a rank correlation of dissimilarity matrices in a Representational Similarity Analysis , or parameters of computational models of decision making ( e . g . the Diffusion Drift Model ) or learning . When evaluating the results of the simulations of Figure 1 , it is important to keep in mind that these results are meant to represent any statistical quantification of any experimental manipulation . Figure 1 illustrates the results of four simulations that differ in the true population mean , µ ( A , B: μ=0; C , D , μ=1 ) and in the number of trials performed per participant ( A , C: 20 trials; B , D: 500 trials ) . For each simulation , we performed inference based on a standard two-sided , one-sample t-test against zero at two levels . First , we applied the standard summary statistic approach: we took the mean across trials for each participant and performed the second-level population t-test on these per-participant mean values . This provides an inference on the population mean , taking into account between-participant variation . This is equivalent to inference on the full hierarchical model in a design where participants are separable ( Holmes and Friston , 1998; Penny and Holmes , 2007; Penny et al . , 2003 ) ( see Materials and methods ) . The modelled population distribution is plotted as a solid curve , coloured according to the result of the population mean inference ( orange for significant population mean , blue for non-significant population mean ) . Second , we performed inference within each participant , applying the t-test on within-participant samples , separately for each participant . The sample mean ± s . e . m . is plotted for each participant ( orange for significant participants , blue for non-significant participants ) . The population mean inference correctly fails to reject the null hypothesis for Figure 1 panels A and B ( ground truth μ=0 ) and correctly rejects the null in panels C and D ( ground truth μ=1 ) . But consider carefully Figure 1B , C . With 500 trials in panel B , 32/50 participants ( orange markers ) show a significant within-participant result . The probability of this happening by chance , if there was no effect in any members of the population , can be calculated from the cumulative density function of the binomial distribution . In this case , it is tiny—for a test with false positive rate α=0 . 05 , and no effect in any individual , p<2 . 2×10-16 ( below 64-bit floating-point precision ) . Compare that to p=0 . 008 for the population t-test in panel C . Thus , the panel B results provide much stronger statistical evidence of a non-zero effect at the population level: the observed results are very unlikely if no individuals in the population have a non-zero effect in this test . This statistical evidence would be ignored in analyses based only on the population mean . Considering inference at the individual level , panel C results ( 11/50 significant ) have p=4 . 9×10-6 if there was no effect within any individuals ( i . e . the proportion of the population showing an effect was zero ) . Thus , even panel C , which simulates an experiment with only 20 trials per participant , provides stronger evidence for a population effect from the within-participant perspective than from the population mean perspective . Obviously , these two different p-values are calculated from two different null hypotheses . The population t-test tests the null hypothesis that the population mean is zero , assuming that individual means follow a normal distribution:H0: μpop=0 while the p-value for the number of significant within-participant tests comes from the null hypothesis that there is no effect in any individual in the population , termed the global null ( Allefeld et al . , 2016; Donhauser et al . , 2018; Nichols et al . , 2005 ) :H0:μi=0 for all i where i indexes all members of the population . These are testing different questions , but importantly both are framed at the population level and both provide a population-level inference . Clearly , the global null itself is quite a blunt tool . The goal of our Bayesian prevalence method is to quantify within-participant effects at the population level in a more meaningful and graded way . We agree that it is important to align ‘the meaning of the quantitative inference with the meaning of the qualitative hypothesis we’re interested in evaluating’ ( Yarkoni , 2020 ) . Often , when the goal of the analysis is to infer the presence of a causal relationship within individuals in the population , the within-participant perspective may be more appropriate . We argue that performing NHST at the individual participant level is preferable for conceptual reasons in psychology and neuroimaging , but also for practical reasons related to the replicability crisis ( see Discussion ) . Although we have so far considered only the global null , the simulations show that the within-participant perspective can give a very different impression of the evidence for a population-level effect from a data set . The results in Figure 1B represent strong evidence of a non-zero effect within many participants , albeit one that is approximately balanced across participants between positive and negative directions . We do not intend to suggest that such two-sided effects are particularly common but would highlight how they are implicitly assumed in traditional population mean inference models . Figure 2 provides further examples of how the within-participant and population mean approaches can diverge in situations with effects in one direction . In fact , if the researcher believes that all effects in an experiment are positive , then the standard null model illustrated here is inappropriate . The traditional population mean t-test is actually providing a fixed-effects analysis based on the global null rather than the random-effects interpretation of generalisation to the population that is normally applied ( Allefeld et al . , 2016 ) . The p-values under the global null are obtained from the cumulative density function of the binomial distribution , based on a within-participant false positive rate α=0 . 05 . However , we can also model the number of above-threshold individuals in a sample when the population prevalence of true positive ( significant ) test results is γ ( see Materials and methods ) . Consider a within-participant test with a false positive rate α . In this case , the distribution of the number of significant test results follows a binomial distribution with success probability θ= ( 1−γ ) α+γ . Here , we present a Bayesian approach to estimate the population prevalence of true positive test results , γ , from this binomial model of within-participant testing . Alternatively , frequentist inference approaches can be used ( Allefeld et al . , 2016; Donhauser et al . , 2018; Friston et al . , 1999a ) . Note that we are not estimating the population prevalence of the ground truth binary status of the tested effect . We could only obtain a lower bound on this quantity because there might be individuals with an effect too small to detect with our test . Therefore , here we focus throughout on the prevalence of true positive test results—that is , the proportion of the population we would expect to give a true positive test result given the specific experimental test procedure considered . The Bayesian approach provides a full posterior distribution for γ , from which we can obtain the maximum a posteriori ( MAP ) estimate , together with measures of uncertainty—for example , highest posterior density intervals ( HPDIs ) or lower bound posterior quantiles . Figure 1E shows the Bayesian posteriors , MAPs , and HPDIs for the four simulated data sets in Figure 1A–D . Even though there is no population mean effect in Figure 1B , the MAP estimate of the prevalence is 0 . 62 ( 96% HPDI: [0 . 47 0 . 76] ) . Given the data , the probability that the population prevalence is greater than 47% is higher than 0 . 96 . Based on this result , we would consider it highly likely that more than 47% of the population would show a true positive effect if tested in the same experiment with 500 trials . Figure 2 demonstrates two situations where Bayesian prevalence gives a different statistical assessment of the presence of an effect in the population compared to a second-level t-test for a non-zero population mean . We simulated a simple EEG experiment with 100 trials repeated in each of 20 participants . Template Gaussian event-related potentials ( ERPs ) are added on each trial with fixed width and uniformly distributed amplitude , with a specific peak time per participant . Both within-participant and second-level t-tests are Bonferroni corrected over the 600 time points considered . Figure 2A shows a simulation where all participants have an effect , with peak times drawn from a uniform distribution over the 100–400 ms range . There is strong evidence for an effect in each participant , and the estimated prevalence is high: 1 [0 . 85 1] ( MAP [96% HPDI] ) . However , because the effects are not aligned in time across participants , there are no time points at which the null hypothesis of zero population mean can be rejected . In the simulation shown in Figure 2B , only 10/20 participants demonstrate an effect , with peak times drawn from a uniform distribution over the range 200–275 ms . Here , effects can be reliably detected in those 10 participants , leading to an estimated prevalence of 0 . 47 [0 . 25 0 . 70] . However , because the effects are not present in all participants , there are no time points when the null hypothesis of zero population mean can be rejected . Interestingly , plotting the prevalence posterior distribution as a function of time does reveal evidence for an effect in the population during the time window of the simulated effect . Often the scientific question of interest might involve comparing an effect between groups of participants or between different experimental conditions in the same set of participants . In the first case , a researcher might , for example , be interested in examining differences in a behavioural or neuroimaging effect between patients and healthy controls , or between participants from two different cultures . In the second case , a researcher might be interested in determining the effect of an intervention on a particular measured effect , testing the same set of participants on two occasions , once with an intervention and once in a control condition . From the population mean perspective , these questions would typically be addressed with a two-sample unpaired t-test for the first case and a paired t-test for the second . From the prevalence perspective , the question would be whether the prevalence of true positive results differs between two sampled populations ( in the first case ) or between two experimental tests ( in the second case ) . We therefore provide additional functions ( see Materials and methods ) to directly estimate the difference in prevalence of true positive test results for these two comparisons , which we term between-group ( same test applied to two different samples ) and within-group ( two different tests applied to a single sample ) . To estimate the difference in prevalence of true positive results for a given test between two samples from separate populations ( e . g . patients vs . healthy controls ) , the input data required is the count of positive results and the total number of participants in each group . We illustrate this with a simulated data set . We specify the true prevalence in the two populations as γ1=0 . 75 and γ2=0 . 25 , respectively , and draw a random sample based on the respective binomial distributions with parameters θi ( see Materials and methods ) . We simulate N1=60 participants in the first group and N2=40 participants in the second group . The results of one such draw give k1=45 , k2=11 positive tests in each group , respectively . In this case , the MAP [96% HPDI] prevalence difference γ1-γ2 , calculated from these four values ( k1 , k2 , N1 , N2 ) , is 0 . 49 [0 . 29 0 . 67] , which closely matches the ground truth ( 0 . 5 ) . Figure 3A and B shows how the between-group posterior prevalence difference estimates scale with the number of participants for three different simulations . To estimate the difference in prevalence of true positive results between two different tests applied to the same sample of participants , the input parameters are the number of participants significant in both tests , the number significant only in each of the two tests , and the total number of participants . We simulate two tests applied to a group of N=50 participants . Each test detects a certain property with false positive rate α=0 . 05 . The ground truth prevalences of true positives for the two tests are γ1=0 . 5 and γ2=0 . 25 , respectively , and the correlation between positive results is ρ12=0 . 2 ( i . e . participants who test positive on one test are more likely to also test positive on the other test ) . The results of one random draw from this model give: ( see Materials and methods ) k11=8 participants with a significant result in both tests; k10=19 participants with a significant result in the first test but not the second; and k01=5 participants with a significant result in the second but not the first . In this case , the MAP [96% HPDI] prevalence difference γ1−γ2 , calculated from these four values ( k11 , k10 , k01 , N ) , is 0 . 28 [0 . 08 0 . 46] , which again matches the ground truth ( 0 . 25 ) . Figure 3C , D shows how the within-group posterior prevalence difference estimates scale with the number of participants for three different ground truth situations , given as [γ1 γ2]ρ12 . Both these approaches are implemented using Monte Carlo methods , and the functions return posterior samples ( Gelman , 2014 ) , which can be used to calculate other quantities , such as the posterior probability that one test or group has a higher prevalence than the other . The posterior log odds in favour of this hypothesis can be computed by applying the logit function to the proportion of posterior samples in favour of a hypothesis . In the between-group example above , the proportion of posterior samples in favor of the hypothesis γ1>γ2 is 0 . 9999987 , corresponding to log posterior odds of 13 . 5 . In the above within-group comparison , the proportion of posterior samples in favour of the hypothesis γ1>γ2 is 0 . 9979451 , corresponding to log posterior odds of 6 . 2 ( each computed from 10 million posterior samples ) . The additional example in Figure 4 demonstrates how between-group prevalence differences can occur between two populations with the same mean . We simulated two groups of participants where group 1 was from a homogenous population with a non-zero mean whereas group 2 comprised participants drawn from two different distributions , one with zero mean and one with a large mean . The two samples ( and populations ) have similar means , with no significant difference between them ( two-sample t-test p=0 . 78 ) . However , considering the prevalence in the two populations clearly reveals their difference ( Figure 4C ) , which is formalised with the posterior distribution of the prevalence difference between the two populations ( Figure 4D ) . In the above , we focused on performing explicit statistical inference within each participant . A possible criticism of this approach is that the within-participant binarisation of a continuous effect size can lose information . If the null distribution is the same for each participant , then the within-participant inference involves comparing each participant’s effect size , Ep , to a common statistical threshold E^ . The prevalence estimation can therefore be interpreted as estimating the population prevalence of participants for which Ep>E^ . In the NHST case , E^ is chosen so that P ( E>E^ ) =α ( usually α=0 . 05 ) , but we can generally consider the prevalence of participants with effects exceeding any value E^ . We therefore estimate the prevalence of Ep>E^ as a function of E^ . This can reveal if a data set provides evidence for population prevalence of a subthreshold within-participant effect , as well as showing how population prevalence decreases for larger effects . Figure 5 demonstrates this approach for the simulated systems of Figure 1 , showing that prevalence results for both right-sided , Ep>E^ , and left-sided , Ep<E^ , effect size thresholds . Note that this approach requires the null distribution to be the same for each participant and requires the false positive rate α to be calculated for each effect size considered . It reveals everything we can learn about the population prevalence of different effect sizes from our data set , exploring beyond the binarisation of the within-participant inference . The asymmetry visible in Figure 5C , D reflects the positive population mean for those simulations . In Figure 6 , three simulated EEG data sets ( c . f . Figure 2 ) show how prevalence as a function of effect size can disambiguate situations where the prevalence of p=0 . 05 NHST rejections does not differ . Each panel ( A–C ) shows a simulation of a different EEG experiment . All three have a similar population prevalence of p=0 . 05 null hypothesis rejections , as shown in the lower-right posterior distributions . However , the prevalence curves as a function of effect size differ in each simulation . For example , in panel A there is no evidence that the population contains individuals with an effect size greater than T ( 99 ) = 12 , whereas the MAP for prevalence of effects greater than 12 is around 50% in panel B . Similarly , the prevalence curve in panel C reflects the larger population variance of the effect in that simulation compared to panel A . While these differences would also be clear from a standard plot of the participant data ( e . g . violin or raincloud plot of per-participant maximal effect sizes ) , the posterior prevalence curves go beyond descriptive visualisations of the specific data sample by quantifying a formal inference to the population , including its associated uncertainty . As in the simulation of Figure 1 , a typical population mean inference is often framed as a two-level summary statistic procedure . At the first level , the effect is quantified within each participant ( e . g . a difference in mean response between two conditions ) . At the second level , the population mean is inferred under the assumption that the effect is normally distributed in the population ( i . e . based on the mean and standard deviation of the measured effect across participants ) . Bayesian prevalence is similarly framed as a two-level procedure . At the first level , a statistical test is applied within each participant , the result of which can be binarized via a within-participant NHST ( e . g . using a parametric test , as in our simulation , or alternatively using non-parametric permutation methods , independently for each participant ) , or via an arbitrary effect size threshold E^ . At the second level , the binary results from the first level ( i . e . the counts of significant participants ) are the input to the Bayesian population prevalence computation . To accompany this paper , we provide code in MATLAB , Python , and R to visualise the full posterior distribution of the population prevalence , as well as extract properties , such as the MAP point estimate and HPDIs . We also provide functions to provide Bayesian estimates of the difference in prevalence between two mutually exclusive participant groups to the same test ( between-group prevalence difference ) , as well as the difference in prevalence between two different tests applied to a single sample of participants ( within-group prevalence difference ) . We suggest reporting population prevalence inference results as the MAP estimate together with one or more HPDIs ( e . g . with probability 0 . 5 or 0 . 96 , see Materials and methods ) . It is important to stress that the second-level prevalence inference does not impose any requirements on the first-level within-participant tests , other than that each test should provide strong control of the same false positive rate α ( see Materials and methods ) . It is not required , for example , that each participant have the same number of trials , degrees of freedom , or within-participant variance . The within-participant test can be parametric ( e . g . a t-test ) or non-parametric ( e . g . based on permutation testing ) . It can be a single statistical test or an inference over multiple tests ( e . g . a neuroimaging effect within a certain brain region ) , provided that the family-wise error rate is controlled at α ( e . g . by using permutation tests with the method of maximum statistics ) . Figure 7 illustrates how Bayesian prevalence inference scales with the number of participants and trials . Figure 7A–C suggests that there are benefits to having larger numbers of participants for the Bayesian prevalence metrics ( including decrease in variance of obtained MAP and HPDI width , increase in prevalence lower bound ) . However , above around 50 participants , these benefits become less pronounced . Figure 7E shows that , on average , inferred prevalence is mostly sensitive to the number of trials per participant ( horizontal contours ) and is invariant to the number of participants ( although variance decreases with increasing N , as in Figure 7A , C and F ) . By comparison , t-test power ( Figure 7D ) is mostly sensitive to the number of participants ( vertical contours ) and is largely invariant to the number of trials above around 100 trials per participant ( Baker et al . , 2020 ) . In sum , compared to the population mean t-test , prevalence exhibits greater sensitivity to the number of trials obtained per participant and less sensitivity to the number of participants . While the problems that underlie the replication crisis are being increasingly recognised , there is currently no consensus as to alternative statistical approaches that are needed to address the underlying problems . Here , we propose that shifting our focus to quantifying and inferring effects within individuals addresses many of the pressing concerns recently highlighted in psychology and neuroimaging ( Amrhein et al . , 2019; Benjamin et al . , 2018; Forstmeier et al . , 2017; Ioannidis , 2005; McShane et al . , 2019 ) . We present a Bayesian approach to estimating population prevalence which is broadly applicable as it places no assumptions on the within-participant tests nor on the distribution of effects in the population . Further , prevalence does not require a Bayesian treatment; frequentist inference approaches can be used instead . The crucial point is to shift our perspective to first evaluate effects within individual participants , who represent the natural replication unit for studies of brains and behaviour . The data shown in Figure 1 were simulated from the standard hierarchical model:yij ∼ N ( μi , σw2 ) μi ∼ N ( μpop , σb2 ) where yij denotes the measurement made on the jth trial ( out of T ) of the ith participant ( out of N ) . μi represents the mean of each individual participant , σw represents a common within-participant standard deviation over trials , σb represents the standard deviation between participants , and μpop represents the overall population mean . This can be written asyij=μpop+ηij+ϵi where ηij~N ( 0 , σw2 ) , and ϵi~N ( 0 , σb2 ) . Note that under this model the distribution of the within-participant means is N ( μpop , σb2+1tσw2 ) . We consider a population of experimental units ( e . g . human participants or individual neurons ) which are of two types: those that have a particular binary effect or property , and those that do not . We consider the population prevalence of the ground truth state of each unit γgt , which is the proportion of the population from which the sample was drawn that have the effect ( 0<γgt<1 ) . If the true status of each individual unit could be directly observed , then the sample could be modelled with a binomial distribution with probability parameter γgt . However , we cannot directly observe the true status of each unit . Instead , we apply to each unit a statistical test following the NHST framework , which outputs a binary result ( which we term here positive vs . negative ) . This test has a false positive rate α , and sensitivity β . Thus , the probability that a randomly selected unit from the population that does not possess the defined effect but will produce a positive test result is α , while the probability that a randomly selected unit that does possess the defined effect will produce a positive test result is β . Under the assumption that the units are independent and α and β are constant across units , the number of positive tests k in a sample of size n can be modelled as a binomial distribution with parameter θ ( Donhauser et al . , 2018; Friston et al . , 1999b; Friston et al . , 1999a; Rogan and Gladen , 1978 ) :P ( X=k\|θ ) = ( nk ) θk ( 1−θ ) n−kθ= ( 1−γgt ) α+γgtβ A major issue with the above approach is that it requires the sensitivity β to be specified and constant across units . β is the probability of a significant result given that the individual has an effect and differs as a function of the ground truth effect size . In general , we do not assume that all members of the population have the same effect size , so it is not possible to specify a common β for all individuals . Therefore , rather than modelling the prevalence of ground truth effects , we consider the prevalence of true positive test results for the particular test procedure we employ , γtp:P ( X=k\|θ ) = ( nk ) θk ( 1−θ ) n−kθ= ( 1−γtp ) α+γtp In this case , the only assumption is that test procedure has the same false positive rate α for every individual , which is easily satisfied by most common parametric and non-parametric statistical tests . Note that this is equivalent to estimating ground truth prevalence with a test with β=1 . In general , a test with lower sensitivity allows inference of a higher prevalence for an observed k because some of the observed negative results will be missed true positive results . Therefore , prevalence of true positives obtained with β=1 is a conservative lower bound on the prevalence of ground truth state ( Allefeld et al . , 2016; Donhauser et al . , 2018; Friston et al . , 1999a ) . Throughout this paper we consider the prevalence of true positives for a given test procedure rather than the prevalence of ground truth state and so omit the subscript tp . This quantifies the proportion of the population that would be expected to provide a true positive test result ( i . e . have a non-null effect that would be detected by the experimental test procedure considered ) . Various frequentist approaches can be used with the above binomial model of statistical testing . First , the maximum likelihood estimate of the population prevalence of true positives can be obtained asγ^=k/n−α1−α Standard bootstrap techniques ( Johnson , 2001 ) can give percentile bootstrap confidence intervals as an indication of uncertainty in this estimate . We can also explicitly test various null hypotheses at the population level . For example , we can test a compound null hypothesis γ<0 . 5 , termed the majority null ( Allefeld et al . , 2016; Donhauser et al . , 2018 ) . This is chosen with the idea that a prevalence of >50% supports a claim that the effect is typical in the population . Other explicit compound nulls of this form can also be tested ( e . g . that γ<0 . 25 or γ<0 . 75 ) . Alternatively , it is possible to infer a lower bound on the population prevalence by finding the largest value γc , such that p ( X>k \| γ<γc ) <0 . 05 ( Allefeld et al . , 2016; Donhauser et al . , 2018 ) . This inferred lower bound provides a more graded output than a binary significance result of testing against a specific compound null ( i . e . the continuous value γc ) . We apply standard Bayesian techniques to estimate the population prevalence parameter of this model ( Gelman , 2014 ) . Assuming a beta prior distribution for θ with shape parameters r , s , together with a binomial likelihood function , the posterior distribution for θ is given by a beta distribution with parameters ( k+r , n-k+s ) , truncated to the interval [α , 1] , where k is the number of participants showing an above-threshold effect out of n tested . In the examples shown here , we use a uniform prior ( beta with shape parameters r=s=1 ) , as in the general case there is no prior information regarding θ . This implies a uniform prior also for γ , so , a priori , we consider any value of population prevalence equally likely . While we default to the uniform prior , the code supports any beta distribution as a prior . Alternative priors could be implemented via Markov chain Monte Carlo methods ( Gelman , 2014 ) together with the models described here . Note that similar Bayesian approaches have been applied in the field of epidemiology , where sometimes multiple complementary diagnostic tests for a disease are applied with or without a gold standard diagnosis in a subset of the sampled units ( Berkvens et al . , 2006; Enøe et al . , 2000; Joseph et al . , 1995 ) . Under the uniform prior , the Bayesian MAP estimate for prevalence proportion of true positives is available analytically and is equivalent to the maximum likelihood estimate:γmap=k/n−α1−α Following McElreath , 2016 , we present 96% HPDIs here to emphasise the arbitrary nature of this value and reduce the temptation to interpret the interval as a frequentist p=0 . 05 inference . We consider here a situation where the same test is applied to units sampled from two different populations . In addition to the prevalence of true positives within each population , we wish to directly estimate the difference in prevalence between the two populations . We denote the prevalence within each population as γ1 , γ2 , respectively . We sample n1 , n2 participants at random from each population and record k1 , k2 , the number of significant within-participant tests in each sample . Assuming independent uniform priors on the prevalences and associated θi variables as above with:θi= ( 1−γi ) α+γi then the posterior distribution for ( θ1 , θ2 ) is given by the product of two truncated beta distributions , with parameters ( ki+1 , ni-ki+1 ) , respectively , both truncated to the interval [α , 1] . The prevalence difference can be obtained as:γ1−γ2= ( θ1−θ2 ) / ( 1−α ) For non-truncated beta distributions , an analytic exact result is available ( Pham-Gia et al . , 1993 ) . This result could be extended to provide an exact distribution for the prevalence difference , but the mathematical expressions involved are quite complex . We find it simpler to employ Monte Carlo methods , which can provide as close an approximation to the exact answer as desired . Here we use Monte Carlo methods to draw samples from the posterior for ( θ1 , θ2 ) , obtaining samples for the prevalence difference with the above expression . We use these samples to numerically compute the MAP and HPDIs . In this situation , we consider that two different test procedures are applied to a single sample of n units . We assume both tests have the same value of α and define:θi=1-γiα+γi=α+1-αγi Here θi is the probability that a randomly selected unit from the population will show a positive result on the ith test , and γi is the population prevalence of true positives associated with the ith test . Now , each unit provides one of four mutually exclusive results based on the combination of binary results from the two tests . k11 represents the number of units that have a positive result in both tests , k10 , k01 represent the number of units that have a positive result only in the first or second test , respectively , and k00 is the number of units that do not show a positive result in either test . So ∑i , jkij=n . We can define analogous variables θ={θij} , representing the population proportion for each of the four combined test outcomes . Note that θij>0 and ∑i , jθij=1 . The marginal success probabilities θi can be expressed as:θ1=θ11+θ10 , θ2=θ11+θ01 and soγ1-γ2= ( θ10-θ01 ) / ( 1-α ) The marginal probabilities θi are subject to the constraintsα<θi<1 and soα<θ11+θ10<1 , α<θ11+θ01<1 Assuming a uniform prior and a multinomial distribution for the kij , the posterior of θ is a truncated Dirichlet distribution with parameters mij=kij+1 subject to the constraints above ( which are analogous to the truncation of the beta posterior distribution in the case of a single test ) . We use a Monte Carlo approach to draw samples from this posterior following a modified stick-breaking process . We use these samples to numerically compute properties like the MAP estimate and HPDIs . To specify a ground truth to simulate data from two tests applied to the same participants ( Figure 3 ) , we require γ1 and γ2 , the population prevalences of the two tested effects , together with ρ12 , the correlation between the presence of the two effects across the population . From this we can calculate γ11 , the proportion of true positive results to both tests asγ11=γ1γ2+ρ12γ11-γ1γ2 ( 1-γ2 ) Similarly , we can define γij representing the population proportions corresponding to the other test result configurations . Then we can generate multinomial samples using the parameters θij computed asθ11=γ11+α2γ00+αγ01+αγ10θ10=α+ ( 1−α ) γ1−θ11θ01=α+ ( 1−α ) γ2−θ11θ00=1−θ11−θ10−θ01 Estimating the prevalence of Ep>E^ proceeds as for prevalence inference based on within-participant NHST . One additional step is the need to calculate α , the false positive rate under the null hypothesis of no effect , for each threshold value E^ . This is simply the probability of Ep>E^ under the null hypothesis . In the examples shown here , we calculate this from the cumulative distribution function of the appropriate t-distribution , but for other tests this could also be estimated non-parametrically . A number of E^ values are selected , linearly spaced over the observed range of the sample . For each of these values the count of the number of participants satisfying the inequality and the α value corresponding to the inequality are used to obtain the Bayesian posterior for prevalence of true positives . Note that this can be applied to either tail Ep>E^ or Ep<E^ . Scripts implementing the simulations and creating the figures from the paper are available at https://github . com/robince/bayesian-prevalence ( copy archived at swh:1:rev:a10f2760930f7638d1c2a73944719e6283aedcec , Ince , 2021 ) . To accompany this paper , we provide functions in MATLAB , Python , and R to calculate the Bayesian prevalence posterior density ( e . g . to plot the full posterior distribution ) , the MAP estimate of the population prevalence , HPDI intervals of the posterior and lower bound quantiles of the posterior , as well as prevalence differences between samples or between tests within a sample . We also provide example scripts that produce posterior plots as in Figure 1E . See https://github . com/robince/bayesian-prevalence .	Scientists use statistical tools to evaluate observations or measurements from carefully designed experiments . In psychology and neuroscience , these experiments involve studying a randomly selected group of people , looking for patterns in their behaviour or brain activity , to infer things about the population at large . The usual method for evaluating the results of these experiments is to carry out null hypothesis statistical testing ( NHST ) on the population mean – that is , the average effect in the population that the study participants were selected from . The test asks whether the observed results in the group studied differ from what might be expected if the average effect in the population was zero . However , in psychology and neuroscience studies , people’s brain activity and performance on cognitive tasks can differ a lot . This means important effects in individuals can be lost in the overall population average . Ince et al . propose that this shortcoming of NHST can be overcome by shifting the statistical analysis away from the population mean , and instead focusing on effects in individual participants . This led them to create a new statistical approach named Bayesian prevalence . The method looks at effects within each individual in the study and asks how likely it would be to see the same result if the experiment was repeated with a new person chosen from the wider population at random . Using this approach , it is possible to quantify how typical or uncommon an observed effect is in the population , and the uncertainty around this estimate . This differs from NHST which only provides a binary ‘yes or no’ answer to the question , ‘does this experiment provide sufficient evidence that the average effect in the population is not zero ? ’ Another benefit of Bayesian prevalence is that it can be applied to studies with small numbers of participants which cannot be analysed using other statistical methods . Ince et al . show that the Bayesian prevalence can be applied to a range of psychology and neuroimaging experiments , from brain imaging to electrophysiology studies . Using this alternative statistical method could help address issues of replication in these fields where NHST results are sometimes not the same when studies are repeated .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "tools", "and", "resources", "neuroscience" ]	2021	Bayesian inference of population prevalence
The brain is capable of large-scale reorganization in blindness or after massive injury . Such reorganization crosses the division into separate sensory cortices ( visual , somatosensory . . . ) . As its result , the visual cortex of the blind becomes active during tactile Braille reading . Although the possibility of such reorganization in the normal , adult brain has been raised , definitive evidence has been lacking . Here , we demonstrate such extensive reorganization in normal , sighted adults who learned Braille while their brain activity was investigated with fMRI and transcranial magnetic stimulation ( TMS ) . Subjects showed enhanced activity for tactile reading in the visual cortex , including the visual word form area ( VWFA ) that was modulated by their Braille reading speed and strengthened resting-state connectivity between visual and somatosensory cortices . Moreover , TMS disruption of VWFA activity decreased their tactile reading accuracy . Our results indicate that large-scale reorganization is a viable mechanism recruited when learning complex skills . The current view of neural plasticity in the adult brain sees it as ubiquitous but generally constrained by functional boundaries ( Hoffman and Logothetis , 2009 ) . At the systems level , experience-driven plasticity is thought to operate within the limits of sensory divisions , where the visual cortex processes visual stimuli and responds to visual training , the tactile cortex processes tactile stimuli and responds to tactile training , and so on . A departure from this rule is usually reported only during large-scale reorganization induced by sensory loss or injury ( Hirsch et al . , 2015; Lomber et al . , 2011; Merabet and Pascual-Leone , 2010; Pavani and Roder , 2012 ) . The ventral visual cortex , in particular , is activated in blind subjects who read Braille ( Büchel et al . , 1998; Reich et al . , 2011; Sadato et al . , 2002 , 1996 ) and lesions of this area impair Braille reading ( Hamilton et al . , 2000 ) . This visual cortex thus has the innate connectivity required to carry out a complex perceptual task – reading – in a modality different than vision . It is unclear , however , to what extent this connectivity is preserved and functional in the normal , adult brain . A growing amount of evidence suggests that some form of cross-modal recruitment of the visual cortex could be possible in the normal healthy adults ( Amedi et al . , 2007; Kim and Zatorre , 2011; Powers et al . , 2012; Saito et al . , 2006; Zangenehpour and Zatorre , 2010; Merabet et al . , 2004; Zangaladze et al . , 1999 ) . Nonetheless , the behavioural relevance of these cortical mechanisms remains unclear , especially for complex stimuli . Notably , several experiments failed to find such cross-modal reorganization in sighted subjects , even after extensive training ( Kupers et al . , 2006; Ptito et al . , 2005 ) . One study found cross-modal plastic changes in subjects that were blindfolded for several days ( Merabet et al . , 2008 ) , but these plastic changes quickly vanished once the subjects removed their blindfolds . Here , for the first time , we show that large-scale , cross-modal cortical reorganization is a viable , adaptive mechanism in the sighted , adult brain . In our experiment , sighted adults followed a 9-month Braille reading course . The resulting cortical changes were tracked using task-based and resting-state functional Magnetic Resonance Imaging ( fMRI ) and manipulated with Transcranial Magnetic Stimulation ( TMS ) . Whole-brain fMRI analysis showed that after the course , activation to tactile Braille reading relative to the tactile control condition peaked within the Visual Word Form Area ( VWFA ) , ( Figure 1D , peak MNI = -45 -58 -12 , Z = 5 . 18 ) , a ventral visual stream ( Van Essen , 2005 ) region known to be involved in visual reading ( Dehaene and Cohen , 2011; Price and Devlin , 2011; Szwed et al . , 2011 ) . Additional activations were observed in the intraparietal sulcus , supramarginal gyrus , and frontal areas ( Table 1 ) . We observed the same increase in VWFA activation when we contrasted tactile Braille activation before and after the course [tactile Braille vs . tactile control x ( after course-before course ) ] ( Figure 1E , peak: MNI = -45 -64 -12 , Z=5 . 02 ) . This contrast also revealed clusters in the middle occipital gyrus , cuneus ( BA18 ) , the left superior temporal gyrus , and frontal areas ( Table 1 ) . To determine whether these changes were due to a general activity increase for all reading conditions , we computed a whole-brain ANOVA interaction analysis of the signal change following the course for reading visual words , visual Braille words and tactile Braille words versus their respective control tasks ( [tactile Braille vs . tactile control - ( visual Braille vs . visual Braille control+visual words vs . visual words control ) ] x ( after course-before course ) ) ( see 'Materials and methods' ) . Relative to visual reading , tactile reading led to increased activation in the VWFA ( Figure 1F; MNI -45 -61 -12 , Z = 4 . 05 ) and other parietal and frontal areas ( Table 1 ) . The results demonstrated that this pattern of activation in visual cortex was observed specifically for tactile reading . Indeed , we found no increase in activation to visual words following the course . The only increases in activation to visual Braille were found in the default mode network nodes in the parietal and prefrontal cortices ( Figure 1—figure supplement 3C , Table 2 , Appendix 1 . 3 ) . Importantly , the analyses mentioned in this section ( Figure 1D–F ) , including the whole-brain ANOVA , did not reveal any activation in the primary and secondary somatosensory cortices , even at an exploratory threshold of p=0 . 01 voxel-wise . Our subjects' progress in tactile reading was not homogeneous , with some subjects being entirely unable to learn Braille at all ( 0 WPM ) and other subjects reaching a speed of 17 WPM ( Appendix 1 . 1 ) . We therefore used regression to ask which fMRI responses to tactile Braille reading were modulated by the subjects’ tactile reading speed . This regression analysis revealed one significant cluster , located in the left inferior occipital gyrus ( Figure 1G; MNI = -45 -76 -12 , Z = 3 . 69 , Table 3 ) . A similar result was obtained when we correlated single letter recognition speed with tactile Braille activations , which further supports the significance of the visual cortex in learning to read Braille ( Table 3 , Appendix 1 . 4 ) . We found no such correlations for other reading speed measures ( e . g . visual Braille reading speed ) or imagery activations ( e . g . imagining tactile Braille; see Appendix 1 . 5 ) . These observations indicate that the ventral visual activations for tactile reading cannot be explained as a by-product of imagery . Visual words , visual Braille and tactile Braille all elicited activity in the VWFA . How similar are the neural representations of these three scripts ? The similarity of neural representations can be studied with multivariate pattern analysis-representational similarity analysis ( MVPA-RSA ) . This method is based on the premise that stimuli that share similar neural representations will generate similar voxel activation patterns ( Kriegeskorte et al . , 2008; Rothlein and Rapp , 2014 ) . Using MVPA-RSA ( 'Materials and methods' ) , we found that the two Braille conditions ( visual and tactile ) had the most similar activation patterns in the VWFA ( Figure 2A , r=0 . 48 ) , despite very large differences in the magnitude of the two activations ( see Figure 2B ) . The correlations of the two Braille conditions with visual words were much weaker ( Figure 2A , 'Materials and methods' and Appendix 1 . 6 ) . Despite a difference in modality ( visual vs . tactile ) , the neural representations of the two versions of Braille script were partially similar and distinct from the well-established representation for visual words . 10 . 7554/eLife . 10762 . 010Figure 2 . Response similarity and region-of-interest ( ROI ) analyses . Response similarity analysis showed that ( A ) the activity patterns for both Braille alphabets were the most similar , whereas the patterns for tactile Braille and visual words differed the most . In the VWFA ( B ) , the response to tactile Braille words changed from a de-activation to a positive activation . The VWFA also showed strong responses to visual Braille words; these , however , did not change significantly following the Braille course . The lateral occipital area ( C ) also saw the emergence of responses to tactile Braille words similar to the VWFA . In contrast , there was no effect of the Braille course in the somatosensory cortex ( D ) . A drop in activation for the control condition was salient in the intraparietal sulcus ( F ) , in which the activation to tactile Braille words remained unchanged , whereas the activation to the tactile control dropped to zero . Arrow thickness and the distance between scripts in ( A ) are proportional to correlation strength . ( * ) p<0 . 001; ( ) p<0 . 01; ( * ) p<0 . 05 . Dashed lines denote interactions . All ROIs are in the left hemisphere . DOI: http://dx . doi . org/10 . 7554/eLife . 10762 . 010 A region-of-interest ( ROI ) analysis was then applied ( see 'Materials and methods' ) . In the VWFA ( Figure 2B; all ROIs are in the left hemisphere ) following the Braille course , the response to tactile Braille words changed from de-activation to positive activation , resulting in a significant difference between tactile words and their control ( interaction: F ( 1 , 28 ) =18 . 5; p<001 ) . This emerging difference was also driven by a decrease in activation to control tactile stimuli . Post-hoc t-tests ( one for tactile words and another for tactile control , before vs . after course ) showed that the two effects were of similar magnitude and neither of them reached statistical significance on their own . The VWFA also showed strong responses to visual Braille words , similar to previously reported responses to novel visual alphabets ( e . g . Szwed et al . , 2014; Vogel et al . , 2014 ) . These responses remained unchanged throughout the course . The lateral occipital area ( Figure 2C ) showed a similar emergence of responses to tactile Braille words after the course as well . Several experiments have shown changes in primary somatosensory ( SI ) cortex activation to tactile stimuli after tactile training , in both humans ( e . g . Pleger et al . , 2003 ) and rodents ( e . g . Guic et al . , 2008 ) . In the current study , however , there was no significant change in activation to tactile Braille words and no after-course differences between tactile Braille words and the tactile control ( Figure 2D ) . In the secondary somatosensory cortex ( MNI: -51 -25 15 ) , we found only a non-specific drop in activation to all tactile stimuli ( F ( 1 , 28 ) =7 . 62 , p=0 . 01 ) , with no difference between tactile Braille words and the tactile control after the course . A drop in activation for the control condition was observed in the posterior attentional network ( intraparietal sulcus , IPS , Figure 2E , t ( 28 ) =2 . 76 , p=0 . 01 ) . Those activation drops in the somatosensory cortices and in IPS are the most likely cause behind the activation drop for control tactile stimuli observed in the VWFA and LO ( Figure 2B–C ) . In the primary motor cortex ( MNI: -39 -25 59 ) , we observed no such drop , which suggests that the finger-movement patterns across sessions and conditions remained unchanged . Thus , the Braille reading course led to an increase in activation to tactile Braille words in visual areas but not in the somatosensory cortex . Learning can impact the resting-state activity of the brain ( e . g . Lewis et al . , 2009 ) . Following the course , we observed an increase in resting-state functional connectivity between the VWFA seed and the left primary somatosensory cortex ( Figure 3A , red; p<0 . 001 , corrected for multiple comparisons; Z=4 . 57; MNI: -57 -24 45; 'Materials and methods' ) . This increase was the only statistically significant positive effect found . The same comparison showed a decrease in the VWFA’s functional connectivity with other visual areas , bilaterally . Furthermore , after the course , the VWFA-S1 functional connectivity level was correlated with the subjects’ progress in tactile Braille reading speed ( in the month preceding the after-course scan , Figure 3; r ( 27 ) =0 . 49 , p=0 . 007 ) . The VWFA thus increased its coupling with the somatosensory cortex while decreasing its coupling with other visual areas . This VWFA-S1 functional connectivity was behaviorally relevant for tactile reading and was likely to be dynamically modulated in a relatively short learning period ( similar to: Lewis et al . , 2009 ) ( see also Appendix 1 . 7 ) . 10 . 7554/eLife . 10762 . 011Figure 3 . Following the tactile Braille course , the VWFA increased its resting-state connectivity with the somatosensory cortex while decreasing its coupling with other visual areas and the motor cortex . The connectivity between the VWFA and the somatosensory cortex was behaviorally relevant for tactile Braille reading . ( A ) Functional connectivity of the VWFA after the tactile Braille course relative to the before-course scan . Red represents increased correlation , and blue represents decreased correlation . The VWFA seed is marked in pink . Thresholds: p = 0 . 001 voxel-wise , p = 0 . 05 cluster-wise . ( B ) Correlation between after-course VWFA – left S1 functional connectivity and progress in tactile Braille reading speed in the last month of the course . DOI: http://dx . doi . org/10 . 7554/eLife . 10762 . 011 Finally , to test the role of the VWFA in Braille reading , we performed a repetitive Transcranial Magnetic Stimulation ( rTMS ) experiment in which nine subjects were tested after the course in tactile Braille reading ( reading speeds: 3–17 WPM ) . rTMS was applied to the VWFA and to two control sites – the lateral occipital area and the vertex . Both the VWFA and the lateral occipital area were localized using the individual subjects’ fMRI results ( 'Materials and methods' ) . Similar to a previous visual reading study ( Duncan et al . , 2010 ) , we chose the lateral occipital area as an additional , negative control site , because TMS to this region evokes muscle contractions indistinguishable from VWFA stimulation . rTMS was applied while subjects performed a lexical decision task on words and pseudowords written in tactile Braille ( see Figure 4A; 'Materials and methods' ) . Based on a previous visual reading study ( Duncan et al . , 2010 ) , we expected that stimulation to the VWFA during the performance of the task would decrease accuracy for lexical decisions on Braille words . As predicted , TMS to the VWFA decreased the accuracy of reading Braille words ( t ( 8 ) =3 . 02 , p=0 . 016 , Figure 4B ) . This TMS result shows that the VWFA is necessary for reading tactile Braille words . 10 . 7554/eLife . 10762 . 012Figure 4 . TMS applied to the Visual Word Form Area selectively decreased the accuracy of Braille word reading . ( A ) Illustration of the experimental design . Subjects read tactile Braille words or pseudowords and performed a lexical decision task based on them . In half of the trials , repetitive TMS was applied . The VWFA and two control sites ( lateral occipital area and vertex ) were tested in separate runs . ( B ) Mean accuracy of reading Braille words and pseudowords is shown for the VWFA and for both control sites , for the TMS and no TMS conditions separately . ( * ) p=0 . 016 . DOI: http://dx . doi . org/10 . 7554/eLife . 10762 . 012 We also tested for effects of TMS on control stimuli ( Braille pseudowords ) and control sites ( lateral occipital area , vertex ) ( Figure 4B ) and none of these tests resulted in significant effects ( VWFA , Braille pseudowords: t ( 8 ) =0 . 03 , p=0 . 977; lateral occipital area , Braille words: t ( 8 ) =0 . 18 , p=0 . 859; lateral occipital area , Braille pseudowords: t ( 8 ) =0 . 03 , p=0 . 979; vertex , Braille words: t ( 8 ) =1 . 26 , p=0 . 243; vertex , Braille pseudowords: t ( 8 ) =0 . 02 , p=0 . 986 ) . However , given the small number of subjects in this part of our study , our TMS experiment was underpowered to statistically verify the specificity of the TMS effect: ANOVAs testing for specificity of our effect of interest showed a trend for an interaction between the stimulus type and TMS for the VWFA ( F ( 1 , 8 ) =3 . 59 , p=0 . 095 ) , no interaction between the stimulus type , TMS , and the stimulation site ( F ( 2 , 16 ) =0 . 41 , p=0 . 580 ) , and no interaction between TMS and the stimulation site for Braille words ( F ( 2 , 16 ) =1 . 5 , p=0 . 253 ) . Several experiments have already indicated that in some contexts , the sighted’s ventral visual cortex can contribute to the perception of tactile ( reviewed in: Amedi et al . , 2005 ) or auditory ( Amedi et al . , 2007 ) stimuli . It is also known that regions higher up in the sensory processing hierarchy , notably the antero-medial parts of the ventral temporal cortex ( MNI y>-40 ) , can host multisensory , abstract object representations ( e . g . Fairhall and Caramazza , 2013; Kassuba et al . , 2014 ) . The left fusiform gyrus in particular is suggested to process object-specific crossmodal interactions ( Kassuba et al . , 2011 ) . Our results demonstrate that the occipitotemporal visual cortex ( VWFA , MNI y≈-60 ) can represent stimuli in a modality other than vision . The fact that TMS to the VWFA can disrupt tactile reading demonstrates the importance of this representation for sensory processing . ROI analysis ( Figure 2B ) revealed that lateral occipital area ( LOA ) presented a pattern of activity increase to tactile words due to the course similar to the VWFA . However , TMS applied to LOA did not disturb tactile reading process ( Figure 4 ) . While the LOA is activated in various visual word recognition tasks ( Duncan et al . , 2009; Wright et al . , 2008 ) , its lesions seems not to affect reading itself ( Cavina-Pratesi et al . , 2015; Milner et al . , 1991; Philipose et al . , 2007 ) . The increase of activity in both LOA and VWFA for tactile reading thus suggest that visual and tactile reading share similar neural correlates along the ventral visual stream . However , the exact function of LOA in reading seems to be more accessory than critical . The ventral visual cortex was also activated when subjects heard auditory words and then imagined reading them in tactile Braille ( Figure 1—figure supplement 2 , Table 4 ) . We also observed robust object activations in the object imagery task ( ventral visual stream , left: MNI -45 -67 -9 , Z=6 . 68 , right: MNI 45 -67 -5 , Z=5 . 94 ) which confirms that subjects successfully engaged in imagery during the experiment . 10 . 7554/eLife . 10762 . 013Table 4 . Summary of main activations in the control experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 10762 . 013ContrastVoxel-wise p tresholdRegionBAHemisphereZ scoreCluster sizeMNI coordinatesTactile Braille Imagine vs Rest after coursep = 0 . 001CerebellumRight>851927-67-24Left3 . 43152-42-64-28Inferior Frontal Gyrus9Left>88307-57827Inferior Parietal Lobule40Left>8-39-4343Medial Frontal Gyrus6Left>8-3-163Middle Frontal Gyrus46Left>8-423223Postcentral Gyrus3Left>8-57-1923Precentral Gyrus6Left>8-39-759Superior Temporal Gyrus22Left>8-548-140Right>839-4047Middle Occipital Gyrus19Left6 . 60152-54-61-12Fusiform Gyrus ( VWFA ) 37Left5 . 95-45-61-12Inferior Temporal Gyrus20Left4 . 22-51-49-16Visual Braille Imagine vs Rest after coursep = 0 . 001Inferior Parietal Lobule40Left>83062-39-434340Right>8123339-4047Middle Frontal Gyrus6Left>83062-24-4516Right7 . 56123327-451Precentral Gyrus6Left>83062-54239CerebellumRight7 . 6313430-67-24Inferior Frontal Gyrus44Right7 . 41123354823Middle Occipital Gyrus19Left7 . 01119-48-58-12Fusiform Gyrus ( VWFA ) 37Left6 . 38-45-58-12Superior Temporal Gyrus42Right4 . 8011366-257Tactile Braille Imagine vs Visual braille imagine after coursep = 0 . 005Inferior Parietal Lobule40Left6 . 846602-57-222340Right4 . 7266-3731CerebellumRight6 . 7596224-52-24Superior Frontal Gyrus10Left6 . 41672-30501910Right5 . 356602274419Precentral Gyrus4Left6 . 21-33-1955Postcentral Gyrus3Right5 . 7260-19232Left4 . 88-57-2547Middle Frontal Gyrus10Right5 . 213350159Left4 . 86672-422635Objects Touchp = 0 . 005CerebellumRight>8121124-52-24Left>8-21-55-24Postcentral Gyrus3Right>8541248-2251Precentral Gyrus6Right>836-13594Left>8-39-2259Middle Occipital Gyrus19Left5 . 1885-51-64-937Right4 . 77121151-6412 There are several reasons to rule out the possibility that the visual activation for tactile reading is a side-effect of mental imagery . First , if the visual activations were only a by-product of mental imagery , TMS to the VWFA would not interfere with Braille reading; yet , it did ( Figure 4B ) . Second , training did not produce any changes in the activations to imagining tactile Braille ( Appendix 1 . 5 ) . Third , activations to imagining Braille in the VWFA did not correlate with tactile Braille reading proficiency ( Appendix 1 . 4 ) . Thus , visual cortex activations for tactile reading cannot be explained as a by-product of visual imagery . Instead , they constitute the signature of a new , tactile script representation emerging in the visual cortex . Cortical reorganization that crosses the sensory boundaries is prominent in blind and deaf humans ( Hirsch et al . , 2015; Pavani and Roder , 2012; Sadato et al . , 2002 ) and in congenitally deaf cats ( Lomber et al . , 2011 ) . In the latter , the deprived auditory areas play a vital role in the deaf cats’ superior visual perception . The above-mentioned studies used the same paradigms in non-deprived humans and cats , but did not find any signs of cross-modal reorganization ( e . g . Figure 7 in Sadato et al . , 2002 ) . One could have thus expected that learning to read by touch would lead to the emergence of a tactile reading area in the somatosensory cortex . Yet , our study produced a different result . Specific responses to tactile reading emerged not in somatosensory areas , but in visual areas . This result might seem incompatible with a large body of data showing that tactile training leads to changes in somatosensory activation to tactile stimuli ( Guic et al . , 2008; Kupers et al . , 2006; Pleger et al . , 2003; Ptito et al . , 2005 ) , or decision-level frontal cortex changes ( Sathian et al . , 2013 ) , but no visual cortex changes . However , all the above-mentioned experiments used simple stimuli , such as gratings , and short learning periods . Our experiment was longer ( 9 months ) and used complex stimuli - entire Braille words . Experiments that study learning-related plasticity at multiple time points ( Lövdén et al . , 2013 ) suggest that at the initial stage of Braille learning , the somatosensory cortex might have increased its response to Braille words . Then , as the effects of early sensory learning consolidated in the somatosensory cortex , the cortical focus of learning shifted elsewhere , in our case the ventral visual stream . Previous studies ( Büchel et al . , 1998; Lomber et al . , 2011; Merabet and Pascual-Leone , 2010; Sadato et al . , 2002 , 1996 ) have suggested that cross-modal plasticity is possible mainly as a result of massive sensory deprivation or injury . Our results demonstrate that given a long training period and a complex task , the normal brain is capable of large-scale plasticity that overcomes the division into separate sensory cortices . The exact mechanisms of this plasticity might of course be different in deprived and non-deprived subjects . Earlier reports of cross-modal plasticity in the normal brain might have already hinted at the possibility of cross-modal plasticity in the normal brain . Indeed , it was shown that some parts of the visual cortex can be activated during auditory or tactile tasks ( Amedi et al . , 2007; Kim and Zatorre , 2011; Powers et al . , 2012; Saito et al . , 2006; Zangenehpour and Zatorre , 2010 ) . Professional pianists , for example , were shown to activate their auditory cortex when viewing mute video recordings of a piano key being pressed ( Haslinger et al . , 2005 ) . The behavioral relevance of such activations , however , was demonstrated only for simple tactile stimuli: grating orientation ( Zangaladze et al . , 1999 ) and distance judgment ( Merabet et al . , 2004 ) . Our study used a controlled , within-subject design and precise behavioral measures supplemented with a causal method , TMS . Its results suggest that large-scale plasticity is a viable mechanism recruited when learning complex skills . This conclusion is congruent with a recent comparative study showing that the morphology of cerebral cortex is substantially less genetically heritable in humans than in chimpanzees ( Gomez-Robles et al . , 2015 ) . Such relaxed genetic control might be the reason for homo sapiens’ increased learning abilities and greater behavioral flexibility . Despite evidence for cross-modal plasticity of the human brain , the dominant view still describes it as necessarily constrained by the sensory boundaries ( e . g . Figure 18–2 in: Kandel et al . , 2012 ) . Our study provides a clear-cut evidence that learning-related neuroplasticity can overcome the sensory division . This calls for a re-assessment of our view of the functional organization of the brain . Thirty-six subjects took part in the first fMRI study ( 32 females , 4 males; mean age = 29 . 17 ) . All were right-handed , fluent in Polish and had normal or corrected-to-normal vision . They were either Braille teachers/professionals ( 14 subjects ) , special education students specializing in blindness and related disabilities ( 11 subjects ) , or close relatives of blind people ( 4 subjects ) . All subjects had or was close to obtain higher education . All subjects showed high motivation to pursue the course . All but three subjects were familiar with Braille visually ( see Appendix 1 . 2 ) . However , all were naive in tactile Braille reading , which was verified by the baseline testing session ( see Appendix 1 . 1 ) . To ensure appropriate statistical power , prior to data collection we decided to recruit at least 30 subjects . The research described in this article was approved by the Commitee for Research Ethics of the Institute of Psychology of the Jagiellonian University ( decisions 28/06/2012 and 12/03/2014 ) . Informed consent and consent to publish were obtained from each of the participants in accord with best-practice guidelines for MRI and TMS research . During the first fMRI examination , one subject was eliminated from the study due to a medical condition discovered by the neurologist supervising the study . Right after the first session , two subjects resigned for personal reasons . Another four subjects resigned during the tactile Braille reading course . Thus , 29 subjects completed the tactile Braille course and were included in the data analysis . At the end of the course , nine of the subjects mentioned above who achieved tactile reading speeds between 3 to 17 WPM participated in the TMS experiment . All of them were female , and their ages ranged from 22 to 36 ( M = 25 . 78 , SD = 4 . 44 ) . For the purpose of this experiment , we developed a made-to-measure tactile Braille course . A detailed description of the course will be published elsewhere . Briefly , the course was based on individual exercises , to be performed while blindfolded . Subjects were instructed to train on one A4 sheet every day ( approximately 20 min ) . The subjects’ progress was monitored by qualified Braille teachers during monthly meetings . The course lasted 9 months . Teaching began with simple tactile recognition exercises . Next , Braille letters were gradually introduced . In the second half of the course , subjects were encouraged to focus mainly on whole-word reading . Although our Braille course was focused on tactile reading , all the subjects checked their exercises visually , which required visual Braille reading training on a daily basis . Thus , we assumed that they would also progress in visual Braille reading . All fMRI data were acquired on the same Siemens MAGNETOM Tim Trio 3T scanner ( Siemens , München , Germany ) . The data from the fMRI reading experiment were collected using a 12-channel head coil . Resting-state fMRI data were acquired with a 32-channel head coil . All data were collected using the same scanning parameters . We used a gradient-echo planar imaging sequence sensitive to blood oxygen level-dependent ( BOLD ) contrast ( 33 contiguous axial slices , phase encoding direction=posterior-anterior , 3 . 6 mm thickness , TR=2190 ms , angle=90° , TE=30 ms , base resolution=64 , phase resolution=200 , matrix=64x64 , no iPAT ) . The fMRI reading experiment was divided into four runs: the first two composed the main reading experiment ( 282 functional images for each run ) ; the latter two composed the control imagery experiment ( 346 images in the first run , and 353 images in the second run ) . The resting-state data were collected in a separate scanning session , in a single run ( 282 volumes ) . In each scanning session , T1-weighted images were also acquired for anatomical localization . To make acquisition conditions as similar as possible between the before- and after-training scans , in the before-course scan we measured each subject’s head position relative to the head coil . Then , in the after-course scan , we reproduced the previously measured position . In addition , we used a standard Siemens Automatic Alignment scout MRI sequence before the two scans . The fMRI reading experiment consisted of two parts: the main experiment ( Figure 1 ) and the control mental imagery experiment ( see Figure 1—figure supplement 1 ) . In both experiments , we used a custom-made fiberglass table for the presentation of tactile stimuli . The table was designed in a way that prevented the subjects from seeing the stimuli that were placed on it . To minimize the time needed for the presentation of tactile stimuli , we used a block design in both parts of the reading experiment . Stimulation was programmed in Presentation ( Neurobehavioral Systems , San Francisco , CA ) . In both experiments , the blocks were separated by 13–20 s rest periods . Each rest period started with an ascending sound , which signaled to the subjects that they should raise their fingers from the fiberglass table . Then , the experimenter switched the cardboard with the stimuli . To equalize all experimental conditions , an empty cardboard was slipped onto the table for visual trials as well . At the end of the rest period , subjects heard a 500 ms descending sound signifying that they should put their fingers back down on the table . ( In the imagery experiment , this sound was preceded by the auditory cue , e . g . 'imagine objects' . ) To prevent them from touching the tactile stimuli prematurely , the subjects were asked to refrain from touching the cardboard until the metronome sound ( main experiment ) or the auditory cue ( control experiment ) was heard , 4–7 s later . The data from the fMRI reading experiment were analyzed using SPM8 software ( www . fil . ion . ucl . ac . uk/spm/software/spm8/ ) . Data preprocessing included: 1 ) slice timing , 2 ) realignment of all EPI images from the before-course and the after-course scans together , 3 ) coregistration of the anatomical image from the first time point to the mean EPI image , 4 ) segmentation of the coregistered anatomical image , 5 ) normalization of all images to MNI space and 5 ) FWHM spatial smoothing ( 5 mm ) . The signal time course for each subject was modelled within a general linear model ( Friston et al . , 1995 ) derived by convolving a canonical hemodynamic response function with the time series of the experimental stimulus categories and estimated movement parameters as regressors . Statistical parametric maps of the t statistic resulting from linear contrasts of each stimulus type minus baseline were generated and stored as separate images for each subject . Contrast images were then entered into an ANOVA model for random group effect analysis . We used first level contrast of each reading condition vs . its respective control after vs . before the tactile Braille course to assess the interaction between the time point ( before and after the course ) and reading condition ( Figure 1F ) . The SPM8 paired t-test was applied in pairwise comparisons of the reading conditions and their respective controls ( e . g . tactile Braille vs . tactile Control ) and in pairwise comparisons across two time points ( before and after the Braille course ) . In addition to the GLM-based activation analysis , we used SPM8 regression to examine 1 ) how tactile Braille reading proficiency modulated activations during tactile Braille reading ( Figure 1G , Figure 1—figure supplement 3D ) and during tactile and visual Braille imagining and 2 ) how visual Braille reading speed influenced activations in tactile Braille reading . We applied a voxel-wise threshold of p<0 . 005 and a p<0 . 05 threshold for cluster extent , corrected for multiple comparisons using REST AlphaSim ( 1000 Monte Carlo simulations ) , across the whole brain , unless stated otherwise . Similar results were observed at a voxel-wise threshold of p<0 . 001 , though not necessarily at cluster level-corrected levels of significance .	According to most textbooks , our brain is divided into separate areas that are dedicated to specific senses . We have a visual cortex for vision , a tactile cortex for touch , and so on . However , researchers suspect that this division might not be as fixed as the textbooks say . For example , blind people can switch their 'leftover' visual cortex to non-visual purposes , such as reading Braille – a tactile alphabet . Can this switch in functional organization also happen in healthy people with normal vision ? To investigate this , Siuda-Krzywicka , Bola et al . taught a group of healthy , sighted people to read Braille by touch , and monitored the changes in brain activity that this caused using a technique called functional magnetic resonance imaging . According to textbooks , tactile reading should engage the tactile cortex . Yet , the experiment revealed that the brain activity critical for reading Braille by touch did not occur in the volunteers’ tactile cortex , but in their visual cortex . Further experiments used a technique called transcranial magnetic stimulation to suppress the activity of the visual cortex of the volunteers . This impaired their ability to read Braille by touch . This is a clear-cut proof that sighted adults can re-program their visual cortex for non-visual , tactile purposes . These results show that intensive training in a complex task can overcome the sensory division-of-labor of our brain . This indicates that our brain is much more flexible than previously thought , and that such flexibility might occur when we learn everyday , complex skills such as driving a car or playing a musical instrument . The next question that follows from this work is: what enables the brain’s activity to change after learning to read Braille ? To understand this , Siuda-Krzywicka , Bola et al . are currently exploring how the physical structure of the brain changes as a result of a person acquiring the ability to read Braille by touch .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Massive cortical reorganization in sighted Braille readers
This paper describes a framework for modelling dopamine function in the mammalian brain . It proposes that both learning and action planning involve processes minimizing prediction errors encoded by dopaminergic neurons . In this framework , dopaminergic neurons projecting to different parts of the striatum encode errors in predictions made by the corresponding systems within the basal ganglia . The dopaminergic neurons encode differences between rewards and expectations in the goal-directed system , and differences between the chosen and habitual actions in the habit system . These prediction errors trigger learning about rewards and habit formation , respectively . Additionally , dopaminergic neurons in the goal-directed system play a key role in action planning: They compute the difference between a desired reward and the reward expected from the current motor plan , and they facilitate action planning until this difference diminishes . Presented models account for dopaminergic responses during movements , effects of dopamine depletion on behaviour , and make several experimental predictions . Neurons releasing dopamine send widespread projections to many brain regions , including basal ganglia and cortex ( Björklund and Dunnett , 2007 ) , and substantially modulate information processing in the target areas . Dopaminergic neurons in the ventral tegmental area respond to unexpected rewards ( Schultz et al . , 1997 ) , and hence it has been proposed that they encode reward prediction error , defined as the difference between obtained and expected reward ( Houk et al . , 1995; Montague et al . , 1996 ) . According to the classical reinforcement learning theory , this prediction error triggers update of the estimates of expected rewards encoded in striatum . Indeed , it has been observed that dopaminergic activity modulates synaptic plasticity in the striatum in a way predicted by the theory ( Reynolds et al . , 2001; Shen et al . , 2008 ) . This classical reinforcement learning theory of dopamine has been one of the greatest successes of computational neuroscience , as the predicted patterns of dopaminergic activity have been seen in diverse studies in multiple species ( Eshel et al . , 2016; Tobler et al . , 2005; Zaghloul et al . , 2009 ) . However , this classical theory does not account for the important role of dopamine in action planning . This role is evident from the difficulties in initiation of voluntary movements seen after the death of dopaminergic neurons in Parkinson’s disease . This role is consistent with the diversity in the activity of dopaminergic neurons , with many of them responding to movements ( da Silva et al . , 2018; Dodson et al . , 2016; Howe and Dombeck , 2016; Jin and Costa , 2010; Lee et al . , 2019; Schultz et al . , 1983; Syed et al . , 2016 ) . The function of dopamine in energizing movements is likely to come from the effects it has on the excitability or gain of the target neurons ( Lahiri and Bevan , 2020; Thurley et al . , 2008 ) . Understanding the role of dopamine in action planning and movement initiation is important for refining treatments for Parkinson’s disease , where the symptoms are caused by dopamine depletion . A foundation for a framework accounting the role of dopamine in both learning and action planning may be provided by a theory called active inference ( Friston , 2010 ) . This theory relies on an assumption that the brain attempts to minimize prediction errors defined as the differences between observed stimuli and expectations . In active inference , these prediction errors can be minimized in two ways: through learning – by updating expectations to match stimuli , and through action – by changing the world to match the expectations . According to the active inference theory , prediction errors may need to be minimized by actions , because the brain maintains prior expectations that are necessary for survival and so cannot be overwritten by learning , e . g . an expectation that food reserves should be at a certain level . When such predictions are not satisfied , the brain plans actions to reduce the corresponding prediction errors , for example by finding food . This paper suggests that a more complete description of dopamine function can be gained by integrating reinforcement learning with elements of three more recent theories . First , taking inspiration from active inference , we propose that prediction errors represented by dopaminergic neurons are minimized by both learning and action planning , which gives rise to the roles of dopamine in both these processes . Second , we incorporate a recent theory of habit formation , which suggests that the habit and goal-directed systems learn on the basis of distinct prediction errors ( Miller et al . , 2019 ) , and we propose that these prediction errors are encoded by distinct populations of dopaminergic neurons , giving rise to the observed diversity of their responses . Third , we assume that the most appropriate actions are identified through Bayesian inference ( Solway and Botvinick , 2012 ) , and present a mathematical framework describing how this inference can be physically implemented in anatomically identified networks within the basal ganglia . Since the framework extends the description of dopamine function to action planning , we refer to it as the DopAct framework . The DopAct framework accounts for a wide range of experimental data including the diversity of dopaminergic responses , the difficulties in initiation of voluntary movements under dopamine depletion , and it makes several experimentally testable predictions . This section first gives an overview of computations taking place during action planning in the DopAct framework , and then summarizes how these computations could be implemented in neural circuits including dopaminergic neurons . The DopAct framework includes two components contributing to planning of behaviour . The first component is a valuation system , which finds the value v of reward that the animal should aim at acquiring in a given situation . A situation of an animal can be described by two classes of factors: internal factors connected with level of reserves such as food , water , etc . to which we refer as ‘reserves’ , and external factors related to the environment , such as stimuli or locations in space , to which we refer as a ‘state’ following reinforcement learning terminology . The value v depends on both the amount of reward available in state s , and the current level of reserves . For example , if animal is not hungry , the desired value is equal to v=0 even if food is available . The second component of the DopAct framework is an actor , which selects an action to obtain the desired reward . This paper focusses on describing computations in the actor . Thus , for simplicity , we assume that the valuation system is able to compute the value v , but this paper does not describe how that computation is performed . In simulations we mostly focus on a case of low reserves , and use a simple model similar to a critic in standard reinforcement learning , which just learns the average value vs of resource in state s ( Sutton and Barto , 1998 ) . Extending the description of the valuation system will be an important direction for future work and we come back to it in Discussion . The goal of the actor is to select an action to obtain the reward set by the valuation system . This action is selected through inference in a probabilistic model , which describes relationships between states , actions and rewards , which we denote by s , a and R . Following reinforcement learning convention , we use R to denote the total reward defined in Equation 1 . 1 of Figure 1A , which includes the current reward r , and the future reward value v computed by the valuation system . The DopAct framework assumes that two systems within the actor learn distinct relationships between the variables , shown in Figure 1A . The first system , shown in orange , learns how the reward depends on the action selected in a given state , and we refer to it as ‘goal-directed’ , because it can infer actions that typically lead to the desired reward . The second system , in blue , learns which actions should generally be chosen in a given state , and we refer to it as ‘habit’ , because it suggests actions without considering the value of the reward currently available . Both goal-directed and habit systems propose an action , and their influence depends on their relative certainty . Figure 1B gives an overview of how the systems mentioned above contribute to action planning , in a typical task . During initial trials , the valuation system ( shown in red ) evaluates the current state s and computes the value of desired reward v , and the goal-directed system selects the action a . At this stage the habit system contributes little to the planning process as its uncertainty is high . As the training progresses , the habit system learns to mimic the choices made by the goal-directed system ( Miller et al . , 2019 ) . On later trials the action is jointly determined by the habit and goal-directed systems ( Figure 1B ) , and their relative contributions depend on their levels of certainty . The details of the above computations in the framework will be described in the next section , and it will be later shown how an algorithm inferring action can be implemented in a network resembling the anatomy of the basal ganglia . But before going through a mathematical description , let us first provide an overview of this implementation ( Figure 1C ) . In this implementation , the valuation , goal-directed and habit systems are mapped on the spectrum of cortico-basal ganglia loops ( Alexander et al . , 1986 ) , ranging from valuation in a loop including ventral striatum , to habit in a loop including the dorsolateral striatum that has been shown to be critical for habitual behaviour ( Burton et al . , 2015 ) . In the DopAct framework , the probability distributions learned by the actor are encoded in the strengths of synaptic connections in the corresponding loops , primarily in cortico-striatal connections . As in a standard implementation of the critic ( Houk et al . , 1995 ) , the parameters of the value function learned by the valuation system are encoded in cortico-striatal connections of the corresponding loop . Analogous to classical reinforcement learning theory , dopaminergic neurons play a critical role in learning , and encode errors in predictions made by the systems in the DopAct framework . However , by contrast to the standard theory , dopaminergic neurons do not all encode the same signal , but instead dopaminergic populations in different systems compute errors in predictions made by their corresponding system . Since both valuation and goal-directed systems learn to predict reward , the dopaminergic neurons in these systems encode reward prediction errors ( which slightly differ between these two systems , as will be illustrated in simulations presented later ) . By contrast , the habit system learns to predict action on the basis of a state , so its prediction error encodes how the currently chosen action differs from a habitual action in the given state . Thus these dopaminergic neurons respond to non-habitual actions in the DopAct framework . We denote the prediction errors in the valuation , goal-directed and habit systems by δv , δg and δh , respectively . The dopaminergic neurons send these prediction errors to the striatum , where they trigger plasticity of cortico-striatal connections . In the DopAct framework , habits are formed through a process in which the habit system learns to mimic the goal-directed system . Unlike in a previous model of habit formation ( Daw et al . , 2005 ) , in the DopAct framework learning in the habit system is not driven by a reward prediction error , but by a signal encoding a difference between chosen and habitual actions . At the start of training , when an action is selected mostly by the goal-directed system , the dopaminergic neurons in the habit system receive an input encoding the chosen action , but the striatal neurons in the habit system are not yet able to predict this action , resulting in a prediction error encoded in dopaminergic activity ( left display in Figure 1D ) . This prediction error triggers plasticity in the striatal neurons of the habit system , so they tend to predict this action in the future ( right display in Figure 1D ) . The systems communicate through an ‘ascending spiral’ structure of striato-dopaminergic projections identified by Haber et al . , 2000 . These Authors observed that dopaminergic neurons within a given loop project to the corresponding striatal neurons , while the striatal neurons project to the dopaminergic neurons in the corresponding and next loops , and they proposed that the projections to the next loop go via interneurons , so they are effectively excitatory ( Figure 1C ) . In the DopAct framework , once the striatal neurons in the valuation system compute the value of the state v , they send it to the dopaminergic neurons in the goal-directed system . In the DopAct framework , dopamine in the goal-directed system plays a role in both action planning and learning , and now an overview of this role is given . In agreement with classical reinforcement learning theory , the dopaminergic activity δg encodes reward prediction error , namely the difference between the reward R ( including both obtained and available reward ) and the expected reward ( Schultz et al . , 1997 ) , but in the DopAct framework the expectation of reward in the goal-directed system is computed on the basis of the current action plan . Therefore , this reward expectation only arises from formulating a plan to achieve it . Consequently , when a reward is available , the prediction error δg can only be reduced to zero , once a plan to obtain the reward is formulated . To gain an intuition for how the goal-directed system operates , let us consider a simple example of a hungry rat in a standard operant conditioning experiment . Assume that the rat has been trained that after pressing a lever a food pellet is delivered ( Figure 2A ) . Consider a situation in which a lever is suddenly made available to the animal . Its sight allows the valuation system to predict that reward is available , and it sends an estimated value of the reward to the goal-directed system . Such input induces a reward prediction error in the goal-directed system , because this system has received information that a reward is available , but has not yet prepared actions to obtain the reward , hence it does not expect any reward for its action . The resulting prediction error triggers a process of planning actions that can get the reward . This facilitation of planning arises in the network , because the dopaminergic neurons in the goal-directed system project to striatal neurons ( Figure 1C ) , and increase their excitability . Once an appropriate action has been computed , the animal starts to expect the available reward , and the dopamine level encoding the prediction error decreases . Importantly , in this network dopamine provides a crucial feedback to striatal neurons on whether the current action plan is sufficient to obtain the available reward . If it is not , this feedback triggers changes in the action plan until it becomes appropriate . Thus the framework suggests why it is useful for the neurons encoding reward prediction error to be involved in planning , namely it suggests that this prediction error provides a useful feedback for the action planning system , informing if the plan is suitable to obtain the reward . It is worth explaining why the reward expectation in the goal-directed system arises already once an action is computed and before it is implemented . It happens in the DopAct framework , because the striatal neurons in the goal-directed system learn over trials to predict that particular pattern of activity of neurons encoding action in the basal ganglia ( which subsequently triggers a motor response ) leads to reward in the future . This mechanism is fully analogous to that in the temporal-difference learning model used to describe classical conditioning , where the reward expectation also arises already after a stimulus , because the striatal neurons learn that the pattern of cortical inputs to the basal ganglia encoding the state ( i . e . the stimulus ) will lead to a reward ( Schultz et al . , 1997 ) . In the goal-directed system of DopAct , an analogous reward prediction is made , but not only on the basis of a state , but on the basis of a combination of state and action . The prediction error in the goal-directed system also allows the animal to learn about the rewards resulting from actions . In the example we considered above such learning would be necessary if the amount of reward changed , for example to two pellets ( Figure 2B ) . On the first trial after such change , a prediction error will be produced after reward delivery . This prediction error can be reduced by learning , so the animal will expect such increased reward in the future trials and no longer produce prediction error at reward delivery . In summary , the prediction errors in the goal-directed system are reduced by both planning and learning , as in active inference ( Friston , 2010 ) . Namely , the prediction errors arising from rewards becoming available are reduced within trials by formulating plans to obtain them , and the prediction errors due to outcomes of actions differing from expectations are reduced across trials by changing weights of synaptic connection encoding expected reward . The next three sections will provide the details of the DopAct framework . For clarity , we will follow Marr’s levels of description , and discuss computations , an algorithm , and its implementation in the basal ganglia network . To illustrate the computations in the framework we will consider a simple task , in which only an intensity of a single action needs to be chosen . Such choice has to be made by animals in classical experiments investigating habit formation , where the animals are offered a single lever , and need to decide how frequently to press it . Furthermore , action intensity often needs to be chosen by animals also in the wild ( e . g . a tiger deciding how vigorously pounce on a prey , a chimpanzee choosing how strongly hit a nut with a stone , or a sheep selecting how quickly eat the grass ) . Let us denote the action intensity by a . Let us assume that the animal chooses it on the basis of the reward it expects R and the stimulus s ( e . g . the size of prey , nut or grass ) . Thus the animal needs to infer an action intensity sufficient to obtain the desired reward ( but not larger to avoid unnecessary effort ) . Let us consider the computation in the DopAct framework during action planning . During planning , the animal has not received any reward yet r=0 , so according to Equation 1 . 1 , the total reward is equal to the reward available R=v . While planning to obtain this reward , the actor combines information from the goal-directed system ( encoding how the reward depends on actions taken in given states ) , and the habit system ( encoding the probability distribution of generally selecting actions in particular states ) . These two pieces information are combined according to Bayes’ theorem ( Equation 3 . 1 in Figure 3 ) , which states that the posterior probability of selecting a particular action given available reward is proportional to the product of a likelihood of the reward given the action , which we propose is represented in the goal-directed system , and a prior , which we propose is encoded by the habit system . In the DopAct framework , an action a is selected which maximizes the probability P ( a\|R , s ) . An analogous way of selecting actions has been used in models treating planning as inference ( Attias , 2003 ) , and it has been nicely summarized by Solway and Botvinick , 2012 'The decision process takes the occurrence of reward as a premise , and leverages the generative model to determine which course of action best explains the observation of reward . ' In this paper , we make explicit the rationale for this approach: The desired amount of resources that should be acquired depends on the levels of reserves ( and a given state ) ; this value is computed by the valuation system , and the actor needs to find the action depending on this reward . Let us provide a further rationale for selecting an action a which maximizes P ( a\|R , s ) , by analysing what this probability expresses . Let us consider the following hypothetical scenario: An animal selected an action without considering the desired reward , that is by sampling it from its default policy P ( a\|s ) provided by the habit system , and obtained reward R . In this case , P ( a\|R , s ) is the probability that the selected action was a . When an animal knows the amount of resource desired R , then instead of just relying on the prior , the animal should rather choose an action maximizing P ( a\|R , s ) , which was the action most likely to yield this reward in the above scenario . One may ask why it is useful to employ the habit system , instead of exclusively relying on the goal-directed system that encodes the relationship between rewards and actions . It is because there may be uncertainty in the action suggested by the goal-directed system , arising for example , from noise in the computations of the valuation system or inaccurate estimates of the parameters of the goal-directed system . According to Bayesian philosophy , in face of such uncertainty , it is useful to additionally bias the action by a prior , which here is provided by the habit system . This prior encodes an action policy that has overall worked in the situations previously experienced by the animal , so it is a useful policy to consider under the uncertainty in the goal-directed system . To make the above computation more concrete , we need to specify the form of the prior and likelihood distributions . We first provide them for the example of choosing action intensity . They are given in Figure 3B , where fx;μ , Σ denotes the probability density of a normal distribution with mean μ and variance Σ . In a case of the prior , we assume that action intensity is normally distributed around a mean given by stimulus intensity scaled by parameter h , reflecting an assumption that a typical action intensity often depends on a stimulus ( e . g . the larger a nut , the harder a chimpanzee must hit it ) . On the other hand , in a case of the probability of reward R maintained by the goal-directed system , the mean of the reward is equal to a product of action intensity and the stimulus size , scaled by parameter q . We assume that the mean reward depends on a product of a and s for three reasons . First , in many situations reward depends jointly on the size of the stimulus , and the intensity with which the action is taken , because if the action is too weak , the reward may not be obtained ( e . g . a prey may escape or a nut may not crack ) , and the product captures this dependence of reward on a conjunction of stimulus and action . Second , in many foraging situations , the reward that can be obtained within a period of time is proportional to a product of a and s ( e . g . the amount of grass eaten by a sheep is proportional to both how quickly the sheep eats it , and how high the grass is ) . Third , when the framework is generalized to multiple actions later in the paper , the assumption of reward being proportional to a product of a and s will highlight a link with classical reinforcement learning . We denote the variances of the distributions of the goal-directed and habit systems by Σg and Σh . The variance Σg quantifies to what extent the obtained rewards have differed from those predicted by the goal-directed system , while the variance Σh describes by how much the chosen actions have differed from the habitual actions . Figure 3C shows an example of probability distributions encoded by the two systems for sample parameters . It also shows a posterior distribution P ( a\|R , s ) , and please note that its peak is in between the peaks of the distributions of the two systems , but it is closer to the peak of a system with smaller uncertainty ( orange distribution is narrower ) . This illustrates how in the DopAct framework , the action is inferred by incorporating information from both systems , but weighting it by the certainty of the systems . In addition to action planning , the animal needs to learn from the outcomes , to predict rewards more accurately in the future . After observing an outcome , the valuation system no longer predicts future reward v=0 , so according to Equation 1 . 1 the total reward is equal to the reward actually obtained R=r . The parameters of the distributions should be updated to increase P ( R\|s ) , so in the future the animal is less surprised by the reward obtained in that state ( Figure 3A ) . Let us describe an algorithm used by the actor to infer action intensity a that maximizes the posterior probability PaR , s . This posterior probability could be computed from Equation 3 . 1 , but note that a does not occur in the denominator of that equation , so we can simply find the action that maximizes the numerator . Hence , we define an objective function F equal to a logarithm of the numerator of Bayes’ theorem ( Equation 4 . 1 in Figure 4 ) . Introducing the logarithm will simplify function F because it will cancel with exponents present in the definition of normal density ( Equation 3 . 3 ) , and it does not change the position of the maximum of the numerator because the logarithm is a monotonic function . For example , the green curve in Figure 4B shows function F corresponding to the posterior probability in Figure 3C . Both green curves have the maximum at the same point , so instead of searching for a maximum of a posterior probability , we can seek the maximum of a simpler function F . During action planning the total reward is equal to reward available , so we set R=v in Equation 4 . 1 , and we find the action maximizing F . This can be achieved by initializing a to any value , and then changing it proportionally to the gradient of F ( Equation 4 . 2 ) . Figure 4B illustrates that with such dynamics , the value of a approaches a maximum of F . Once a converges , the animal may select the action with the corresponding intensity . In summary , this method yields a differential equation describing an evolution of a variable a , which converges to a value of a that maximizes P ( a\|R , s ) . After obtaining a reward , R is equal to the reward obtained , so we set R=r in Equation 4 . 1 , and the values of parameters are changed proportionally to the gradient of F ( Equations 4 . 3 ) . Such parameter updates allow the model to be less surprised by the rewards ( as aimed for in Figure 3A ) , because under certain assumptions function F expresses 'negative free energy' . The negative free energy ( for the inference problem considered in this paper ) is defined as F=ln⁡PRs-KL , where KL is the Kullback-Leibler divergence between P ( a\|R , s ) and an estimate of this distribution ( a detailed definition and an explanation for why F given in Equation 4 . 1 expresses negative free energy for an analogous problem is given by Bogacz , 2017 ) . Importantly , since KL≥0 , the negative free energy provides a lower bound on P ( R\|s ) ( Friston , 2005 ) . Thus changing the parameters to increase F , rises the lower bound on P ( R\|s ) , and so it tends to increase P ( R\|s ) . Let us derive the details of the algorithm ( general form of which is given in Figure 4A ) for the problem of choosing action intensity . Let us start with considering a special case in which both variance parameters are fixed to Σg=Σh=1 , because then the form of the algorithm and its mapping on the network are particularly beautiful . Substituting probability densities of likelihood and prior distributions ( Equations 3 . 2-3 . 3 ) for the case of unit variances into Equation 4 . 1 ( and ignoring constants 1/2π ) , we obtain the expression for the objective function F in Equation 5 . 1 ( Figure 5A ) . We see that F consists of two terms , which are the squared prediction errors associated with goal-directed and habit systems . The prediction error for the goal-directed system describes how the reward differs from the expected mean , while the prediction error of the habit system expresses how the chosen action differs from that typically chosen in the current state ( Equations 5 . 2 ) . As described in the previous section , action intensity can be found by changing its value according to a gradient of F ( Equation 4 . 2 ) . Computing the derivative of F over a , we obtain Equation 5 . 3 , where the two colours indicate terms connected with derivatives of the corresponding prediction errors . Finally , when the reward is obtained , we modify the parameters proportionally to the derivatives of F over the parameters , which are equal to relatively simple expressions in Equations 5 . 4 . Figure 5A illustrates the key feature of the DopAct framework , that both action planning and learning can be described by the same process . Namely in both planning and learning , certain variables ( the action intensity and synaptic weights , respectively ) are changed to maximize the same function F ( Equations 5 . 3 and 5 . 4 ) . Since F is a negative of the sum of prediction errors ( Equation 5 . 1 ) , both action planning and learning are aimed at reducing prediction errors . The key elements of the algorithm in Figure 5A naturally map on the known anatomy of striato-dopaminergic connections . This mapping relies on three assumptions analogous to those typically made in models of the basal ganglia: ( i ) the information about state s is provided to the striatum by cortical input , ( ii ) the parameters of the systems q and h are encoded in the cortico-striatal weights , and ( iii ) the computed action intensity is represented in the thalamus ( Figure 5B ) . Under these assumptions , Equation 5 . 3 describing an update of action intensity can be mapped on the circuit: The action intensity in the model is jointly determined by the striatal neurons in the goal-directed and habit systems , which compute the corresponding terms of Equation 5 . 3 , and communicate them by projecting to the thalamus via the output nuclei of the basal ganglia . The first term δgqs can be provided by striatal neurons in the goal-directed system ( denoted by G in Figure 5B ) : They receive cortical input encoding stimulus intensity s , which is scaled by cortico-striatal weights encoding parameter q , so these neurons receive synaptic input qs . To compute δgqs , the gain of the striatal neurons in the goal-directed system needs to be modulated by dopaminergic neurons encoding prediction error δg ( this modulation is represented in Figure 5B by an arrow from dopaminergic to striatal neurons ) . Hence , these dopaminergic neurons drive an increase in action intensity until the prediction error they represent is reduced ( as discussed in Figure 2 ) . The second term hs in Equation 5 . 3 can be computed by a population of neurons in the habit system receiving cortical input via connection with the weight h . Finally , the last term -a simply corresponds to a decay . In the DopAct framework , dopaminergic neurons within each system compute errors in the predictions about the corresponding variable , i . e . reward for the goal-directed system , and action for the habit system . Importantly , in the network on Figure 5B this computation can be performed locally , i . e . the dopaminergic neurons receive inputs encoding all quantities necessary to compute their corresponding errors . In the habit system , the prediction error is equal to a difference between action a and expectation hs ( blue Equation 5 . 2 ) . Such error can be easily computed in a network of Figure 5B , where the dopaminergic neurons in the habit system receive effective input form the output nuclei equal to a ( as they receive inhibition equal to -a ) , and inhibition hs from the striatal neurons . In the goal-directed system , the expression for prediction error is more complex ( orange Equation 5 . 2 ) , but importantly , all terms occurring in the equation could be provided to dopaminergic neurons in the goal-directed system via connections shown in Figure 5B ( qs could be provided by the striatum , while a thorough an input from the output nuclei which have been reported to project to dopaminergic neurons [Watabe-Uchida et al . , 2012] ) . Once the actual reward is obtained , changing parameters proportionally to prediction errors ( Equations 5 . 4 ) can arise due to dopaminergic modulation of the plasticity of cortico-striatal connections ( represented in Figure 5B by arrows going from dopamine neurons to parameters ) . With such a modulation , learning could be achieved through local synaptic plasticity: The update of a weight encoding parameter h ( blue Equation 5 . 4 ) is simply proportional to the product of presynaptic ( s ) and dopaminergic activity ( δh ) . In the goal-directed system , orange Equation 5 . 4 corresponds to local plasticity , if at the time of reward the striatal neurons encode information about action intensity ( see definition of G in Figure 5C ) . Such information could be provided from the thalamus during action execution . Then the update of synaptic weight encoding parameter q will correspond to a standard three-factor rule ( Kuśmierz et al . , 2017 ) involving a product of presynaptic ( s ) , postsynaptic ( a ) and dopaminergic activity ( δg ) . The model can be extended so that the parameters Σg and Σh describing variances of distributions are encoded in synaptic connections or internal properties of the neurons ( e . g . leak conductance ) . In such an extended model , the action proposals of the two systems are weighted according to their certainties . Figure 6A shows the general description of the algorithms which is analogous to that in Figure 5A . The action intensity is driven by both goal-directed and habit systems , but now their contributions are normalised by the variance parameters . For the habit system this normalization is stated explicitly in Equation 6 . 2 , while for the goal-directed system it comes from a normalization of prediction error by variance in orange Equation 6 . 3 ( it is not necessary to normalize habit prediction error by variance because the contribution of the habit system is already normalized in Equation 6 . 2 ) . There are several ways of including the variance parameters in the network , and one of them is illustrated in Figure 6B ( see caption for details ) . The updates of the variance parameters ( Equations 6 . 5 ) only depend on the corresponding prediction errors and the variance parameters themselves , so they could be implemented with local plasticity , if the neurons encoding variance parameters received corresponding prediction errors . Figure 6C provides a complete description of the dynamics of the simulated model . It parallels that in Figure 6B , but now explicitly includes time constants for update of neural activity ( τ , τδ ) , and learning rates for synaptic weights ( α with corresponding indices ) . As described in the Materials and methods , a simple model of the valuation system based on standard temporal-difference learning was employed in simulations ( because the simulations corresponded to a case of low level of animal’s reserves ) . Striatal neurons in the valuation system compute the reward expected in a current state on the basis of parameters wt denoting estimates of reward at time t after a stimulus , and following standard reinforcement learning we assume that these parameters are encoded in cortico-striatal weights . The dopaminergic neurons in the valuation system encode the prediction error similar to that in the temporal-difference learning model , and after reward delivery , they modulate plasticity of cortico-striatal connections . The Method section also provides details of the implementation and simulations of the model . To illustrate how the model mechanistically operates and to help relate it to experimental data , we now describe a simulation of the model inferring action intensity . On each simulated trial the model selected action intensity , after observing a stimulus , which was set to s=1 . The reward obtained depended on action intensity as shown in Figure 7A , according to r=5tanh⁡3a/5-a . Thus , the reward was proportional to the action intensity , transformed through a saturating function , and a cost was subtracted proportional to the action intensity , that could correspond to a price for making an effort . We also added Gaussian noise to reward ( with standard deviation σr=0 . 5 ) to account for randomness in the environment , and to action intensity to account for imprecision of the motor system or exploration . Figure 7AB shows how the quantities encoded in the valuation system changed throughout the learning process . The pattern of prediction errors in this figure is very similar to that expected from the temporal difference model , as the valuation system was based on that model . The stimulus was presented at time t=1 . On the first trial ( left display ) the simulated animal received a positive reward at time t=2 ( dashed black curve ) due to stochastic nature of the rewards in the simulation . As initially the expectation of reward was low ( dashed red curve ) , the reward triggered a substantial prediction error ( solid red curve ) . The middle and right plots show the same quantities after learning . Now the prediction error was produced after the presentation of the stimulus , because after seeing the stimulus a simulated animal expected more reward than before the stimulus . In the middle display the reward received at time t=2 was very close to the expectation , so the prediction error at the time of the reward was close to 0 . In the right display the reward happened to be lower than usual ( due to noise in the reward ) , which resulted in a negative prediction error . Note that the pattern of prediction errors in the valuation system in Figure 7B resembles the famous figure showing the activity of dopaminergic neurons during conditioning ( Schultz et al . , 1997 ) . Figure 7C shows the prediction errors in the actor and action intensity on the same trials that were visualised in Figure 7B . Prediction errors in the goal-directed system follow a similar pattern as in the valuation system in the left and middle displays in Figure 7C , that is before the behaviour becomes habitual . The middle display in Figure 7C shows simulated neural activity that was schematically illustrated in Figure 2A: As the valuation system detected that a reward was available ( see panel above ) , it initially resulted in a prediction error in the goal-directed system , visible as an increase in the orange curve . This prediction error triggered a process of action planning , so with time the green curve representing planned action intensity increased . Once the action plan has been formulated , it provided a reward expectation , so the orange prediction error decreased . When an action became habitual after extensive training ( right display in Figure 7C ) , the prediction error in the goal-directed system started to qualitatively differ from that in the valuation system . At this stage of training , the action was rapidly computed by the habit system , and the goal-directed system was too slow to lead action planning , so the orange prediction error was lower . This illustrates that in the DopAct framework reward expectations in the goal-directed system can arise even if an action is computed by the habit system . The prediction error in the habit system follows a very different pattern than in other systems . Before an action became habitual , the prediction errors in the habit system arose after the action has been computed ( middle display in Figure 7C ) . Since the habit system has not formed significant habits on early trials , it was surprised by the action , and this high value of blue prediction error drove its learning over trials . Once the habit system was highly trained ( right display in Figure 7C ) it rapidly drove action planning , so the green curve showing planned action intensity increased more rapidly . Nevertheless , due to the dynamics in the model , the increase in action intensity was not instant , so there was a transient negative prediction error in the habit system while an action was not yet equal to the intensity predicted by the habit system . The prediction error in the habit system at the time of action execution depended on how the chosen action differed from a habitual one , rather than on the received reward ( e . g . in the right display in Figure 7C , δh>0 because the executed action was stronger than the planned one due to motor noise , despite reward being lower than expected ) . Figure 7D shows how the parameters in the model evolved over the trials in the simulation . The left display shows changes in the parameters of the three systems . A parameter of the valuation system correctly converged to the maximum value of the reward available in the task w1≈2 ( i . e . the maximum of the curve in Figure 7A ) . The parameter of the habit system correctly converged to h≈2 , i . e . typical action intensity chosen over trials ( shown by a green curve in the right display of Figure 7D ) . The parameter of the goal-directed system converged to a vicinity of q≈1 , which allows the goal-directed system to expect the reward of 2 after selecting an action with intensity 2 ( according to orange Equation 3 . 2 the reward expected by the goal-directed system is equal to aqs≈2×1×1=2 ) . The right display in Figure 7D shows how the variance parameters in the goal-directed and habit systems changed during the simulation . The variance of the habit system was initialised to a high value , and it decreased over time , resulting in an increased certainty of the habit system . Dopaminergic neurons in the model are only required to facilitate planning in the goal-directed system , where they increase excitability of striatal neurons , but not in the habit system . To illustrate it , Figure 7E shows simulations of a complete dopamine depletion in the model . It shows action intensity produced by the model in which following training , all dopaminergic neurons were set to 0 . After 119 trials of training , on the 120th trial , the model was unable to plan an action . By contrast , after 359 training trials ( when the uncertainty of the habit system has decreased – see the blue curve in right display of Figure 7D ) , the model was still able to produce a habitual response , because dopaminergic neurons are not required for generating habitual responses in the model . This parallels the experimentally observed robustness of habitual responses to blocking dopaminergic modulation ( Choi et al . , 2005 ) . This section shows that the model is able to reproduce two key patterns of behaviour that are thought to arise from interactions between different learning systems , namely the resistance of habitual responses to reward devaluation ( Dickinson , 1985 ) , and Pavlovian-instrumental transfer ( Estes , 1943 ) . In experiments investigating devaluation , animals are trained to press a level ( typically multiple times ) for reward , for example food . Following this training the reward is devalued in a subgroup of animals , e . g . the animals in the devaluation group are fed to satiety , so they no longer desire the reward . Top displays in Figure 8A replot experimental data from one such study ( Dickinson et al . , 1995 ) . The displays show the average number of lever presses made by trained animals during a testing period in which no reward was given for lever pressing . The dashed and solid curves correspond to devaluation and control groups , and the two displays correspond to groups of animals trained for different periods , that is trained until they received 120 or 360 rewards respectively . Figure 8A illustrates two key effects . First , all animals eventually reduced lever pressing with time , thus demonstrating extinction of the previously learned responses . Second , the effect of devaluation on initial testing trials depended on the amount of training . In particular , in the case of animals that received moderate amount of training ( top left display ) the number of responses in the first bin was much lower for the devaluation group than control group . By contrast , highly trained animals ( top right display ) produced more similar numbers of responses in the first bin irrespective of devaluation . Such production of actions despite their consequence being no longer desired is considered as a hallmark of habit formation . The model can also produce insensitivity to devaluation with extensive training . Although the experimental tasks involving pressing levers multiple times is not identical to choosing intensity of a single action , such tasks could be conceptualized as a choice of the frequency of pressing a lever , that could also be described by a single number a . Furthermore , the average reward rate experienced by an animal in paradigms typically used in studies of habit formation ( variable interval schedules that will be explained in Discussion ) may correspond to a non-monotonic function similar to that in Figure 7A , because in these paradigms the reward per unit of time increases with frequency of lever press only to a certain point , but beyond certain frequency , there is no benefit of pressing faster . To simulate the experiment described above , the model was trained either for 120 trials ( bottom left display in Figure 8A ) or 360 trials ( bottom right display ) . During the training the reward depended on action as in Figure 7A . Following this training , the model was tested on 180 trials on which reward was not delivered , so in simulations r=-a reflecting just a cost connected with making an effort . To simulate devaluation , the expectation of reward was set to 0 . Bottom displays in Figure 8A show the average action intensity produced by the model , and they reproduce qualitatively the key two effects in the top displays . First , the action intensity decreased with time , because the valuation and goal-directed systems learned that the reward was no longer available . Second , the action intensity just after devaluation was higher in the highly trained group ( bottom right display ) than in moderately trained group ( bottom left display ) . This effect was produced by the model because after 360 trials of training the variance Σh in the habit system was much lower than after 120 trials ( right display in Figure 7D ) , so after the extended training , the action intensity was to a larger extent determined by the habit system , which was not affected by devaluation . The model can be easily extended to capture the phenomenon of Pavlovian-instrumental transfer . This phenomenon was observed in an experiment that consisted of three stages ( Estes , 1943 ) . First , animals were trained to press a lever to obtain a reward . Second , the animals were placed in a cage without levers , and trained that a conditioned stimulus predicted the reward . Third , the animals were placed back to a conditioning apparatus , but no reward was given for lever pressing . Top display in Figure 8B shows the numbers of responses in that third stage , and as expected they gradually decreased as animals learned that no reward was available . Importantly , in the third and fifth intervals of this testing phase the conditioned stimulus was shown ( highlighted with pink background in Figure 8B ) , and then the lever pressing increased . Thus the learned association between the conditioned stimulus and reward influenced the intensity of actions produced in the presence of the stimulus . The bottom display of Figure 8B shows the action intensity produced by the model in simulations of the above paradigm . As described in Materials and methods , the valuation system learned the rewards associated with two states: presence of a lever , and the conditioned stimulus . During the first stage ( operant conditioning ) , the reward expectation computed by the valuation system drove action planning , while in the second stage ( classical conditioning ) , no action was available , so the valuation system generated predictions for the reward without triggering action planning . In the third stage ( testing ) , on the highlighted intervals on which the conditioned stimulus was present , the expected reward v was increased , because it was a sum of rewards associated with both states . Consequently , the actor computed that a higher action intensity was required to obtain a bigger reward , because the goal-directed system assumes that the action intensity is proportional to the mean reward ( orange Equation 3 . 2 ) . In summary , the model explains the Pavlovian-instrumental transfer by proposing that the presence of the conditioned stimulus increases the reward expected by the valuation system , which results in actor selecting higher action intensity to obtain this anticipated reward . This section shows how models developed within the DopAct framework can also describe more complex tasks with multiple actions and multiple dimensions of state . We consider a task involving choice between two options , often used in experimental studies , as it allows illustrating the generalization , and at the same time results in a relatively simple model . This section will also show that the models developed in the framework can under certain assumptions be closely related to previously proposed models of reinforcement learning and habit formation . To make dimensionality of all variables and parameters explicit , we will denote vectors with a bar and matrices with a bold font . Thus s- is a vector where different entries correspond to intensities of different stimuli in an environment , and a- is a vector where different entries correspond to intensities of different actions . The model is set up such that only one action can be chosen , so following a decision , ai=1 for the chosen action i , while for other actions aj≠i=0 . Thus symbol a- still denotes action intensity , but the intensity of an action only takes binary values once an action has been chosen . Equation 9 . 1 in Figure 9A shows how the definitions of the probability distributions encoded by the goal-directed and habit systems can be generalized to multiple dimensions . Orange Equation 9 . 1 states that the reward expected by the goal-directed system has mean a-TQs- , where Q is now a matrix of parameters . This notation highlights the link with the standard reinforcement learning , where the expected reward for selecting action i in state j is denoted by Qi , j: Note that if a- and s- are both binary vectors with entries i and j equal to 1 in the corresponding vectors , and all other entries equal to 0 , then a-TQs- is equal to the element Qi , j of matrix Q . In the model , the prior probability is proportional to a product of three distributions . The first of them is encoded by the habit system and given in blue Equation 9 . 1 . The expected action intensity encoded in the habit system has mean Hs- , and this notation highlights the analogy with a recent model of habit formation ( Miller et al . , 2019 ) where a tendency to select action i in state j is also denoted by Hi , j . Additionally , we introduce another prior given in Equation 9 . 2 , which ensures that only one action has intensity significantly deviating from 0 . Furthermore , to link the framework with classical reinforcement learning , we enforce a third condition ensuring that action intensity remains between 0 and 1 ( Equation 9 . 3 ) . These additional priors will often result in one entry of a- converging to 1 , while all other entries decaying towards 0 due to competition . Since in our simulations we also use a binary state vector , the reward expected by the goal-directed system will often be equal to Qi , j as in the classical reinforcement learning ( see paragraph above ) . Let us now derive equations describing inference and learning for the above probabilistic model . Substituting probability densities from Equations 9 . 1 and 9 . 2 into the objective function of Equation 4 . 1 , we obtain Equation 9 . 4 in Figure 9B . To ensure that action intensity remained between 0 and 1 ( Equation 9 . 3 ) , ai was set to one of these values if it exceeded the range during numerical integration . To obtain the equations describing action planning or learning , we need to compute derivatives of F over vectors or matrices . The rules for computing such derivatives are natural generalizations of the standard rules and they can be found in a tutorial paper ( Bogacz , 2017 ) . During planning , the action intensity should change proportionally to a gradient of F , which is given in Equation 9 . 5 , where the prediction errors are defined in Equations 9 . 6 . These equations have an analogous form to those in Figure 6A , but are generalized to matrices . The only additional element is the last term in Equation 9 . 5 , which ensures competition between different actions , i . e . a1 will be decreased proportionally to a2 , and vice versa . During learning , the parameters need to be updated proportionally to the corresponding gradients of F , which are given in Equations 9 . 7 and 9 . 8 . Again , these equations are fully analogous to those in Figure 6A . Both action selection and learning in the above model share similarities with standard models of reinforcement learning and a recent model of habit formation ( Miller et al . , 2019 ) . To see which action is most likely to be selected in the model , it is useful to consider the evolution of action intensity at the start of a trial , when ai≈0 , because the action with a largest initial input is likely to win the competition and be selected . Substituting orange Equation 9 . 6 into Equation 9 . 5 and setting ai=0 , we obtain Equation 9 . 9 in Figure 9C . This equation suggests that probabilities of selecting actions depend on a sum of inputs form the goal-directed and habit systems weighted by their certainty , analogously as in a model by Miller et al . , 2019 . There are also similarities in the update rules: if only single elements of vectors a- and s- have non-zero values ai=1 and sj=1 , then substituting Equations 9 . 6 into 9 . 7 and ignoring constants gives Equations 9 . 10 . These equations suggest that the parameter Qi , j describing expected reward for action i in state j is modified proportionally to a reward prediction error , as in classical reinforcement learning . Additionally , for every action and current state j the parameter describing a tendency to take this action is modified proportionally to a prediction error equal to a difference between the intensity of this action and the intensity expected by the habit system , as in a model of habit formation ( Miller et al . , 2019 ) . The similarity of a model developed in the DopAct framework to classical reinforcement learning , which has been designed to maximize resources , highlights that the model also tends to maximize resources , when animal’s reserves are sufficiently low . But the framework is additionally adaptive to the levels of reserves: If the reserves were at the desired level , then R=0 during action planning , so according to Equation 9 . 9 , the goal-directed system would not suggest any action . Let us now consider how the inference and learning can be implemented in a generalized version of the network described previously , which is shown in Figure 10A . In this network , striatum , output nuclei and thalamus include neural populations selective for the two alternative actions ( shown in vivid and pale colours in Figure 10A ) , as in standard models of action selection in the basal ganglia ( Bogacz and Gurney , 2007; Frank et al . , 2007; Gurney et al . , 2001 ) . We assume that the connections between these nuclei are within the populations selective for a given action , as in previous models ( Bogacz and Gurney , 2007; Frank et al . , 2007; Gurney et al . , 2001 ) . Additionally , we assume that sensory cortex includes neurons selective for different states ( shown in black and grey in Figure 10A ) , and the parameters Qi , j and Hi , j are encoded in cortico-striatal connections . Then , the orange and blue terms in Equation 9 . 5 can be computed by the striatal neurons in goal-directed and habit systems in exactly analogous way as in the network inferring action intensity , and these terms can be integrated in the output nuclei and thalamus . The last term in Equation 9 . 5 corresponds to mutual inhibition between the populations selective for the two actions , and such inhibition could be provided by inhibitory projections that are presents in many different regions of this circuit , e . g . by co-lateral projections of striatal neurons ( Preston et al . , 1980 ) or via a subthalamic nucleus , which has been proposed to play role in inhibiting non-selected actions ( Bogacz and Gurney , 2007; Frank et al . , 2007; Gurney et al . , 2001 ) . The prediction error in the goal-directed system ( orange Equation 9 . 6 ) could be computed locally , because the orange dopaminergic neurons in Figure 10A receive inputs encoding all terms in the equation . During learning , the prediction error in the goal-directed system modulates plasticity of the corresponding cortico-striatal connections according to orange Equation 9 . 7 , which describes a standard tri-factor Hebbian rule ( if following movement the striatal neurons encode chosen action , as assumed in Figure 5C ) . The prediction error in the habit system ( blue Equation 9 . 6 ) is a vector , so computing it explicitly would also require multiple populations of dopaminergic neurons in the habit system selective for available actions , but different dopaminergic neurons in the real brain may not be selective for different actions ( da Silva et al . , 2018 ) . Nevertheless , learning in the habit system can be approximated with a single dopaminergic population , because the prediction error δ-h has a characteristic structure with large redundancy . Namely , if only one entry in the vectors a- and s- is equal to 1 and other entries to 0 , then only one entry in δ-h corresponding to the chosen action is positive , while all other entries are negative ( because parameters Hi , j stay in a range between 0 and 1 when initialized within this range and updated according to blue Equation 9 . 7 ) . Hence , we simulated an approximate model just encoding the prediction error for the chosen action ( Equation 10 . 1 ) . With such a single modulatory signal , the learning rules for striatal neurons in the habit system have to be adjusted so the plasticity has opposite directions for the neurons selective for the chosen and the other actions . Such modified rule is given in Equation 10 . 2 and corresponds to tri-factor Hebbian learning ( if striatal neurons in the habit system have activity proportional to a- during learning , as we assumed for the goal-directed system ) . Thanks to this approximation , the prediction error and plasticity in the habit system take a form that is more analogous to that in the goal-directed system . When the prediction error in the habit system is a scalar , the learning rule for the variance parameter ( blue Equation 9 . 8 ) becomes the same as in the model in the previous section ( cf . blue Equation 6 . 5 ) . Materials and method section provides the description of the valuation system in this model , and describes details of the simulations . To illustrate predictions made by the model , we simulated it in a probabilistic reversal task . On each trial , the model was 'presented' with one of two 'stimuli' , that is one randomly chosen entry of vector s- was set to 1 , while the other entry was set to 0 . On the initial 150 trials , the correct response was to select action 1 for stimulus 1 and action 2 for stimulus 2 , while on the subsequent trials , the correct responses were reversed . The mean reward was equal to 1 for a correct response and 0 for an error . In each case , a Gaussian noise ( with standard deviation σr=0 . 5 ) was added to the reward . Figure 11A shows changes in action intensity and inputs from goal-directed and habit systems as a function of time during planning on different trials within a simulation . On an early trial ( left display ) the changes in action intensity were primarily driven by the goal-directed system . The intensity of the correct action converged to 1 , while it stayed at 0 for the incorrect one . After substantial training ( middle display ) , the changes in action intensity were primarily driven by the faster habit system . Following a reversal ( right display ) one can observe a competition between the two systems: Although the goal-directed system had already learned the new contingency ( solid orange curve ) , the habit system still provided larger input to the incorrect action node ( dashed blue curve ) . Since the habit system was faster , the incorrect action had higher intensity initially , and only with time , the correct action node received input from the goal-directed system , and inhibited the incorrect one . Figure 11B shows how parameters in the model changed over trials . Left display illustrates changes in sample cortico-striatal weights in the three systems . The valuation system rapidly learned the reward available , but after reversal this estimate decreased , as the model persevered in choosing the incorrect option . Once the model discovered the new rule , the estimated value of the stimulus increased . The goal-directed system learned that selecting the first action after the first stimulus gave higher rewards before reversal , but not after . The changes in the parameters of the habit system followed those in the goal-directed system . The right display shows that the variance estimated by the habit system initially decreased , but then increased several trials after the reversal , when the goal-directed system discovered the new contingency , and thus the selected actions differed from the habitual ones . Figure 11C shows an analogous pattern in dopaminergic activity , where the neurons in the habit system signalled higher prediction errors following a reversal . This pattern of prediction errors is unique to the habit system , as the prediction errors in the goal-directed system ( orange curve ) fluctuated throughout the simulation following the fluctuations in reward . The increase in dopaminergic activity in the habit system following a reversal is a key experimental prediction of the model , to which we will come back in Discussion . Let us consider the mechanisms of reversal in the model . Since the prediction errors in the habit system do not directly depend on rewards , the habit system would not perform reversal on its own , and the goal-directed system is necessary to initiate the reversal . This feature is visible in simulations , where just after the reversal the agent was still selecting the same actions as before , so the habits were still being strengthen rather weakened ( the blue curve in left display of Figure 11B still increased for ~20 trials after the reversal ) . When the goal-directed system learned that the previously selected actions were no longer rewarded , the tendency to select them decreased , and other actions had higher chances of being selected due to noise ( although the amount of noise added to the choice process was constant , there was a higher chance for noise to affect behaviour , because the old actions were now suggested only by the habit rather than both systems ) . Once the goal-directed system found that the actions selected according to new contingency gave rewards , the probability of selecting action according to the old contingency decreased , and only then the habit system slowly unlearned the old habit . It is worth adding that the reversal was made harder by the fact that a sudden change in reward increased the uncertainty of the goal-directed system ( the orange curve in the right display of Figure 11B increased after reversal ) , which actually weakened the control by that system . Nevertheless , this increase of uncertainty was brief , because the goal-directed system quickly learned to predict rewards in the new contingency and regained its influence on choices . The DopAct framework combines elements from four theories: reinforcement learning , active inference , habit formation , and planning as inference . For each of the theories we summarize key similarities , and highlight the ways in which the DopAct framework extends them . As in classical reinforcement learning ( Houk et al . , 1995; Montague et al . , 1996 ) , in the DopAct framework the dopaminergic neurons in the valuation and goal-directed systems encode reward prediction errors , and these prediction errors drive learning to improve future choices . However , the key conceptual difference of the DopAct framework is that it assumes that animals aim to achieve a desired level of reserves ( Buckley et al . , 2017; Hull , 1952; Stephan et al . , 2016 ) , rather than always maximize acquiring resources . It has been proposed that when a physiological state is considered , the reward an animal aims to maximize can be defined as a reduction of distance between the current and desired levels of reserves ( Juechems and Summerfield , 2019; Keramati and Gutkin , 2014 ) . Under this definition , a resource is equal to such subjective reward only if consuming it would not bring the animal beyond its optimal reserve level . When an animal is close to the desired level , acquiring a resource may even move the animal further from the desired level , resulting in a negative subjective reward . As the standard reinforcement learning algorithms do not consider physiological state , they do not always maximize the subjective reward defined in this way . By contrast , the DopAct framework offers flexibility to stop acquiring resources , when the reserves reach the desired level . The DopAct framework relies on a key high-level principle from the active inference theory ( Friston , 2010 ) that the prediction errors can be minimized by both learning and action planning . Furthermore , the network implementations of the proposed models share a similarity with predictive coding networks that the neurons encoding prediction errors affect both the plasticity and the activity of its target neurons ( Friston , 2005; Rao and Ballard , 1999 ) . A novel contribution of this paper is to show how these principles can be realized in anatomically identified networks in the brain . The DopAct framework shares a feature of a recent model of habit formation ( Miller et al . , 2019 ) that learning in the habit system is driven by prediction errors that do not depend on reward , but rather encode the difference between the chosen and habitual actions . The key new contribution of this paper is to propose how such learning can be implemented in the basal ganglia circuit including multiple populations of dopaminergic neurons encoding different prediction errors . Similarly as in the model describing goal-directed decision making as probabilistic inference ( Solway and Botvinick , 2012 ) , the actions selected in the DopAct framework maximize a posterior probability of action given the reward . The new contribution of this paper is making explicit the rationale for why such probabilistic inference is the right thing for the brain to do: The resource that should be acquired in a given state depends on the level of reserves , so the inferred action should depend on the reward required to restore the reserves . We also proposed a detailed implementation of the probabilistic inference in the basal ganglia circuit . It is useful to discuss the relationship of the DopAct framework to several other theories . The tonic level of dopamine has been proposed to determine the vigour of movements ( Niv et al . , 2007 ) . In our model selecting action intensity , the dopaminergic signals in the valuation and goal-directed systems indeed influence the resulting intensity of movement , but in the DopAct framework , it is the phasic rather than tonic dopamine that determines the vigour , in agreement with recent data ( da Silva et al . , 2018 ) . It has been also proposed that dopamine encodes incentive salience of the available rewards ( Berridge and Robinson , 1998; McClure et al . , 2003 ) . Such encoding is present in the DopAct framework , where the prediction error in the goal-directed system depends on whether the available resource is desired by an animal . To relate the DopAct framework to experimental data , we need to assume a particular mapping of different systems on anatomically defined brain regions . Thus we assume that the striatal neurons in valuation , goal-directed , and habit systems can be approximately mapped on ventral , dorsomedial , and dorsolateral striatum . This mapping is consistent with the pattern of neural activity in the striatum , which shifts from encoding reward expectation to movement as one progresses from ventral to dorsolateral striatum ( Burton et al . , 2015 ) , and with increased activity in dorsolateral striatum during habitual movements ( Tricomi et al . , 2009 ) . This mapping is also consistent with the observation that deactivation of dorsomedial striatum impairs learning which action leads to larger rewards ( Yin et al . , 2005 ) , while lesion of dorsolateral striatum prevents habit formation ( Yin et al . , 2004 ) . Furthermore , we will assume that dopaminergic neurons in valuation , goal-directed , and habit systems can be mapped on a spectrum of dopaminergic neurons ranging from ventral tegmental area ( VTA ) to substantia nigra pars compacta ( SNc ) . VTA is connected with striatal regions we mapped on the valuation system , while SNc with those mapped on the habit system ( Haber et al . , 2000 ) , so we assume that δv and δh are represented in VTA and SNc respectively . Such mapping in consistent with lesions to SNc preventing habit formation ( Faure et al . , 2005 ) . The mapping of the dopaminergic neurons from the goal-directed system is less clear , so let us assume that these neurons may be present in both areas . The key prediction of the DopAct framework is that the dopaminergic neurons in the valuation and goal-directed systems should encode reward prediction errors , while the dopaminergic neurons in the habit system should respond to non-habitual actions . This prediction can be most directly compared with the data in a study where rewards and movements have been dissociated . That study employed a task in which mice could make spontaneous movements and rewards were delivered at random times ( Howe and Dombeck , 2016 ) . It has been observed that a fraction of dopaminergic neurons had increased responses to rewards , while a group of neurons responded to movements . Moreover , the reward responding neurons were located in VTA while most movement responding neurons in SNc ( Howe and Dombeck , 2016 ) . In that study the rewards were delivered to animals irrespectively of movements , so the movements they generated were most likely not driven by processes aiming at achieving reward ( simulated in this paper ) , but rather by other inputs ( modelled by noise in our simulations ) . To relate this task to the DopAct framework , let us consider the prediction errors likely to occur at the times of reward and movement . At the time of reward the animal was not able to predict it , so δv>0 , δg>0 , but it was not necessarily making any movements δh=0 , while at the time of a movement the animal might have not expected reward δv=δg=0 , but might have made non-habitual movements δh>0 . Hence the framework predicts separate groups of dopaminergic neurons to produce responses at times of reward and movements , as experimentally observed ( Howe and Dombeck , 2016 ) . Furthermore , the peak of the movement related response of SNc neurons was observed to occur after the movement onset ( Howe and Dombeck , 2016 ) , which suggests that most of this dopaminergic activity was a response to a movement rather than activity initiating a movement . This timing is consistent with the role of dopaminergic neurons in the habit system , which compute a movement prediction error , rather than initiate movements . While discussing dopaminergic neurons , one has to mention the influential studies showing that VTA neurons encode reward prediction error ( Eshel et al . , 2016; Schultz et al . , 1997; Tobler et al . , 2005 ) . So for completeness , let us reiterate that in the DopAct framework the valuation system is similar to the standard temporal difference learning model , hence it inherits the ability to account for the dopaminergic responses to unexpected rewards previously explained with that model ( Figure 7B ) . The DopAct framework also makes predictions on dopaminergic responses during movements performed to obtain rewards . In presented simulations , such responses were present in all systems ( Figure 7B–C ) , and indeed responses to reward-directed movements were observed experimentally in both VTA and SNc ( Engelhard et al . , 2019; Schultz , 1986 ) . The framework predicts that the responses to movements should be modulated by the magnitude of available reward in the valuation and goal-directed systems , but not in the habit system . This prediction can be compared with data from a task in which animals could press one of two levers that differed in magnitude of resulting rewards ( Jin and Costa , 2010 ) . So for this task , the framework predicts that the dopaminergic neurons in the valuation and goal-directed systems should respond differently depending on which lever was pressed , while the dopaminergic response in the habit system should depend just on action intensity but not reward magnitude . Indeed , a diversity of dopaminergic neurons have been observed in SNc , and the neurons differed in whether their movement related response depended on reward available ( Figure 4j in the paper by Jin and Costa , 2010 ) . In the DopAct framework , the activity of dopaminergic neurons in the goal-directed system is normalized by the uncertainty of that system . Analogous scaling of dopaminergic activity by an estimate of reward variance is also present in a model by Gershman , 2017 . He demonstrated that such scaling is consistent with an experimental observation that dopaminergic responses adapt to the range of rewards available in a given context ( Tobler et al . , 2005 ) . In the DopAct framework the role of dopamine during action planning is specific to preparing goal-directed but not habitual movements ( Figure 7E ) . Thus the framework is consistent with an observation that blocking dopaminergic transmission slows responses to reward-predicting cues early in training , but not after extensive training , when the responses presumably became habitual ( Choi et al . , 2005 ) . Analogously , the DopAct framework is consistent with an impairment in Parkinson’s disease for goal-directed but not habitual choices ( de Wit et al . , 2011 ) or voluntary but not cue driven movements ( Johnson et al . , 2016 ) . The difficulty in movement initiation in Parkinson’s disease seems to depend on whether the action is voluntary or in response to a stimulus , so even highly practiced movements like walking may be difficult if performed voluntarily , but easier in response to auditory or visual cues ( Rochester et al . , 2005 ) . Such movements performed to cues are likely to engage the habit system , because responding to stimuli is a hallmark of habitual behaviour ( Dickinson and Balleine , 2002 ) . Finally , let us discuss a feature of the DopAct framework related to the dynamics of competition between systems during action planning . Such competition is illustrated in the right display of Figure 11A , where after a reversal , the faster habit system initially prepared an incorrect action , but later the slower goal-directed system increased the intensity of the correct action . Analogous behaviour has been shown in a recent study , where human participants were extensively trained to make particular responses to given stimuli ( Hardwick et al . , 2019 ) . After a reversal , they tended to produce incorrect habitual actions when required to respond rapidly , but were able to produce the correct actions given sufficient time . Since the mechanisms of habit formation in the DopAct framework fundamentally differ from a theory widely accepted by a computational neuroscience community ( Daw et al . , 2005 ) , this section is dedicated to comparing the two accounts , and discussing the properties of the habit system in the framework . An influential theory suggests that two anatomically separate systems in the brain underlie goal-directed and habitual behaviour and a competition between them is resolved according to uncertainty of the systems ( Daw et al . , 2005 ) . The DopAct framework agrees with these general principles but differs from the theory of Daw et al . , 2005 in the nature of computations in these systems , and their mapping on brain anatomy . Daw et al . , 2005 proposed that goal-directed behaviour is controlled by a cortical model-based system that learns the transitions between states resulting from actions , while habitual behaviour arises from a striatal model-free system that learns policy according to standard reinforcement learning . By contrast , the DopAct framework suggests that goal-directed behaviour in simple lever-pressing experiments does not require learning state transitions , but such behaviour can be also supported by a striatal goal-directed system that learns expected rewards from actions in a way similar to standard reinforcement learning models . So in the DopAct framework it is the goal-directed rather than habit system that learns according to reward prediction error encoded by dopaminergic neurons . Furthermore , in the DopAct framework ( following the model by Miller et al . , 2019 ) habits arise simply from repeating actions , so their acquisition is not directly driven by reward prediction error , unlike in the model of Daw et al . , 2005 . The accounts of habit formation in the DopAct framework and the model of Daw et al . , 2005 make different predictions . Since the theory of Daw et al . , 2005 assumes that a system underlying habitual behaviour learns with standard reinforcement learning , it predicts that striatal neurons supporting habitual behaviour should receive reward prediction error . However , the dopaminergic neurons that have been famously shown to encode reward prediction error ( Schultz et al . , 1997 ) are located in VTA , which does not send major projections to the dorsolateral striatum underlying habitual behaviour . These striatal neurons receive dopaminergic input from SNc ( Haber et al . , 2000 ) , and it is questionable to what extent dopaminergic neurons in SNc encode reward prediction error . Although such encoding has been reported ( Zaghloul et al . , 2009 ) , studies which directly compared the activity of VTA and SNc neurons demonstrated that neurons encoding reward prediction error are significantly more frequent in VTA than SNc ( Howe and Dombeck , 2016; Matsumoto and Hikosaka , 2009 ) . So the striatal neurons underlying habitual behaviour do not seem to receive much of the teaching signal that would be expected if habit formation arose from the processes of reinforcement learning proposed by Daw et al . , 2005 . By contrast , the DopAct framework assumes that the habit system learns on the basis of a teaching signal encoding how the chosen action differs from the habitual one , so it predicts that SNc neurons should respond to non-habitual movements . It has indeed been observed that the dopaminergic neurons in SNc respond to movements ( Howe and Dombeck , 2016; Schultz et al . , 1983 ) , but it has not been systematically analysed yet if these responses preferentially encode non-habitual movements ( we will come back to this key prediction in the next section ) . It is worth discussing how the habits may be suppressed if previously learnt habitual behaviour is no longer appropriate . In the DopAct framework , old habits die hard . When the habitual behaviour is no longer rewarded , the negative reward prediction errors do not directly suppress the behaviour in the habit system . So , as mentioned at the end of the Results section , in order to reverse behaviour , the control cannot be completely taken over by the habit system , but the goal-directed system needs to provide at least some contribution to action planning to initiate the reversal when needed . Nevertheless , simulations presented in this paper show that for certain parameters the control of habit system may be released when no longer required , and the model can reproduce the patterns of behaviour observed in extinction experiments ( Figure 8 ) . However , simulations by Miller et al . , 2019 show that their closely related model can sometimes persist in habitual behaviour even if it is not desired . Therefore , it is possible that there may exist other mechanisms that may help the goal-directed system to regain control if habitual behaviour ceases to be appropriate . For example , it has been proposed that a sudden increase in prediction errors occurring when environment changes may attract attention and result in the goal-directed system taking charge of animals’ choices ( FitzGerald et al . , 2014 ) . Finally , let us discuss the relationship of the DopAct framework to an observation that habits are more difficult to produce in variable ratio schedules than variable interval schedules ( Dickinson et al . , 1983 ) . In the variable ratio schedules a lever press is followed by a reward with a fixed probability p . By contrast in the variable interval schedule a lever press is followed by a reward only if the reward is 'available' . Just after consuming a reward , lever pressing has no effect , and another reward may become “available” as time goes on with a fixed probability per unit of time . An elegant explanation for why habit formation depends on the schedule has been provided by Miller et al . , 2019 , and a partially similar explanation can be given within the DopAct framework , as we now summarize . Miller et al . , 2019 noticed that reward rate as a function of action frequency follows qualitatively different relationships in different schedules . In particular , in the variable ratio schedule the expected number of rewards per unit time is directly proportional to number of lever presses , i . e . E ( r ) =pa . By contrast , in the variable interval schedule , the reward rate initially increases with the number of level presses , but beyond some frequency there is little benefit of responding more often , so the reward rate is a nonlinear saturating function of action frequency . The model selecting action intensity in the DopAct framework assumes a linear dependence of mean reward on action intensity ( orange Equation 3 . 2 ) , so in the variable ratio schedule , it will learn q=p , and then predict mean reward accurately no matter what action intensity is selected . By contrast , in the variable interval schedule the predictions will be less accurate , because the form of the actual dependence of reward on action frequency is different to that assumed by the model . Consequently , the reward uncertainty of the goal-directed system Σg is likely to be lower in the variable ratio than variable interval schedule . This decreased uncertainty makes the goal-directed system less likely to give in to the habit system , resulting in less habitual behaviour in the variable ratio schedule . We start with describing two most critical predictions of the DopAct framework , testing of which may validate or falsify the two key assumptions of the framework , and next we discuss other predictions . The first key prediction of the DopAct framework is that the dopaminergic neurons in the habit system should respond to movements more , when they are not habitual , e . g . at an initial phase of task acquisition or after a reversal ( Figure 11C ) . This prediction could be tested by monitoring the activity of dopaminergic neurons projecting to dorsolateral striatum in a task where animals are trained to perform a particular response for sufficiently long that it becomes habitual , and then the required response is reversed . The framework predicts that these dopaminergic neurons should have higher activity during initial training and in a period after the reversal , than during the period when the action is habitual . The second key prediction follows from a central feature of the DopAct framework that the expectation of the reward in the goal-directed system arises from forming a motor plan to obtain it . Thus the framework predicts that the dopaminergic responses in the goal-directed system to stimuli predicting a reward should last longer if planning actions to obtain the reward takes more time , or if an animal is prevented from making a response . One way to test this prediction would be to optogenetically block striatal neurons expressing D1 receptors in the goal-directed system for a fixed period after the onset of a stimulus , so the action plan cannot be formed . The framework predicts that such manipulation should prolong the response of dopaminergic neurons in that system . Another way of testing this prediction would be to employ a task where goal-directed planning becomes more efficient and thus shorter with practice . The framework predicts that in such tasks the responses of dopaminergic neurons in the goal-directed system during action planning should get briefer with practice , and their duration should be correlated with reaction time across stages of task acquisition . The DopAct framework also predicts distinct patterns of activity for different populations of dopaminergic neurons . As already mentioned above , dopaminergic neurons in the habit system should respond to movements more , when they are not habitual . When the movements become highly habitual , these neurons should tend to more often produce brief decreases in response ( Figure 7C , right ) . Furthermore , when the choices become mostly driven by the habit system , then dopaminergic neurons in the goal-directed system should no longer signal reward prediction error after stimulus ( Figure 7C , right ) . By contrast , the dopaminergic neurons in the valuation system should signal reward prediction error after stimulus even once the action becomes habitual ( Figure 7B ) . Patterns of prediction errors expected from the DopAct framework could also be investigated with fMRI . Models developed within the framework could be fitted to behaviour of human participants performing choice tasks . Such models could then generate patterns of different prediction errors ( δv , δg , δh ) expected on individual trials . Since prediction errors encoded by dopaminergic neurons are also correlated with striatal BOLD signal ( O'Doherty et al . , 2004 ) , one could investigate if different prediction errors in the DopAct framework are correlated with BOLD signal in different striatal regions . In the DopAct framework dopaminergic neurons increase the gain of striatal neurons during action planning , only in the goal-directed but not in the habit system . Therefore , the framework predicts that the dopamine concentration should have a larger effect on the slope of firing-Input curves for the striatal neurons in the goal-directed than the habit system . This prediction may seem surprising , because striatal neurons express dopaminergic receptors throughout the striatum ( Huntley et al . , 1992 ) . Nevertheless , it is consistent with reduced effects of dopamine blockade on habitual movements ( Choi et al . , 2005 ) that are known to rely on dorsolateral striatum ( Yin et al . , 2004 ) . Accordingly , the DopAct framework predicts that the dopaminergic modulation in dorsolateral striatum should primarily affect plasticity rather than excitability of neurons . This paper described a general framework for understanding the function of dopaminergic neurons in the basal ganglia , and presented simple models capturing only a subset of experimental data . To describe responses observed in more realistically complex tasks , models could be developed following a similar procedure as in this paper . Namely , a probabilistic model could be formulated for a task , and a network minimizing the corresponding free-energy derived , simulated and compared with experimental data . This section highlights key experimental observations the models described in this paper are unable to capture , and suggests directions for developing models consistent with them . The presented models do not mechanistically explain the dependence of dopamine release in ventral striatum on motivational state such as hunger or thirst ( Papageorgiou et al . , 2016 ) . To reproduce these activity patterns , it will be important to extend the framework to describe the computations in the valuation system . It will also be important to better understand the interactions between the valuation and goal-directed systems during the choice of action intensity . In the presented model , the selected action intensity depends on the value of the state estimated by the valuation system , and conversely , the produced action intensity influences reward and thus the value learned by the valuation system . In the presented simulations the parameters ( e . g . learning rates ) were chosen such that the model learned to select action intensity giving highest reward , but such behaviour was not present for all parameter values . Hence it needs to be understood how the interactions between the valuation and goal-directed systems need to be set up so the model robustly finds the action intensity giving the maximum reward . The models do not describe how the striatal neurons distinguish whether dopaminergic prediction error should affect their plasticity or excitability , and for simplicity , in the presented simulations we allowed the weights to be modified only when reward was presented . However , the same dopaminergic signal after a stimulus predicting reward may need to trigger plasticity in one group of striatal neurons ( selective for a past action that led to this valuable state ) , and changes in excitability in another group ( selective for a future action ) . It will be important to further understand the mechanisms which can be employed by striatal neurons to appropriately react to dopamine signals ( Berke , 2018; Mohebi et al . , 2019 ) . The models presented in this paper described only a part of the basal ganglia circuit , and it will be important to include also other elements of the circuit . In particular , this paper focussed on a subset of striatal neurons expressing D1 receptors , which project directly to the output nuclei and facilitate movements , but another population expressing D2 receptors projects via an indirect pathway and inhibits movements ( Kravitz et al . , 2010 ) . Computational models suggest that these neurons predominantly learn from negative feedback ( Collins and Frank , 2014; Mikhael and Bogacz , 2016; Möller and Bogacz , 2019 ) and it would be interesting include their role in preventing unsuitable movements in the DopAct framework . The basal ganglia circuit also includes a hyperdirect pathway , which contains the subthalamic nucleus . It has been proposed that a function of the subthalamic nucleus is to inhibit non-selected actions ( Gurney et al . , 2001 ) , and the hyperdirect pathway may support the competition between actions that is present in the framework . The subthalamic nucleus has also been proposed to be involved in determining when the planning process should finish and action should be initiated ( Frank et al . , 2007 ) . For simplicity , in this paper the process of action planning has been simulated for a fixed interval ( until time t=2 in Figures 7 and 11 ) . It will be important to extend the framework to describe the mechanisms initiating an action . If actions were executed as soon as a motor plan is formed , the increase in the habit prediction error would be briefer than that depicted in Figure 7C . In such an extended model the valuation and goal-directed systems would also need to be modified to learn to expect reward at a particular time after the action . The presented models cannot reproduce the ramping of dopaminergic activity , observed as animals approached rewards ( Howe et al . , 2013 ) . To capture these data , the valuation system could incorporate synaptic decay that has been shown to allow standard reinforcement learning models to reproduce the ramping of prediction error ( Kato and Morita , 2016 ) . It has been also observed that dopaminergic neurons respond not only to unexpected magnitude of reward , but also when the type of reward differs from that expected ( Takahashi et al . , 2017 ) . To capture such prediction errors , the framework could be extended to assume that each system tries to predict multiple dimensions of reward or movement ( cf Gardner et al . , 2018 ) . Finally , dopaminergic neurons also project to regions beyond basal ganglia , such as amygdala , which plays a role in habit formation ( Balleine et al . , 2003 ) , and cortex , where they have been proposed to modulate synaptic plasticity ( Roelfsema and van Ooyen , 2005 ) . It would be interesting to extend the DopAct framework to capture dopamine role in learning and action planning in these regions . We first describe the valuation system , and then provide details of the model in various simulated scenarios . The valuation system was based on the standard temporal difference model ( Montague et al . , 1996 ) . Following that model we assume that the valuation system can access information on how long ago a stimulus was presented . In particular , we assume that time can be divided into brief intervals of length I . The state of the environment is represented by a column vector s-v with entries corresponding to individual intervals , such that sv , 1=1 if the stimulus has been present in the current interval , sv , 2=1 if the stimulus was present in the previous interval , etc . Although more realistic generalizations of this representation have been proposed ( Daw et al . , 2006; Ludvig et al . , 2008 ) , we use this standard representation for simplicity . Figure 12A lists equations describing the valuation system , which are based on temporal difference learning but adapted to continuous time . According to Equation 12 . 1 , the estimate of the value of state s converges in equilibrium to v=w-s-v , where w- denotes a row vector of parameters describing how much reward can be expected after stimulus appearing in a particular interval . Equation 12 . 2 describes the dynamics of the prediction error in the valuation system , which converges to a difference between total reward ( r+v ) and the expectation of that reward made at a previous interval ( vt-I ) , as in the standard temporal difference learning ( Sutton and Barto , 1998 ) . The weight parameters are modified proportionally to the prediction error as described by Equation 12 . 3 , where αv is a learning rate , and e- are eligibility traces associated with weights w- , which describe when the weights can be modified . In basic reinforcement learning e-=s-vT , i . e . a weight can only be modified if the corresponding state is present . Equation 12 . 4 describes the dynamics of the eligibility traces , and if one ignored the first term on the right , it would converge to e-=s-vT . The first term on the right of Equation 12 . 4 ensures that the eligibility traces persist over time , and parameter λ describes what fraction of the eligibility traces survives from one interval to the next ( Ludvig et al . , 2008 ) . Such persistent eligibility traces are known to speed up learning ( Sutton and Barto , 1998 ) . The first term on the right of Equation 12 . 4 includes an eligibility trace from time t-I-3τ , that is from a time slightly further than one interval ago , to avoid the influence of transient dynamics occurring at the transition between intervals . It is also ensured in the simulations that parameters w- do not become negative , as the desired reward value v computed by the valuation system should not be negative . Thus if any element of w- becomes negative , it is set to 0 . Finally , Equation 12 . 5 describes the dynamics of the reward signal r , which follows the actual value to reward r0 . This dynamics has been introduced so that the reward signal rises with the same rate as the value estimate ( the same time constant is used in Equations 12 . 1 and 12 . 5 ) , and these quantities can be subtracted to result in no prediction error when the reward obtained is equal to that predicted by the valuation system . In simulations involving selection of action intensity , the time represented by the valuation system was divided into intervals of I=0 . 2 . The stimulus was presented at time t=1 , while the reward was given at time t=2 , thus the valuation system represented the value of 5 time intervals ( i . e . vectors w- , s-v and e- had 5 elements each ) . The parameters controlling retention of eligibility trace was set to λ=0 . 9 . The state provided to the actor was equal to s=1 from time t=1 onwards . We assumed that the intensity of action executed by the agent was equal to the inferred action intensity plus motor noise with standard deviation σa=1 ( this random number was added to action intensity at time t=2 ) . During intervals in which rewards were provided ( from t=2 onwards ) the parameters were continuously updated according to Equations 6 . 8-9 . In simulations the learning rates were set to: αv=0 . 5 , αg=0 . 05 , αh=0 . 02 , αΣg=0 . 05 , αΣh=0 . 1 . The time constants were set to: τ=0 . 05 , τδ=0 . 02 , and the differential equations were solved numerically using Euler method with integration step 0 . 001 . The model parameters were initialized to: vi=q=0 . 1 , h=0 , Σg=1 and Σh=100 . To simulate devaluation , the expectation of reward was set to 0 by setting vi=q=0 , as a recent modelling study suggests that such scaling of learned parameters by motivational state is required for reproducing experimentally observed effects of motivational state on dopaminergic responses encoding reward prediction error ( van Swieten and Bogacz , 2020 ) . In the simulations of Pavlovian-instrumental transfer , the valuation system was learning the values of two states corresponding to the presence of the lever and the conditioned stimulus . Thus the state vector s-v had 10 entries , where the first 5 entries were set to 1 at different intervals after 'lever appearance' , while the other 5 entries were set to 1 at different intervals after conditioned stimulus . Consequently , the vector of parameters of the valuation system w- also had 10 entries . The simulations of the first stage ( operant conditioning ) consisted of 100 trials in which the model was trained analogously as in the simulations described in the above paragraph . At this stage only first 5 entries of vector s-v could take non-zero values , and hence only the first 5 entries of w- were modified . The state provided to the actor was equal to s=1 when 'lever appeared' that is from time t=1 onwards . The simulations of the second stage ( classical conditioning ) consisted of 100 trials in which only the valuation system was learning . At this stage , the conditioned stimulus was presented at time t=1 , and the reward r=1 was given at time t=2 , thus sv , 6=1 for t∈1 , 1 . 2 , sv , 7=1 for t∈1 . 2 , 1 . 4 , etc . The simulations of the third stage ( testing ) consisted of 60 trials in which only negative reward accounting for effort r=-a was given . On trials 21-30 and 41-50 , both 'lever and conditioned stimulus were presented' , that is sv , 1=sv , 6=1 for t∈1 , 1 . 2 , etc . , while on the other trials only the 'lever was presented' . The model was simulated with the same parameters as described in the previous paragraph , except for modified values of two learning rates αg=0 . 015 , αh=0 . 005 , to reproduce the dynamics of learning shown by experimental animals . In all simulations in this paper , a constraint ( or a 'hyperprior' ) on the minimum value of the variance parameters was introduced , such that if Σg or Σh decreased below 0 . 2 , it was set to 0 . 2 . Analogously , as in the previous section , we first describe the valuation system , and then provide the details of the simulations . In the simulations of choice , we used a simplified version of the valuation system , which for each state j learns a single parameter wj ( rather than the vector of parameters encoding the reward predicted in different moments in time ) . The equations describing this simplified valuation system are shown in Figure 12B . According to Equation 12 . 6 , the estimate of the value of state s converges in equilibrium to v=w-s- . Following reward delivery , parameters wj are modified according to Equation 12 . 7 , where v is taken as the estimated value at the end of simulation of the planning phase on this trial . In order to simulate the actor , its description has been converted to differential equations in analogous way as in Figure 6C . At the end of the planning phase , Gaussian noise with standard deviation σa=2 was added to all entries of the action vector ( to allow exploration ) , and the action with the highest intensity was 'chosen' by the model . Subsequently , for the chosen action i the intensity was set to ai=1 , while for the other action it was set to ak≠i=0 . For simplicity we did not explicitly simulate the dynamics of the model after the delivery of reward r , but we computed the prediction errors in the goal-directed and habit system in an equilibrium ( orange Equation 9 . 6 and Equation 10 . 1 ) , and updated the parameters . In simulations the learning rate in the valuation system was set to αv=0 . 5 on trials with δv>0 , and to αv=0 . 1 when δv≤0 . Other learning rates were set to: αg=0 . 1 , αh=0 . 05 , αΣ=0 . 01 . The remaining parameters of the simulations had the same value as in the previous section .	In the brain , chemicals such as dopamine allow nerve cells to ‘talk’ to each other and to relay information from and to the environment . Dopamine , in particular , is released when pleasant surprises are experienced: this helps the organism to learn about the consequences of certain actions . If a new flavour of ice-cream tastes better than expected , for example , the release of dopamine tells the brain that this flavour is worth choosing again . However , dopamine has an additional role in controlling movement . When the cells that produce dopamine die , for instance in Parkinson’s disease , individuals may find it difficult to initiate deliberate movements . Here , Rafal Bogacz aimed to develop a comprehensive framework that could reconcile the two seemingly unrelated roles played by dopamine . The new theory proposes that dopamine is released when an outcome differs from expectations , which helps the organism to adjust and minimise these differences . In the ice-cream example , the difference is between how good the treat is expected to taste , and how tasty it really is . By learning to select the same flavour repeatedly , the brain aligns expectation and the result of the choice . This ability would also apply when movements are planned . In this case , the brain compares the desired reward with the predicted results of the planned actions . For example , while planning to get a spoonful of ice-cream , the brain compares the pleasure expected from the movement that is currently planned , and the pleasure of eating a full spoon of the treat . If the two differ , for example because no movement has been planned yet , the brain releases dopamine to form a better version of the action plan . The theory was then tested using a computer simulation of nerve cells that release dopamine; this showed that the behaviour of the virtual cells closely matched that of their real-life counterparts . This work offers a comprehensive description of the fundamental role of dopamine in the brain . The model now needs to be verified through experiments on living nerve cells; ultimately , it could help doctors and researchers to develop better treatments for conditions such as Parkinson’s disease or ADHD , which are linked to a lack of dopamine .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "neuroscience" ]	2020	Dopamine role in learning and action inference
Dynamic remodeling of the intrahepatic biliary epithelial tissue plays key roles in liver regeneration , yet the cellular basis for this process remains unclear . We took an unbiased approach based on in vivo clonal labeling and tracking of biliary epithelial cells in the three-dimensional landscape , in combination with mathematical simulation , to understand their mode of proliferation in a mouse liver injury model where the nascent biliary structure formed in a tissue-intrinsic manner . An apparent heterogeneity among biliary epithelial cells was observed: whereas most of the responders that entered the cell cycle upon injury exhibited a limited and tapering growth potential , a select population continued to proliferate , making a major contribution in sustaining the biliary expansion . Our study has highlighted a unique mode of epithelial tissue dynamics , which depends not on a hierarchical system driven by fixated stem cells , but rather , on a stochastically maintained progenitor population with persistent proliferative activity . Tissue growth , maintenance , and remodeling play central roles in ensuring the structural and functional integrity of adult organs and are achieved through the coordinated actions of cell proliferation and differentiation . In these processes , the location , arrangement and timing of cell proliferation are tightly regulated in tissue-specific and context-dependent manners ( Barker et al . , 2010 ) . Selected and dedicated populations of adult stem cells that continuously provide progenies to replenish aged or damaged cells can be found in some tissues , while multiplication of differentiated cells also plays an active role in normal tissue turnover and regeneration in other tissues . Elucidating the mechanisms that regulate the spatiotemporal pattern of cell proliferation in intact tissues will provide fundamental insights into organ homeostasis and regeneration . The liver epithelium comprises the parenchymal cells , or hepatocytes , and the biliary epithelial cells ( BECs ) , also known as cholangiocytes . In the adult liver , these cells are maintained in a quiescent state in which they stable epithelial sheets and tubules . They enter dynamic regeneration processes once the organ suffers tissue loss or various types of injury . In particular , the regeneration process that responds to chronic injury involves drastic changes in the morphology and phenotype of liver epithelial tissues , and in both human pathologies and animal models often accompanies a phenomenon called the ductular reaction ( Gouw et al . , 2011; Michalopoulos and Khan , 2015 ) . The ductular reaction has been histologically characterized as the ectopic emergence and expansion of BEC-marker-positive cells in the liver parenchymal region . After using newly established imaging approaches to capture three-dimensional ( 3D ) tissue morphology in situ , we recently reported that such a phenomenon results from dynamic and adaptive structural changes in intra-hepatic biliary tree architecture ( Kaneko et al . , 2015 ) . Thus , ductular reaction essentially represents drastic and complex remodeling of the biliary epithelial tissue , which is likely to be regulated by a sophisticated mechanism that controls BEC proliferation . Notably , ductular reaction not only serves as an attractive model for studies of the mechanisms of tissue remodeling and cell proliferation , but also is a pathophysiologically relevant regenerative response of the liver to counter various types of injury stimuli ( Michalopoulos , 2014 ) . Several lines of evidence involving knock-out mice for regulatory signals , such as TWEAK , HGF/c-Met and FGF7 , have collectively demonstrated that the suppression or failure of ductular reaction leads to exacerbated liver injury and severe defects in liver regeneration ( Ishikawa et al . , 2012; Takase et al . , 2013; Lu et al . , 2015 ) . However , the in vivo behavior and manner of growth of BECs during biliary tree remodeling remains largely unknown . For many years , ductular reaction has been regarded and studied as a model that represents the activation of adult liver stem/progenitor cells , which may reside in the biliary tree and which can differentiate into hepatocytes or BECs ( Duncan et al . , 2009; Miyajima et al . , 2014 ) . Although in vitro studies have demonstrated the presence of clonogenic cells in the biliary compartment that are highly proliferative under culture conditions ( Miyajima et al . , 2014 ) , the in vivo existence and behavior of such proliferative BEC subpopulations remain unclear . This was partly because many recent studies employing genetic-lineage tracing approaches in vivo have focused on the trans-differentiation capacities of BECs and of hepatocytes , rather than on the mode of proliferation of BECs themselves ( Grompe , 2014; Michalopoulos and Khan , 2015 ) Recently , single-cell approaches have been applied to the field of stem cell research ( Etzrodt et al . , 2014 ) . Among these , in vivo quantitative single-cell tracing has successfully revealed the presence of stem/progenitor cell populations and their unique features in various organs , providing fundamental insights into the cellular basis of tissue homeostasis , regeneration and tumorigenesis ( Doupé et al . , 2010; Driessens et al . , 2012; Hara et al . , 2014 ) . This technique is designed to reveal the rules of cellular dynamics that underlie tissue growth by tracking a population of single cells comprehensively and by deducing the characteristics of proliferative capacity , cell fate and behavior through statistical analyses involving mathematical simulation . A clonal cell tracing study on BECs was recently reported ( Tarlow et al . , 2014b ) that focused solely on the clonal differentiation potential of both hepatocyte and BEC lineages , but not on the clone size ( representing the clonal growth potential ) . Thus , the exact cell numbers of BEC-derived clones were not quantified . The biliary tree has a highly complex and fine structure ( Kaneko et al . , 2015 ) , thus it is practically impossible to count or even estimate the number of BECs that reside in it using conventional histological analysis in tissue sections . Hence , it was essential to develop a 3D imaging method that can provide detailed and reliable tissue structure images and thus allow quantitative assessment . Here , we aimed to elucidate the basic mechanisms that underlie the morphological transformation of the biliary tree , a key process in liver regeneration . In order to study the precise cellular dynamics in the context of complex tissue structures with branching morphology , we introduced three new methods and strategies: high-resolution 3D imaging , quantitative single-cell tracing , and computational simulation . Using our newly established platform to visualize , label and trace BECs in liver tissue , we first revealed that the expansion and remodeling of the biliary tree in a mouse model of chronic liver injury was predominantly driven by the intrinsic growth of biliary epithelial tissue . We performed quantitative single-cell tracing of BECs in vivo to elucidate the underlying cellular behavior , which was further characterized by mathematical modeling and computational simulation . The results highlighted hitherto unrecognized heterogeneity among BECs and the mode of their proliferation , constituting the basis for the drastic structural transformation of the biliary epithelial tissue in regenerating liver in vivo . In order to achieve specific and permanent labeling of BECs , we employed a mouse strain in which a tamoxifen-inducible variant of Cre ( CreERT2 ) is knocked-in to the Prominin1 ( Prom1 ) locus , hereafter referred to as the Prom1-CreERT2 mouse ( Zhu et al . , 2009 ) . Prom1 , better known as the surface antigen CD133 , has been reported to be an 'oval cell/liver progenitor cell ( LPC ) marker' in injured liver ( Rountree et al . , 2007; Dorrell et al . , 2011 ) , but it is also expressed in BECs under normal conditions ( Suzuki et al . , 2008 ) . We examined the expression pattern of the endogenous Prom1 gene product in the normal adult mouse liver , and confirmed that it was expressed in essentially the same manner as CK19 and EpCAM , which are robust and reliable markers of BECs ( Figure 1a , b ) . 10 . 7554/eLife . 15034 . 003Figure 1 . Visualization and lineage labeling of BECs using Prominin1 ( Prom1 ) expression . ( a ) Immunofluorescent ( IF ) staining for CK19 and Prom1 in the adult mouse liver ( scale bars , 100 μm ) . ( b ) Representative expression pattern of Prom1 and EpCAM by fluorescence-activated cell sorting ( FACS ) analysis . Each dot or point represents an individual cell . The colors ( pseudo-colors ) indicate the density of dots ( i . e . , cells ) , corresponding to increasing numbers of events from blue to red . The outlying ( and outlined ) group highlights the EpCAM+ Prom1+ double positive population , indicating that these markers are co-expressed with each other . Successive gates were applied for DAPI- , forward and side scatter ( FSC/ SSC ) , and pulse width ( not shown ) . ( c ) Prom1-CreERT2;R26R-tdTomato mice were used to allow the detection of Prom1+ cells on the basis of LacZ expression . X-gal staining was performed on tissue sections ( left panel ) and whole liver samples ( middle panels ) . The latter was cleared with benzyl-alcohol and benzyl-benzoate ( BABB ) after staining . The intra-hepatic biliary tree , as well as the extra-hepatic bile duct ( white arrowhead ) , and the base of the gallbladder ( white arrow ) were visualized . Liver sections were also stained with anti-LacZ and anti-CK19 antibodies ( right panel ) . ( d and e ) Lineage labeling in the Prom1-CreERT2;R26R-tdTomato mouse liver after tamoxifen administration ( scale bars , 100 μm ) . ( d ) Immunostaining of liver section . ( e ) Representative FACS plot pattern of labeled cells . Successive gates were applied for DAPI- , FSC/SSC , pulse width and EpCAM+ ( not shown ) . All experiments were performed with at least four biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 00310 . 7554/eLife . 15034 . 004Figure 1—figure supplement 1 . Quantification of lineage-labeled cells in Prom1-CreERT2; R26R-tdTomato mice . ( a ) BEC labeling ratio in the Prom1-CreERT2;R26R-tdTomato mice was evaluated by FACS analysis . Left panel shows the experimental design . Two weeks after tamoxifen administration , the mice were subjected to analysis either directly ( Before injury ) or after 8 weeks of thioacetamide ( TAA ) injury . Non-parenchymal cells were collected from the dissociated liver and stained with anti-EpCAM antibody . FACS gates were applied sequentially as follows: DAPI- ( live cells ) , FSC and SSC , pulse width , and EpCAM+ . No labeled cell was detected in the mice treated with vehicle alone ( No tamoxifen ) . Before and after tamoxifen administration , tdTomato+ cells were detected only in the EpCAM+ population ( not shown ) . ( b ) Quantification of the labeled BEC populations . No significant difference was observed between those before and after the TAA injury . The data represent the mean ± SEM for the results of five different experiments for each condition . p=0 . 70 ( un-paired Student’s t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 00410 . 7554/eLife . 15034 . 005Figure 1—figure supplement 2 . EdU uptake assay in lineage-labeled cells in Prom1-CreERT2;R26R-tdTomato mice . EdU uptake assay was performed to examine the proliferative characteristics of lineage-labeled and non-labeled BECs . After tamoxifen administration ( 10 mg/20 g body weight ) , the Prom1-CreERT2;R26R-tdTomato mice were subjected to TAA injury for 8 weeks , injected intraperitoneally with EdU ( 1 . 5 mg/20 g body weight ) , and then analyzed . For FACS analysis , successive gates were applied for DAPI- , FSC/SSC , pulse width , and EpCAM+ ( not shown ) . No significant difference was observed between the labeled ( tdTomato+ ) and non-labeled ( tdTomato- ) BECs . The data represent the mean ± SD for five mice each . p=0 . 58 ( paired Student’s t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 00510 . 7554/eLife . 15034 . 006Figure 1—figure supplement 3 . No labeled cells were detected in the absence of tamoxifen in Prom1-CreERT2;R26R-tdTomato mice . ( a and b ) Leaky labeling in the Prom1-CreERT2;R26R-tdTomato mice was evaluated by FACS analysis . Left panel shows the experimental design . The mice were analyzed in the presence ( a ) or absence ( b ) of TAA injury after 8 weeks . Non-parenchymal cells were collected from the dissociated liver and stained with anti-EpCAM antibody . FACS gates were applied sequentially as follows: DAPI- ( live cells ) , FSC/SSC , pulse width , and EpCAM+ . No labeled cell was detected in either condition . Experiments were performed with three biological replicates and representative results are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 006 In the Prom1-CreERT2 strain , a nuclear localization signal-conjugated LacZ ( nLacZ ) gene was also knocked-in to the same Prom1 locus . X-gal staining experiments using liver sections showed that nLacZ was expressed in a BEC-specific manner ( Figure 1c , left and right panels ) . We also performed whole-mount X-gal staining of the entire liver of the Prom1-CreERT2 mice . This resulted in 3D visualization of finely branching , tree-like architecture spreading throughout the organ ( Figure 1c , middle and right panels ) , the pattern of which matches well with the biliary tree structure that we recently revealed using an ink-casting technique ( Kaneko et al . , 2015 ) . The staining pattern was also consistent with 3D images of CK19 immunostaining ( see below ) . We crossed the Prom1-CreERT2 mice with a Cre-inducible fluorescent reporter mouse strain ( R26R-tdTomato ) for permanent labeling and tracing of BECs ( Madisen et al . , 2010 ) . We first administered a relatively high dose of tamoxifen ( 10 mg/20 g mouse body weight ) into Prom1-CreERT2;R26R-tdTomato mice and analyzed their livers to evaluate labeling specificity and efficiency . Immunostaining images ( Figure 1d ) and FACS plot ( Figure 1e and Figure 1—figure supplement 1 ) showed that all of the labeled cells were included in the EpCAM+ BEC population , indicating that the Prom1-CreERT2;R26R-tdTomato mouse can be used for specific labeling of BECs . The labeling efficiency at this dose of tamoxifen was approximately 30% ( Figure 1—figure supplement 1 ) . The labeling seemed to occur at random in the BEC population: the labeled tdTomato+ cells were distributed ubiquitously among EpCAM+ BECs with no apparent relation to features of the biliary structure , which will be described later in more detail . We also compared the Prom1-CreERT2 lineage-labeled and non-labeled cells in terms of their proliferative characteristics upon liver injury; using a 5-ethynyl-2’-deoxyuridine ( EdU ) incorporation assay , and found no significant difference between them ( Figure 1—figure supplement 2 ) , showing that their proliferative capacities are indistinguishable . In addition , the BEC lineage labeling rate of around 30% did not change significantly , even after ductular reaction was induced by a chronic injury model ( Figure 1—figure supplement 1 ) . We also confirmed that no leakiness of labeling occurred in the absence of tamoxifen administration ( Figure 1—figure supplement 3 ) . These results indicate that the BEC labeling in our system occurs in an un-biased manner , and that the labeled cells faithfully represent the entire BEC population . As shown in Figure 1c , the intrahepatic biliary epithelial tissue exhibits a complex yet ordered tree-like structure , which cannot be readily recognized by conventional histological analyses of thin tissue sections . We modified and improved a 3D immunostaining and imaging protocol using thick sections that we had recently reported ( Kaneko et al . , 2015 ) , and established a new fluorescent 3D imaging platform with a single-cell resolution . As schematically depicted in Figure 2a , we cut liver samples into thick sections ( 200−500 μm thickness ) and subjected them to immunostaining and optical clearing . The 3D image obtained by this method consistently recapitulated the basic structural unit of the biliary tree obtained by other protocols ( Kaneko et al . , 2015; Takashima et al . , 2015 ) ( Figure 2b ) . One or two duct tubes run alongside the portal vein ( PV ) ( Figure 2b; PV is not shown ) , and many finer branches , which we call ductules hereafter , protrude from the duct and wrap around the PV ( Figure 2c ) . This structural unit was observed around the entire biliary tree ( Figure 2—figure supplement 2a ) . 10 . 7554/eLife . 15034 . 007Figure 2 . Pre-existing BECs contribute to the nascent biliary epithelial tissue structure upon injury . ( a ) Schematic illustration of the 3D imaging method used to observe biliary tree structures . ( b ) 3D imaging view of the normal biliary tree structure revealed by anti-CK19 immunostaining . Z-stacked images were acquired by confocal microscopy and reconstructed by IMARIS software ( normal shading mode ) . ( c ) Schematic model for the biliary tree structure under the normal condition . ( d ) Experimental scheme . ( e–h ) 3D reconstructed images of the biliary tree revealed by anti-CK19 immunostaining ( green ) , showing the distribution of the BEC lineage-labeled cells ( red ) in the expanded biliary structure . Serial z-stacked confocal images were tiled ( 3 x 3 tiles ) automatically by automatic positioning stage and Olympus fluoview software . Data are displayed as maximum-intensity projections . A region indicated by a white box in the left panel is magnified in the middle and right panels . Scale bars represent 100 μm . ( f ) White arrows indicate that pre-existing BECs ( tdTomato+ cells ) are extending outward . ( g ) White arrows indicate a branch of the biliary tree that connects the biliary duct around the PV with newly formed biliary branches around the CV . ( h ) White arrows indicate clusters of labeled cells that are located around the CV . White arrowheads indicate that the duct compartment around the PV shows a uniform mosaic pattern . All experiments were performed with at least five biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 00710 . 7554/eLife . 15034 . 008Figure 2—figure supplement 1 . The level and distribution pattern of the ductular reaction in a microscopic view is highly diversified within a liver . ( a ) Progression of the ductular reaction is not uniform microscopically . Sections were prepared from different regions of the same liver sample derived from a mouse treated with TAA for 6 weeks , and then immunostained with anti-CK19 antibody to reveal the biliary tree structure . Note that the level and distribution pattern of the ductular reaction appear considerably diversified even in the same liver when we focus on relatively small areas . ( b ) The area of CK19+ regions in different CV units ( >18 per mouse ) was quantified by using Volocity software and plotted to show highly divergent levels of biliary tree expansion in each liver . Three different mice treated with TAA for 6 weeks were analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 00810 . 7554/eLife . 15034 . 009Figure 2—figure supplement 2 . Macroscopic view of the ductular reaction upon TAA injury over time . The entire tissue structure of the biliary tree was visualized by whole-mount X-gal staining of the liver from the Prom1-CreERT2 mice harboring the nLacZ gene knocked-in to the Prom1 locus . Whole-liver samples were harvested , fixed , stained and cleared with clearing reagent . Images were acquired with a macro zoom microscope ( Olympus MVX10 ) . Green signals represent nuclei of the Prom1+ cells . Scale bars , 1 mm . ( a ) Prom1+ signals revealed the finely branching biliary tree structure under the normal condition . The duct and ductule units can be clearly recognized , delineating the PV . ( b and c ) Upon TAA-induced injury , fine branches of the biliary tree extend from around the PV outward to the liver parenchyma . ( d and e ) After extensive branching and proliferation , BECs ( Prom1+ cells ) occupied the entire liver lobule . Note that the ductular reaction proceeds in a relatively stereotyped manner throughout the liver . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 009 We chose the thioacetamide ( TAA ) -induced chronic liver injury protocol as an experimental model system with which to study the dynamic morphological changes and cellular behavior of the biliary epithelial tissue . TAA is known to induce localized cell death of hepatocytes , specifically around the central vein ( CV ) . Continuous administration of this drug causes chronic inflammation and bridging fibrosis , and eventually leads to tumor formation , reminiscent of the progression of fibrotic and cirrhotic liver disease in humans ( Yeh et al . , 2004; De Minicis et al . , 2013 ) . In addition to this pathophysiological relevance , pilot test experiments comparing several injury models also revealed that TAA caused the least autofluorescence in liver specimens ( data not shown ) , which is suitable and advantageous for obtaining high-quality imaging data that can be used to perform analyses on detailed tissue and cellular structures . Before proceeding with single-cell clonal tracing of BECs , we conducted tracing experiments in a condition where BECs were labeled en masse to analyze: the pattern of 3D changes in the morphology of the biliary structure; and the relative contribution of the pre-existing BECs to the expanded structures . Beginning 2 weeks after a single administration of a high dose of tamoxifen , the Prom1-CreERT2;R26R-tdTomato mice were continuously administered TAA ( Figure 2d ) . We first confirmed the spatial distribution of the labeled cells under normal conditions at the onset of the injury protocol . As shown in Figure 2e , the BEC labeling occurred evenly in both the duct and ductule compartments , and in an un-biased mosaic pattern independent of other morphological features such as the duct size or branch locations . We then analyzed the fate of labeled cells and branching morphogenesis over 8 weeks of the injury ( Figure 2f–h ) . The biliary tree structure began to undergo dynamic remodeling by TAA 2 weeks , with ductules , which reside omni-directionally around the PV in the normal condition , extending uni-directionally toward the CV . The labeled tdTomato+ cells were clearly observed in the extending ductular compartment , indicating that the pre-existing biliary epithelial tissue , presumably the ductule , has undergone the morphological change ( Figure 2f , white arrows ) . At TAA 4 weeks , the biliary extension that reached deep into the liver lobule had begun to form intricately branched structures around the CV area . Many tdTomato+ cells were observed in these newly formed branches . Remarkably , none of the expanded branches lost their connectivity with the main duct in the peri-portal area ( Figure 2g , white arrows ) . This indicates that the ductular reaction is neither the migration of detached BECs nor ectopic emergence of BEC-like cells at a distant area in the liver parenchyma , but rather an extension of branching architecture from the pre-existing biliary tissue . Morphological changes further continued along with the disease progression , and at TAA 8 weeks , drastically expanded BECs appeared to make mesh-like network structures ( Figure 2h ) . We observed changes to the distribution pattern of the labeled BECs during the course of the injury progression . The labeled tdTomato+ cells remained evenly distributed in a mosaic fashion in the main duct structure around the PV ( Figure 2h , white arrowheads ) , but they displayed apparently uneven distribution in the newly formed structures . That is , the expanded compartment around the CV was composed of several clusters of tdTomato+ cells ( Figure 2h , white arrows ) , implicating clonal expansion of a BEC subpopulation therein . Notably , we found two opposite features , diversity and uniformity , of the ductular reaction that were visible at the micro- and macro- scale , respectively . When subjected to microscopic observation , biliary tree within each defined region of interest exhibited diversity in terms of growth speed and structural features , particularly at its periphery , even in 3D images . Quantitative analysis of BEC distribution in tissue sections supported this notion ( Figure 2—figure supplement 1 ) . Taking advantage of the LacZ expression in BECs of the Prom1-CreERT2 mouse strain , we also examined the TAA-induced biliary tissue remodeling at the macroscopic scale ( Figure 2—figure supplement 2 ) . Organ-wide visualization of the entire biliary tree structure enabled us to grasp the landscape of the ductular reaction , in which whole-tissue remodeling follows an ordered and uniform pattern with respect to both spatial and temporal changes . The results of the BEC lineage tracing in the 3D tissue architecture strongly suggested that pre-existing BECs made a major contribution to the ductular reaction in the TAA injury model , at least within the initial period of 8 weeks . It should be noted that , in the present experimental setting , virtually all of the labeled cells were contained within the CK19+ cell population and that labeled hepatocytes were rarely detected ( Figure 2h ) . This result is consistent with recent reports by many other groups who have used lineage tracing experiments , in that the cells that are induced and expanded by the ductular reaction , or LPCs , do not show stem or progenitor cell-like activity that contributes to new hepatocytes in most , if not all , models of mouse liver injury ( Grompe , 2014; Tarlow et al . , 2014b; Yanger et al . , 2014 ) . On the basis of the results of the 3D tracing of BECs , we assumed that pre-existing BECs were the main source of the newly formed biliary structure . However , several studies have recently reported that hepatocytes are capable of converting to become BEC-like cells upon liver injury ( Michalopoulos et al . , 2005; Sekiya and Suzuki , 2012; Yanger et al . , 2013; Nagahama et al . , 2014; Tanimizu et al . , 2014; Tarlow et al . , 2014a ) . We thus sought to employ a complimentary lineage-tracing strategy to directly evaluate the contribution of hepatocytes as an alternative source of the expanded biliary structure upon TAA-induced ductular reaction . For lineage tracing of hepatocytes , we used a recombinant adeno-associated virus vector pseudoserotyped with capsid 8 ( rAAV2/8 ) , which expresses an improved version of the Cre recombinase ( iCre ) gene under the control of a hepatocyte-specific promoter . The rAAV2/8 vector is well known to target hepatocytes in the mouse liver in a highly specific and efficient manner , and has been used in many studies to transfer genes into hepatocytes in vivo . Importantly , it does not transduce BECs in the adult mouse liver ( Yanger et al . , 2014 ) . We injected rAAV2/8-iCre into the R26R-tdTomato mice to permanently label hepatocytes ( Figure 3a ) . FACS analysis of liver cells isolated from these mice confirmed that almost all hepatocytes were labeled ( more than 97%; Figure 3b ) , while EpCAM+ BECs were not labeled ( less than 0 . 1%; Figure 3—figure supplement 1b ) . We also confirmed the specificity and efficiency of the rAAV2/8-iCre-mediated labeling by using immunohistological analysis ( Figure 3c ) . 10 . 7554/eLife . 15034 . 010Figure 3 . Lineage tracing of hepatocytes . ( a ) R26R-tdTomato mice were used in combination with rAAV2/8-iCre for the labeling of hepatocytes . rAAV2/8-iCre is designed to transduce only hepatocytes . ( b ) Representative image of FACS analysis of hepatocytes labeled by rAAV2/8-iCre . These histogram images show the result of serial purification gates ( FSC/SSC , pulse width , DAPI- ) . ( c ) Adult R26R-tdTomato mice were injected with rAAV2/8-iCre ( 1x1011 vg /mouse ) . 2 weeks after injection , the mice were sacrificed and the livers were stained with anti-EpCAM and anti-Spp1 antibody ( scale bar , 100 um ) . ( d and e ) Mice were injected with rAAV2/8-iCre ( 1011 vg /mouse ) and then subjected to a 3 , 5-diethoxycarbonyl-1 , 4-dihydrocollidine ( DDC ) or TAA injury model . tdTomato+ Spp1+ EpCAM- cells were only observed in DDC-fed mouse liver sections ( white arrows ) . Analysis was done with 5 mice per each injury model . More than 6 sections were made per mouse . ( f ) 3D imaging was performed with WT mice ( normal state , DDC for 8 weeks , TAA for 8 weeks ) . Acquired z-stack data is displayed as maximum-intensity projection after contrast adjustment with IMARIS software . In the DDC liver , Spp1+ EpCAM- cells were observed ( white arrow ) around main biliary tubular structures that were composed of EpCAM+ cells . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01010 . 7554/eLife . 15034 . 011Figure 3—figure supplement 1 . FACS analysis of hepatocyte-derived cells in the TAA and DDC models . The experimental design is described on the left side of the figure . Hepatocytes of R26R-tdTomato mice were labeled with rAAV2/8-iCre . Dissociated liver cells were stained with anti-MIC1-1C3 antibody , anti-EpCAM antibody , anti-CD45 antibody and anti-Prom1 antibody . The results for Prom1 are not shown because they were essentially the same as those for EpCAM . All experiments were done more than three times to confirm reproducibility . Sequential FACS gates were applied; DAPI- ( live cells ) , FSC/SSC , and pulse width . ( a ) All EpCAM+ cells were included in the MIC1-1C3+ cell population . ( b ) Before injury , almost no MIC1-1C3+ or EpCAM+ cells were labeled . ( c ) After DDC injury , a tdTomato+ population emerged in the MIC1-1C3+ population . Meanwhile almost no labeled cells emerged in the EpCAM+ population . ( d ) After TAA injury , almost no tdTomato+ cells appeared in MIC1-1C3+ population or in the EpCAM+ population . Experiments were performed with three biological replicates and representative results are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01110 . 7554/eLife . 15034 . 012Figure 3—figure supplement 2 . FACS quantification of labeled ratio in the Prom1-CreERT2; R26R-tdTomato mice before/after DDC injury . ( a and b ) Labeling ratio in EpCAM+or MIC1-1C3+ cell populations in the Prom1-CreERT2;R26R-tdTomato mice was evaluated by FACS . The panels on the left show the experimental design . The mice were subjected to analysis DDC injury . Non-parenchymal cells were collected from the dissociated liver and stained with anti-EpCAM antibody or anti-MIC1-1C3 antibody . FACS gates were applied sequentially as follows: DAPI- ( live cells ) , FSC/SSC , pulse width , CD45- , and EpCAM+ or MIC1-1C3+ . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 012 Here , lineage-tracing of hepatocytes was performed in both the TAA and the DDC injury models , as the latter was used in recent reports to convincingly demonstrate the contribution of hepatocytes to the proliferating ductules ( Tanimizu et al . , 2014; Tarlow et al . , 2014a ) . We analyzed the expression of several molecular markers that are associated with BECs/LPCs , including EpCAM , CK19 , Prom1 , Spp1 ( also known as osteopontin ) , and MIC1-1C3 . Upon DDC administration for 8 weeks , we detected many Spp1+ cells in the tdTomato-labeled population , but we did not detect any EpCAM+ tdTomato+ cells in this population ( Figure 3d ) . This indicates that while hepatocytes can indeed gain a 'biliary' phenotype in terms of Spp1 expression , they are still not fully converted to BECs as defined by EpCAM expression . FACS analysis using the MIC1-1C3 surface antigen instead of Spp1 also revealed that hepatocytes contributed to the MIC1-1C3+ 'biliary' population but not to that of the EpCAM+ subset ( Figure 3—figure supplement 1c ) . We also examined the expression patterns of Prom1 and CK19 and found that they were essentially the same as that of EpCAM , but they were different from those of Spp1 and MIC1-1C3 ( data not shown ) . These results are consistent with the report by Tarlow et . al . ( Tarlow et al . , 2014a ) showing that Spp1+ MIC1-1C3+ proliferating ductular cells derived from hepatocytes upon DDC injury ( which they named hepPDs ) are distinct from those derived from pre-existing BECs ( which they named bilPDs ) with respect to the gene expression profile . Specifically , such hepatocyte-derived cells ( hepPDs ) show little or no expression of CK19 , EpCAM , or Prom1 . We also estimated the contribution of hepPDs to the biliary population using our BEC labeling system , in which a decrease in the labeling ratio of BECs would be expected if a large number of hepatocytes converted into BECs . We found no significant decrease of labeling ratio after DDC administration in EpCAM+ cells . The ratio might be slightly decreased in MIC1-1C3+ cells , but this remains to be confirmed as statistically significant ( Figure 3—figure supplement 2 ) . Considering the relatively small contribution of the hepatocyte-derived cells to the entire MIC1-1C3+ population upon DDC injury ( 1 . 88% in Figure 3—figure supplement 1 ) , the result overall is consistent with that of the hepatocyte-tracing experiments using the AAV system . In the TAA model ( as in the DDC model ) , we did not detect the emergence of hepatocyte-derived EpCAM+ cells ( Figure 3e ) . Intriguingly , even hepatocyte-derived Spp1+ cells were not detected , suggesting that hepPDs are not induced under this injury condition . Consistently , FACS analysis detected no MIC1-1C3+ cells or EpCAM+ cells in the tdTomato+ population ( Figure 3—figure supplement 1d ) , further confirming the notion that hepatocytes do not convert into either BECs ( or bilPDs ) or BEC-like cells ( or hepPDs ) in the course of TAA-induced liver injury , even after 8 weeks . Again , the expression patterns of Prom1 and CK19 were the same as that of EpCAM , but different from those of Spp1 and MIC1-1C3 ( data not shown ) . Thus , Spp1 and MIC1-1C3 mark a different and broader cell population from that defined by the expression of CK19 , EpCAM and Prom1 . We concluded that hepatocyte-derived duct-like cells ( hepPDs ) do emerge in the DDC model but not in the TAA model , and that they are distinguishable from bilPDs . This finding may be of pathophysiological relevance , in that the DDC model is a cholestatic liver injury model that primarily targets the biliary epithelial system , whereas the TAA model is a hepatotoxic injury model with less severe cholestatic disease phenotypes . To gain an insight into the spatial relationship between hepPDs ( Spp1+ EpCAM- ) and the biliary tissue structure composed of bilPDs ( Spp1+ EpCAM+ cells ) , we further applied the 3D immunostaining and imaging analyses . 3D images confirmed that hepPDs were barely detected under the normal physiological condition or upon TAA-induced injury . In the DDC model , however , hepPDs emerged around several parts of the pre-existing biliary architecture ( Figure 3f ) . The tissue architecture composed of hepPDs was different from that composed of bilPDs: the former did not form any single and contiguous branch structure but rather formed a subsidiary structure , located just around the pre-existing branched structures of bilPDs . The results of these fate-tracing and 3D-imaging analyses confirmed that although hepatocytes are capable of being converted into the biliary epithelial state in a context-dependent manner , the main contributors to the ductular reaction ( in terms of gene expression , cell number , and 3D structure ) are pre-existing BECs . These findings also indicate that we can focus solely on the fate and behavior of pre-existing BECs when attempting to determine the cellular basis of the ductular reaction , particularly when using the TAA model . We have to note that this study does not refute the existence of hepatocyte-to-biliary phenotypic conversion , nor its contribution to liver regeneration . We wish to point out that hepPDs and bilPDs are different and their significance should be discussed separately . Based upon the notion that the nascent biliary structure is formed upon TAA injury through intrinsic cell growth in the biliary epithelial tissue , we set out to perform quantitative single-BEC tracing experiments in vivo in order to elucidate the cellular basis of this dynamic tissue remodeling process ( Figure 4a , b ) . Prom1-CreERT2;R26R-tdTomato mice were administered a very low dosage of tamoxifen to label BECs at a very low frequency . 3D imaging and FACS analyses showed that the labeling was introduced at a rate of less than 0 . 2% of total BECs , and that the labeled cells located singly and apart from each other ( Figure 4c , d ) . After single-cell labeling , mice were subjected to the TAA protocol and the size of single-cell-derived clonal progenies were analyzed during the course of injury progression . Of note , the labeling index of 0 . 2% did not change even after the injury period , which is consistent with the notion that the labeling was introduced in an unbiased manner ( Figure 4—figure supplement 1 ) . After 6 weeks of injury , there were still many BECs that had rarely divided and that remained as single cells or merely as clusters composed of a few cells . At the same time , we also observed larger colonies composed of dozens of cells , indicating that some BECs had undergone several rounds of cell division ( Figure 4e , f ) . This clearly indicates that the BEC population does not proliferate uniformly as a whole upon injury , but rather , BECs exhibit heterogeneity with regard to their proliferative capacity in vivo . 10 . 7554/eLife . 15034 . 013Figure 4 . BECs exhibit heterogeneity in terms of proliferative capacity in vivo . ( a ) Schematic diagram showing the rationale for quantitative in vivo single-BEC tracing . ( b ) Experimental design . ( c ) Upon administration of a very low dosage of tamoxifen ( 0 . 25 mg/kg body weight ) , liver samples were stained with anti-CK19 antibody and Hoechst33342 . BEC labeling was introduced at the single-cell level ( white arrows ) . A 3D image and an optical section corresponding to the same visual field are shown in the left and right panels , respectively ( scale bars , 100 μm ) . ( d ) Quantification of the BEC-labeling efficiency after the low-dosage tamoxifen injection . For FACS analysis , successive gates were applied for DAPI- , FSC/SSC , pulse width and EpCAM+ ( not shown ) . A representative plot pattern for 4 biological replicates is shown . ( e and f ) 3D images of labeled colonies after 6 weeks of TAA injury . Thick sections were stained with anti-CK19 antibody and 3D images were acquired with tdTomato+ colonies ( white arrows ) using confocal microscopy . The data are shown as maximum intensity projections . ( e ) Duct compartment around the PV area . ( f ) Peripheral ductule compartment around the CV area ( scale bars , 100 μm ) . ( g ) Distribution of the quantified colony size at TAA 6 weeks ( n = 5 mice , mean ± SD ) . The colonies were classified into two categories ( duct and peripheral ductule ) as described in the 'Materials and methods' section . ( h ) Relative numbers of colonies categorized by colony size as depicted in the legend to the right ( left stacked bar chart ) , and the relative contribution of cell amounts from each colony category ( right stacked bar chart ) ( calculated as follows: 100 x ( sum of the cell numbers in a colony size ) / ( sum of all the counted cell numbers ) ) . ( i ) Scatter plot of the colony size distribution over time . Data from five mice were pooled for each time point ( total colony numbers counted were 257 , 272 , 304 , 307 and 310 for the 0 , 2 , 4 , 6 and 8 week samples , respectively ) . Horizontal lines show the mean of colony size . Images shown in panels ( c ) , ( e ) , and ( f ) are representative data for at least 5 biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01310 . 7554/eLife . 15034 . 014Figure 4—figure supplement 1 . Labeling ratio in the Prom1-CreERT2; R26R-tdTomato mice was not changed after 8 weeks of TAA injury . Left panel shows the experimental design . Mice were subjected to TAA injury after the low-dosage tamoxifen injection ( 0 . 25 mg/kg body weight ) . Before injury , the labeling ratio of BECs was less than 0 . 2% ( around 0 . 18% ) as shown in Figure 4d . Changes of labeling ratio were quantified after TAA injury . For FACS analysis , successive gates were applied for DAPI- , FSC/SSC , pulse width and EpCAM+ ( not shown ) . Experiments were performed with 3 biological replicates and representative results are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 014 Taking advantage of the 3D imaging , we quantified the exact size ( i . e . , the numbers of constituent cells ) of each single-cell-derived clone in the liver tissue . We classified the biliary epithelial tissue into two distinct portions according to their tube diameter sizes and locations , namely 'duct' and 'peripheral ductule' , as we anticipated a putative relationship between the tissue architecture and cell proliferation capacity . This highlighted two important features of colony size distribution . First , the duct compartment contained no large colonies of more than five cells , while the peripheral ductule compartment did contain large colonies ( Figure 4e–g ) . Notably , we found no large colonies that were directly connected to the duct cells adjacent to PV . Thus , it is not likely that some proliferative cells reside in a fixed position in the duct compartment and continuously supply progeny toward the periphery . Rather , the colony-initiating , proliferative cells may reside in the peripheral ductule compartment and relocate their position upon parenchymal injury . The second feature of colony size distribution relates to the pattern if this distribution . We initially assumed that the distribution of colony size would have a bimodal shape , with one peak corresponding to a non-proliferative population and the other to a proliferative one . However , the empirical data showed a unimodal distribution pattern with long tail ( Figure 4g and i ) . Although the proportion of large colonies was seemingly low , these colonies actually incorporated a relatively large proportion of the total number of proliferated cells ( Figure 4h ) . The results thus highlight a proliferative and expandable subpopulation of BECs that should make a major contribution to the growth of biliary epithelial tissue . We further sought to reveal the relationship between cell proliferation and tissue structure in detail . In many tissues , proliferating cells , or stem cells , are spatially arranged within a specific area that has a characteristic tissue structure; for example , intestinal stem cells reside in the crypt bottom of the intestinal epithelium ( Barker et al . , 2010 ) . We searched 3D images of the biliary tree for any structural feature corresponding to the location of cycling cells by immunostaining for the cell-proliferation marker Ki67 . Most of the Ki67+ BECs localized to the peripheral area of the biliary tree in the TAA injured liver , whereas only a few Ki67+ BECs were seen in the duct around the PV area ( Figure 5a and Figure 5—figure supplement 1 ) . This distribution pattern fits well with the results of the clonal cell tracing experiments . In the peripheral ductule compartment , the Ki67+ cells were widely scattered . No further sign of a defined stem cell niche , such as a cluster or aligned arrangement of Ki67+ cells , was observed . At 2 weeks of TAA administration , Ki67+ BECs were already enriched in the peripheral ductule region , and scattered therein , rather than in the ducts ( Figure 5b ) . This suggests that the mode of proliferation is maintained over time . We also examined the distribution pattern of proliferating BECs by continuous labeling of cycling cells using 5-bromo-2’-deoxyuridine ( BrdU ) incorporation , and obtained the same result ( Figure 5c ) . 10 . 7554/eLife . 15034 . 015Figure 5 . Proliferating BECs are scattered in the peripheral branching architectures in the biliary tree . ( a ) 3D images of the biliary tree ( CK19 immunostaining; green ) and the cell cycle marker Ki67 ( magenta ) in TAA-injured liver samples . Middle and right panels show magnified views of the region of interest ( ROI ) 1 shown in the left panel , where CK19+ area and Ki67+ BECs therein were extracted using the IMARIS surface protocol . Distribution patterns of the CK19+ area and the Ki67+ CK19+ cells were calculated using the IMARIS vantage protocol after the signals were projected onto the background , and depicted in 2D box-and-whisker plots . ( b ) Liver samples at TAA 2 weeks were analyzed as in ( a ) . Biliary structure is classified into duct compartment ( shown in blue in the center image ) and ductule ( green ) . ( c ) Proliferating cells were labeled by continuous administration of BrdU for 8 days in the course of the TAA injury and were analyzed by anti-BrdU immunostaining ( magenta ) . BrdU incorporation was observed in BECs residing in the peripheral ductule compartment ( white arrows ) , but rarely in those in the duct compartment . All experiments were performed with at least 3 biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01510 . 7554/eLife . 15034 . 016Figure 5—figure supplement 1 . Distribution of Ki67+ BECs in the duct unit . 3D images corresponding to the ROI 2 area in Figure 5a represent the duct compartment of a mouse liver at TAA 6 weeks . Signals of 3D immunofluorescence for CK19 and Ki67 ( magenta ) were converted into 3D graphics using the surface protocol of the IMARIS software . The duct and ductule compartments are colored in blue and green in the left panel , respectively . Only one Ki67+ nuclei was found in the duct compartment , whereas there are many in the ductules . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 016 The quantitative single-cell tracing data reveal heterogeneity among BECs with regard to their proliferative capacity in vivo . To understand the mechanistic basis for such heterogeneity , we sought to construct a simple growth model that could potentially explain and be used to simulate the proliferative behavior of BECs upon injury . For this purpose , it was necessary to collect information about cellular events in a short time scale . We hypothesized that BECs could be classified into two states based on their temporal proliferative capacities , and if this was the case , then the relationship between these states could be examined . We designed an experiment to investigate the temporal proliferative state and cell fates of BECs using two nucleotide analogs , BrdU and EdU , like that employed in a previous study on pancreatic progenitor cells ( Teta et al . , 2007 ) . First , mice were administered with BrdU continuously for 8 days to label in vivo as many proliferating cells as possible ( 1st label ) . After a short interval , mice were then administered the other nucleotide analog , EdU , for pulse labeling ( 2nd label ) ( Figure 6a ) . The rationale for this consecutive and double labeling experiment is as follows . If BECs can be classified into two distinct states: an actively and continuously dividing state and a quiescent state , then those BECs that are subjected to the 2nd labeling ( corresponding to the actively and continuously dividing cells ) should also have undergone the 1st labeling and thus are mostly observed as the 1st and 2nd label double-positive population . Conversely , if BECs are able to convert between the proliferative and quiescent states reversibly , they can divide at arbitrary time points and hence the 1st and 2nd labelings will occur independently . In this case , significant numbers of the 2nd label single-positive cells should be observed ( Figure 6b ) . This experimental strategy can thus allow us to reveal the mode of cell proliferation in terms of the cells' potential for short-term transition between the proliferative and quiescent states . We must note that the definition of proliferative capacity in this particular set of experiments is based on relatively short-term cell behavior , and the proliferative or quiescent state of a current cell does not apply to its progeny . In other words , BECs are classified by their transient cell division manner , but they are not restricted to this state permanently . 10 . 7554/eLife . 15034 . 017Figure 6 . BECs do not proliferate uniformly and can be subdivided into those in the proliferative state and those in the quiescent state . ( a ) Schematic diagram for the experimental design . At week 8 of the TAA injury model , mice were given BrdU via drinking water ( 0 . 8 mg/ml ) for 8 days ( 1st label ) . After an interval of a further 8 days , the mice were intraperitoneally injected with EdU ( 2 mg/20 g body weight ) for pulse labeling ( 2nd label ) . ( b ) Schematic diagram depicting two possible growth modes . In the top model , proliferative and quiescent cell populations can be distinguished by temporal state . This growth mode will give an experimental result in which the 2nd label+ cells are included within the 1st label+ cells . In the second model ( bottom ) , the cells change their growth state in an unregulated manner , resulting in an un-biased labeling pattern . ( c ) Immunofluorescent staining results for BrdU and EdU incorporation together with CK19 immunostaining . Representative regions of interest ( ROI 1–4 ) in the left panel are shown in the middle and right panels . In the magnified images , the boundaries of the CK19+ areas are delineated in gray lines . White arrowheads , EdU+ BrdU- cells; white arrows , non-labeled cells; yellow arrows , BrdU+ EdU+double-positive cells . Scale bar represents 100 μm . ( d ) Quantification of the BrdU+ EdU+ double-labeled cells . Data represents the mean ± SEM for 5 mice . ( e ) Comparison of the incidence of the BrdU+ EdU+ cells between the experimentally obtained data and the result predicted by the assumption that the BrdU and EdU labelings occur independently . p-value was calculated by two-tailed paired Student’s t-test . All experiments were performed with 5 biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 017 We observed many BrdU and EdU double-positive cells in the biliary tissue structure , and these cells constituted the major part of the EdU-labeled BEC population ( Figure 6c , d ) . By contrast , among non-BEC ( CK19- ) populations , the majority of the EdU+ cells were BrdU- , confirming that the presence of the double-positive cells was not attributable to insufficient washout of the 1st label . We also compared the proportion of BECs that were double-positive in our experiment with that theoretically predicted on the assumption that BECs proliferate in an unregulated manner ( Figure 6b , lower panel ) , and found that their difference was statistically significant ( Figure 6e ) . These results suggest that the proliferation pattern of BECs conforms to a model in which BECs can take two different states and change from one to the other irreversibly ( Figure 6b , upper panel ) . By combining our data from quantitative single-cell tracing ( Figure 4 ) with that on short-term proliferation behavior ( Figure 6 ) , we finally sought to establish a mathematical model that can explain the dynamics of biliary epithelial tissue growth in vivo . In the process of fitting a putative tissue growth model , we employed computational simulation by Markov chain Monte Carlo methods using R software to obtain numerous outputs for different sizes of BEC colonies , which eventually converged into a specific distribution pattern . Many rounds of simulation , validation , and modification of different parameters in various combinations were performed until we obtained a simulation result that fitted well with the empirical data from clonal tracing in vivo . ( See 'Materials and methods' for details ) . Considering the results of our nucleotide double-labeling experiments ( Figure 6 ) , we assumed a simple growth model driven by two cell states , continuously proliferating cells and quiescent cells . As stated before , we recognized that the most distinctive feature of the quantitative clonal tracing data was the heterogeneity in BECs with regard to proliferative activity , as manifested by the uni-modality and long-tailed pattern of the colony size distribution ( Figure 4g and i ) . To construct a model that recapitulates this feature , we hypothesized that there must be a factor that accounts for the heterogeneous proliferative capacities of each cell . Thus , we incorporated a stochastic cell behavior into our two-state growth model , in that a cell in the proliferative state can alter its growth state to the quiescent state in a probabilistic manner ( Figure 7a ) . We use the term 'stochastic' here in exactly the same manner as in previous studies ( Doupé et al . , 2010; Driessens et al . , 2012 ) : to describe that the state of each cell is unpredictable rather than being fixed or pre-determined , with the total proportion of cells in each divergent state being balanced . It is important to note that the concept of stochastic growth does not mean that the transition process occurs in a non-regulated manner . There are indeed factors that regulate cellular growth state and behavior , although the switching and maintenance of each cell's state , timing of each cell division , and duration of each cell cycle 'appear' to be stochastic . Such stochastic behavior has already been shown to exist in progenitor cells in the interfolliculer epidermis ( IFE ) ( Doupé et al . , 2010; Driessens et al . , 2012 ) , and we defined parameters on the basis of that IFE model ( Figure 7a ) . As expected , the model with the stochastic feature produced a widespread distribution in colony size , with each single proliferating cell yielding colonies of various sizes along the time course ( Figure 7b ) . 10 . 7554/eLife . 15034 . 018Figure 7 . Mathematical modeling predicts the stochastic nature of the BEC proliferation . ( a ) Schematic diagram of the proposed growth model . We set five parameters . The parameter p represents the initial ratio of the cells in the proliferative state compared to total cells; m represents the probability that a cell will enter the cell cycle within one day . The frequency of each cell fate is defined by two parameters , r and s: r affects stability of colony size , whereas s represents imbalance of cell fate selection between proliferative and quiescent states . l represents the probability that a cell will be lost due to cell death ( expressed as lost cells per total number of cells per day ) . ( b–e ) Monte Carlo simulation was performed using R software . ( b ) Consecutive changes in colony size over time are shown in the graph on the left . Each line represents a colony derived from a single proliferative cell . The scatter plot on the right corresponds to the colony-size distribution after 50 days of growth . ( c ) Quantitative data for Ki67+ BECs upon TAA injury revealed by immunostaining . Day 0 samples correspond to livers under normal conditions . A significant increase in Ki67+ cells among the BEC population was observed between 4 and 6 days after the start of injury ( p<0 . 005 , two-tailed paired Student’s t-test ) . Data represent mean ± SEM for 4 mice . ( d and e ) A simulated pattern for distribution of colony size and changes in this distribution for colonies derived from a single proliferative cell . Results are shown for a simulation when the parameters were set as follows: m = 0 . 175 , r = 0 . 15 , s = 0 . 06 , and l = 0 . 001 ( simulated cell number = 2000 ) . ( f ) The quantified in vivo data alongside simulation results . The data simulating changes in colony size were constructed from the data for a single proliferative cell ( d and e ) compensated for the presence of quiescent cells ( the initial ratio p=0 . 465 ) . The experimental data were derived from the data sets used in Figure 4i . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01810 . 7554/eLife . 15034 . 019Figure 7—figure supplement 1 . Discordance observed during the model-fitting process before taking account of the timing of the start of BEC proliferation upon injury . Simulation data obtained from a model that is not adjusted for the proliferation start timing ( gray bars ) did not fit the experimental data for BEC colony formation in vivo ( black bars ) . Early in the model-fitting process , the optimum model with the parameters set to best fit the data for colony-size distributions at 4 , 6 and 8 weeks of injury showed a considerable discrepancy in fitting the data for 2 weeks of injury ( black arrowheads and arrows ) . In particular , relatively large colonies were found only in the simulation data ( black arrows ) . simulated data for 8 days of injury , albeit using the same model and parameters ( white bars ) , showed better fit with the experimental data for 2 weeks of injury than did simulated data for 2 weeks of injury . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 01910 . 7554/eLife . 15034 . 020Figure 7—figure supplement 2 . TAA causes death of hepatocytes around CV , but not death of BECs . Cell death of BECs was analyzed using a cell-death detection method by in vivo administration of EthD-3 . After staining with EthD-3 in vivo , liver tissue sections were prepared and immunostaining was performed with anti-CK19 antibody and anti-Hnf4a antibody . White arrows indicate dead cells . All signals of EthD-3 were merged with Hnf4a . Scale bar represents 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 020 During the model-fitting process , the overall simulation result almost adequately fitted the experimental in vivo clonal tracing data , but there was an inevitable discordance , specifically at the time point of 2-week injury ( Figure 7—figure supplement 1 ) . A simulation result for 8-day injury period fitted significantly better with the experimental data for 2-week tracing than did simulated data for 2 weeks of injury , implying a gap of several days between the timing of colony formation in vivo and in silico . The initial simulation model was based on the assumption that the BECs began to enter the proliferative mode on the day when the injury was induced . We reasoned that the gap might be due to an incorrect estimation of the timing of the commencement of BEC proliferation upon injury , as there must be a lag between the timing of the injury application and that of the concomitant induction of tissue growth . To test this hypothesis , we performed additional immunostaining experiments using Ki67 and found that BECs first entered the cell cycle in vivo 6 days after the TAA injury was started ( Figure 7c ) . This result was consistent with our prediction , and we modified the parameter for the starting point in our revised simulation model accordingly . This modification allowed the obtained patterns of simulation to fit the experimentally obtained in vivo data completely over the time course examined ( Figure 7d–f ) . We further evaluated our model in an additional experiment . Our model predicted that the rate of 'cell loss' should be minimal in the BEC population ( model parameter l = 0 . 001 , which means that only 0 . 1% of BECs would be lost in a day ) . We evaluated the occurrence of cell death in TAA injury by an in vivo cell-death detection method ( Edwards et al . , 2007 ) . As expected , almost no cell death was detected in the CK19+ area ( Figure 7—figure supplement 2 ) , thus supporting our model prediction . Taken together , our results strongly suggest that the proliferation dynamics of BECs upon TAA injury conform to the following stochastic model: ( i ) upon liver injury , some BECs are activated to change their growth state to the proliferative state , while the others remain quiescent; ( ii ) those BECs that are in the proliferative state can stochastically and irreversibly convert back to the quiescent state during the course of injury; ( iii ) the highly proliferative BEC subpopulation is stochastically maintained and produces a large number of progenies , thereby making a major contribution to the expansion and remodeling of biliary epithelial tissue . Understanding the relationship between the mode of cell proliferation and the resultant structural organization within a tissue is one of the fundamental issues in stem cell biology . Here , we sought to reveal this relationship by focusing on the ductular reaction , a unique and dynamic remodeling of the intrahepatic biliary tree that occurs upon chronic liver injury . Our results have delineated a progressing morphological transformation of the peripheral ductular structure in the TAA model , which is primarily achieved by a tissue-intrinsic cell expansion within the pre-existing biliary epithelium . Intriguingly , the expansion of biliary epithelial tissue is dictated by heterogeneous and stochastic behavior of BECs . Thus , a subset of proliferating BECs that is stochastically maintained , rather than a pre-determined and hard-wired progenitor population , exhibits clonogenic growth potential in vivo and plays a central role in the remodeling of biliary epithelial tissue . In the present and recent studies , we have shown that the biliary epithelial tissue comprises of clearly distinct structural units ( Figure 2b , c ) ( Kaneko et al . , 2015 ) . In many instances , the architecture of the biliary tree is illustrated as a simple monolayer tube with a luminal structure that corresponds to the ducts , with little or no attention being paid to the presence of peripheral ductules . However , it is the ductule compartment , rather than the ducts , that makes a key contribution to tissue remodeling , as exemplified by the fact that the majority of the proliferating BECs reside in the peripheral ductule region ( Figures 4g , 5 and Figure 5—figure supplement 1 ) . Structural and functional units found in many different organs , such as the intestinal crypt and the hair follicle , exhibit fairly ordered structure in that they are reiteratively aligned in quasi-two-dimensional sheet-like tissue structures . By contrast , the biliary tree , especially the ductule compartment , spreads omni-directionally and it is quite difficult to capture its structure comprehensively and reproducibly . Therefore , it is necessary to use 3D-imaging methods to capture the tissue structures and cellular growth dynamics therein accurately and precisely . In order to observe detailed features and connectivity within the biliary tree structure in the liver , we have established a novel 3D staining and imaging method . There are substantial gains to be made by adapting the method so that it can be performed easily and effectively with minimum time and cost , especially for experiments such as quantitative single cell tracing analysis in which large numbers of samples must be handled . We have thoroughly optimized our 3D imaging protocol at various steps to render it easy and cost-effective; our protocol does not require expensive reagents or specialized equipment , and thus provides a versatile method for studying various tissue structures in the liver , as well as in other organs . Using quantitative single-cell tracing , we revealed cellular heterogeneity in the biliary epithelial tissue in vivo . More specifically , we found two levels of heterogeneity among the BEC population . First , the biliary epithelial tissue can be subdivided into two classes , namely the duct and peripheral ductule compartments , on the basis of morphological and geographical characteristics . At the population level , the BECs in the peripheral ductule compartment have higher proliferative tendency than those in the ducts ( Figures 4g , 5 and Figure 5—figure supplement 1 ) . Second , among the BECs in the peripheral ductule compartment , there exists further heterogeneity in terms of the proliferative behavior . Not all of the ductule cells exhibit a proliferative response upon injury , and the size of the colonies varies even among the proliferative population . Both of these levels of heterogeneity can be explained by our mathematical model . The model predicts the initial ratio of the proliferating BEC population to be p=0 . 465 , meaning that the majority of BECs do not divide in response to the injury stimulus . This is consistent with the notion that the duct compartment cells , which constitute the major part of the BEC population under the normal condition , do not or rarely divide upon injury ( Figure 5 ) . The proliferative heterogeneity in the peripheral ductule compartment is well represented in the model as change in stochastic growth state ( i . e . , proliferating vs . non-proliferating states ) of BECs . The dual nucleotide labeling experiment showed that the proliferative state of BECs switched irreversibly to a quiescent state ( Figure 6 ) . This feature has also been reported in a previous study of epithelial tissue in vivo ( Doupé et al . , 2010; Driessens et al . , 2012 ) , but is not consistent with a report that described the manner of cell division and cell fate decision in an in vitro model ( Spencer et al . , 2013 ) . We think that this discrepancy may be due to the role of BECs in vivo . During the remodeling and growth process in biliary epithelial tissue , BECs must form a functional tubular structure . We assume that BECs may gain a type of steady-state phenotype ( or , become more matured and differentiated cells ) in order to generate and maintain a functional and robust epithelial tubular structure . From this point of view , it is reasonable to assume that their proliferation state changes irreversibly in vivo . Unfortunately , we do not yet have an established marker that can clearly distinguish BECs in different stages in the course of their functional differentiation and maturation , which prevented us from directly evaluating this possibility . Nevertheless , it could be supported in part by our findings that revealed the relationship between the proliferative capacity of BECs and structural features of the biliary tree ( Figure 4g ) , as BECs constituting different tissue structures may be in distinct differentiation stages . The opposing concepts of 'stochastic' versus 'deterministic' cell-fate decisions have been established as an important paradigm in stem-cell biology ( O'Neill and Schaffer , 2004; Enver et al . , 2009 ) . It should be noted that the concept of stochastic versus deterministic regulation , which relates to Figures 4 and 7 , does not refute the existence of temporal proliferative sub-states , as shown in the dual nucleotide labeling experiment ( Figure 6 ) . Here , the deterministic model refers to restriction of a cell to a state that is permanently fixed . In both models , cells are divided into sub-populations according to their transient proliferative state , rather than according to their permanent state . Thus , the existence of a proliferative state does not refute the stochastic model , and the existence of a temporal proliferative state does not simply mean that cell proliferation follows the deterministic model . Nevertheless , the mechanisms that govern the maintenance and regulation of cells are defined differently in the stochastic and deterministic models . In the stochastic model , tissue stem cells , for example , have non-uniform cell fate and the consequences of cell fate selection are not predictable , whereas in the deterministic model , the stem cells give rise to their progeny uniformly and stably . The stochastic growth model first emerged in the 1960s , in order to explain the heterogeneous behavior of transplanted hematopoietic stem cells ( Siminovitch et al . , 1963; Till et al . , 1964 ) . Since then , many studies have revealed the existence and role of stochastic behaviour in various phenomena in living things ( Samoilov et al . , 2006 ) . In our simulation , we used a stochastic feature to model not only the cell-fate decision process but also the regulation of the cell cycle length ( Figure 7 ) . In our mathematical simulation , we defined quiescent cells as those that do not divide at all , thus we do not need to consider their cell cycle length . We did , however , have to consider the cell cycle length of proliferative cells . We first ran many simulations with a uniform cell cycle length for all cells . We found , however , that the best fit was achieved when we allowed variation in cell-cycle length among the cells . This variation was defined as the parameter m ( see 'Materials and methods' for details ) . Interestingly , this result agrees with the notion proposed by Spencer et al . ( Spencer et al . , 2013 ) , confirmed by their single-cell experimental system , that there is substantial cell-to-cell variation in cell cycle lengths and delay between mitogen re-stimulation and CDK activation . They also reported that cells had heterogeneous inter-mitotic times , ranging from 20 hr to >50 hr . This fits well with our stochastic proliferation model whose definitions include varying inter-mitotic times . In the stochastic model , the number and state of stem cells are maintained unstably . This fits the structural features of the biliary tree well in many regards . First , as the injury progressed , the biliary tree formed numerous arborization and branch ends around the injured area ( Figure 2e–h and Figure 2—figure supplement 2 ) . In addition , the location of cycling cells appeared to be scattered at the 3D level ( Figure 5a ) . Such frequent remodeling and sporadic proliferation are well suited to the stochastic growth model , which can maintain a spatiotemporally flexible population of proliferating cells . Second , the stochastic model is consistent with diversity in tissue structure at the micro scale . In microscopic view , the stochastic system shows a degree of instability because the cell state may fluctuate and the output may be unpredictable . This feature nicely explains the dynamics of biliary remodeling , in that the detailed shape and degree of expansion of the extended biliary tree varied in each small area of interest ( Figure 2h ) . This is why the architecture of the biliary tree looks complicated when observed by 2D liver section . Thus , the instability shown in the stochastic model matches the microscopic structural diversity of the biliary tree . Third , we focus on a macroscopic feature . In our stochastic model , each cell behaves in a flexible manner , but the total percentages of cells that take each cell-fate decision is strictly defined and follows a particular value at the population level ( Figure 7a ) . This feature of the stochastic model makes the simulation outcome converge into an ordered pattern ( Figure 7c ) . In other words , the stochastic growth mode provides a stable , robust and consistent outcome in a macroscopic view . This macroscopic feature corresponds to the biliary tree architecture in the macroscopic view , as the biliary tree appears to follow a uniform pattern in both the spatial and temporal terms when it is observed at a large scale ( Figure 2—figure supplement 2 ) . In conclusion , the stochastic behavior of BECs plays a fundamental role in establishing a flexible and adaptive tissue remodeling system in the biliary epithelium , which underlies the robust regenerative capacity of the liver upon injury . Studies in other systems , including hematopoietic stem cells and epidermal homeostasis , have revealed similar models and concepts ( Siminovitch et al . , 1963; Doupé et al . , 2010 ) , implying that a common and fundamental mechanism governs stochastic cell behavior . Knowledge on the mode of tissue growth provided by this study applies to systems beyond the hepatobiliary system , and should provide significant insights into tissue dynamics and homeostasis in many other organs . All animal experiments were conducted in accordance with the Guideline for the Care and Use of Laboratory Animals of The University of Tokyo , under the approval of the Institutional Animal Care and Use Committee of the Institute of Molecular and Cellular Biosciences , The University of Tokyo ( approval numbers 2501 , 2501–1 , 2609 and 2706 ) . Prom1-CreERT2 knock-in mice ( Zhu et al . , 2009 ) and R26R-tdTomato reporter mice ( Madisen et al . , 2010 ) were purchased from Jackson Laboratory and maintained on a C57BL/6J background . Wild-type C57BL/6J mice were purchased from CLEA Japan ( Japan ) . 6- to 10-week-old mice , including both males and females , were used unless otherwise specified . For liver injury models , mice were administered with 0 . 3% ( wt/vol ) TAA in drinking water or fed a diet containing 0 . 1% 3 , 5-diethoxycarbonyl-1 , 4-dihydrocollidine ( DDC ) . Tamoxifen was dissolved in corn oil and administered via oral gavage . In labeling experiments with nucleotide analogs , BrdU was administered via drinking water ( 0 . 8 mg/ml ) and EdU for pulse labeling via intra-peritoneal injection . For the hepatocyte labeling experiment , rAAV2/8-iCre was packaged in HEK293 cells according to the protocol described previously ( Kok et al . , 2013 ) . The iCe-expressing plasmid vector was constructed by replacing the GFP insert in the pAM-LSP1-eGFP vector with the iCre gene . iCre was taken from pDIRE , which was a gift from Rolf Zeller ( Addgene plasmid # 26745 , [Osterwalder et al . , 2010] ) . The titered rAAV2/8-iCre was injected by intraperitoneal injection ( 1x1011 vector genome / mice ) . Liver cell preparation and FACS analyses were done as previously described ( Okabe et al . , 2009 ) . Non-parenchymal cells were obtained from the mouse liver by the two-step collagenase liver digestion method and used for FACS analysis . Mice were laparotomized under isoflurane inhalation anesthesia , and 15 ml of pre-warmed ( at 37°C ) Liver Perfusion Medium ( 17701–0381–038 , Life technologies , Waltham , MA ) was perfused through the liver from the portal vein . The liver was then perfused with 20 ml of collagenase solution containing collagenase type IV ( C5138-5G , Sigma Aldrich , St . Louis , MO ) and fetal bovine serum ( FBS , S1820-500 , Biowest , France ) . The liver was harvested and placed into Dulbecco's Modified Eagle Medium ( D5796 , Life technologies ) after removing the gallbladder and extra-hepatic bile duct , and minced gently with surgical knives . Roughly digested liver samples were filtered through a cell strainer ( 70 μm ) . The remaining undigested clots were further treated with digestion solution containing collagenase type IV , DNase I ( DN25-5G , Sigma Aldrich ) , and pronase ( 10165921001 , Roche , Switzerland ) . After removing the parenchymal fraction ( hepatocytes ) by centrifugation , the non-parenchymal fraction in the supernatant was treated with hemolysis buffer to remove red blood cells . The remaining cells were incubated with anti-FcR antibody , followed by staining with the antibodies listed in Table 1 . Stained cells were suspended in phosphate-buffered saline ( PBS ) containing 3% FBS and 4' , 6-diamidino-2-phenylindole ( DAPI , D3571 , Life technologies ) for dead cell staining . 10 . 7554/eLife . 15034 . 021Table 1 . List of antibodies used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 15034 . 021AntibodyCompany/SourceHost animalMethodDilutionProminin1/CD133 ( APC-conjugated ) Biolegend ( San Diego , CA ) RatFACS1:100Prominin1/CD133 ( purified ) eBioscience ( Santa Clara , CA ) RatIF1:100EpCAM ( FITC-conjugated ) ( Okabe et al . , 2009 ) RatFACS1:200EpCAM ( purified ) BD Pharmingen ( Franklin Lakes , NJ ) RatIF1:200CD45 ( APC-conjugated or APC-Cy7-conjugated ) BiolegendRatFACS1:200MIC1-1C3STEMGENT ( Lexington , MA ) RatFACS1:200Spp1R&D systems ( Minneapolis , MN ) GoatIF1:200Cytokeratin 19 ( Tanimizu et al . , 2003 ) RabbitIF1:200Ki67eBioscienceRatIF1:200LacZAbcam ( United Kingdom ) ChickenIF1:200BrdUAbcamRatIF1:200Hnf4aSanta Cruz ( Dallas , TX ) GoatIF1:200 In the EdU uptake experiments , we used the Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry Assay Kit ( C10634 , Life technologies ) and Fixable Viability Stain 450 ( 562247 , BD Biosciences , Franklin Lakes , NJ ) according to the manufacturers’ instructions . Data were acquired using a FACSCanto II cell analyzer ( BD Biosciences ) or the MoFlo XDP ( Beckman Coulter , Brea , CA ) . Final data analyses were performed with FlowJo software . Whole-liver samples from adult mice were fixed with paraformaldehyde ( PFA ) as follows . The liver was perfused via the portal vein sequentially with 10 ml of PBS , 10 ml of 2% PFA , and 10 ml of 4% PFA . Upon harvest , the liver was cut into several blocks , placed into a 15 ml tube containing 4% PFA , and incubated for 12 hr . After fixation , the liquid was changed to 20% sucrose in PBS . The liver blocks were embedded in Tissue-Tek O . C . T . compound ( Sakura Finetek , Japan ) and snap frozen . Frozen samples were cut into 10-μm thickness using a cryostat-microtome ( HM525 , Microm International , Germany ) and stained with the antibodies described in Table 1 . For BrdU immunodetection , sections were heat-treated in TE buffer in an autoclave ( 120°C , 5 min ) for antigen retrieval . For EdU detection , sections were treated with Click iT Plus EdU Alexa Fluor 488 Imaging Kit ( C10637 , Life technologies ) . Nuclei were counterstained with Hoechst33342 ( H1399 , Life technologies ) . The mouse liver samples were pre-fixed using the protocol described for the conventional 2D staining . Note that the fixation by perfusion steps were essential to keep the tissue structure intact and to prevent soluble proteins from leaking out . The frozen liver samples were cut into thick sections ( 200–500 μm ) using the cryostat-microtome . At this point , the samples set on the stage of the microtome were briefly warmed ( by touching with gloved fingers ) immediately before sectioning to prevent the samples from cracking . Samples were placed in PBS in disposable tubes , washed with PBS twice more , and then incubated with blocking/permeabilization reagent ( 3% FBS , 0 . 02% sodium azide , and 0 . 2% Triton X-100 in PBS ) for 30 min at room temperature ( RT ) . The same blocking/permeabilization reagent was also used in the following staining process for antibody dilution . We typically stained 10 thick sections in a 2 ml tube with 500 µl of diluted antibody solution . The samples were incubated in the antibody solution on a rocking or shaking device at 4°C for 2 days . The tubes were inverted by hand once a day to facilitate thorough mixing and uniform staining . The samples were washed with PBS twice and then transferred into a new 50 ml tube with 40 ml of PBS to wash out excess primary antibodies by incubating on a rocking device at 4°C for 2 days . After thorough washing , samples were treated with fluorescence-conjugated secondary antibodies using the same process as that used for primary antibody staining . Nucleotide staining dye , such as Hoechst33342 , was mixed into the secondary antibody solution if necessary . Samples were then washed thoroughly with PBS again . After staining and washing , samples were treated with the tissue-clearing reagent SeeDB ( Ke et al . , 2013 ) . The SeeDB reagent contains fructose as the main component and is safe , inexpensive , and easy to handle . We compared various different types of clearing reagents and obtained the best results in liver tissue imaging with SeeDB . To perform quantitative single-cell tracing analyses based on tdTomato fluorescence , we stained liver cell nuclei by treating sectioned samples with SYTOX Green ( in PBS containing 0 . 1% Triton X-100 ) on a rocking device overnight at RT . The samples were washed once with PBS and then directly placed into SeeDB . For data acquisition , we used confocal microscopes ( FV1000 or FV1200 , Olympus , Japan ) with a 30x silicone immersion lens ( UPLSAPO30XS , Olympus ) . We classified BECs into two compartments , duct and ductule , based on two criteria . First , location within the biliary tree structure was considered . In injured liver , the duct still runs alongside the PV , while the ductule is expanded to the outer parenchymal area . Second , we measured the diameter of the internal luminal space of the biliary structure . Biliary epithelial tubular structures with a luminal diameter of more than 8 μm were defined as the duct compartment , while the other tubular structures with smaller diameter were defined as the ductule . For organ-wide 3D visualization of the biliary tree , we performed X-gal staining of whole-liver samples derived from the Prom1-CreERT2 mice in which nLacZ is also knocked-in to the Prom1 locus . The liver was perfused with 10 ml of ice-cold PBS containing 2 mM MgCl2 , then with 10 ml of fixative solution ( 0 . 2% PFA , 0 . 1 M HEPES , 2 mM MgCl2 , 5 mM EGTA , pH 7 . 3 ) . The liver was incubated with fixative solution for 48 hr at 4°C with a daily change of the solution . The fixed liver was then treated with detergent buffer ( 0 . 1 M phosphate buffer , pH 7 . 3 , 2 mM MgCl2 , 0 . 01% sodium deoxycholate , 0 . 02% Nonidet p-40 ) for 24 hr at 4°C . Next , the liver was treated with staining buffer ( 1 mg/ml X-gal in detergent buffer ) for 48 hr at 4°C ( from this step on , sample tubes were wrapped with foil for shading ) and further for 12 hr at 37˚°C . At all incubation steps , samples were put on a rocking device . After washing out staining buffer with PBS , the liver was dehydrated with ethanol and then cleared with a 2:1 benzyl benzoate:benzyl alcohol ( BABB ) solution . For evaluation of cell death in injured liver , we performed a cell death detection assay ( Edwards et al . , 2007 ) with some modification . 200 µl of EthD-3 ( 0 . 2 mg/ml in PBS , PK-CA707-40050 , Takara , Japan ) was injected intravenously to stain the nuclei of dead cells in living mice . After 15 min , mice were sacrificed and PBS was perfused via the portal vein to drain the blood that contained excess EthD-3 . Then the liver was processed using the 2D staining protocol described above . In all animal experiments , the samples represent biological replicates derived from different mouse individuals . Representative data were supported by at least three biological replicates . Detailed sample size was estimated by considering the means and variation data from preliminary experiments . No randomization or blinding process was performed . The F-test was used to check the homoscedasticity of the data , and the Kolmogorov-Smirnov test to check whether the data follow a Gaussian distribution . Significance tests were performed as described in the legends to each figure using Prism software ( Graph pad , San Diego , CA ) . In order to reveal the cellular behavior that underlies biliary tissue growth and remodeling , we traced the fate ( i . e . , the clone size of the progeny ) of each single cell in vivo , made a simple growth model , and simulated it by computational methods . The program code is available at the following address . http://www . iam . u-tokyo . ac . jp/cytokine/research/simulation_code_ver_201 . txt	Cell proliferation – the process by which cells multiply – plays an important role in many biological processes , including tissue growth , maintenance and remodeling . In these processes , the way cells proliferate is reportedly related to their roles in the tissue and the structures that they form . The biliary tree , a piping system that exists to drain the bile produced in the liver , forms a complex , tree-like , tubular structure . The biliary tree is essential for healthy livers to work well , and has been known to grow and change its structure quite dynamically during an injury or while the liver regenerates . However , it was not clear how biliary tree cells behave as the biliary tree grows and remodels itself . Does each cell behave in the same way ? And how does cell growth relate to changes in the structure of the biliary tree ? Kamimoto et al . have now developed new methods to observe detailed three-dimensional tissue structures and to trace the behavior of single cells . Using these techniques to study a mouse model whose liver was injured by toxic chemicals revealed the behavior of biliary cells as they responded to the injury . None of the biliary cells proliferated uniformly , and there were some peculiar cells that proliferated quite vigorously compared to the others . Kamimoto et al . then made a mathematical model that could explain cell behavior and tissue remodeling at different scales . This showed that the activity of those peculiar , rapidly proliferating cells was maintained by chance as the biliary tree expanded . These findings help us understand how the biliary tissue grows and the liver regenerates . They may also provide us with a clue to understanding the nature of the behavior of living things , which is sometimes seemingly ordered and robust , and sometimes unpredictable and mysterious . It remains to be seen whether the new model can be applied to other types of tissues or in other species . Further work is also needed to investigate which genes and proteins are involved in controlling the behavior of cells in the growing biliary tissue .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2016	Heterogeneity and stochastic growth regulation of biliary epithelial cells dictate dynamic epithelial tissue remodeling
Loss of ESCRT function in Drosophila imaginal discs is known to cause neoplastic overgrowth fueled by mis-regulation of signaling pathways . Its impact on junctional integrity , however , remains obscure . To dissect the events leading to neoplasia , we used transmission electron microscopy ( TEM ) on wing imaginal discs temporally depleted of the ESCRT-III core component Shrub . We find a specific requirement for Shrub in maintaining septate junction ( SJ ) integrity by transporting the claudin Megatrachea ( Mega ) to the SJ . In absence of Shrub function , Mega is lost from the SJ and becomes trapped on endosomes coated with the endosomal retrieval machinery retromer . We show that ESCRT function is required for apical localization and mobility of retromer positive carrier vesicles , which mediate the biosynthetic delivery of Mega to the SJ . Accordingly , loss of retromer function impairs the anterograde transport of several SJ core components , revealing a novel physiological role for this ancient endosomal agent . Developmental and physiological functions of epithelia rely on a set of cellular junctions , linking cells within the tissue to a functional unit . While E-cadherin-based adherens junctions ( AJs ) provide adhesion and mechanical properties , formation of the paracellular diffusion barrier depends on tight junctions ( TJs ) . Proteins of the conserved claudin family play a key role in establishing and regulating TJ permeability in the intercellular space by homo- and heterophilic interactions with Claudins of neighboring cells ( Günzel and Yu , 2013 ) . Arthropods , such as Drosophila , do not possess TJs but a functionally similar structure in ectoderm-derived epithelia termed pleated septate junction ( pSJ , SJ hereafter ) , characterized by protein dense septa lining the intercellular space in electron micrographs ( Gilula et al . , 1970 ) . Structure and function of Drosophila SJs depend on a convoluted multiprotein complex containing at least a dozen components . Three claudins , among them Megatrachea ( Mega ) , have been shown to be required for SJ formation and barrier function in flies ( Behr et al . , 2003; Nelson et al . , 2010; Wu et al . , 2004 ) . Besides claudins , several transmembrane proteins ( TMPs ) such as Neurexin-IV ( NrxIV ) , Neuroglian ( Nrg ) or ATPα contribute to the formation of the stable SJ core complex , which is characterized by low mobility within the membrane ( Baumgartner et al . , 1996; Genova and Fehon , 2003; Oshima and Fehon , 2011 ) . At the intracellular side of the junction , cytoplasmic proteins such as Coracle ( Cora ) , Varicose ( Vari ) , and Discs large ( Dlg ) associate with the transmembrane components , contributing to the formation of a stable fence-like scaffold ( Bachmann et al . , 2008; Lamb et al . , 1998; Laval et al . , 2008; Woods and Bryant , 1991; Wu et al . , 2007 ) . While junction formation during embryogenesis requires the SJ localized cytoplasmic protein Dlg , this basolateral cell polarity factor is not a structural part of the immobile junction core complex ( Oshima and Fehon , 2011; Woods et al . , 1996 ) . This explains the functional separation of barrier formation and apicobasal polarity despite the close association of Dlg-complex components with the SJ . Albeit growing knowledge about the structural composition of SJs , the intracellular events required for assembly and maintenance of SJ complexes remain largely unknown . Specifically , how proliferative tissues , such as the imaginal disc epithelium , maintain SJ integrity is not well established . It was recently shown that newly synthesized SJ components integrate into the junction from the apical side ( in between AJ and SJ ) in a ‘conveyor belt-like’ fashion ( Babatz et al . , 2018; Daniel et al . , 2018 ) . In addition , SJ components are frequently associated with endosomal compartments , suggesting a role for the endosomal system in coordinating transport and turnover of SJ complexes ( Nilton et al . , 2010; Tempesta et al . , 2017; Tiklová et al . , 2010 ) . Consistently , endocytosis is required to concentrate SJ components at the junctional region during embryogenesis ( Tiklová et al . , 2010 ) . This suggests that passage of SJ TMP components ( or the whole SJ protein complex ) through the endosomal system may be a requirement for SJ formation , with the underlying mechanisms remaining poorly characterized . The endosomal system fulfils a plethora of physiological functions by tightly regulating the intracellular transport of TMPs and membranes within the cell . Following endocytosis from the plasma membrane , TMPs enter the endosomal system where they undergo cargo specific sorting . This process provides separation of proteins destined for degradation from those that exit the endosomal system to be recycled . Two evolutionary conserved endosomal sorting machineries , the endosomal sorting complex required for transport ( ESCRT ) and the retromer complex , mediate cargo sorting into the degradative and recycling pathway , respectively ( Cullen and Steinberg , 2018 ) . To coordinate these opposing transport activities , the endosomal system comprises a highly dynamic membrane network governing retromer-dependent tubulation for recycling and ESCRT-mediated generation of intraluminal vesicles ( ILV ) for degradation . Endocytosed proteins can evade ESCRT-dependent packaging into ILVs by exiting the maturing endosome ( ME ) through tubular retrieval domains induced by specialized recycling machineries such as retromer . Initially characterized as a regulator of endosome-to-Golgi cargo retrieval in yeast , this endosomal agent comprises two subcomplexes that cooperatively drive cargo sorting into tubular recycling carriers ( Carlton et al . , 2004; Horazdovsky et al . , 1997; Seaman et al . , 1997 ) . Similar to the ESCRT machinery , cargo clustering and membrane deformation is performed by distinct functional units within the retromer pathway . Motif-based cargo recognition and aggregation is mediated by the endosomally localized Vps26:Vps29:Vps35 complex , which has been termed cargo-selective complex ( CSC ) ( Lucas et al . , 2016; Nothwehr et al . , 2000; Seaman et al . , 1997 ) . Since the ancient CSC does not possess membrane bending activity , cooperation with tubulating factors such as proteins of the SNX-BAR ( Sorting Nexin-Bin/Amphiphysin/Rvs ) family is required for recycling carrier generation ( Cullen , 2008 ) . Proteins containing the curved BAR-domain can assemble into regular helical coats on endosomes , thereby inducing cytoplasm faced tubulation ( Frost et al . , 2009 ) . Concerted action of CSC stably complexed with SNX-BAR proteins to retrieve endosomal cargo was initially characterized as the classical retromer pathway in yeast ( Horazdovsky et al . , 1997; Seaman et al . , 1998 ) . In metazoans however , retromer function is not restricted to SNX-BAR-dependent pathways . Specifically , cooperations of CSC with SNX3 or SNX27 ( both lacking BAR-domains ) emerged as alternative routes for endosomal retrieval ( Harterink et al . , 2011; Lauffer et al . , 2010; Steinberg et al . , 2013 ) . Proteomic data from mammalian cells suggest that surface levels of well over 100 TMPs depend on retromer and many of these proteins seemingly interact with CSC or SNX27 ( Steinberg et al . , 2013 ) . Recently , Drosophila has proven invaluable for assessing and confirming the physiological relevance of some of these putative retromer cargos in vivo ( Strutt et al . , 2019 ) . Cargo proteins within the endosomal system that do not undergo recycling can enter the degradative trafficking route starting with their sorting into ILVs . Generation of ILVs at the limiting membrane of MEs requires the canonical ESCRT function , which is performed by four in sequence acting complexes ( ESCRT-0 , -I , -II , III ) and the ATPase Vps4 ( Babst et al . , 2002b; Babst et al . , 2002a; Babst et al . , 1998; Bilodeau et al . , 2002; Katzmann et al . , 2001 ) . Ubiquitination of TMPs serves as the primary degradative sorting signal and sequestration of TMPs into ILVs is an essential prerequisite to complete lysosomal degradation . Several ESCRT components such as Vps27/Hrs ( ESCRT-0 ) and Vps23/TSG101 ( ESCRT-I ) possess ubiquitin interacting motifs , which allow them to bind and cluster ubiquitinated TMPs ( Bilodeau et al . , 2002; Katzmann et al . , 2001; Shih et al . , 2002 ) . Consequently , local concentration of ubiquitinated cargo by ESCRT complexes establishes a degradative subdomain at the endosomal membrane that is spatially separated from the retrieval subdomain ( Norris et al . , 2017; Raiborg et al . , 2002; Raiborg et al . , 2001 ) . While ESCRT-0-II complexes provide cargo recognition and clustering , the membrane-deforming activity required to bud and abscise ILVs into the endosomal lumen depends on ESCRT-III components , which polymerize into helical arrays at the endosomal membrane ( Hanson et al . , 2008; Saksena et al . , 2009 ) . The most abundant ESCRT-III component is the highly conserved yeast Snf7/Vps32 , encoded by the gene shrub ( shrb ) in Drosophila ( Sweeney et al . , 2006; Teis et al . , 2008 ) . Unlike upstream ESCRT components , ESCRT-III proteins only transiently assemble into a heterooligomeric complex at the endosomal membrane ( Teis et al . , 2008; Wollert et al . , 2009 ) . In consequence of ESCRT activity , the maturing endosome accumulates cargo-containing ILVs and is recognized in electron micrographs as a multi-vesicular body ( MVB ) . The ESCRT/MVB pathway ends with Vps4-dependent dissociation of ESCRT-III components from the endosomal membrane . This step is required for the release of the nascent ILV and subsequent rounds of ILV formation ( Babst et al . , 1998; Wollert et al . , 2009 ) . Loss of ESCRT function was initially studied in yeast cells in which it led to the emergence of an aberrant pre-vacuolar endosomal organelle , termed class E compartment ( Raymond et al . , 1992 ) . This defective endosomal structure is characterized by accumulation of degradative cargo and a failure to fuse with the vacuole/lysosome ( Doyotte et al . , 2005; Raymond et al . , 1992 ) . The physiological relevance of ESCRT-mediated degradative TMP trafficking is particularly evident in Drosophila imaginal disc tissue . Here , loss of ESCRT function induces severe overgrowth , multilayering , apoptosis , and invasive behavior of the tissue; a phenotype attributed to mis-regulation of cellular signaling pathways , such as the Jak/Stat- , Jun-Kinase- , and Notch pathways ( Herz et al . , 2006; Moberg et al . , 2005; Thompson et al . , 2005; Vaccari and Bilder , 2005 ) . Consequently , ESCRT components were classified as endocytic neoplastic tumor suppressor genes ( nTSG ) in Drosophila ( Hariharan and Bilder , 2006 ) . While induction of over-proliferation and apoptosis in nTSG mutants have been extensively characterized , the events leading to loss of cell polarity and ultimately neoplastic transformation of the tissue remain poorly understood . Here , we have analyzed the integrity of cellular junctions in an ESCRT-depleted wing imaginal disc epithelium to gain insight into the initial events leading to neoplastic transformation . To our surprise , preceding neoplastic overgrowth , we found a strong and specific reduction in the density of SJ . We show that ESCRT and retromer functions are required for anterograde transport of SJ components . By dissecting the intracellular trafficking itinerary of the claudin Megatrachea , we reveal that biosynthetic delivery of this core SJ component depends on a complex basal to apical transcytosis route relying on ESCRT and retromer functions . To analyze the impact of ESCRT loss of function on junctional integrity , transmission electron microscopy ( TEM ) was used on wing imaginal discs that have been depleted of ESCRT function . We devised an RNAi-based approach allowing spatiotemporal knockdown of the ESCRT-III component Shrub . By using hhGal4 and the temperature sensitive Gal4 Repressor ( tubGal80ts ) , we specifically inhibited Shrub protein expression in the posterior compartment by expressing a UAS-shrub-RNAi construct ( Sweeney et al . , 2006 ) for specified durations . After 32 hr of RNAi expression , Shrub protein was effectively reduced in the posterior compartment as visualized by antibody staining ( Figure 1A ) . ESCRT loss of function is known to induce high levels of apoptosis in imaginal disc tissue ( Herz et al . , 2006; Thompson et al . , 2005; Vaccari and Bilder , 2005 ) . Since apoptotic cells disassemble their junctions prior to extrusion from the tissue ( Brancolini et al . , 1997; Steinhusen et al . , 2001 ) , we co-expressed the viral caspase inhibitor p35 ( Hay et al . , 1994 ) with shrub-RNAi to preserve tissue integrity and allow unambiguous analysis of junctional structures . Interestingly , while 48 hr expression of p35 + shrub-RNAi yielded discs with no apparent morphological defects seen in a thin section ( Figure 1B ) , discs after 65 hr of expression were disorganized and multi-layered in the posterior compartment as reported for shrub null mutant clonal eye disc tissue ( Figure 1C; Vaccari et al . , 2009 ) . This indicates that our assay is able to reproduce the hallmarks of ESCRT-mediated neoplasia . We used the 48-hr stage to analyze junctions by TEM prior to neoplasia . In the wildtype anterior control compartment , the membrane basal to the AJ was lined with the ladder-like electron-dense structures that represent the SJ ( Figure 1E ) . In the posterior ( shrub-RNAi ) compartment however , only few electron-dense structures basally to the AJ were detected and an obvious ladder pattern was rarely seen ( Figure 1E´ ) . Interestingly , we did not find AJ appearance to differ noticeably between wildtype and shrub-RNAi compartments ( Figure 1E/E´ ) . We sought to quantify SJ integrity by measuring the total length of electron-dense structures basally to the AJ in a region of interest ( ROI ) with specified length of 2 μm . In anterior control compartments , roughly 32% of membrane within ROIs was lined with SJ ( Figure 1D ) . This value decreased to about 15% in posterior compartments , indicating that 48 hr expression of p35 + shrub-RNAi reduces SJ density in wing discs by more than 50% ( Figure 1D ) . Due to the lack of apparent neoplastic transformation at the 48 hr stage , we conclude that reduction in SJ density does not reflect indirect effects resulting from ESCRT induced epithelial to mesenchymal transition ( EMT ) , but rather suggests a direct involvement of ESCRT in maintaining SJ density in wing imaginal discs . We turned to fluorescence microscopy to analyze the subcellular localization of junctional proteins upon Shrub depletion . Again , we used depletion for 48 hr , which does not show apparent neoplastic overgrowth and thus preserves epithelial monolayer organization . We used antibodies against E-cadherin and the claudin Mega to reveal AJ and SJ , respectively . Both proteins localized almost exclusively at the apical membrane in wildtype cells of the anterior compartment of the disc ( Figure 1F , left arrowheads ) . However , in the Shrub-depleted posterior compartment , Mega was not detected at the SJs , while E-cad showed wildtype AJ localization ( Figure 1F , right arrowheads ) . The absence of Mega from SJs is also visible in maximum intensity projections of the junctional region ( Figure 1G ) . Importantly , Mega accumulated in vesicle like structures at the basal side of the cells , suggesting that it might be trapped in intracellular compartments ( Figure 1F , arrow ) . This basal fraction of Mega colocalized with the endosomal marker Rab7 , suggesting that Mega was trapped within maturing endosomes ( inset in Figure 1H´ ) . We reasoned that junctional loss of Mega together with the decreased SJ density revealed by TEM analysis point toward defects in SJ integrity/function in shrub-RNAi tissue . Indeed , dye exclusion experiments suggest that epithelial barrier function is compromised in shrub-RNAi expressing wing discs ( Figure 1—figure supplement 1 ) . Results from another experiment also support this notion: The GPI-anchored SJ protein Lachesin ( Lac ) is required for barrier function and localized at the outer membrane leaflet ( Llimargas et al . , 2004 ) . We visualized endogenous Lac localization by using a GFP protein trap line in p35 + shrub-RNAi expressing discs . While Lac::GFP was restricted to SJ in the anterior control compartment , the protein spread more laterally in shrub-RNAi tissue ( Figure 1I , arrowheads ) . Interestingly , this phenotype resembles mis-localization of SJ proteins in mutants of individual complex components during junction formation in the embryo . For example , in mega deficient embryos , NrxIV fails to concentrate at the SJ and localizes along the entire lateral membrane ( Behr et al . , 2003 ) . We conclude that Shrub function maintains integrity of the SJ complex required for containment of Lac within the junction of the wing disc epithelium . Strikingly , when we depleted posterior wing disc compartments of Mega , we could only detect a faint lateral spreading of Lac::GFP basal to the SJ ( Figure 1—figure supplement 2 ) . This suggested that in wing discs , Shrub depletion more severely affects integrity of the SJ complex with respect to Lac::GFP confinement to the junction compared to Mega depletion . This made us wonder whether junction levels of other SJ core components might be affected upon Shrub depletion . Indeed , when we analyzed subcellular localization of the SJ core component ATPα upon 48 hr of shrub-RNAi expression , we found a strong and specific reduction of its junctional level in the Shrub-depleted compartment ( Figure 1—figure supplement 3 ) . This result indicates that the function of Shrub in regulating junctional membrane levels is not limited to Mega . The lateral spreading of Lac::GFP within the membrane in Shrub-depleted tissue also suggested that its mobility might be increased . We measured fluorescence recovery after photobleaching ( FRAP ) of Lac::GFP to test this hypothesis . Consistent with a very stable and immobile SJ complex ( Oshima and Fehon , 2011 ) , Lac::GFP fluorescence recovery at the junction was low in the anterior control compartment , with GFP signal intensities barely reaching 20% of the pre-bleach levels half an hour after photobleaching ( Figure 1J , green graph ) . In the shrub-RNAi compartment however , Lac::GFP fluorescence recovery at the SJ was quick and reached a plateau at roughly 40% of pre-bleach levels after a few minutes . This suggests that a fraction of Lac::GFP molecules shows increased mobility within the SJ membrane region in shrub-RNAi expressing cells , in line with a defective barrier/fence function of the SJ complex . Altogether , these results indicate that SJ integrity critically depends on ESCRT function and suggest that Shrub is required for intracellular transport of Mega from an endosomal compartment toward the junction . The loss of Mega from the SJ and concomitant accumulation in basal aggregates cannot easily be explained with a role of Shrub in Mega degradative trafficking but rather suggest that its export from the endosomal system might be impaired in Shrub-depleted tissue . Consistent with a non-degradative ESCRT role in transport of Mega , we did not find Mega to accumulate intracellularly when we interfered with endosomal maturation or prevented endolysosomal fusion downstream of ESCRT function , in contrast to the canonical ESCRT cargo Notch ( Figure 2—figure supplement 1 ) . These results suggest that Mega undergoes very little ( if any ) lysosomal turnover in wing imaginal discs and point toward an ESCRT function that is distinct from the degradative MVB pathway in trafficking of Mega . We hypothesized that biosynthetic delivery of Mega to the SJ might require an endosomal recycling pathway depending on ESCRT for its proper function . To test this idea , we depleted imaginal discs of proteins known to regulate endosomal recycling and analyzed the impact on junctional Mega levels . We found that expression of an RNAi construct targeting the retromer CSC component Vps35 led to a significant reduction of Mega at the SJ ( Figure 2—figure supplement 1 ) , which led us to investigate the function of retromer in transport of SJ components . We generated null mutant clones of the retromer core component Vps35 ( Vps35MH20 , Franch-Marro et al . , 2008 ) in wing discs and analyzed the subcellular distribution of junctional proteins . While Mega membrane levels at the SJ were reduced in Vps35MH20 tissue , junctional E-cad levels were unaffected ( Figure 2A ) . This result is analogous to the phenotype seen upon shrub-RNAi expression ( Figure 1 ) . We also found reduced membrane levels of the SJ core components ATPα , NrxIV and Nrg in Vps35MH20 clones , visualized by using endogenously tagged GFP protein trap lines ( Figure 2B–D ) . These data suggest that retromer has a general function in SJ protein trafficking that is not restricted to Mega . In line with this , quantification of junctional signal revealed that all of the affected SJ core components show similarly reduced levels in Vps35MH20 tissue ( roughly 50% compared to wildtype , Figure 2H ) . Additionally , we found Vps35 to be required for regulating membrane levels of SJ components in several tissues , such as eye/leg imaginal discs and pupal wings , pointing to a common requirement for the retromer pathway in SJ protein transport ( Figure 2—figure supplement 2 ) . Surprisingly , loss of NrxIV::GFP from the SJ membrane in Vps35MH20 mutant pupal wing tissue was more severe compared to the phenotype seen in third instar wing discs , suggesting that SJs within a tissue that undergoes morphogenetic changes and does not strongly proliferate might be even more sensitive toward loss of retromer function ( Figure 2—figure supplement 2 ) . Clones mutant for the Vps35 interaction partner Vps26 phenocopied Vps35MH20 tissue with regard to Mega membrane levels , supporting a general function of the CSC in transport of SJ components ( Figure 2E ) . Interestingly , in contrast to these structural SJ core components , the levels of the associated cytoplasmic scaffolding proteins Dlg and Cora were unaffected in Vps35MH20 mutant tissues ( Figure 2F , G ) . This is consistent with Vps35MH20 clones maintaining intact apicobasal polarity and supports a role of retromer in trafficking specific structural components of SJs . Nevertheless , our extended analysis with further SJ components revealed that the GPI-anchored proteins Lachesin , Contactin , as well as the cytoplasmic Varicose , show reduced SJ levels in Vps35MH20 clones ( Figure 2—figure supplement 3 ) . This indicates that the retromer CSC regulates apical levels of a large subset of SJ components . Importantly , while membrane levels of the SJ core components Mega and ATPα were reduced within retromer clones ( Figure 2A–B ) , levels of the SJ-associated TMP FasIII were not affected ( Figure 2—figure supplement 3 ) . These data mirror the phenotypes observed upon Shrub depletion ( Figure 1F–G and Figure 1—figure supplement 3 ) and hint at a common pathway in apical delivery of a subset of SJ proteins . Next , we quantified SJ density detected by TEM in wing discs depleted of Vps26 and found a reduction of electron-dense septa by about 60% ( Figure 2—figure supplement 4 ) . Consistent with compromised SJ integrity , a diffusible dye readily infiltrated Vps26-depleted tissue , indicating defective barrier function ( Figure 1—figure supplement 1 ) . We conclude that the retromer CSC regulates membrane levels of many SJ core components , thereby contributing to junction integrity and function . In optical sections of Vps35MH20 clones , endogenously tagged Mega::YFP protein shows reduced overall levels throughout the mutant tissue ( Figure 2—figure supplement 3 ) . This suggests that misrouting of Mega into the degradative pathway may occur upon loss of Vps35 function , as has been shown for many retromer cargos ( Franch-Marro et al . , 2008; Harterink et al . , 2011; Pocha et al . , 2011 ) . Consistently , when retromer function and endo-lysosomal fusion are simultaneously impaired , Mega::YFP shows reduced SJ levels and accumulates in intracellular vesicles , hinting at a failure to be degraded ( Figure 2—figure supplement 1 ) . This result supports the idea that misrouting of Mega into the degradation pathway occurs when CSC function is compromised . We conclude that Mega behaves like a bona fide retromer cargo . To gain further insight into the CSC-dependent pathway required for SJ delivery of Mega , we investigated null mutant clones of known retromer-associated factors ( SNX-BAR , Snx3 , Snx27 , and the WASH-complex component Fam21 ) . Surprisingly , we failed to find any of these proteins to be required for the regulation of Mega membrane levels , raising questions on the mechanism governing CSC-dependent transport of Mega ( Figure 2—figure supplement 5 ) . Nevertheless , based on above findings , we propose a novel function of the retromer CSC in delivery of SJ core components to the junction , thereby contributing to SJ homeostasis in the proliferative wing disc epithelium . The above results indicate that retromer is required for regulating membrane levels of Mega and other SJ components . Therefore , trapping of Mega in basally localized endosomal compartments upon shrub-RNAi expression could be a consequence of defective retromer-dependent endosomal export in this ESCRT deficient situation . We reasoned that loss of ESCRT function might alter organization and function of retromer-dependent carrier vesicles , thereby affecting cargo flux . To test this hypothesis , we analyzed the subcellular distribution of retromer components in Shrub-depleted cells . We visualized the retromer CSC by using an endogenously tagged Vps35 allele ( Vps35::RFP , Koles et al . , 2016 ) . In anterior control compartments , Vps35::RFP was found in vesicular structures throughout the cell , but with significantly higher abundance in the apical cytoplasm ( Figure 3A–D , arrows ) . Interestingly , this polarized apical localization is consistent with an apical transport hub at the junction level in wing disc cells that is characterized by enrichment of more than half of the Drosophila Rab GTPases ( Dunst et al . , 2015 ) . In contrast to its concentration within the apical hub , little Vps35::RFP was detected in the basal cytoplasm of wildtype tissue ( Figure 3A´ ) . Upon shrub-RNAi expression however , Vps35::RFP strongly accumulated basally ( Figure 3A´ , arrowhead ) while apical hub localization was almost completely abolished ( Figure 3A ) . Therefore , loss of Shrub function appears to re-distribute Vps35::RFP positive vesicles from the apical hub into the basal cytoplasm where they accumulate . This apical to basal shift of retromer CSC positive vesicles is also evident in optical cross sections of wing discs expressing RNAi constructs for several ESCRT components ( Figure 3B–D ) . Importantly , RNAi directed against the ESCRT-I component TSG101 or Vps4 , but not ESCRT-0 , phenocopied shrub-RNAi expression with regard to Vps35::RFP subcellular localization ( Figure 3B–D , Figure 3—figure supplement 1 ) . This indicates that several ESCRT complexes are required for maintaining apical hub localization of the CSC in the wing disc epithelium . We found strong colocalization of Mega and Vps35::RFP on basal endosomal aggregates of shrub-RNAi expressing cells ( Figure 3—figure supplement 2 ) . This is in line with Mega being trapped in aberrant Rab7 and CSC positive endosomal compartments residing in the basal cytoplasm of ESCRT deficient cells . Our analysis of these basal aggregates revealed that besides containing Mega , they are enriched in ubiquitinated cargo and are coated with retromer components , Hrs and endosomal GTPases Rab5 and Rab7 ( Figure 3—figure supplement 2 ) . We conclude that these aggregates likely represent Drosophila ‘class E-like’ compartments , which is supported by TEM analysis of basally localized endosomal structures within shrub-RNAi expressing wing disc tissue ( Figure 3—figure supplement 3 ) . While the regularly shaped and sized wildtype MVBs were absent in Shrub-depleted cells , we found a variety of enlarged , abnormal membranous compartments that are reminiscent of class E compartments in mammalian cells ( Doyotte et al . , 2005; Stuffers et al . , 2009 ) . We speculate that these irregular compartments aberrantly cluster endosomal machineries and cargos , thereby interfering with endosome function . Consequently , the SJ phenotype seen in shrub-RNAi tissue ( Figure 1 ) could be explained by basal displacement of retromer-dependent carriers , which are required for targeting SJ components to the junctional region . The apical bias of Vps35::RFP positive vesicles in wildtype cells suggests a polarized movement of CSC carriers along the apicobasal axis with relatively long dwell times at the apical hub . We reasoned that in wing discs , a majority of CSC-dependent cargo might be released preferentially at the apical pole of the cells . We turned to live imaging to study movement of CSC positive vesicles in wildtype and Shrub-depleted cells . We visualized CSC using Vps35::RFP or a transgene encompassing the genomic region of Vps26 fused to a C-terminal EGFP-tag ( Wang et al . , 2014 ) . In anterior control compartments , Vps26-EGFP localized in a vesicular pattern throughout the cells with increased abundance at the apical hub ( Figure 3E , arrow ) . In the posterior Shrub-depleted cells however , Vps26-EGFP localization at the apical hub was reduced and basal accumulation was evident ( Figure 3E ) . Therefore , live imaging of Vps26-EGFP recapitulates the apical to basal shift of Vps35::RFP in ESCRT-depleted fixed tissue ( Figure 3B–D ) . We used Vps35::RFP/Vps26-EGFP to record time series of CSC positive carrier vesicles within wildtype and Shrub-depleted tissue . We applied the Fiji plugin TrackMate ( Tinevez et al . , 2017 ) to map individual trajectories of Vps35::RFP/Vps26-EGFP vesicles over a timeframe of 5 min ( Figure 3E' ) . The results show that the CSC carriers in anterior control cells were highly mobile along the apicobasal axis with a subset of vesicles moving rapidly and relatively long distances between the apical and basal poles ( Figure 3E' , arrows ) . This suggests that CSC carriers , although preferentially residing at the apical hub , are highly mobile and occasionally shuttle between the apical and basal poles of wing disc cells . In contrast , long-distance movement of Vps35::RFP/Vps26-EGFP vesicles was severely reduced in the posterior Shrub-depleted compartment ( Figure 3E' and F ) . Importantly , the large basal CSC aggregates ( class E-like compartments ) were mostly immobile , showing no apparent movement along the apicobasal axis ( Figure 3E' , arrowheads ) . These results indicate that ESCRT function is required for the mobility of CSC positive endosomes . We hypothesize that certain retromer cargos might rely on CSC shuttling between the apical and basal poles for efficient transport . Mega::YFP vesicular structures overlapped extensively with Vps26 not only at the junctional level but also in vesicles with close proximity to the basal pole of the cells ( arrowheads in Figure 3G ) . Colocalization analysis revealed that 71 . 9% of Mega::YFP vesicles along the apicobasal axis were positive for Vps26 ( n = 3 discs/128 vesicles ) . This suggests that Mega and possibly other SJ components shuttle along the apicobasal axis in CSC positive carrier vesicles . We confirmed this by live imaging of Mega::YFP together with Vps35::RFP , which revealed extensive co-mobility of this retromer component with the vesicular fraction of Mega::YFP ( Animation 1 ) . Thus , Mega is moving along the apicobasal axis in CSC decorated vesicles . Together , the results reveal that ESCRT function regulates the mobility and apical hub localization of retromer CSC positive endosomes in wing imaginal discs . We conclude that by trapping CSC on aberrant endosomal compartments , loss of ESCRT function impairs retromer-dependent export of Mega from the endosome , consequently depleting its SJ pool in wing disc cells . Endosomal trafficking by retromer could occur via several different pathways . While certain retromer cargos are transported directly from endosomes to the plasma membrane , the classical retromer route involves cargo recycling in a detour via the Golgi . Besides these two endosomal recycling pathways , retromer has also been shown to regulate transcytosis from one membrane domain to another ( Vergés et al . , 2004 ) . Prior to all retromer and ESCRT-dependent trafficking events , endocytosis of cargo is required for subsequent sorting within the endosomal system . We reasoned that blocking endocytosis of Mega might reveal the membrane domain from which it is internalized into the endosomal system and aid in understanding how the retromer pathway is involved in its trafficking . We suppressed clathrin-mediated endocytosis by RNAi induced depletion of clathrin heavy chain ( Chc ) for 32 hr in the posterior compartment . While depletion of Chc did not have an apparent effect on junctional E-cad levels ( Figure 4A' ) , Mega::YFP levels were strongly reduced at the SJ ( Figure 4A" ) . Since membrane proteins accumulate at their site of endocytosis when the internalization process is inhibited , this result argues against Mega undergoing clathrin-dependent endocytosis at the apical membrane . Importantly , we found Mega::YFP accumulating at the most basal region of the lateral membrane ( Figure 4A"' ) , suggesting that Mega is continuously removed from a basal membrane pool by endocytosis . These results also suggest that Mega , prior to accumulating at the SJ , undergoes clathrin-mediated endocytosis at the most basal section of the basolateral membrane ( for the sake of simplicity , we will refer to this as the basodistal membrane ) . Next , we interfered with dynamin-dependent endocytosis by expressing a dominant negative version of Shibire , the Drosophila dynamin homolog ( UAS-ShiK44 , Moline et al . , 1999 ) under the control of hhGal4 . An expression of ShiK44 for 16 hr was sufficient to cause strong accumulation of Notch at the apical membrane , indicating that dynamin-dependent endocytosis of apically internalized membrane proteins is effectively impaired ( Figure 4B ) . In line with diminished uptake of Notch into the endosomal system , we found reduced abundance of intracellular Notch vesicles in the posterior compartment ( Figure 4B , arrowheads ) . In contrast to apically endocytosed Notch , Mega::YFP seemed to be reduced at the SJ level upon 16 hr of ShiK44 expression ( Figure 4C ) . In addition , it accumulated at the basal pole of the cells ( Figure 4C , arrow ) . Reduced SJ levels accompanied by basodistal membrane accumulation of Mega::YFP were also found upon depleting the tissue of Rab5 by RNAi ( Figure 4D ) . This common phenotype suggests that a critical step in SJ supply of Mega is its internalization from the basodistal membrane , hinting at a transcytosis-like mechanism with basal to apical direction . We confirmed the close association of Mega with Chc in wing imaginal discs by detecting colocalization of Mega::YFP and Chc in vesicular structures in wildtype tissue ( Figure 4E ) , which is consistent with Chc being part of the Mega proteome in Drosophila embryos ( Jaspers et al . , 2012 ) . Importantly , Mega::YFP vesicles in proximity to the basodistal membrane were Chc positive , in line with Mega being endocytosed basally ( Figure 4E , arrows ) . Colocalization analysis revealed that throughout the entire cell , 55 . 2% of Mega::YFP vesicles stained positive for Chc ( n = 4 discs/87 vesicles ) . This Chc positivity rate was slightly increased to 63 . 01% in focal planes at the basal cell pole ( n = 3 discs/73 vesicles ) , consistent with Mega::YFP entering the endosomal system by clathrin-dependent endocytosis at the basodistal membrane . Together , these data support a transcytosis-like mechanism from the basodistal domain of the lateral membrane to the SJ , which is required to supply the junction with newly synthesized Mega . This process depends on the endocytic and early endosomal machinery as interference with dynamin , clathrin and Rab5 function leads to reduced SJ levels of Mega accompanied by its accumulation at the basodistal membrane . These results suggest that Mega needs to traverse through the endosomal system prior to reaching the SJ and argue against a conventional recycling function of retromer in this process . This notion is further supported by our comparative analysis of Mega trafficking to that of Crumbs ( Crb ) , an established retromer recycling cargo in wing discs ( Pocha et al . , 2011 ) . It revealed distinct phenotypes upon Vps35 loss or inhibition of endocytosis , suggesting that these proteins do not traverse a common retromer-dependent pathway ( Figure 4—figure supplement 1 ) . We conclude that Mega is not undergoing apical recycling but rather requires basal to apical transcytosis for junctional delivery . Thus , our data suggest that transcytosis of SJ components is not limited to initial SJ formation in the embryo ( Tiklová et al . , 2010 ) , but is also required in a proliferative epithelium to maintain the junctional pool of the SJ component Mega . The complex anterograde trafficking of Mega to the SJ prompted us to investigate it in more detail . We generated an HA-tagged UAS-Mega construct , allowing us to analyze its delivery to the SJ upon overexpression . Continuous hhGal4-driven expression of UAS-HA-Mega led to its integration into SJ , which we confirmed by colocalization with Lac::GFP ( Figure 5A" ) . Interestingly , in contrast to endogenous Mega , we detected a fraction of HA-Mega at the most basal part of the epithelium ( Figure 5A ) . This suggests that the basodistal membrane pool of Mega that we observed upon endocytosis block is also detectable upon Mega overexpression . We also found increased abundance of intracellular vesicles , consistent with elevated trafficking of Mega within the cells ( Figure 5B ) . HA-Mega colocalized with Vps35::RFP on vesicles ( Figure 5B , arrows ) , reminiscent of endogenous Mega::YFP shuttling in Vps35::RFP positive endosomal carriers ( Animation 1 ) . Hence , we conclude that over-expression of UAS-HA-Mega recapitulates the hallmarks of Mega intracellular transport . While the majority of HA-Mega signal upon continuous overexpression was detected at the SJ level ( Figure 5C , arrowhead ) , HA-Mega was also found in lateral and basal membrane regions ( Figure 5C , arrow ) . Strikingly , the basodistal pool of HA-Mega was characterized by an intense membrane localization pattern resembling the staining of the junctional region ( Figure 5C' ) . The previous experiments suggested that Mega is continuously endocytosed from the basodistal membrane to supply the SJ pool . We therefore reasoned that basodistal membrane accumulation might be a short-lived intermediate step in Mega transport toward the SJ . To proof this assumption , we conducted a pulse-chase experiment using the Gal4/Galt80ts system and found that HA-Mega localization was much more confined to the SJ ( similar to endogenous Mega localization ) when its expression was halted for 14 hr after continuous expression ( Figure 5D ) . After this chase , the basodistal membrane pool was almost completely diminished ( Figure 5D' ) . This is consistent with a transient localization of Mega at the basodistal membrane prior to endocytosis and subsequent targeting to the SJ . Next , we investigated the roles of ESCRT and retromer in HA-Mega trafficking . Strikingly , when we expressed HA-Mega in shrub-RNAi tissue , it failed to reach the SJ altogether and instead was trapped within aberrant endosomal compartments in the basal cytoplasm ( Figure 5E , arrow ) . Colocalization with Vps35::RFP confirmed that HA-Mega accumulated in basal retromer CSC positive compartments ( Figure 5F , arrows ) , consistent with the results obtained from Mega staining in shrub-RNAi tissue ( Figure 3—figure supplement 2 ) . When we expressed HA-Mega in CSC-depleted tissue ( Vps35-RNAi ) , a similar , albeit weaker trapping of HA-Mega in basally localized vesicular compartments was observed ( Figure 5G , arrows ) . In contrast to exclusive localization of HA-Mega in basal aggregates upon shrub-RNAi expression ( Figure 5F , arrows ) , we also detected a pool of HA-Mega at the SJ level ( Figure 5G ) . However , imaging the SJ plane revealed that apical HA-Mega was mostly vesicular with no apparent membrane staining ( Figure 5I ) . This is in strong contrast to HA-Mega localization in wildtype tissue , characterized by a distinct junctional honeycomb pattern in both the disc proper and the overlaying peripodial disc cells ( Figure 5J ) . These results suggest that HA-Mega fails to integrate into the SJ in retromer or ESCRT-depleted wing disc cells . Recognition of proteins by the ESCRT machinery requires cargo ubiquitination , serving as a sorting signal for the ILV pathway ( Bilodeau et al . , 2002; Katzmann et al . , 2001; Shih et al . , 2002 ) . To test whether HA-Mega requires ubiquitination for its integration into SJ , we devised a UAS-HA-Mega construct in which all intracellular lysines ( K ) were changed to arginines ( R ) ( UAS HA-MegaK2R ) . This construct should be devoid of any potential ubiquitination , thereby preventing a possible recognition by ESCRT and sorting into the ILV pathway . However , upon expression in wing disc cells , HA-MegaK2R targeting to the junctional membrane was not impaired and the subcellular localization of the construct was indistinguishable from wildtype HA-Mega ( Figure 5—figure supplement 1 ) . This result suggests that Mega transcytosis and SJ integration does not require ubiquitination and that direct interaction with ESCRT components and sorting into ILVs is dispensable during anterograde transport of Mega . These data are compatible with a model of indirect regulation of Mega trafficking by ESCRT that is based on the requirement for ESCRT function in maintaining apical hub localization and mobility of the retromer CSC . Based on above findings , we conclude that a functional endosomal system is critically required for Mega transcytosis from the basodistal to the apical membrane , which is an essential prerequisite for its delivery to the SJ . Importantly , our data indicate a crucial role for ESCRT and retromer in a pathway that regulates anterograde trafficking of an apically localized membrane protein . The previous experiments suggested that Mega , despite residing in apical SJs , requires crucial trafficking steps at the basodistal plasma membrane prior to its integration into the junction . We therefore wondered whether locally restricted synthesis of Mega in the basal part of the cell might account for the transient pool at the basodistal plasma membrane observed upon endocytosis block or Mega overexpression . We used RNA fluorescence in situ hybridization to reveal the subcellular localization of Mega transcripts . To verify RNA probe specificity , we expressed UAS-HA-Mega using hhGal4 , which should strongly increase transcript abundance and consequently fluorescence signal intensity in the posterior compartment . Consistently , we found intense Mega RNA staining in the posterior compartment , confirming the specificity of our probe ( Figure 6A ) . We also detected some apical fluorescence signal that is notably visible in the epithelial fold of the hinge region ( Figure 6A , B , red asterisk ) . Very likely , this fraction represents unspecific signal from fluorophores accumulating in the extracellular space between the peripodial membrane and the disc proper , since the signal intensity did not differ between the wildtype anterior and the HA-Mega overexpressing posterior compartment ( Figure 6B , red asterisk ) . In contrast , the specific intracellular signal representing Mega transcript was markedly increased in the posterior compartment ( Figure 6B ) . Strikingly , endogenous , as well as overexpressed Mega RNA , was almost exclusively detected in dotted structures residing in the basal cytoplasm ( Figure 6B , arrows and arrowheads , respectively ) . The basal subcellular localization of Mega RNA is in strong contrast to that of the protein , which almost exclusively localizes in SJs at the apical pole of the cells ( Figure 6C ) . Thus , the discrepancy in RNA and protein localization may explain the necessity for a transcytosis route from the basodistal membrane to the SJ . We reasoned that Mega mRNA is translated basally , which may result in its initial targeting to the basodistal membrane following passage through the Golgi . Consistent with this idea , we found robust colocalization of HA-Mega with the Golgi marker Golgin84 ( Riedel et al . , 2016 ) at the basal pole of the cells ( Figure 6D , arrows ) . In contrast , little colocalization was found in the apical region in proximity to the SJ . This is surprising , since the majority of Golgi stacks appeared to reside apically ( Figure 6D' ) . To further study the early biosynthetic trafficking of Mega , we induced a short hhGal4-driven expression of UAS-HA-Mega for 2 hr . Only single cells within the posterior compartment showed detectable expression of HA-Mega ( Figure 6E , arrows ) . Strikingly , we occasionally detected cells containing exclusively basal vesicular HA-signal that overlapped with Golgin84 ( Figure 6F , arrows ) . Other cells showed extensive spreading of HA-Mega along the apicobasal axis and colocalization with Lac::GFP at the apical membrane , indicating integration into the SJ ( Figure 6G–I , arrow ) . These data are consistent with the first appearance of HA-Mega post biosynthesis in basal Golgi stacks , followed by transport to the basodistal membrane and subsequent transcytosis toward the SJ . Together , the results suggest that basal subcellular localization of Mega mRNA supports its local translation and subsequent secretion of Mega protein from basal Golgi stacks . Thus , the transient basodistal membrane pool of Mega is likely a consequence of this locally restricted secretion and represents newly synthesized protein that has not yet entered the transcytosis pathway required for anterograde delivery to the SJ . Our data reveal a critical requirement for ESCRT in a transport pathway that depends on retromer-mediated transcytosis to deliver newly synthesized Mega to its apical destination . Defects in endosomal retrieval upon ESCRT inactivation have been previously described in other systems , such as yeast or mammalian cells and , thus , appear to represent a common feature of the pleiotropic ESCRT deficient phenotype . In yeast , the endosome-to-Golgi retrieval of the sorting receptor Vps10p and its cargo carboxypeptidase Y ( CPY ) depends on retromer function ( Seaman et al . , 1997; Seaman et al . , 1998 ) . ESCRT mutant strains accumulate CPY in class E compartments from which retrieval to the Golgi is blocked ( Babst et al . , 1997; Piper et al . , 1995; Raymond et al . , 1992 ) . Similarly , the mammalian retromer cargo mannose 6-phosphate receptor ( M6PR ) also failed to recycle from endosomes to the Golgi in HeLa cells depleted of TSG101/ESCRT-I function ( Doyotte et al . , 2005 ) . In this study , the authors suggested that generation of class E compartments occurs at the expense of endosomal tubules ( Doyotte et al . , 2005 ) . Consistently , the retromer-associated tubulation factor SNX1 and its yeast homolog Vps5p were found on the rims of mammalian and yeast class E compartments , respectively ( Doyotte et al . , 2005; Seaman et al . , 1998 ) . Together with our finding of CSC accumulation on Drosophila class E-like compartments ( Figure 3 , Figure 3—figure supplement 2 ) , this suggests that ESCRT deficient endosomes remain coated with retromer components but fail to export specific cargo . While we cannot rule out the possibility of ESCRT components directly cooperating with retromer to form recycling tubules ( note that SNX-BAR , Snx3 , and Snx27 are not required for Mega transport; see Figure 2—figure supplement 5 ) , we favor an indirect mechanism linking ESCRT and retromer in this transport pathway . Our analysis of the aberrant endosomal compartments induced upon Shrub depletion revealed that they are enriched in endosomal organizers such as Rab5 and Rab7 , which could potentially interfere with retromer-dependent export when their activity at the limiting membrane is unrestrained ( Figure 3—figure supplement 2 ) . While the role of Rab7 in endosomal recruitment of the CSC is well established , the necessity for Rab7 GDP/GTP cycling during retromer-dependent carrier generation is still under debate ( Jia et al . , 2016; Jimenez-Orgaz et al . , 2018; Seaman et al . , 2009 ) . Rab7 and its GTPase-activating protein ( GAP ) Tbc1d5 are interaction partners of the CSC and can modulate its capability to retrieve endosomal cargo ( Jia et al . , 2016; Seaman et al . , 2009 ) . For example , interfering with Rab7-GTP hydrolysis by Tbc1d5 depletion yielded defects in retromer-dependent transport in HeLa cells ( Jia et al . , 2016 ) . Strikingly , under these conditions , retromer cargo was trapped in CSC-coated endosomes , paralleling our observation of Mega subcellular localization upon ESCRT depletion ( Jia et al . , 2016 ) . Similarly , by exposing the interplay between the CSC component Vps29 , Tbc1d5 , and Rab7 in adult Drosophila brains , the authors of a recent study reported the capability of endosomal Rab7 to interfere with retromer CSC function in vivo ( Ye et al . , 2020 ) . While the exact mechanism rendering retromer dysfunctional at Drosophila class E compartments remains to be determined , our data support the mounting pool of evidence that ESCRT is required for multiple endosomal retrieval pathways . It is therefore likely that aspects of the pleiotropic ESCRT phenotype in metazoans stem from defective export of proteins from the endosomal system . For example , in Drosophila , leaky SJ could support ESCRT-mediated neoplastic transformation by permitting diffusion of signaling molecules within the imaginal disc tissue . We here found that biosynthetic delivery of Mega depends on a transcytosis-like mechanism from the basodistal to the apical plasma membrane . This long-distance transport required sequential action of endocytic ( clathrin , dynamin , Rab5 ) and endosomal ( ESCRT , retromer CSC ) machineries . Importantly , the finding that overexpressed HA-Mega is unable to reach the SJ in absence of retromer and ESCRT function ( Figure 5 ) is in agreement with biosynthetic delivery of Mega relying on endosomal function . Therefore , although we cannot exclude the possibility that Mega transiently passes the Golgi after endocytosis at the basodistal membrane , we favor the unconventional transcytosis model . Strikingly , while retromer-dependent endosomal recycling has been extensively documented , only one mammalian cell culture study implicated retromer in transcytosis from one membrane domain to another ( Vergés et al . , 2004 ) . Thus , SJ delivery of Mega in imaginal discs represents a novel physiological role of retromer to study this process in vivo . The finding that CSC-mediated anterograde transport of Mega is independent of retromer-associated sorting nexins ( Figure 2—figure supplement 5 ) indicates that this transcytosis pathway is distinct from many established CSC-dependent routes and suggests that it may require unknown cofactors ( or does not require endosomal tubulation ) . Our analysis of Vps35 clones in pupal wings or leg imaginal discs revealed that in these tissues , clones completely devoid of the SJ core component NrxIV occur frequently ( Figure 2—figure supplement 2 ) . Similarly , shrub mutant clones in the pupal notum were entirely lacking junctional ATPα ( Roland Le Borgne , personal communication , July 2020 ) . This is in contrast to surface levels of SJ components in Vps35 mutant wing discs , which were consistently reduced by about 50% ( Figure 2 ) . This provokes the hypothesis that a parallel endosomal export pathway for SJ components may exist in wing discs that could partially compensate for loss of retromer . However , we think this is unlikely since overexpressed HA-Mega fails to reach the SJ not only upon Shrub but also upon Vps26 depletion ( Figure 5E–J ) . One has to keep in mind that this experiment specifically monitors delivery of newly synthesized HA-Mega while the Vps35 clonal analysis assesses the impact of retromer loss of function on pre-existing SJ . Thus , in a clonal situation , a ‘thinning out’ of junctions is expected with consecutive rounds of cell division , which could explain incomplete phenotypic expressivity in wing disc Vps35 clones . Nevertheless , it remains to be determined why in leg discs and pupal wings , SJ appear to be more sensitive toward CSC loss ( Figure 2—figure supplement 2 ) . During metamorphosis , wing imaginal disc cells undergo drastic morphogenetic changes to form the pupal wing epithelium , a process known to require AJ remodeling ( Classen et al . , 2005 ) . It is therefore possible that analogously to AJ , SJ may also be actively remodeled during pupal wing formation . This could explain the strong requirement for retromer function in maintaining SJ integrity in a tissue undergoing morphogenetic changes . Paralleling our findings described here , a previous study showed that embryonic SJ formation depends on endocytosis and subsequent redistribution of junction components from the lateral membrane to the SJ ( Tiklová et al . , 2010 ) . Thus , transcytosis of SJ components is a mechanism likely required for both initial SJ formation as well as maintenance of SJ integrity in non-embryonic tissues in Drosophila . While we did not assess the roles of ESCRT and retromer CSC in initial SJ formation , it is conceivable that they are already required for transporting SJ components during embryogenesis . Consistently , shrub mutant embryos display a defective epithelial barrier function , suggesting they fail to form functional SJ ( Dong et al . , 2014 ) . The reason for SJ maintenance to rely on such an elaborate trafficking of its components remains to be determined . It has been suggested that SJ components form stable complexes prior to integration into the junction . Potentially , essential post-translational modifications of certain SJ components required for complex formation may occur exclusively at the basodistal membrane or during the passage through the endosomal system . If transient localization of SJ components at the basodistal membrane is a prerequisite for efficient SJ core complex formation , depositing transcripts for structural components such as Mega in the basal cytoplasm would shorten the route individual SJ components need to pass prior to SJ complex formation at the basodistal membrane domain . Alternatively , the finding that Mega mRNA predominantly localizes in the basal cytoplasm ( Figure 6 ) provides the foundation for another hypothesis: It is widely accepted that apical and basolateral cargos undergo motif-based sorting leading to secretion toward the respective membrane domains ( Stoops and Caplan , 2014 ) . Basal subcellular localization of Mega transcripts could potentially reflect an apical/basolateral sorting divergence at the mRNA level . Accordingly , basal translation and exocytosis of Mega ( induced by a putative basal/basolateral sorting signal ) may lead to its targeting toward the basodistal membrane , despite the fact that the SJ resides apically in wing disc cells . Thus , transcytosis may serve as an adaptation for redistribution of cargos to their destined membrane domain when they are initially secreted to a different one due to early sorting signals . Since the columnar wing imaginal disc cells have a very elongated shape and possess Golgi stacks all along the apicobasal axis , it is conceivable that certain Golgi stacks residing at the apical and basal poles are specialized in secreting apical and basolateral cargos , respectively . Although this is highly hypothetical , systematic analysis of transcript localization for apical ( e . g . E-cad ) and basolateral ( e . g . SJ components ) cargos could reveal a potential spatial separation of distinct secretory routes already at the mRNA level . In wing imaginal discs , a similarly complex transcytosis route ( albeit with an apical to basal direction ) has been described for the signaling molecule Wingless , which is translated apically , transiently presented at the apical membrane and finally transcytosed toward the basal membrane where it is secreted ( Yamazaki et al . , 2016 ) . Thus , distinct transcytosis pathways in the wing disc epithelium provide a mechanism for targeting certain proteins to their site of action , specifically when the protein is translated far away from its terminal destination . In this study , we unravel a novel physiological retromer function in regulating surface levels of a claudin and other structural SJ components ( e . g . Nrg , ATPα , Lac , NrxIV , Cont ) in several Drosophila tissues ( Figure 2 and Figure 2—figure supplement 2 ) . Presently , we do not know how the SJ components are selected for this retromer-dependent pathway , and whether it requires physical interaction with CSC components . Since SJ proteins may traverse the endosomal system in complex , the vast number of different components brings about a plethora of possible interaction sites . Importantly , a mass-spectrometry-based study of the Mega interactome did not detect any retromer CSC components or associated factors but confidently found SJ core components as well as clathrin ( Jaspers et al . , 2012 ) . While the interaction mode of CSC and SJ proteins remains to be determined , our data reflect the assumption that several SJ core components represent novel putative retromer cargos in Drosophila . Strikingly , among the proteins affected by retromer loss of function , many possess mammalian homologs ( e . g . NrxIV/CNTNAP2 , ATPα/ATP1A1 , Nrg/NRCAM ) . This suggests they could represent a novel set of conserved retromer cargos . Indeed , several lines of evidence suggest that ESCRT/retromer-mediated transport of SJ components may be evolutionary conserved from Drosophila to mammals . Depletion of ESCRT-I component TSG101 in mammalian epithelial cells led to a reduction of trans-epithelial resistance ( TER ) , indicating defects in TJ-mediated barrier function ( Dukes et al . , 2011 ) . Additionally , Claudin-1 , an essential TJ component , continuously underwent endocytosis and recycling back to the plasma membrane in several mammalian cell lines in a process requiring ESCRT function ( Dukes et al . , 2011 ) . Importantly , the mechanism behind reduced recycling and entrapment of Claudin-1 in ubiquitin-positive aberrant endosomes upon interference with ESCRT function remained elusive ( Dukes et al . , 2011 ) . Thus , it is unknown how the export of Claudin-1 from the endosomal system in mammalian cells is achieved ( Dukes et al . , 2011 ) . By revealing an ESCRT-dependent function of the retromer CSC in claudin endosomal export in Drosophila , our data may provide an explanation for a possibly conserved trafficking pathway of claudins . In support of this , Claudin-1 and Claudin-4 membrane levels were significantly reduced in a mass spectrometry-based surface proteome study of Vps35-depleted human cells ( Steinberg et al . , 2013 ) . Furthermore , the TJ protein Zonula occludens-2 ( ZO-2 ) was strongly enriched in a Vps26 interactome , suggesting that a presumptive retromer function in TJ maintenance in mammalian cells may not be limited to claudins , similar to our findings in Drosophila presented here ( Steinberg et al . , 2013 ) . It remains to be determined whether a physiological role of retromer in mammalian TJ maintenance occurs also in vivo . An increasing amount of tools , such as conditional Vps35 knockout mice ( de Groot et al . , 2013 ) , will enable analysis of this putatively conserved retromer function in mammalian systems and reveal any possible implications in development and/or disease . A complete list of all stocks used in this study is found in the key resources table above . Flies were raised on standard cornmeal/molasses/yeast diet and kept at room temperature . Crossings were raised on 25°C , except for experiments containing the temperature sensitive tubGal80ts ( McGuire et al . , 2004 ) . Those flies were kept at 18°C ( permissive temperature ) to inhibit Gal4/UAS-mediated expression ( Brand and Perrimon , 1993 ) and shifted to 29°C ( restrictive ) for specific time spans to activate UAS-based expression . Flp/FRT system ( Xu and Rubin , 1993 ) induced clones were either generated by expression of Ubx-FLP or using hs-FLP with an 70 min heatshock in the first instar larval stage ( 24–48 hr after egg laying ) . For generation of UAS and tubP expressed HA-tagged Mega constructs , the clone LD14222 from Drosophila Genomics Resource Center ( DGRC ) was used as cDNA source . The HA-tag followed by a 3xGlycin linker was fused to the N-terminus of the Mega open reading frame by PCR using the following primers: fw-NotI-ATG-HA-3xGly-Mega GgcggccgcatgTACCCATACGAcGTTCCAGAcTACGCTggcgggggcCGCGAACTTAACAAGCAGCAG rev-Mega-XhoI ctgcactcgagTTATATGTAGCCCTGCAGGC . The generated PCR fragment was restriction digested with NotI and XhoI enzymes ( New England Biolabs ) and subsequently cloned into pUAST-attB and pTUB vectors . The tubulinP-based plasmid derived from a pCaSpeR4-tubulin-QF#7 vector ( addgene ) with SV40 polyA 3' UTR . The UAS-HA-MegaK2R construct was designed analogous to UAS-HA-Mega with all intracellular lysines exchanged to arginines . The cDNA for this construct was synthesized by BaseClear B . V . Generation of transgenic flies was performed by attB/attP specific genomic integration into the landing sites 51C ( for 2nd chromosome ) and 86Fb ( for 3rd chromosome ) ( Bischof et al . , 2007 ) . Injection of embryos was either performed in house or by BestGene Inc ( CA 91709 , U . S . A ) . A complete list of all antibodies used in this study is found in the key resources table above . Late L3 larvae or pupae were dissected in phosphate-buffered saline ( PBS , pH 7 . 4 ) on ice and immediately fixed for 30 min in 4% paraformaldehyde in PBS . Following three 10 min wash steps in PBT ( 0 . 3% Triton X-100 in PBS ) , tissue was blocked with 5% normal goat serum ( NGS , Jackson ImmunoResearch ) in PBT and subsequently incubated with primary antibodies in 5% NGS in PBT for 2 hr . After three 15 min wash steps with PBT , discs were incubated with fluorochrome-conjugated antibodies ( Alexa-488 , –568 , −647 from Thermo Fisher Scientific ) for 2 hr in 5% NGS in PBT . For nuclear staining , Hoechst 33258 ( Sigma Aldrich ) was used at a concentration of 1:10 , 000 . Discs or pupae were mounted in Vectashield ( Vector Laboratories ) and imaged with a Zeiss AxioImager Z1 wide field microscope equipped with a Zeiss Apotome . For SJ length measurements , ROIs were manually assigned with Fiji and the total length of electron-dense SJ within ROIs was measured to yield a SJ length/ROI length ratio . Statistical analysis was performed with GraphPad Prism 7 . 0 d . Late L3 instar larvae were dissected in PBS and immediately mounted on coverslips in Schneider's Drosophila medium ( Pan Biotech ) . Double-sided tape containing a punchhole was used both as a spacer to avoid tissue damage during mounting and as a short-term imaging chamber that restricted wing disc movement . Imaging was performed with a Zeiss LSM 880 laser scanning microscope equipped with an Airyscan detector . For time series , frames were acquired in 1–4 s intervals and line switching was used for dual channel acquisition . Wing discs were mounted as described above ( see Live Imaging ) . Photobleaching and imaging were performed on a Zeiss LSM 880 confocal microscope using a gallium arsenide phosphide ( GaAsP ) detector and 40x objective with a numeric aperture of 1 . 2 . A 15x digital zoom was applied to yield a detection area of 14 . 17 × 14 . 17 micron , equivalent to 512 × 512 pixels . In the focal plane of the junctions , a bicellular SJ ROI ( 0 . 83 × 0 . 83 microns ) was bleached with 100% excitation laser intensity and 10 repeats . The laser dwell time for each pixel was 4 . 1 μs . After bleaching , the tissue was imaged in 2 min intervals for 30 min and any potential drifting ( x- , y- , z-axis ) was corrected manually . Images were merged into a time series with Fiji and the mean gray values within the ROI were plotted against time/min ( Figure 1J ) . Statistical analysis was performed with Microsoft Excel and GraphPad Prism 7 . 0 d . The Fiji plugin TrackMate ( Tinevez et al . , 2017 ) was used on bleach corrected ( exponential fit ) time series obtained by live imaging to analyze movement of vesicles positive for either Vps26-EGFP or Vps35::RFP . A Dog detector with estimated blob diameter of 2 micron and threshold 1 were used as a fixed setup . Tracks were analyzed by the simple LAP tracker with following parameters: Linking max distance: 2 . Gap closing distance: 2 . Gap closing max frame gap: 20 . Mean velocity was used as the parameter for comparing vesicles within wildtype and Shrub-depleted tissue . Statistical analysis was performed with GraphPad Prism 7 . 0 d . Wing discs were dissected in 0 . 1M phosphate buffer ( 0 . 1M PO4 ) on ice and immediately fixed with 2 . 5% glutaraldehyde in 0 . 1M PO4 for 1 hr . Following five 5 min washsteps with 0 . 1M PO4 , tissue was postfixed with osmium tetroxide ( 2% in 0 . 1M PO4 ) for 1 hr on ice . After three 5 min washsteps with 0 . 1M PO4 and further washsteps with ddH2O , the specimens were gently dehydrated with ethanol in concentrations ranging from 50 to 100% . Acetone was used as an intermediate solvent to support Epoxy Embedding Medium ( Epon , Sigma Aldrich ) infiltration into the tissue . Specimens were stored over night in 100% Epon and subsequently embedded and sectioned . Semi-thin sections were stained with Richardson blue . Ultra-thin sections were contrasted with 2% uranyl acetate and lead citrate prior to imaging . Sections were analyzed with an EM 902 ( Zeiss ) microscope at 80 KV . For generation of Mega RNA antisense probe , the Mega/Pickel cDNA containing LD14222 clone from DGRC was linearized by EcoRI digestion following T7 polymerase-dependent in vitro transcription and digoxigenin ( DIG ) labeling . RNA probes were purified with a NucleoSpin Gel and PCR clean-up kit before use in the hybridization reaction . For FISH , late L3 larvae were dissected in PBS on ice and fixed for 20 min in 3 . 7% formaldehyde in PBS . Specimens were handled according to standard protocols . Briefly , larval tissue was incubated with RNA probes in formamide-based hybridization buffer at 65°C over night . To detect probes , HRP-conjugated anti-DIG antibodies ( Perkin Elmer ) were used , followed by a tyramide signal amplification reaction ( TSA plus Cyanine 3 Fluorescence Kit , Perkin Elmer ) . Wing discs were mounted in Vectashield ( Vector Laboratories ) and imaged with a Zeiss AxioImager Z1 wide field microscope equipped with a Zeiss Apotome . Late L3 larvae were dissected in PBS and briefly rinsed ( 30 s ) in PBS containing 10 kDa Texas-dsRed labeled Dextran ( Molecular Probes D-1828 ) in a concentration of 1 mg/ml . Following two quick wash steps in PBS , imaginal discs were dissected from the carcasses and immediately mounted in Vectashield ( Vector Laboratories ) on microscope slides . To avoid tissue damage due to compression , double-sided tape was used as a spacer between slide and cover slip . Images were acquired on a Zeiss AxioImager Z1 wide field microscope immediately after mounting ( within a timeframe of 5 min ) . For measuring colocalization of Mega::YFP with either Vps26 or Chc , dual channel images of single fluorescent planes ( either lateral views or basal planes ) were adjusted for brightness in FiJi . Colocalization was manually determined for each individual Mega::YFP vesicle using the FiJi cell counter plugin .	Proteins are large molecules responsible for a variety of activities that cells needs to perform to survive; from respiration to copying DNA before cells divide . To perform these roles proteins need to be transported to the correct cell compartment , or to the cell membrane . This protein trafficking depends on the endosomal system , a set of membrane compartments that can travel within the cell and act as a protein sorting hub . This system needs its own proteins to work properly . In particular , there are two sets of proteins that are crucial for the endosomal systems activity: a group of proteins known as the ESCRT ( endosomal sorting complex required for transport ) machinery and a complex called retromer . The retromer complex regulates recycling of receptor proteins so they can be reused , while the ESCRT machinery mediates degradation of proteins that the cell does not require anymore . In the epithelia of fruit fly larvae – the tissues that form layers of cells , usually covering an organ but also making structures like wings – defects in ESCRT activity lead to a loss of tissue integrity . This loss of tissue integrity suggests that the endosomal system might be involved in transporting proteins that form cellular junctions , the multiprotein complexes that establish contacts between cells or between a cell and the extracellular space . In arthropods such as the fruit fly , the adherens junction and the septate junction are two types of cellular junctions important for the integrity of epithelia integrity . Adherens junctions allow cells to adhere to each other , while septate junctions stop nutrient molecules , ions and water from leaking into the tissue . The role of the endosomal system in trafficking the proteins that form septate junctions remains a mystery . To better understand the role of the endosomal system in regulating cell junctions and tissue integrity , Pannen et al . blocked the activity of either the ESCRT or retromer in wing imaginal discs – the future wings – of fruit fly larvae . Pannen et al . then analyzed the effects of these endosomal defects on cellular junctions using an imaging technique called transmission electron microscopy . The results showed that both ESCRT and retromer activities are necessary for the correct delivery of septate junction components to the cell membrane . However , neither retromer nor ESCRT were required for the delivery of adherens junction proteins . These findings shed light on how retromer and the ESCRT machinery are involved in the epithelial tissue integrity of fruit fly larvae through their effects on cell junctions . Humans have their own versions of the ESCRT , retromer , and cell junction proteins , all of which are very similar to their fly counterparts . Since defects in the human versions of these proteins have been associated with a variety of diseases , from infections to cancer , these results may have implications for research into treating those diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2020	The ESCRT machinery regulates retromer-dependent transcytosis of septate junction components in Drosophila
Non-centrosomal microtubule arrays assemble in differentiated tissues to perform mechanical and transport-based functions . In this study , we identify Caenorhabditis elegans NOCA-1 as a protein with homology to vertebrate ninein . NOCA-1 contributes to the assembly of non-centrosomal microtubule arrays in multiple tissues . In the larval epidermis , NOCA-1 functions redundantly with the minus end protection factor Patronin/PTRN-1 to assemble a circumferential microtubule array essential for worm growth and morphogenesis . Controlled degradation of a γ-tubulin complex subunit in this tissue revealed that γ-tubulin acts with NOCA-1 in parallel to Patronin/PTRN-1 . In the germline , NOCA-1 and γ-tubulin co-localize at the cell surface , and inhibiting either leads to a microtubule assembly defect . γ-tubulin targets independently of NOCA-1 , but NOCA-1 targeting requires γ-tubulin when a non-essential putatively palmitoylated cysteine is mutated . These results show that NOCA-1 acts with γ-tubulin to assemble non-centrosomal arrays in multiple tissues and highlight functional overlap between the ninein and Patronin protein families . Differentiated cells assemble non-centrosomal microtubule arrays to perform structural , mechanical , and transport-based functions ( Keating and Borisy , 1999; Bartolini and Gundersen , 2006 ) . Examples include the neuronal microtubule arrays that structure axons and dendritic arbors ( Kuijpers and Hoogenraad , 2011 ) , longitudinal arrays of parallel microtubules in syncytial myotubes ( Warren , 1974; Tassin et al . , 1985 ) , and non-centrosomal arrays in epithelial cells ( Keating and Borisy , 1999; Bartolini and Gundersen , 2006 ) . In simple epithelia , cells build arrays of parallel microtubules that run along their apical–basal axis ( Keating and Borisy , 1999; Bartolini and Gundersen , 2006; Brodu et al . , 2010; Feldman and Priess , 2012 ) , whereas desmosomal cell–cell junctions organize microtubule arrays that form around the periphery of stratified epithelial cells in mouse skin ( Lechler and Fuchs , 2007; Sumigray et al . , 2012 ) . The radial organization of centrosomal arrays arises from the fact that microtubules are nucleated , and their nascent minus ends capped and anchored , by centrosomally targeted protein complexes . Similarly , assembly of non-centrosomal microtubule arrays is likely to involve targeting of microtubule nucleating , as well as minus-end protection and/or anchoring factors , to non-centrosomal sites . Important current goals include identifying the factors that control the assembly of non-centrosomal arrays and determining the extent of overlap between the mechanisms utilized at centrosomes and non-centrosomal sites in different tissues . Complexes containing γ-tubulin , a specialized tubulin isoform implicated in microtubule nucleation ( Zheng et al . , 1995; Oegema et al . , 1999; Kollman et al . , 2011 ) , are thought to contribute to the assembly of both centrosomal and non-centrosomal arrays . During the differentiation of Drosophila tracheal epithelial cells , both γ-tubulin complexes , and the center of microtubule nucleation in regrowth experiments , transition from centrosomes to the apical cell surface ( Brodu et al . , 2010 ) . In Caenorhabditis elegans , γ-tubulin is also targeted to the cell surface in the embryonic epidermis and germline , and the apical cell surface in the intestinal epithelium ( Zhou et al . , 2009; Fridolfsson and Starr , 2010; Feldman and Priess , 2012 ) . Ninein is a large coiled-coil protein that localizes to the sub-distal appendages of mother centrioles ( Mogensen et al . , 2000 ) , where it is thought to anchor centrosomal microtubules ( Dammermann and Merdes , 2002; Delgehyr et al . , 2005 ) . During the differentiation of mouse cochlear epithelial cells , ninein re-localizes from centrosomes to the apical surface ( Mogensen et al . , 2000; Moss et al . , 2007 ) ; ninein re-localization also occurs during the differentiation of stratified epithelial cells in the mouse epidermis , where it targets to desmosomal junctions ( Lechler and Fuchs , 2007 ) . Inhibition of the core desmosomal component , desmoplakin , disrupts ninein targeting and formation of the peripheral non-centrosomal microtubule array ( Lechler and Fuchs , 2007 ) , but direct evidence that ninein is important for array formation is currently lacking . The Patronin/CAMSAP/Nezha family of minus end-associated proteins , conserved among animals with differentiated tissues ( Baines et al . , 2009 ) , are also implicated in the formation of non-centrosomal arrays ( Akhmanova and Hoogenraad , 2015 ) . Members of this protein family are thought to be involved in protecting microtubule minus ends from depolymerizing kinesins ( Goodwin and Vale , 2010; Hendershott and Vale , 2014; Jiang et al . , 2014 ) . Drosophila and C . elegans each have one family member ( Patronin and PTRN-1 , respectively ) , whereas vertebrates have three ( calmodulin-regulated spectrin-associated protein or CAMSAP1-3 ) . Although initially identified in cultured epithelial cells ( Meng et al . , 2008; Jiang et al . , 2014 ) , the main in vivo phenotypes associated with knockdown of Patronin/CAMSAP/Nezha family members have been in neurons ( Chuang et al . , 2014; King et al . , 2014; Marcette et al . , 2014; Richardson et al . , 2014; Yau et al . , 2014 ) . As outlined above , γ-tubulin and Patronin respectively harbor minus-end nucleation and protection activities , and ninein is proposed to anchor microtubules . Mechanistic work has also raised the possibility of functional redundancies between minus end-associated factors . For example , in addition to being a microtubule nucleator , γ-tubulin complexes can cap microtubule minus ends ( Keating and Borisy , 2000; Wiese and Zheng , 2000 ) . Similarly , CAMSAP-tubulin stretches may function as seeds that allow microtubule regrowth ( Tanaka et al . , 2012; Jiang et al . , 2014 ) , and both ninein and Patronin family members localize to junctional complexes ( Lechler and Fuchs , 2007; Meng et al . , 2008 ) where they could serve anchoring functions . Hence , another important open question is the extent to which minus end-associated factors function collaboratively or redundantly during microtubule array assembly in vivo . Here , we characterize the C . elegans protein NOCA-1 ( non-centrosomal array 1 ) , a protein we identified in a prior high-content screen because its inhibition phenocopied the effect of γ-tubulin removal on germline morphology ( Green et al . , 2011 ) . We show that NOCA-1 shares homology with vertebrate ninein and identify isoforms that are necessary and sufficient for NOCA-1 function in three different tissues . We explore the functional relationship between NOCA-1 , γ-tubulin , and Patronin/PTRN-1 in the assembly of non-centrosomal microtubule arrays . In the larval epidermis , NOCA-1 functions with γ-tubulin in parallel to Patronin/PTRN-1 to assemble a circumferential microtubule array required for larval development . In the germline and embryonic epidermis , NOCA-1 functions independently of Patronin to promote assembly of microtubule arrays required for nuclear positioning . Cumulatively , our results suggest that NOCA-1 functions together with γ-tubulin to direct the assembly of non-centrosomal arrays in multiple tissues and highlight functional overlap between the ninein and Patronin families of microtubule cytoskeleton-controlling proteins . The noca-1 locus is large ( 23 kb ) and more complex than typical for C . elegans genes , encoding eight alternatively spliced isoforms that share a common 466 amino acid C-terminal domain with a predicted coiled-coil region ( Figure 1A ) . Sequence homology searches identified similarity between this C-terminal domain of nematode NOCA-1 proteins and vertebrate nineins ( Figure 1A and Figure 1—figure supplement 1 ) . Ninein ( NIN ) and the related ninein-like protein ( NINL ) are homologous in their N- and C-termini but differ in their central region . The domain common to NOCA-1 isoforms is homologous to the ninein-specific central region that is absent in ninein-like protein ( Figure 1A and Figure 1—figure supplement 1 ) . This ninein-specific region resides within a larger domain suggested to be required for the microtubule anchoring function of centrosomal ninein ( Delgehyr et al . , 2005 ) . We refer to the C-terminal domain of NOCA-1 common to all isoforms as the ninein homology domain ( NHD ) . 10 . 7554/eLife . 08649 . 003Figure 1 . NOCA-1 is a protein with homology to vertebrate ninein that functions redundantly with PTRN-1/Patronin to promote larval development and viability . ( A ) Schematics of the noca-1 locus , encoded NOCA-1 isoforms , and a short human ninein isoform showing the region with homology to NOCA-1 ( alignment in Figure 1—figure supplement 1A ) . The region of ninein absent from ( dark green ) or with low homology to ( light green ) ninein-like protein is underlined . Red line above the NOCA-1 isoforms shows the region deleted in the ok3692 allele . ( B ) Immunoblot of NOCA-1 in lysates from control , noca-1∆ , and noca-1 ( RNAi ) worms . ( C ) Top: schematic of the Caenorhabditis elegans Patronin homolog , PTRN-1 . Bottom: immunoblot of PTRN-1 in lysates from control and ptrn-1∆ worms . ( D ) Images of control and mutant worms 72 hr post L1 recovery ( snapshots from Video 1 ) . Arrowheads mark dead worms . ( E ) Plot of percentage of normal-sized adults , small uncs , and dead worms 72 hr post L1 for the indicated genotypes . n is number of worms analyzed in 3–5 independent experiments . ( F ) Plots of body length ( left ) and % living worms ( right ) vs time for worms with the indicated genotypes . ( G ) Left: Coomassie blue staining of recombinant proteins purified from baculovirus-infected insect cells . Right top: schematic of flow-cell-based kinesin gliding assay . Right center: kymographs showing microtubule gliding in the presence of indicated GFP-tagged proteins . Right bottom: plot of frequencies of plus end , minus end , or side binding . ( H ) Left: flow chart of microtubule co-sedimentation experiment . Right: immunoblots probing for NOCA-1 or PTRN-1 ( top and center ) and Coomassie blue staining showing tubulin ( bottom ) after sedimentation . Markers are in kDa . Coiled-coil predictions were performed using Paircoil2 ( 28 aa window , 0 . 025 threshold ) . Error bars are SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00310 . 7554/eLife . 08649 . 004Figure 1—figure supplement 1 . NOCA-1 has homology to vertebrate ninein . ( A ) Alignment of nematode NOCA-1 homologs with vertebrate nineins . Mafft WS sequence alignment of the indicated regions from nematode NOCA-1s ( C . elegans , GI 32567236; Ascaris suum , GI 541046681; Brugia malayi , GI 671417611 ) and vertebrate nineins ( Chinchilla lanigera isoform 6 , GI 533123118; Homo sapiens isoform X4 , GI 530403936; Equus prezewalskii isoform X4 , GI 664719818 ) . Color-coding is based on the BLOSUM62 matrix . Green asterisks mark the a and d positions of a predicted coiled coil ( Paircoil2 ) in the C . elegans sequence . ( B ) The homology between nematode NOCA-1 homologs and vertebrate nineins was discovered in an NCBI BLAST using the conserved region of NOCA-1 from B . malayi as the query . This is one of the best ways to identify non-nematode homologs of C . elegans proteins , since Brugia sequences tend to be among the least divergent for nematode species . The BLAST using the Brugia sequence identified all of the nematode NOCA-1 homologs ( red , pink , and green text ) , along with the C . lanigera , Heterocephalus glaber , and Fukomys damrensis nineins ( black and blue text ) . ( C ) Reverse BLAST of aa 1549–1801 of C . lanigera ninein isoform X4 against all nematode sequences yielded B . malayi NOCA-1 as the top hit ( E value = 2e−07 ) and Loa loa NOCA-1 as the second hit ( E value = 2e−04 ) . Other coiled-coil proteins were also detected , but with substantially less significant E values ( i . e . , Toxocara canis myosin II , E value = 3 . 3; C . Briggsae HCP-2 , E value = 6 . 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00410 . 7554/eLife . 08649 . 005Figure 1—figure supplement 2 . Expanded view of the immunoblot for NOCA-1 in lysates from control and noca-1 ( RNAi ) worms shown in the right panel of Figure 1B . Black arrows , NOCA-1 isoforms . Markers are in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00510 . 7554/eLife . 08649 . 006Figure 1—figure supplement 3 . Construction of a deletion allele for the gene encoding C . elegans Patronin , PTRN-1 . Schematic showing the strategy used to generate a null ptrn-1 deletion allele . Briefly , a double-stranded break was generated by injecting a plasmid-expressing Mos1 transposase into a strain with a Mos1 transposon insertion in a ptrn-1 intron ( ttTi21011 ) . A repairing plasmid was co-injected with the transposase to induce homology-based repair that resulted in deletion of the majority of ptrn-1 coding sequence , including the transcription start site . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00610 . 7554/eLife . 08649 . 007Figure 1—figure supplement 4 . Hydrodynamic analysis of purified NOCA-1 and PTRN-1 proteins . ( A ) Coomassie blue stained gel of recombinant NOCA-1 proteins purified from baculovirus-infected insect cells . ( B ) Coomassie blue stained gels of purified GFP fusions with DmPatronin , PTRN-1 , NOCA-1LICR+NHD , and NOCA-1NHD subjected to sucrose gradient sedimentation ( left ) or gel filtration ( right ) . ( C ) Plots of fraction number vs sedimentation coefficient ( left ) or elution time vs Stokes radius ( right ) for the standards used to estimate the S value and Stokes radius of the test proteins . R2 is the coefficient of determination for linear regression . ( D ) The sedimentation coefficient ( S ) and Stokes radius ( Rs ) for each test protein were estimated using the standard curves in ( B ) , and the molecular weight was calculated by Mw = 4205 × S × Rs ( Siegel and Monty , 1966 ) . The ‘Monomer Mw’ was calculated from the amino acid sequence of each protein . The Smax is the S value assuming the protein is a smooth sphere , calculated by Smax = 0 . 00361 × M2/3 . The ratio Smax/S for all test proteins are ∼2 . 0 , indicating these proteins are moderately elongated in solution ( Erickson , 2009 ) . ( E ) Left: Coomassie blue-stained gel of recombinant MBP fusions purified from baculovirus-infected insect cells . Right: elution profiles of purified proteins on a Superose 6 size-exclusion column in H100 buffer ( 25 mM Hepes-NaOH pH 7 . 6 , 100 mM NaCl and 1 mM DTT ) . ( F ) Coomassie blue-stained gel of supernatant ( S ) or pellet ( P ) samples from a microtubule co-sedimentation assay performed in H100 buffer . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00710 . 7554/eLife . 08649 . 008Figure 1—figure supplement 5 . Purified NOCA-1 binds to microtubules in aggregated forms . ( A ) Left: schematic of flow-cell based microtubule anchoring assay . Middle: maximum intensity projections of time lapse fluorescence confocal images of microtubules bound by the indicated proteins . Right: plots of frequencies of microtubule side and end binding . ( B ) Top: schematic of the kinesin gliding assay ( left ) and the plot of frequency of microtubules ( MTs ) decorated with GFP puncta . Center: fluorescence confocal images of the coverslip surface of flow cells containing rhodamine-labeled microtubules ( red ) and 60 nM GFP-fused test proteins ( green ) . BRB80 buffer: 80 mM Pipes-KOH pH 6 . 8 , 1 mM MgCl2 , and 1 mM EGTA . Bottom: kymographs showing microtubule gliding in the presence of indicated GFP tagged proteins . Scale bars , 10 μm or as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 008 NOCA-1 isoforms can be partitioned into two groups based on their sequence features: three short isoforms ( d , e and g ) that contain the NHD , and five long isoforms ( a , b , c , f and h ) that contain the NHD as well as an additional 205 shared amino acids that we will refer to as the Long Isoform Common Region ( LICR ) . Each isoform also has a unique N-terminal extension ( Figure 1A , rainbow colors ) that varies in length from 18 to 251 amino acids . Thus , all NOCA-1 isoforms contain a common C-terminal domain with homology to the central ninein-specific region of vertebrate ninein . To examine the in vivo functions of NOCA-1 , we began by analyzing the phenotype of a noca-1 deletion that affects all isoforms by removing the NHD ( ok3692; Figure 1A ) . Immunoblotting with an antibody against the NHD coiled-coil recognized four major species that were absent or strongly reduced in extracts from noca-1∆ and noca-1 ( RNAi ) worms ( Figure 1B and Figure 1—figure supplement 2 ) , indicating that at least four isoforms are expressed at detectable levels . Consistent with our prior work ( Green et al . , 2011 ) , noca-1∆ worms were sterile and exhibited germline phenotypes equivalent to γ-tubulin depletion confirming that NOCA-1 has an essential role in assembly of the germline microtubule array . However , aside from germline abnormalities , noca-1∆ adult worms appeared morphologically normal and did not exhibit motility defects ( Figure 1D and Video 1 ) . 10 . 7554/eLife . 08649 . 012Video 1 . NOCA-1 and PTRN-1 redundantly perform a function essential for larval development . Worms with the indicated genotypes were filmed using an eyepiece camera ( DinoEye ) mounted on a dissection scope 72 hr after release from a synchronized L1 stage . Playback is 2× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 012 We found it surprising that deletion of NOCA-1 , which has eight isoforms and a critical role in the germline , had such a limited effect on development . Since NOCA-1 has homology to ninein , which has been proposed to anchor microtubules at centrosomes ( Mogensen et al . , 2000; Delgehyr et al . , 2005 ) , we considered whether it might function redundantly with Patronin , another microtubule minus end-associated protein . To test this , we used a transposon-based method to generate a null mutant in ptrn-1 , which encodes the only C . elegans Patronin family member ( Figure 1C and Figure 1—figure supplement 3; Frøkjær-Jensen et al . , 2010; Chuang et al . , 2014 ) . A polyclonal antibody against the PTRN-1 C-terminus recognized a single band of ∼130 kD that was absent in ptrn-1∆ worms ( Figure 1C ) . Like noca-1∆ worms , ptrn-1∆ worms appeared morphologically normal ( Chuang et al . , 2014; Marcette et al . , 2014; Richardson et al . , 2014; Figure 1D and Video 1 ) . However , in contrast to noca-1∆ worms , ptrn-1∆ worms were fertile , indicating that PTRN-1 function is not required in the germline . In striking contrast to the two single mutants , noca-1∆; ptrn-1∆ worms exhibited severe developmental defects . Double mutant worms grew slowly , and ∼60% ruptured and died during the first 3 days of post-embryonic development , largely at L4 and early adult stages ( Figure 1D–F and Video 1 ) . The 40% that survived were small and uncoordinated ( Small Unc; Figure 1E ) . We conclude that NOCA-1 and PTRN-1 are redundantly required for larval development and viability . Patronin family members bind to microtubule minus ends ( Meng et al . , 2008; Goodwin and Vale , 2010; Hendershott and Vale , 2014; Jiang et al . , 2014 ) . To determine if this is also true for C . elegans PTRN-1 , we expressed and purified recombinant GFP fusions with full-length PTRN-1 and DmPatronin , as a control , from insect cells ( Figure 1G ) . Employing a kinesin gliding assay to define polarity at physiological ionic strength ( 100 mM KCl ) , we observed puncta of GFP::PTRN-1 and GFP::DmPatronin at the leading end of gliding microtubules , indicating binding to minus ends ( Figure 1G ) . Thus , C . elegans PTRN-1 possesses the minus end recognition activity predicted based on its homology to Patronin family proteins . Both NOCA-1 and PTRN-1 were detected in the pellet after microtubule sedimentation from C . elegans extracts ( Figure 1H ) , indicating that NOCA-1 possesses either a direct or indirect microtubule-binding activity . To determine if purified NOCA-1 binds directly to microtubules , we purified GFP-tagged NOCA-1NHD and NOCA-1LICR+NHD from insect cells ( Figure 1—figure supplement 4A ) . Hydrodynamic analysis in 500 mM salt indicated that both NOCA-1 fusions were dimeric , whereas GFP-tagged PTRN-1 and DmPatronin were monomeric ( Figure 1—figure supplement 4B–D ) . Unfortunately , lowering the ionic strength to physiological levels caused both NOCA-1LICR+NHD and NOCA-1NHD to precipitate . Adding detergents or stabilizers , such as glycerol or sucrose , did not circumvent this problem; however , we were able to generate an MBP::NOCA-1NHD::GFP fusion that was soluble at physiological ionic strength . While the ability of this soluble fusion to co-sediment with microtubules was negligible ( Figure 1—figure supplement 4E , F ) , we did observe that aggregated forms of NOCA-1 fusion proteins associated with microtubules . When small aggregates of GFP::NOCA-1NHD or GFP::NOCA-1LICR+NHD were analyzed in a coverslip-anchorage assay , analogous to that performed previously for Patronin ( Goodwin and Vale , 2010; Figure 1—figure supplement 5A ) , they anchored microtubules by binding to their ends ( Figure 1—figure supplement 5A ) . Similarly , dilution of MBP::NOCA-1NHD::GFP into a classical microtubule assay buffer caused it to form small aggregates that bound along the lengths of microtubules ( Figure 1—figure supplement 5B ) . These results hint that NOCA-1 may associate directly with microtubules , although significant additional work will be necessary to overcome the limitations imposed by the low solubility of purified NOCA-1 in order to rigorously assess microtubule interactions in vitro . The failure of larval development in the noca-1∆; ptrn-1∆ double mutant indicated that NOCA-1 and PTRN-1 function in parallel to promote larval growth and morphogenesis . Mitotic spindle assembly in the early embryo and embryonic viability were not affected by either single or double inhibitions of NOCA-1 and PTRN-1 ( Figure 2—figure supplement 1 ) , indicating that their redundant function essential for larval development is likely in a differentiated tissue , and not in the formation of centrosomal microtubule arrays required for cell division . To identify this tissue , we expressed PTRN-1::GFP under different tissue-specific promoters . PTRN-1::GFP expressed from its endogenous promoter ( Pptrn-1 ) rescued the synthetic lethality of the noca-1∆; ptrn-1∆ double mutant ( Figure 2A ) and localized in multiple tissues , including the larval/adult epidermis , neurons , intestine , and pharynx ( Figure 2—figure supplement 2 ) . Selective expression of PTRN-1::GFP in the larval/adult epidermis ( Pdpy-7 ) rescued the lethality and morphology/movement phenotypes of noca-1∆; ptrn-1∆ mutants , whereas no rescue was observed following expression in neurons ( Prgef-1 ) or the pharynx and intestine ( Ppha-4 ) ( Figure 2A and Figure 2—figure supplement 2 ) . Transgenes encoding the NOCA-1d and e isoforms or only the d isoform expressed from their endogenous promoters rescued larval development in noca-1∆; ptrn-1∆ worms , whereas transgenes encoding the abcfgh ( Figure 2—figure supplement 3 ) or e isoforms did not . The short NOCA-1d isoform consists of the NHD and a short unique N-terminal extension ( Figure 1A ) . The N-terminal extension was not required for function , since expression of an NHD::GFP fusion under the Pptrn-1 promoter was sufficient to rescue the double mutant phenotype ( Figure 2—figure supplement 3 ) . These results indicate that the NHD of NOCA-1 is sufficient to function redundantly with PTRN-1 in the larval/adult epidermis to support organismal growth and development . 10 . 7554/eLife . 08649 . 016Figure 2 . NOCA-1 and PTRN-1 control assembly of a circumferential microtubule array required for the integrity of the larval/adult epidermis . ( A ) Left: plots of the percentage of normal-sized adults , small uncs , and dead worms 72 hr post L1 for noca-1∆; ptrn-1∆ worms-expressing PTRN-1::GFP under the control of the indicated promoters or with noca-1 transgenes directing expression of the indicated isoforms from their own promoters . n is number of worms analyzed in 3–5 independent experiments . Right: schematics of noca-1 transgenes . Note that the data for noca-1∆; ptrn-1∆ worms in both plots are the same as in Figure 1E . ( B ) Left: schematics illustrating the organization of the larval epidermis . The body epidermis ( gold in 3D view ) is a large , thin multinucleated syncytial cell that covers the majority of the worm's body; rows of seam cells ( pink ) are embedded within the body epidermis in rows that run along either side of the worm . Right: maximum intensity projection of fluorescence confocal image of GFP::β-tubulin and mCherry::Histone in the larval epidermis of an L3 stage worm ( n = 20 ) . ( C ) Schematic and fluorescence confocal images of L3 stage worms of the indicated genotypes expressing GFP::β-tubulin . Right: plot of microtubule bundle density in worms of the indicated genotypes . ( D ) Left: fluorescence confocal images of L3 stage worms expressing EB1::GFP . Right top: schematic of the imaged region . Right bottom: plots of EB1 comet density and microtubule growth rate in worms of the indicated genotypes . ( E ) Top: schematic of early adult worm expressing DLG-1::GFP , which marks the junctions between the body epidermis and the seam cell syncytia . Bottom: fluorescence confocal images of control and noca-1∆; ptrn-1∆ worms expressing DLG-1::GFP . ( F ) Left: schematic of the permeability assay . Right: DIC and fluorescence images of worms after treatment with Hoechst . Statistics , one-way ANOVA followed by Dunnett's multiple comparisons test . p-values are the probability of obtaining the observed results assuming the test group is the same as control . Error bars are SEM . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01610 . 7554/eLife . 08649 . 017Figure 2—figure supplement 1 . Both NOCA-1 and PTRN-1 are dispensable for mitotic divisions . ( A ) Schematic and fluorescence confocal images of the spindle at early anaphase of the first mitotic division in C . elegans embryos from the indicated genotypes . ( B ) Plot of percent embryonic lethality for hermaphrodites with indicated perturbations . N = number of worms , n = number of embryos . The low level lethality observed for noca-1 ( RNAi ) embryos is likely a secondary consequence of the effect of NOCA-1 depletion on germline structure . ( C ) Plot of the duration of the first C . elegans embryonic division for the indicated perturbations . NEBD , nuclear envelope break-down . n is the number of scored embryos . Statistics , one-way ANOVA followed by Dunnett's multiple comparisons test . p-values indicate the probability of obtaining the observed or more extreme results assuming the test group is the same as control . Error bars are SEM . Scare bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01710 . 7554/eLife . 08649 . 018Figure 2—figure supplement 2 . Expression of PTRN-1::GFP in multiple tissues . Schematic and fluorescence confocal images of PTRN-1::GFP expressed from transgenes under the control of the endogenous ptrn-1 promoter ( left panels , Pptrn-1 , n = 9 ) or promoters specific for the larval/adult epidermis ( Pdpy-7; n = 7 ) , the pharynx and intestine ( Ppha-4; n = 20 ) or neurons ( Prgef-1; n = 7 ) . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01810 . 7554/eLife . 08649 . 019Figure 2—figure supplement 3 . NOCA-1 immunoblot in lysate from noca-1Δ worms expressing noca-1abcfgh and the NOCA-1d isoform-specific region is dispensable for its function in the larval epidermis . ( A ) Left: schematics of the noca-1 locus and noca-1abcfgh transgene . Right: immunoblot of NOCA-1 in lysates from control worms and noca-1Δ worms expressing noca-1abcfgh . Markers are in kDa . ( B ) Left: schematic illustrating the analyzed truncation . Right: plot of percentage of normal-sized adults , small uncs , and dead worms 72 hr post L1 for the indicated genotypes . n is number of worms analyzed in 3–5 independent experiments . Note that the data for noca-1∆; ptrn-1∆ worms shown for comparison is the same as that in Figures 1E , 2A . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01910 . 7554/eLife . 08649 . 020Figure 2—figure supplement 4 . Illustration of seam cell fusion event at mid-L4 stage . Schematic and fluorescence confocal images of worms expressing the epithelial junction marker DLG-1::GFP before ( n = 13 ) and after ( n = 13 ) the seam cells fuse at the mid-L4 stage . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 02010 . 7554/eLife . 08649 . 021Figure 2—figure supplement 5 . Time course of larval permeability in control and mutant backgrounds . ( A ) DIC and fluorescent images of worms after incubation in Hoechst dye . ( B ) Plot of percentage of permeable worms for the indicated genotypes at the indicated developmental stages . n is the number of analyzed worms . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 02110 . 7554/eLife . 08649 . 022Figure 2—figure supplement 6 . Microtubule bundles in the post-embryonic epidermis co-align with cuticle annuli . Top: schematic of the microtubule array in the post-embryonic epidermis . Bottom left: DIC image showing the cuticle annuli and fluorescence confocal image showing the microtubule bundles . Bottom right: line-scan plots of the center panel images . 16 worms between the L3 and adult stages were imaged . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 022 The larval/adult epidermis ( the worm's ‘skin’ ) is composed of a single , multinuclear syncytial cell ( hyp7 ) that covers the majority of the worm's body ( gold in 3D schematic in Figure 2B ) . Embedded in this cell are two lateral rows of seams cells that run along either side of the worm's body . The seam cells fuse to form syncytia at the mid-L4 stage ( Chisholm and Hsiao , 2012; Figure 2B and Figure 2—figure supplement 4 ) . Other syncytial cells cover the head and tail . We visualized the microtubule array in the syncytial epidermis by co-expressing GFP::β-tubulin and mCherry::histone under control of the dpy-7 promoter ( Figure 2B ) . As previously reported ( Priess and Hirsh , 1986; Costa et al . , 1997 ) , the epidermal microtubule array is composed of regularly spaced circumferential bundles that appear as lines perpendicular to the larva/worm body axis in longitudinal sectional views ( Figure 2B ) . The density of microtubule bundles along the length of the worm was not significantly different from controls in the noca-1∆ mutant and was only slightly reduced in the ptrn-1∆ mutant ( Figure 2C and Video 2 ) . In contrast , significantly fewer microtubule bundles were observed in the noca-1∆; ptrn-1∆ double mutant ( Figure 2C and Video 2 ) . We conclude that NOCA-1 and Patronin/PTRN-1 redundantly control the assembly of a circumferential microtubule array required for larval development . 10 . 7554/eLife . 08649 . 023Video 2 . NOCA-1 and PTRN-1 function in parallel to control microtubule array formation in the larval epidermis . Timelapse fluorescence confocal microscopy was used to acquire images of the head epidermal region of control , noca-1Δ , ptrn-1Δ , and noca-1Δ; ptrn-1Δ worms expressing GFP::β-tubulin . Images were acquired at 1 s intervals . Playback is 6× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 023 To investigate the impact of NOCA-1 and PTRN-1 on microtubule dynamics , we took advantage of the fact that similarly structured microtubule arrays form in the larval epidermis in the presence of NOCA-1 only , PTRN-1 only , or in the presence of both proteins ( Figure 2C ) , and imaged microtubules and growing microtubule ends marked by EB1 comets ( Akhmanova and Steinmetz , 2008 ) . When only PTRN-1 was present ( noca-1∆ ) , microtubules appeared less dynamic than in wild type , whereas microtubules exhibited apparently normal dynamics when only NOCA-1 was present ( ptrn-1∆; Video 2 ) . Consistent with this impression , the density of EB1 comets was substantially reduced when only PTRN-1 was present ( noca-1∆ ) but was comparable to controls when only NOCA-1 was present ( ptrn-1∆; Figure 2D and Video 3 ) . EB1 signal was observed along the lattice of the bundles and only occasionally in comets when only PTRN-1 was present , possibly due to the reduced number of growing microtubule ends . The microtubule growth rate , measured by tracking of EB1 comets , was also reduced by ∼20% compared to controls in worms expressing PTRN-1 only ( noca-1∆ ) but not in worm expressing NOCA-1 only ( ptrn-1∆; Figure 2D ) . These results suggest that although either NOCA-1 or PTRN-1 can support the assembly of a circumferential microtubule array in the larval epidermis , the presence of NOCA-1 makes the arrays significantly more dynamic . 10 . 7554/eLife . 08649 . 024Video 3 . NOCA-1 makes the microtubule arrays in the larval/adult epidermis more dynamic . Timelapse fluorescence confocal microscopy was used to acquire images of the dorsal or ventral side of larval body epidermis in control , noca-1Δ , and ptrn-1Δ worms expressing EB1::GFP ( marks growing microtubule ends ) . Images were acquired at 1-s intervals . Playback is 6× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 024 To determine if the circumferential microtubule array maintains the structure of the epidermis , we analyzed two features in noca-1∆; ptrn-1∆ double mutants: localization of the apical junction marker DLG-1::GFP ( McMahon et al . , 2001 ) and integrity of the cuticle , which is secreted by the epidermis to function as an environmental barrier ( Page and Johnstone , 2007 ) . DLG-1::GFP outlines the junctions between the body epidermis and the seam cell syncytia that are embedded along the left and right sides of the worm ( Figure 2—figure supplement 4 ) . In wild-type worms , parallel lines of DLG-1::GFP are observed running along the entire body length . In contrast , in noca-1∆; ptrn-1∆ double mutants , seam cell syncytia were frequently branched/broken ( 71%; n = 17 ) as well as disconnected from the head epidermis ( 68%; n = 22; Figure 2E ) . This result suggests that the circumferential microtubule array in the body epidermis could have a role in positioning the seam cells prior to fusion . However , since the Pdpy-7 promoter also directs expression in the seam cells , we also cannot rule out that the fusion defect results from direct effects on the seam cells or their capacity to fuse . In addition to seam cell defects , the cuticles of noca-1∆; ptrn-1∆ mutant worms became permeable to the normally excluded Hoechst dye , beginning ∼24 hr after the L1 larval stage ( Figure 2F and Figure 2—figure supplement 5 ) . These defects in the epidermis and cuticle likely underlie the rupture phenotype with extrusion of internal tissues observed in noca-1∆; ptrn-1∆ mutant worms ( Figure 1D ) . We conclude that NOCA-1 and PTRN-1 function in parallel to promote the assembly of a circumferential array of microtubule bundles that is required for the morphology and integrity of the larval/adult epidermis . Our analysis placed NOCA-1 and PTRN-1 in parallel pathways controlling assembly of an essential circumferential microtubule array in the embryonic epidermis . Imaging of a GFP fusion with NOCA-1 in the larval epidermis revealed that it had a localization pattern very similar to that of γ-tubulin; NOCA-1 and γ-tubulin were both observed in puncta in the epidermal syncytium where the microtubule bundles are present ( Figure 3A , magnified insets ) and also concentrated along the junctions between the epidermal body syncytium and the seam cells ( Figure 3A , B ) . The localization pattern of PTRN-1::GFP was distinct . Consistent with prior work ( Jiang et al . , 2014 ) PTRN-1::GFP was observed in stretches as well as puncta in the body syncytium . PTRN-1::GFP was also observed in puncta within the seam cells but did not accumulate along the seam cell junctions . In double label images of NOCA-1d::GFP or PTRN-1::GFP with tagRFP::β-tubulin , many puncta of both proteins were observed coincident with the microtubule bundles in the body epidermis ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 08649 . 013Figure 3 . The γ-tubulin complex functions coordinately with NOCA-1 and in parallel to PTRN-1 to promote larval development and viability . ( A ) Top: schematic of the imaged region . Bottom: fluorescence confocal images of L3 stage worms expressing NOCA-1::GFP ( n = 27 ) , γ-tubulin::GFP ( n = 6 ) , or PTRN-1::GFP ( n = 17 ) . Insets below are magnified eightfold . Arrowheads point to examples of stretches observed in worms expressing PTRN-1::GFP . Note that the vertical lines in the images are cuticle auto-fluorescence due to high laser power and long exposure times required to visualize the GFP puncta/stretches . ( B ) Top: schematic of the imaged region . Bottom: fluorescence confocal images of L3-stage worms co-expressing NOCA-1d::GFP ( n = 12 ) or PTRN-1::GFP ( n = 4 ) with γ-tubulin::mCherry . ( C ) Schematic outlining the method used to specifically degrade the essential γ-tubulin complex component GIP-2::GFP in the epidermis . ( D ) Top: schematics and fluorescence confocal images of L4 stage worms expressing GIP-2::GFP with or without Pdpy-7::GFP nanobody::ZIF-1 ( epiDEG ) . Bottom: plots of normalized GIP-2::GFP fluorescence intensity in the epidermis or germline from worms with indicated genotypes . ( E ) Images of control and mutant worms 72 hr post L1 recovery ( snapshots from Video 4 ) . Arrowheads mark dead worms . ( F ) Plot of percentage of normal-sized adults , larval arrest , small uncs , and dead worms 72 hr post L1 for the indicated genotypes . n is total number of worms analyzed in 1 ( control ) , 2 ( gip-2::gfp; epiDEG and gip-2::gfp; epiDEG ;noca-1∆ ) , or 3 ( gip-2::gfp; epiDEG; ptrn-1∆ ) independent experiments . ( G ) Plots of body length ( left ) and % living worms ( right ) vs time for worms with the indicated genotypes . ( H ) Schematic describing two parallel pathways for assembly of a functional microtubule array in larval epidermis . Statistics , Student's t-test . p-values are the probability of obtaining the observed results assuming the test group is the same as control . Error bars are SEM . Scale bars , 10 µm or as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01310 . 7554/eLife . 08649 . 014Figure 3—figure supplement 1 . NOCA-1 and PTRN-1 localize along microtubules in the larval epidermis . Schematic showing the imaged region and fluorescence confocal images of NOCA1d::GFP ( n = 23 ) or PTRN-1::GFP ( n = 10 ) along with tagRFP::β-tubulin in the epidermis of L3 stage worms . Both NOCA-1d and PTRN-1 are observed in puncta ( sometimes PTRN-1 as short stretches ) , many of which co-localize with microtubule bundles . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01410 . 7554/eLife . 08649 . 015Figure 3—figure supplement 2 . Strategy to selectively inhibit the γ-tubulin complex in the larval/adult epidermis of C . elegans . To selectively inhibit the γ-tubulin complex in the larval epidermis , we used a CRISPR/Cas9-based approach to insert a sequence encoding GFP downstream of the gip-2 gene at its endogenous locus on Chr I and introduced the ‘epiDEG’ transgene encoding a fusion of an anti-GFP nanobody with ZIF-1 , which recruits the target to a Cullin2-based ubiquitin ligase , on Chr II . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 015 Given their similar localization patterns and the fact that knockdown of NOCA-1 and γ-tubulin resulted in an essentially identical defect in the germline ( Green et al . , 2011 and Figure 4 below ) , we wanted to test whether γ-tubulin functioned in microtubule generation pathways with NOCA-1 , PTRN-1 , or both in the larval epidermis . Since γ-tubulin is essential for cell division , analyzing its role in the larval epidermis required eliminating γ-tubulin function after the tissue is already formed . To achieve this , we developed a method based on two previously described protein degradation methods ( Caussinus et al . , 2012; Armenti et al . , 2014 ) for tissue-specific degradation of a functional GFP-fused target protein . Since fluorescently tagged γ-tubulin fusions were not fully functional ( not shown ) , we used a CRISPR/Cas9-mediated strategy ( Dickinson et al . , 2013 ) to insert a C-terminal GFP tag in the endogenous locus of gip-2 ( Figure 3—figure supplement 2 ) , which encodes an essential component of the C . elegans γ-tubulin complex ( Hannak et al . , 2002 ) . Endogenously tagged GIP-2 fully supported the essential functions of the γ-tubulin complex , as indicated by the normal development of worms homozygous for the insertion . To specifically degrade GIP-2::GFP in the epidermis , we expressed a GFP nanobody::ZIF-1 fusion under an epidermal promoter ( Pdpy-7; Figure 3—figure supplement 2 ) . This fusion , which we call epiDEG , serves as a GFP-to-ligase adapter that recognizes GFP-fused target proteins and brings them to the ECS ( Elongin-C , Cul2 , SOCS-box family ) E3 ubiquitin ligase complex for ubiquitination and proteasome-mediated degradation ( Figure 3C; DeRenzo et al . , 2003 ) . Quantification revealed that the GIP-2::GFP signal in the larval epidermis was reduced by >80% compared to controls in epiDEG worms whereas the signal in the germline was unaffected ( Figure 3D ) , indicating the GFP-mediated degradation is efficient and tissue specific . The gip-2::gfp; epiDEG animals grew slightly slower than wild-type worms and a small percentage of them were arrested at early larval stage ( Figure 3E–G and Video 4 ) , possibly because the dpy-7 promoter-driven epiDEG may cause some degradation of GIP-2::GFP in the dividing seam cells . However , the majority of gip-2::gfp; epiDEG animals exhibited normal development . 10 . 7554/eLife . 08649 . 009Figure 4 . NOCA-1 isoform h functions in the germline to assemble a non-centrosomal microtubule array for nuclear positioning . ( A ) Left: schematic showing the germline and location of the imaged region . Middle: fluorescence confocal image of the germline in a worm expressing GFP::β-tubulin . Inset to the right is magnified 3 . 3-fold . Right: Schematic of the region highlighted in the inset , illustrating the organization of the microtubule arrays in the compartments that hold the nuclei near the cell surface and prevent them from falling into the rachis . ( B ) Left top: schematic illustrating the structure of the syncytial germline . Left bottom: fluorescence confocal images of germlines in control ( n = 14 ) , γ-tubulin ( RNAi ) ( n = 7 ) , noca-1∆ ( n = 12 ) , and ptrn-1∆ ( n = 11 ) worms expressing a GFP-tagged plasma membrane marker and mCherry-tagged histone H2B . Frequencies of disorganized germlines with nuclei falling out of their compartments were 100% in γ-tubulin ( RNAi ) and noca-1Δ worms and 0% in control and ptrn-1∆ worms . Right: plot of brood size for worms of the indicated genotypes . ( C ) Left: schematic illustrating microtubule organization in the germline . Right: fluorescence confocal images of germlines in control ( n = 22 ) , γ-tubulin ( RNAi ) ( n = 10 ) , noca-1Δ ( n = 13 ) and ptrn-1Δ ( n = 7 ) worms expressing GFP::β-tubulin . Frequencies of the nuclear fall-out phenotype were 100% in γ-tubulin ( RNAi ) and noca-1Δ worms and 0% in control and ptrn-1Δ worms . ( D ) Left: schematic showing the location of the imaged region . Middle: fluorescence confocal images of growing microtubule ends marked by EB1::GFP in the germline . Right: plot of EB1 comet density in worms depleted of the indicated proteins by RNAi . ( E ) Left: immunoblot of NOCA-1 in lysates from control and noca-1h ( RNAi ) worms . Middle: fluorescence confocal images of germlines in control ( n = 13 ) and noca-1h ( RNAi ) ( n = 10 ) worms expressing a GFP-tagged plasma membrane marker and mCherry::histone . Frequencies of disorganized germlines with nuclear fallout were 100% in noca-1h ( RNAi ) and 0% in control worms . Right: plot of brood size for control and noca-1h ( RNAi ) worms . ( F ) Top: schematic illustrating the RNAi-resistant noca-1h::gfp transgene . Bottom: brood size plot for worms subjected to the indicated perturbations . ( G ) Left: schematic showing NOCA-1h and the two analyzed truncations . Germline expression was driven by the noca-1h promoter . Middle: immunoblot of lysates prepared from worms with the indicated genotypes . The asterisk marks a non-specific band . Right: Plot of brood size for worms subjected to indicated perturbations . Statistics in B and D , one-way ANOVA followed by Dunnett's multiple comparisons test . Statistics in E , F and G , Student's t-test . p-values are the probability of obtaining the observed results assuming the test group is the same as control . Error bars are SEM . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 00910 . 7554/eLife . 08649 . 010Figure 4—figure supplement 1 . Ectopic germline expression of PTRN-1 does not substitute the germline function of NOCA-1 . ( A ) Schematic and fluorescence confocal images of the germline in control ( n = 6 ) , noca-1 ( RNAi ) ( n = 5 ) , and noca-1Δ ( n = 6 ) worms ectopically expressing PTRN-1::GFP under control of the Pnoca-1h germline promoter . ( B ) Plot of brood size for worms with indicated perturbations . Statistics , one-way ANOVA followed by Dunnett's multiple comparisons test . p-values indicate the probability of obtaining the observed or more extreme results assuming the test group is the same as control ( no transgene , no RNAi ) . Error bars are SEM . Scare bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01010 . 7554/eLife . 08649 . 011Figure 4—figure supplement 2 . Sequence of the RNAi-resistant region of the NOCA-1h::GFP transgene and localization of NOCA-1h::GFP in the germline . ( A ) Schematic of the RNAi-resistant NOCA-1h::GFP single-copy transgene showing the sequence of the RNAi-resistant region . The codons in the RNAi-resistant region were shuffled to prevent targeting of the transgene by dsRNA directed against the corresponding region of the endogenous gene , while maintaining amino acid sequence . Capital letters indicate the altered codons . ( B ) Left: schematic showing the location of the imaged region of the germline . Right: fluorescence confocal image showing the membrane localization of NOCA-1h::GFP ( n = 5 ) . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 01110 . 7554/eLife . 08649 . 025Video 4 . Epidermal degradation of GIP-2::GFP synergizes with ptrn-1Δ but not noca-1Δ . Worms with the indicated genotypes were filmed using an eyepiece camera ( DinoEye ) mounted on a dissection scope 72 hr after release from a synchronized L1 stage . Playback is 2× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 025 Having established a method to selectively degrade an essential γ-tubulin complex subunit in the larval epidermis , we tested whether this perturbation of γ-tubulin enhanced the noca-1∆ or ptrn-1∆ phenotypes . We found that noca-1∆; gip-2::gfp; epiDEG animals exhibited the same mild phenotypes observed in gip-2::gfp; epiDEG animals . In contrast , more than 70% of ptrn-1∆; gip-2::gfp; epiDEG; animals ruptured and died at late L4 to early adult stages ( Figure 3E–G and Video 4 ) . The 30% survivors were mostly small and uncoordinated or arrested as larva . This striking difference between the effects of inhibiting the γ-tubulin complex in the noca-1∆ and ptrn-1∆ mutants suggests that the γ-tubulin complex functions together with NOCA-1 and in parallel to PTRN-1 for non-centrosomal microtubule array generation in the larval epidermis ( Figure 3H ) . A major phenotypic difference between noca-1∆ and ptrn-1∆ worms is that the former are sterile whereas the latter are fertile ( Figure 4B ) . The C . elegans germline is a syncytial structure in which nuclei in various stages of meiotic prophase are housed in membrane-bound compartments that are open on one side towards the common cytoplasmic core , called the rachis . Non-centrosomal microtubule arrays assemble within the compartments that hold the nuclei near the surface and prevent them from dropping into the rachis ( Figure 4A; Zhou et al . , 2009 ) . Within the rachis there are also microtubules that flow with the streaming cytoplasm into the forming oocytes ( Wolke et al . , 2007 ) . Imaging germline architecture in worms expressing a GFP-tagged plasma membrane probe along with mCherry-histone or GFP::β-tubulin revealed that noca-1 deletion resulted in an essentially identical phenotype to γ-tubulin depletion; in both cases , nuclei fell out of their compartments and formed clumps in the rachis center , indicating a dramatic failure in the function of the microtubule arrays in the compartments ( Figure 4B , C ) . In contrast , germline structure in ptrn-1∆ worms was similar to that in controls ( Figure 4B , C ) . Since compartment structure collapsed as the nuclei fell into the rachis , we could not assess the impact of loss of NOCA-1 or γ-tubulin on the dynamics of the arrays within the compartments . However , we were able to measure the density of growing microtubule plus ends , measured as the number of EB1 comets , in a fixed area of the rachis , which was reduced to a similar extent by NOCA-1 and γ-tubulin inhibitions ( Figure 4D and Video 5 ) . 10 . 7554/eLife . 08649 . 026Video 5 . Depletion of γ-tubulin or NOCA-1 reduces growing microtubule ends in the germline . Timelapse fluorescence confocal microscopy was used to acquire images of a central plane of the pachytene region of the germline in worms expressing EB1::GFP ( marks growing microtubule ends ) . Images of control , γ-tubulin ( RNAi ) , and noca-1 ( RNAi ) worms were acquired at 1-s intervals . Playback is 6× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 026 Consistent with lack of an effect of ptrn-1∆ , PTRN-1 is not expressed in the germline ( not shown ) ; in addition , forcing PTRN-1 expression in the germline did not rescue noca-1∆ sterility ( Figure 4—figure supplement 1 ) . Selectively depleting the longest NOCA-1 isoform ( NOCA-1h ) using a dsRNA targeting its unique N-terminal extension disrupted germline architecture and led to sterility ( Figure 4E ) , and expressing NOCA-1h from an RNAi-resistant transgene under its own promoter rescued both phenotypes ( Figure 4F and Figure 4—figure supplement 2 ) , indicating that NOCA-1h is both necessary and sufficient for germline function . Expression of a NOCA-1 truncation that included the NHD and the long isoform common region ( NOCA-1NHD+LICR ) under the same promoter also rescued the effects of depleting NOCA-1h on the germline , whereas expression of the NHD alone did not ( NOCA-1NHD; Figure 4G ) . Thus , in the germline , NOCA-1 function requires the LICR in addition to the NHD , but the h isoform specific region is not essential . We conclude that , in the germline , γ-tubulin and NOCA-1h act independently of PTRN-1 to direct assembly of non-centrosomal microtubule arrays that position nuclei . In the germline , NOCA-1h co-localizes with γ-tubulin to the surface of the compartments but does not co-localize with γ-tubulin at centrosomes ( Figure 5A , B , red arrows point to centrosomes ) . This result raised the possibility that NOCA-1 promotes non-centrosomal microtubule array formation by recruiting γ-tubulin to the cell surface . We tested this possibility by imaging γ-tubulin::mCherry in noca-1∆ worms . Although compartment structure is disrupted in noca-1∆ worms , γ-tubulin::mCherry was still clearly observed on the compartment surfaces indicating that NOCA-1 is not required to recruit γ-tubulin to that location ( Figure 5C ) . 10 . 7554/eLife . 08649 . 027Figure 5 . γ-tubulin-dependent and independent mechanisms target NOCA-1 to the plasma membrane in the germline . ( A ) Left: schematic of region imaged in A–F . Right: fluorescence confocal images of the germline in worms co-expressing NOCA-1::GFP and γ-tubulin::mCherry ( n = 10 ) . Arrow points to a centrosome . ( B ) Fluorescence confocal images of a germline in a worm co-expressing GFP::SPD-5 ( a centrosome marker ) and γ-tubulin::mCherry ( n = 13 ) . Arrows point to centrosomes . ( C ) Fluorescence confocal images of γ-tubulin::mCherry in the germline of control ( n = 11 ) and noca-1∆ ( n = 8 ) worms . ( D ) Fluorescence confocal images of the germline in control ( n = 16 ) and γ-tubulin ( RNAi ) ( n = 10 ) worms co-expressing NOCA-1h::GFP and an mCherry-tagged plasma membrane marker . ( E ) Fluorescence confocal images of the germline from control ( n = 25 ) and γ-tubulin ( RNAi ) ( n = 23 ) worms expressing NOCA-1LICR+NHD::GFP and an mCherry-tagged plasma membrane marker . ( F ) Fluorescence confocal images of the germline in worms expressing NOCA-1hC10A::GFP and an mCherry-tagged plasma membrane marker that were depleted of endogenous NOCA-1 by RNAi . Images are shown for control worms ( n = 17 ) or worms that were also depleted of γ-tubulin ( n = 20 ) or α-tubulin ( n = 18 ) . ( G ) Top: schematic illustrating the RNAi-resistant NOCA-1hC10A::GFP transgene . Bottom: brood size plot for worms subjected to indicated perturbations . ( H ) Schematic summarizing the mechanisms that target NOCA-1h to the cell surface in the germline . Statistics , Student's t-test . p-values are the probability of obtaining the observed results assuming the test group is the same as control . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 02710 . 7554/eLife . 08649 . 028Figure 5—figure supplement 1 . The isoform specific region of NOCA-1h localizes to the plasma membrane through a putative palmitoylation . ( A ) Left: schematic of imaged region on the top . Right: fluorescence confocal images of worms expressing NOCA-1h ( 1-251 ) ::GFP with ( n = 13 ) or without ( n = 25 ) the C10A mutation . Scale bar , 10 μm . ( B ) Immunoblot of worms expressing indicated transgenes using GFP antibody . Markers are in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 028 We next tested if NOCA-1 required γ-tubulin to localize to the surface of germline compartments . As full-length NOCA-1h and NOCA-1LICR+NHD , which lack the h isoform-specific region , are both functional , we analyzed the localization of both in control and γ-tubulin-depleted germlines . Surprisingly , NOCA-1LICR+NHD required γ-tubulin to localize to compartment surfaces whereas full-length NOCA-1h did not ( Figure 5D , E ) . This result suggested that the non-essential isoform-specific region of NOCA-1h harbors a γ-tubulin-independent cell surface targeting activity ( Figure 5H ) . Consistent with this idea , a GFP fusion with the h isoform specific region localized to compartment surfaces , and this localization was dependent on a predicted palmitoylation site ( cysteine 10; Figure 5—figure supplement 1 ) . Mutation of this predicted palmitoylation site in the full-length protein ( NOCA-1hC10A ) did not compromise NOCA-1h function but rendered its localization γ-tubulin dependent ( Figure 5F–H , Figure 5—figure supplement 1 ) . This result explains why NOCA-1h localization at compartment surfaces was not eliminated by γ-tubulin depletion and implicates a potential lipid modification in providing a redundant means for NOCA-1 targeting to the membrane . γ-tubulin could direct NOCA-1 localization to the cell surface either through a direct interaction or indirectly through nucleated microtubules . To distinguish these two possibilities , we disrupted microtubule assembly by using RNAi to deplete α-tubulin . While this disrupted germline structure to a comparable extent to inhibition of NOCA-1 or γ-tubulin , cell surface targeting of NOCA-1hC10A was still observed ( Figure 5F ) . This result suggests that an interaction between NOCA-1NHD+LICR and the γ-tubulin complex may contribute to recruitment of NOCA-1 to the cell surface . However , we have not yet detected an interaction in immunoprecipitations from C . elegans extracts or yeast two-hybrid experiments with NOCA-1 and γ-tubulin complex components , indicating that additional work is needed to understand precisely how γ-tubulin promotes the cell surface recruitment of NOCA-1 . Based on these results , we conclude that the γ-tubulin complex recruits NOCA-1 to the cell surface , where they are both required to generate functional non-centrosomal microtubule arrays that position nuclei within compartments . Our prior work suggested that NOCA-1 is also involved in assembly of non-centrosomal microtubule arrays that position nuclei in the embryonic epidermis ( Green et al . , 2011 ) . Imaging of noca-1∆ mutant embryos expressing GFP::β-tubulin suggested a reduction in the number of microtubules in the embryonic epidermis ( Figure 6A ) . Consistent with this , EB1 comet density was also reduced ∼twofold in noca-1∆ embryos ( Figure 6B and Video 6 ) . The microtubule arrays in these cells have previously been implicated in nuclear migration ( Fridolfsson and Starr , 2010; Starr and Fridolfsson , 2010 ) . Defects in nuclear migration lead to the presence of nuclei in the larval dorsal cord , which is not observed in wild type ( Figure 6C; Fridolfsson and Starr , 2010 ) . A clear nuclear migration defect was observed in noca-1∆ mutants , whereas no defect was observed in ptrn-1∆ mutants ( Figure 6C ) . NOCA-1 co-localizes with γ-tubulin in the embryonic epidermis ( Figure 6D ) . Although we do not yet have a tissue-specific knockdown system to determine if γ-tubulin is required for assembly of this array , these results suggest that NOCA-1 functions with γ-tubulin independently of PTRN-1 in the embryonic epidermis as it does in the germline . 10 . 7554/eLife . 08649 . 029Figure 6 . NOCA-1 , but not PTRN-1 , is required for the function of a non-centrosomal microtubule array that positions nuclei in the embryonic epidermis . ( A ) Left: schematic showing the imaged region of the dorsal embryonic epidermis . Right: maximum intensity projections of fluorescence confocal images of the dorsal epidermis in control ( n = 4 ) and noca-1∆ ( n = 5 ) embryos expressing GFP::β-tubulin . Images were captured and displayed using identical settings . ( B ) Left and Middle: schematic and images of control ( n = 16 ) and noca-1∆ ( n = 10 ) embryos expressing EB1::GFP to mark growing microtubule ends . Right: plot of EB1 comet density in control and noca-1∆ embryos . ( C ) Left: schematic illustrating nuclear migration in the developing dorsal epidermis of C . elegans embryos . Right: plot of the number of nuclei in the dorsal cord for worms with indicated genotypes . ( D ) Left: schematic showing location of the imaged region . Right: images of C . elegans embryos co-expressing NOCA-1::GFP and γ-tubulin::mCherry ( n = 14 ) . ( E ) Left: schematic illustrating noca-1 transgenes expressing different isoform subsets . 2 . 4 kb of 5′ UTR and 1 . 2 kb of 3′ UTR were used in all transgenes . Right: plot of nuclei number in dorsal cord for worms with indicated genotypes . Plbp-1 is an epidermis specific promoter . Data for control and noca-1∆ are the same as in ( C ) . ( F ) Left: schematic of the two analyzed truncations . Embryonic epidermis expression was driven by Plbp-1 . Right: GFP immunoblot of worm lysates prepared from worms with indicated genotypes . ‘’ marks a non-specific band . ( G ) Plot of nuclei number in dorsal cord for worms with indicated genotypes . Note that data for control and noca-1∆ are the same as in ( C ) and ( E ) . Error bars are SEM . Statistics in C , E and G , one-way ANOVA followed by Dunnett's multiple comparisons test . Statistics in B , Student's t-test . p-values indicate the probability of obtaining the observed results assuming the test group is the same as control . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 02910 . 7554/eLife . 08649 . 033Video 6 . Deletion of NOCA-1 reduces growing microtubules in the embryonic epidermis . Timelapse fluorescence confocal microscopy was used to acquire images of the dorsal epidermis in C . elegans embryos expressing EB1::GFP ( Plbp-1::EB1::GFP ) . Images of embryos from control and noca-1 ( RNAi ) worms were acquired at 1 s intervals . Playback is 6× realtime . DOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 033 In the embryonic epidermis , a noca-1 transgene encoding the abcfgh isoforms ( Figure 2—figure supplement 3A ) rescued nuclear migration , whereas a comparable transgene with a stop codon that specifically blocks expression of the b isoform ( acfgh ) did not ( Figure 6E ) . Expression of the b isoform under an epidermal promoter rescued nuclear migration , identifying NOCA-1b as necessary and sufficient for NOCA-1 function in the embryonic epidermis ( Figure 6E ) . A truncation analysis revealed that although expression of NOCA-1NHD appeared to partially suppress the nuclear migration defect , expression of NOCA-1NHD+LICR was required for full rescue . We conclude that , as in the germline , a long NOCA-1 isoform that includes the LICR as well as the NHD is required to direct the PTRN-1-independent assembly of a functional non-centrosomal microtubule array that positions nuclei in the embryonic epidermis . A common region of all 8 NOCA-1 isoforms shares homology with a region of vertebrate ninein that has been implicated in microtubule anchoring at centrosomes ( Delgehyr et al . , 2005 ) ; this region is absent in the homologous ninein-like protein that is also present in vertebrates . Like NOCA-1 , ninein has been shown to re-localize to the cell surface during the assembly of non-centrosomal microtubule arrays in simple and stratified epithelia ( Mogensen et al . , 2000; Lechler and Fuchs , 2007; Moss et al . , 2007 ) , suggesting a role in the assembly of non-centrosomal microtubule arrays . Our results show that in the germline and embryonic epidermis NOCA-1 and γ-tubulin are required independently of Patronin/PTRN-1 , whereas in the larval/adult epidermis , the NOCA-1/γ-tubulin pathway and the Patronin-dependent pathway redundantly support microtubule generation ( Figure 7A ) . We expect that our analysis of NOCA-1 may inform studies of vertebrate ninein , mutations in which have been implicated in the human disorders microcephalic primordial dwarfism and spondyloepimetaphyseal dysplasia ( Dauber et al . , 2012; Grosch et al . , 2013 ) . The functional overlap between the ninein and Patronin families of microtubule cytoskeleton-associated proteins observed in the larval/adult epidermis may also aid future analysis of these two protein classes in vertebrates . Our data suggest that NOCA-1 functions together with γ-tubulin to promote the formation of non-centrosomal microtubule arrays in multiple tissues . We identified three NOCA-1 isoforms that are each necessary and sufficient to promote the assembly of different non-centrosomal microtubule arrays ( Figure 7A ) . This pattern suggests that the remaining five NOCA-1 isoforms will function with γ-tubulin in the assembly of microtubule arrays in tissues that we have not yet characterized; some of these may also act in parallel to PTRN-1 . Importantly , the isoform-specific regions were not essential for NOCA-1 function in the three different contexts analyzed , suggesting that these regions primarily reflect use of alterative promoters/transcriptional start sites . In the germline , the tissue-specific isoform region directed non-essential , γ-tubulin-independent membrane localization , potentially via palmitoylation of a cysteine residue in the extreme N-terminus . Whether this residue is indeed palmitoylated will need to be addressed in future work . In the tissues we analyzed , NOCA-1 co-localized with γ-tubulin ( except at centrosomes ) and NOCA-1 inhibition phenocopied inhibition of γ-tubulin , blocking the key functions of the arrays and leading to a similar reduction in the number of EB1-marked growing microtubule ends . In the germline , where we were able to analyze localization dependencies , γ-tubulin localized to the cell surface independently of NOCA-1 . Understanding how γ-tubulin is recruited to non-centrosomal sites is an important question , as SPD-5 , the major pericentriolar material matrix component that is thought to recruit γ-tubulin to centrosomes , is not recruited to non-centrosomal sites ( Figure 5B; Feldman and Priess , 2012 ) . In contrast to the NOCA-1-independent targeting of γ-tubulin complexes , a functional version of NOCA-1 lacking the putative palmitoylation site , required γ-tubulin for its cell surface targeting . Depleting α-tubulin , while having a comparable effect to γ-tubulin removal on germline structure , did not disrupt NOCA-1 targeting . This result suggests that NOCA-1 may be recruited to the surface via an interaction with γ-tubulin rather than the microtubules that it nucleates , although we cannot fully exclude a contribution from residual microtubules in the α-tubulin depletion . Our functional analysis raises the important mechanistic question of how γ-tubulin and NOCA-1 act together . One model is that γ-tubulin complexes at non-centrosomal sites recruit NOCA-1 , which in turn activates their nucleating activity , leading to generation of new microtubules . Structural work on γ-tubulin containing complexes has suggested that their activation may be coupled to interaction with factors that recruit them to specific sites ( Kollman et al . , 2011 ) . Since C . elegans , like budding yeast , has components of the γTuSC ( γ-tubulin small complex ) but not the γTuRC ( γ-tubulin ring complex ) , one possibility is that NOCA-1 would drive assembly of the γTuSC into larger γTuRC-like complexes as proposed for γTuSC-anchoring factors in budding yeast ( Figure 7B; Kollman et al . , 2011 ) . A second model is that NOCA-1 is recruited by γ-tubulin to generate a structure that stabilizes and/or anchors nascent microtubule minus ends generated by γ-tubulin's nucleating activity ( Figure 7B ) . Discriminating between these and other possibilities will require solving the challenge of analyzing purified NOCA-1 at physiological ionic strengths , which would enable better reconstitution of the interaction between NOCA-1 and microtubules ( whether direct or indirect ) in vitro and also enable analysis under conditions that include γ-tubulin-mediated nucleation . NOCA-1 functions independently of PTRN-1 in some tissues , and in parallel to PTRN-1 in the larval/adult epidermis , where the NHD of NOCA-1 and PTRN-1 function redundantly to generate a circumferential array of microtubule bundles immediately juxtaposed to the plasma membrane ( Priess and Hirsh , 1986; Costa et al . , 1997 ) . Imaging the dynamics of these bundles revealed that , despite the functional redundancy in supporting growth and morphogenesis , the microtubule arrays formed in the presence of NOCA-1 or PTRN-1 alone were distinct . When PTRN-1 was removed and only NOCA-1 was present , the microtubule growth rate and the number of growing EB1-marked microtubule ends were similar to controls . In contrast , removal of NOCA-1 led to a dramatic effect , causing a threefold reduction in the number of growing EB1-marked microtubule ends ( Figure 2D ) . At the same time , the microtubules appeared to be less dynamic , and the appearance of the arrays combined with an ∼20% reduction in growth rate suggests that there may be a small shift in the monomer/polymer balance towards more polymer . One possibility is that these effects result from the differences in the persistence of NOCA-1/γ-tubulin vs Patronin-based structures at microtubule minus ends . For example , NOCA-1/γ-tubulin might release microtubule minus ends more readily , perhaps leading to minus end depolymerization and shorter microtubules , whereas Patronin stretches might be less likely to be released leading to longer microtubules . Differences in microtubule length and minus-end dynamics could , in turn , affect plus-end dynamics . Alternatively , as has previously been proposed for γ-tubulin complexes ( Oakley et al . , 2015 ) , it is possible that NOCA-1/γ-tubulin and Patronin-based structures affect plus-end dynamics by promoting the loading of different microtubule dynamicity factors . In this vein , the effect of NOCA-1 removal on EB1::GFP localization is particularly interesting . When NOCA-1 is removed , increased amounts of EB1::GFP are observed along the length of the microtubules and an increase is also observed in EB1 comet length ( Figure 2D and Video 3 ) . It would be very interesting if NOCA-1/γ-tubulin vs Patronin-based structures at minus ends impacted the loading of factors that affect EB1 clearance from microtubules . The differences in the effects of NOCA-1 vs Patronin depletion raise the possibility that the choice between NOCA-1/ninein and/or PTRN-1/Patronin family members in different tissues may be related to the dynamicity ( or lack thereof ) required for the functions of different types of microtubule arrays . It will be particularly interesting to analyze NOCA-1 and PTRN-1 in the nervous system , where PTRN-1 has already been shown to support normal neuronal morphology and contribute to microtubule assembly and axon regeneration ( Chuang et al . , 2014; Marcette et al . , 2014; Richardson et al . , 2014 ) . The field is still in the early stages of investigating the question of redundancy between microtubule minus end-associated factors with respect to nucleating , stabilizing , and anchoring nascent minus ends . In vertebrate epithelial cells , Patronin/CAMSAP-mediated microtubule assembly has been reported to be independent of γ-tubulin-mediated nucleation and to potentially even compete with it ( Tanaka et al . , 2012 ) . In contrast , in rat hippocampal neurons , γ-tubulin has been proposed to nucleate microtubules that are subsequently stabilized by CAMSAP2 ( Yau et al . , 2014 ) . Our results in the C . elegans larval epidermis where NOCA-1 and Patronin/PTRN-1 are in parallel pathways with respect to microtubule generation , suggest that γ-tubulin cooperates with NOCA-1 but not with Patronin/PTRN-1 . Whether Patronin/PTRN-1 promotes the assembly microtubules on its own in this context or functions together with other factors such as severing proteins ( Roll-Mecak and Vale , 2006; Lindeboom et al . , 2013 ) will be important to address in the future . In summary , our work has shown that NOCA-1 , a protein with homology to vertebrate ninein , functions together with γ-tubulin in the generation of microtubules in non-centrosomal microtubule arrays . Our results shed light on non-centrosomal microtubule array formation in diverse tissues in a whole organism and also reveal functional overlap between the ninein and Patronin families of microtubule cytoskeleton-regulating proteins . The C . elegans strains used in this study are listed in Table 1 . All worm strains were maintained at 20°C on standard NGM plates seeded with OP-50 bacteria . The noca-1 ( ok3692 ) allele is balanced with a translocation balancer ( nT1[qIs51] ) . However , as the noca-1 locus is slightly outside of the balanced region ( ∼2 cM from the translocation junction; MacQueen et al . , 2005 ) , worms containing nT1 balanced noca-1 ( ok3692 ) were maintained by singling individual worms at each generation from the progeny of worms yielding the proper phenotypic distribution ( 4/5 fertile worms with pharyngeal GFP and 1/5 sterile worms without pharyngeal GFP ) . 10 . 7554/eLife . 08649 . 031Table 1 . C . elegans strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 031Strain #GenotypeN2wild type ( ancestral ) OD522unc-119 ( ed3 ) III; ltSi62[pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]IIOD523unc-119 ( ed3 ) III; ltSi63[pOD1111/pSW009; CEOP3608 TBG-1::GFP; cb-unc-119 ( + ) ]IIOD528unc-119 ( ed3 ) III; ttTi22935 V ( Mos1 insertion ) OD723noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD726ltSi77[pOD1112/pSW032; Plbp-1::mCherry; cb-unc-119 ( + ) ]VOD747unc-119 ( ed3 ) III; ttTi21011 XOD752unc-119 ( ed3 ) III; ItSi182[pOD1237/pSW055; Pnoca-1::noca-1abcfgh; cb-unc-119 ( + ) ]IIOD758unc-119 ( ed3 ) III ? ; ItSi182[pOD1237/pSW055; Pnoca-1::noca-1abcfgh; cb-unc-119 ( + ) ]II; noca-1 ( ok3692 ) VOD843unc-119 ( ed3 ) III ? ; ltIs38 [pAA1; pie-1/GFP::PH ( PLC1delta1 ) ; unc-119 ( + ) ]; ltIs37 [pAA64; pie-1/mCHERRY::his-58; unc-119 ( + ) ] IV; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD851unc-119 ( ed3 ) III ? ; ltSi62[pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]II; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD854ptrn-1 ( lt1::cb-unc-119+ ) XOD866ltSi219[pOD1248/pSW076; Pmex-5::GFP::PH ( PLC1delta1 ) ::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]IOD868ltSi220[pOD1249/pSW077; Pmex-5::GFP::tbb-2::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]IOD891noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD907ltSi222[pOD1250/pSW078; Plbp-1::GFP::tbb-2::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]I; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD909ltSi222[pOD1250/pSW078; Plbp-1::GFP::tbb-2::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]I; ltSi77[pOD1112/pSW032; Plbp-1::mCherry; cb-unc-119 ( + ) ]VOD911ltSi220[pOD1249/pSW077; Pmex-5::GFP::tbb-2::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]I; ptrn-1 ( lt1::cb-unc-119+ ) XOD952unc-119 ( ed3 ) III; ltSi246[pOD1270/pSW082; Pnoca-1::noca-1abcfgh::superfolderGFP; cb-unc-119 ( + ) ]IIOD961ltSi249[pOD1274/pSW098; Pdlg-1delta7::dlg-1::GFP::unc-54-3′ UTR; cb-unc-119 ( + ) ]IOD1011ltSi220[pOD1249/pSW077; Pmex-5::GFP::tbb-2::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]I; noca-1 ( ok3692 ) V/nT1 ( IV;V ) OD1222ItSi182[pOD1237/pSW055; Pnoca-1::noca-1abcfgh; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V; ptrn-1 ( lt1::cb-unc-119+ ) XOD1223unc-119 ( ed3 ) III; ltSi364[pOD1330/pSW147; Pnoca-1h::noca-1h ( 1-251 ) ::superfolderGFP; cb-unc-119 ( + ) ]IIOD1225unc-119 ( ed3 ) III; ltSi366[pOD1332/pSW149; Pnoca-1h::noca-1h ( 457-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]IIOD1227unc-119 ( ed3 ) III; ltSi368[pOD1334/pSW151; Pnoca-1h::noca-1h ( 252-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]IIOD1233ltSi369[pOD1335/pSW152; Pnoca-1h::noca-1h ( RNAi resistant ) ::superfolderGFP; cb-unc-119 ( + ) ]IIOD1339unc-119 ( ed3 ) III; ltSi417[pOD1342/pSW159; Pnoca-1de::noca-1de::mCherry; cb-unc-119 ( + ) ]IIOD1345ltSi417[pOD1342/pSW159; Pnoca-1de::noca-1de::mCherry; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1347ltSi419[pOD1465/pSW177; Pnoca-1h::ptrn-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1359ltSi716[pOD1935/pDC208; Pmex-5::EBP-2::GFP::tbb-2_3′ UTR; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD1394ltSi443[pOD1471/pSW182; Pnoca-1h::noca-1h ( 1-251 ) ::superfolderGFP ( C10A ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1426ltSi449[pOD1461/pSW173; Plbp-1::EBP-2::GFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD1442ltSi458[pOD1477/pSW188; Pnoca-1d::noca-1d ( cDNA ) ::mCherry; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1443ltSi459[pOD1478/pSW189; Pnoca-1e::noca-1e ( cDNA ) ::mCherry; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1446ltSi461[pOD1340/pSW157; Pnoca-1::noca-1abcgh ( STOP in the first exon of isoform f ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1504ltSi449[pOD1461/pSW173; Plbp-1::EBP-2::GFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) ? III; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1505ltSi449[pOD1461/pSW173; Plbp-1::EBP-2::GFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) ? III; ltSi77[pOD1112/pSW032; Plbp-1::mCherry; cb-unc-119 ( + ) ]VOD1510ltSi249[pOD1274/pSW098; Pdlg-1delta7::dlg-1::GFP::unc-54-3′ UTR; cb-unc-119 ( + ) ]I; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1511ltSi249[pOD1274/pSW098; Pdlg-1delta7::dlg-1::GFP::unc-54-3′ UTR; cb-unc-119 ( + ) ]I; ptrn-1 ( lt1::cb-unc-119+ ) XOD1512ltSi249[pOD1274/pSW098; Pdlg-1delta7::dlg-1::GFP::unc-54-3′ UTR; cb-unc-119 ( + ) ]I; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1516ltSi458[pOD1477/pSW188; Pnoca-1d::noca-1d ( cDNA ) ::mCherry; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) ? III; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1517ltSi459[pOD1478/pSW189; Pnoca-1e::noca-1e ( cDNA ) ::mCherry; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) ? III; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1521ltSi461[pOD1340/pSW157; Pnoca-1::noca-1abcgh ( STOP in the first exon of isoform f ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) ? III; noca-1 ( ok3692 ) VOD1558ltSi518[pOD1338/pSW155; Pnoca-1::noca-1acfgh ( STOP coden in the first exon of isoform b ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1578ltSi523[pOD1339/pSW156; Pnoca-1::noca-1abfgh ( STOP coden in the first exon of isoform c ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1580ltSi518[pOD1338/pSW155; Pnoca-1::noca-1acfgh ( STOP coden in the first exon of isoform b ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) VOD1600ltSi523[pOD1339/pSW156; Pnoca-1::noca-1abfgh ( STOP coden in the first exon of isoform c ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) VOD1605ltSi531[pOD1337/pSW154; Pnoca-1::noca-1bcfgh ( STOP coden in the first exon of isoform a ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1606ltSi531[pOD1337/pSW154; Pnoca-1::noca-1bcfgh ( STOP coden in the first exon of isoform a ) ; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) VOD1652ltSi540[pOD1343/pSW160; Pnoca-1de::noca-1de::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1653ltSi541[pOD1505/pSW210; Pdpy-7::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1654ltSi542[pOD1506/pSW211; Pptrn-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1690ltSi561[pOD1508/pSW213; Pptrn-1::noca-1h ( 457-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1691ltSi562[pOD1509/pSW214; Pptrn-1::noca-1h ( 252-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1708ltSi568[pOD1518/pSW223; Pmex-5::mCherry::PH::tbb-2_3′ UTR; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD1709ltSi569[oxTi185; pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD1727ltSi569[oxTi185; pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]I; ltSi246[pOD1270/pSW082; Pnoca-1::noca-1abcfgh::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? OD1731ltSi568[pOD1518/pSW223; Pmex-5::mCherry::PH::tbb-2_3′ UTR; cb-unc-119 ( + ) ]I; ltSi369[pOD1335/pSW152; Pnoca-1h::noca-1hRR::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? OD1737ltSi542[pOD1506/pSW211; Pptrn-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1739ltSi561[pOD1508/pSW213; Pptrn-1::noca-1h ( 457-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1740ltSi562[pOD1509/pSW214; Pptrn-1::noca-1h ( 252-922 ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1741ltSi570[pOD1527/pSW232; Pdpy-7::GFP::tbb-2::mCHerry::his-11; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD1742ltSi419[pOD1465/pSW177; Pnoca-1h::ptrn-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1780ltSi570[pOD1527/pSW232; Pdpy-7::GFP::tbb-2::mCHerry::his-11; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1781ltSi570[pOD1527/pSW232; Pdpy-7::GFP::tbb-2::mCHerry::his-11; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) III ? ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1782ltSi570[pOD1527/pSW232; Pdpy-7::GFP::tbb-2::mCHerry::his-11; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1864ltSi598[pOD1553/pSW252; Plbp-1::noca-1b::superfolderGFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1865ltSi599[pOD1554/pSW253; Plbp-1::noca-1h ( 252-922 ) ::superfolderGFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1866ltSi600[pOD1555/pSW254; Plbp-1::noca-1h ( 457-922 ) ::superfolderGFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1867ltSi601[pOD1542/pSW244; Ppha-4int1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1869ltSi603[pOD1544/pSW246; Prgef-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD1908ltSi598[pOD1553/pSW252; Plbp-1::noca-1b::superfolderGFP::opLinker::mCHerry::histone; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1909ltSi599[pOD1554/pSW253; Plbp-1::noca-1h ( 252-922 ) ::superfolderGFP::opLinker::mCHerry::histone; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1910ltSi600[pOD1555/pSW254; Plbp-1::noca-1h ( 457-922 ) ::superfolderGFP::opLinker::mCHerry::histone; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD1911ltSi601[pOD1542/pSW244; Ppha-4int1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1913ltSi603[pOD1544/pSW246; Prgef-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD1914ltSi219[pOD1248/pSW076; Pmex-5::GFP::PH ( PLC1delta1 ) ::operon_linker::mCHerry::his-11; cb-unc-119 ( + ) ]I; ptrn-1 ( lt1::cb-unc-119+ ) XOD2006ltSi541[pOD1505/pSW210; Pdpy-7::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ; ptrn-1 ( lt1::cb-unc-119+ ) XOD2074ltSi670[pSW268/pOD1786; Pmex-5::noca-1h ( 252-922 ) ::superfolderGFP::opLinker::mCHerry::PH; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD2111ltSi673[pSW279/pOD1787; Pdpy-7::tagRFP::tbb-2; cb-unc-119 ( + ) ]I; unc-119 ( ed3 ) IIIOD2113ltSi673[pSW279/pOD1787; Pdpy-7::tagRFP::tbb-2; cb-unc-119 ( + ) ]I; ltSi540[pOD1343/pSW160; Pnoca-1de::noca-1de::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? OD2114ltSi673[pSW279/pOD1787; Pdpy-7::tagRFP::tbb-2; cb-unc-119 ( + ) ]I; ltSi542[pOD1506/pSW211; Pptrn-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD2115ltSi569[oxTi185; pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]I; ltSi540[pOD1343/pSW160; Pnoca-1de::noca-1de::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? OD2116ltSi569[oxTi185; pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]I; ltSi542[pOD1506/pSW211; Pptrn-1::PTRN-1 ( cDNA ) ::superfolderGFP; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? OD2396mcIs46[pCL08 ( dlg-1::RFP ) ; cb-unc-119 ( + ) ] ? ; mcSi53[Pdpy-7::EB1::GFP; cb-unc-119 ( + ) ]II; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) ;OD2397mcIs46[pCL08 ( dlg-1::RFP ) ; cb-unc-119 ( + ) ] ? ; mcSi53[Pdpy-7::EB1::GFP; cb-unc-119 ( + ) ]II; ptrn-1 ( lt1::cb-unc-119+ ) XOD2435ltSi569[oxTi185; pOD1110/pSW008; CEOP3608 TBG-1::mCherry; cb-unc-119 ( + ) ]I; ltSi202[pVV103; Pspd-2::GFP::SPD-5 reencoded; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD2442ltSi794[pOD1988/pSW302; Pdpy-7::vhhGFP4::ZIF-1::unc-54_3′ UTR; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) IIIOD2509gip-2 ( lt19[gip-2::GFP]::loxP::cb-unc-119 ( + ) ::loxP ) I; unc-119 ( ed3 ) IIIOD2624gip-2 ( lt19[gip-2::GFP]::loxP::cb-unc-119 ( + ) ::loxP ) I; ltSi794[pOD1988/pSW302; Pdpy-7::vhhGFP4::ZIF-1::unc-54_3′ UTR; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; noca-1 ( ok3692 ) V/nT1[qIs51] ( IV;V ) OD2625gip-2 ( lt19[gip-2::GFP]::loxP::cb-unc-119 ( + ) ::loxP ) I/hT2[bli-4 ( e937 ) let- ? ( q782 ) qIs48] ( I;III ) ; ltSi794[pOD1988/pSW302; Pdpy-7::vhhGFP4::ZIF-1::unc-54_3′ UTR; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; ptrn-1 ( lt1::cb-unc-119+ ) XOD2626gip-2 ( lt19[gip-2::GFP]::loxP::cb-unc-119 ( + ) ::loxP ) I; ltSi794[pOD1988/pSW302; Pdpy-7::vhhGFP4::ZIF-1::unc-54_3′ UTR; cb-unc-119 ( + ) ]II; unc-119 ( ed3 ) III ? ; ptrn-1 ( lt1::cb-unc-119+ ) XML1654mcIs46[pCL08 ( dlg-1::RFP ) ; cb-unc-119 ( + ) ] ? ; mcSi53[Pdpy-7::EB1::GFP; cb-unc-119 ( + ) ]II A transposon-based deletion strategy ( MosDEL; Frøkjær-Jensen et al . , 2010 ) was used to make the null ptrn-1Δ allele ( ptrn-1 ( lt1::cb-unc-119+ ) ; Figure 1—figure supplement 3 ) . Briefly , a repair plasmid containing the Cb-unc-119 selection marker and appropriate homology arms ( pOD1877 , 50 ng/µl ) was co-injected with a plasmid encoding the Mos1 transposase ( pJL43 . 1 , Pglh-2::Mos1 transposase , 50 ng/µl ) and three plasmids encoding fluorescent markers for negative selection ( pCFJ90 [Pmyo-2::mCherry , 2 . 5 ng/µl] , pCFJ104 [Pmyo-3::mCherry , 5 ng/µl] and pGH8 [Prab-3::mCherry , 10 ng/µl] ) into the strain OD747 . After 1 week , moving progeny lacking fluorescent markers were identified and ptrn-1 deletion was confirmed in their progeny by PCR spanning both homology regions . A similar transposon-based strategy ( MosSCI; Frøkjær-Jensen et al . , 2008 ) was used to generate all of the transgenes used in this study . To make the noca-1h::superfolderGFP ( superfolder GFP is a folding-improved GFP version; see Pédelacq et al . , 2006 ) transgene RNAi resistant , a 999-bp region close to the 3′-end of the noca-1 coding sequence was re-encoded by codon shuffling ( Figure 4—figure supplement 2 ) . Depending on which Mos1 insertion site was used , transgenes were cloned into pCFJ151 ( ChrII insertion , ttTi5605; UniI insertion , oxTi185; UniIV insertion , oxTi177 ) , pCFJ352 ( ChrI insertion , ttTi4348 ) , or were cloned de novo ( assembly of multiple linear DNA fragments obtained by PCR [Gibson et al . , 2009]; ChrV insertion , ttTi22935 ) . In most cases , an improved transposase plasmid using a stronger promoter ( pCFJ601 , Peft-3::Mos1 transposase , 50 ng/µl ) and an additional negative selection marker pMA122 ( Phsp-16 . 41::peel-1 , 10 ng/µl ) were used in the injection mix . Single copy transgenes were generated by injecting a mixture of repairing plasmid , transposase plasmid , and selection markers into strains EG6429 ( ttTi5605 , Chr II ) , EG6701 ( ttTi4348 , Chr I ) , EG8078 ( oxTi185 , Chr I ) , or EG8081 ( oxTi177 , Chr IV ) . After 1 week , progeny of injected worms were heat shocked at 34°C for 2–4 hr to induce the expression of PEEL-1 , in order to kill extra chromosomal array containing worms ( Seidel et al . , 2011 ) . Moving worms without fluorescent markers were identified and transgene integration was confirmed in their progeny by PCR spanning both homology regions . A CRISPR/Cas9-based method ( Dickinson et al . , 2013 ) was used to generate the endogenously tagged gip-2::GFP strain . Briefly , a repairing plasmid containing the Cb-unc-119 selection marker and appropriate homology arms ( 678 bp at the 3′-end of gip-2 coding sequence and 750 bp for the gip-2 3′ UTR; pOD1999 , 20 ng/µl ) was co-injected with two plasmids modified from pDD162 by inserting two different guide RNA sequences ( 5′-AGTTCAGTCAAGAGCTCGAA-3′ and 5′-TTATTATGTCTTTTGGGTAT-3′; the plasmid also encodes the Cas9 protein; 50 ng/µl for each ) , three plasmids encoding fluorescent markers for negative selection ( pCFJ90 [Pmyo-2::mCherry , 2 . 5 ng/µl] , pCFJ104 [Pmyo-3::mCherry , 5 ng/µl] and pGH8 [Prab-3::mCherry , 10 ng/µl] ) and one plasmid encoding a heat shock-inducible toxin ( pMA122 , Phsp-16 . 41::peel-1 , 10 ng/µl ) into the strain HT1593 . After 1 week , progeny of injected worms were heat shocked at 34°C for 2 hr to induce the expression of PEEL-1 , in order to kill extra chromosomal array containing worms ( Seidel et al . , 2011 ) . Moving worms without fluorescent markers were identified and GFP insertion was confirmed in their progeny by PCR spanning both homology regions . Homozygous noca-1∆ embryos ( Figure 6A , B ) were obtained from heterozygous noca-1∆ mothers , because noca-1∆ mutants are completely sterile . To distinguish the homozygous noca-1∆ embryos from its wild-type and heterozygous siblings , we used MosSCI ( Frøkjær-Jensen et al . , 2008 ) to insert a reporter transgene ( cytoplasmic mCherry driven by an epidermal promoter Plbp-1 ) at a site only 0 . 06 cM from the noca-1 locus , so that the wild-type noca-1+ is tightly linked to the mCherry reporter . To obtain noca-1∆ embryos , the reporter strain was mated with the nT1[qIs51] balanced noca-1∆ strain ( both strains expressing GFP::β-tubulin in the embryonic epidermis ) to get heterozygous Plbp-1::mCherry ( noca-1+ ) /noca-1∆ worms . Embryos from these worms were dissected out , mounted on 2% agarose pad and imaged ( see ‘Light microscopy’ section ) . Seven splicing isoforms of noca-1 have been annotated in Wormbase . We noticed that two EST clones ( yk322h12 and yk639c8 , Wormbase ) spanned noca-1 and its 5′ neighboring gene K03H4 . 2 , suggesting the existence of an additional previously unannotated noca-1 isoform . 5′-RACE ( Invitrogen , Carlsbad , CA; 18374-058 ) using three gene-specific primers: 5′-gcttccattgaaatgagacgat-3′ , 5′-gacgaagaatgtctcgactgg-3′ and 5′-tggcttggtgttgaatgaga-3′ ( within the exons shared by all known isoforms ) revealed an eighth isoform ( noca-1h ) . The sequence of full-length noca-1h amplified from C . elegans cDNA is inserted below . >noca-1 , isoform h atgctcaaacaactattggctttgacttgcatgcacaaaaaagataaaaataagcttgcaataactgctggaaccgcagaatgttcgaacagatctcctcaaaattcaccgggatcttcctctgaaggcgctgcagacgaatctctaaatcagagtgttgctattccggaagaagctcatctgaacacttcacagtttatttcacttcccctctccgacgtctcatttgaagccgctgcatctcaaaatcgagctacaccgattgattttggtacacgagaagtgaaagaagatgacgatgttctcagtgacactggtcgtcgtcgaagcgttaacttaataacgccttctcctattccagaagaaaccgaggataacttaacagaaacgcctattcctgtagttgaacacattccaagaagtgcaatttttgaacctttcaatcacgaaaattctcctttgttctccgtgaaggcacgtaagaaagctcatgaataccgctccaacgattcaactctcagtccttcatcatcttccaacaatgacgacagtatccggattgacagtatccgtgtccgttcatcaaaatctgcaacgaataatcaactgaaaggacggcttacaccaatactaggagggtcccttcgcccgattccaaaaaaaaggaaccgagtcgctttcaacggaaattctacatttgtcgcaccggagagactatgcttggaagttgataaaatacatcaagatcgttttcgcctccgtaaacgtggagatacgtcccgtcgagatgcagtggaagctggattcgaaccgagagatactgttccacgatgtcattcaacacagtcgttgagagatgttcaacgtgttcgatcatacaacaattcacagtttcaggccagtgatctttcactcaatccaaatggaagtattcgtgctgcttgtgacagtacaagtggatctgtcgccccaacagcagttgtaaatcctgcccggaatcatgtcatttcacatcgacaacaacatcatacaagctacgagaaggatcttattccccatcataacattgatgtggatcgtcgccgtagtttgcaagctctcaatggctcatctgctctctatcaactaaataatggcggttcaccgaatggagtgagatctcaattttcaccttcggatctttctatccatacaccagttcatcatgttggaagtcgagttcgagtgtccagtgtcaaccagatttgcgattcgaacagtgctccacaattcagtatcgatcaacgccgcagtgttcacaacattggaaatccggttcgaaattcgtttgtggatggaataaaaactacatcgactccaaaaaatcagatagcagttgctccactggctcacaaaagtagacatttgagtgaatctcgagatgagatgcgtggcggtgcagaacgacgtggcagtggtggtcaaatgaatttaccagcctacactaattatcttatacgccattctggagaagagcgtcttgtggatggaccggtcactaatgccagcgatgctcggattgcttatcttgaaaaacgaatccgagaacttgaactgacacaaaaagaacagagctctcattcaacaccaagccagtcgagacattcttcgtctaaatcgtctcatttcaatggaagcagtaacttgtctacaagcgaacaactccgattgcaagaaatgagcgatgagttggcgaacaaggatcgtaaagttacatctttggaatcgaagcttctgaaagcttatcaaagaattgaacgactgaacgaggagtacgacggaaaaataaaaaatctgatgtatgacagtgaacgtgctcgcgacgatctcactcgatgtgttgataagattcagcaattggaaaacgaacttgatgagacacgagctgcagtacaaaatggagatcatgcaaatgaacaggaatatcatgagttacgagataagatctggaaacaagaacgtgaacttcaagagagtcgtacgttgcttactcgtttgcgagaaaaagaagcagaatttgagagaatgcgatcagagaaaggatatcttgagttgaagaatgagaatctcaacaagaaattggaagcgaaaaagcgagcagttgaagaactcgaaagaagtgtttcgactcttcgattggagcaaactatttgccagcaatcatgctcatctggatcaacaccgcttgctgatgagatggagattatgtcagatatccgaccatcactcgccagaccatacaccaaggctcattcgacactcgggtcccacaatatgtcaccactatcgcactcaaagtccagtggattaacgaagagtttttcgaattttgcgctcaactcatctaaacagcgtgatgatatcaccgccaatatgagccgatcgattcgtgaacaaaaccgtcacataacaatgtgtagagctatggttgtttgtctgaaggatacggtagaccgaatggcacgtggagagaatcctgatgttgctcgtctgctcggtgtcaagttgaatgtgatgtctgaaagtgaaatggaagatgatgaagatcatgaggctgatgcatcacaaccgttttcaatgatgtctgctgaatcagcgctctcgaagcaatgcggaaaactcgctgatctcgataaagacctagatacaattcgctgtcaactcgcagattggcatggtcaaacaaatgcagaaggagatggtgatcgtgatgtatgcagagttcaatag CSS-Palm 4 . 0 with the default medium threshold ( Ren et al . , 2008; http://csspalm . biocuckoo . org/ ) identified Cysteine 10 of NOCA-1h as a putative palmitoylation site ( score 38 . 139 vs the cut-off score of 3 . 717 ) . Single-stranded RNAs ( ssRNAs ) were synthesized in 50 µl T3 and T7 reactions ( MEGAscript , Invitrogen ) using cleaned DNA templates generated by PCR from N2 cDNA using the oligos in Table 2 . Reactions were cleaned using the MEGAclear kit ( Invitrogen ) , and the 50 µl T3 and T7 reactions were mixed with 50 µl of 3× soaking buffer ( 32 . 7 mM Na2HPO4 , 16 . 5 mM KH2PO4 , 6 . 3 mM NaCl , 14 . 1 mM NH4Cl ) , denatured at 68°C for 10 min , and then annealed at 37°C for 30 min to generate dsRNA . 10 . 7554/eLife . 08649 . 032Table 2 . Oligos used for dsRNA productionDOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 032GeneOligonucleotide 1Oligonucleotide 2Templatemg/mlT09E8 . 1 ( noca-1 ) AATTAACCCTCACTAAAGGggcgaacaaggatcgtaaagTAATACGACTCACTATAGGctgcatttgtttgaccatgcN2 cDNA1 . 8T09E8 . 1h ( noca-1h ) AATTAACCCTCACTAAAGGgcttgcaataactgctggaaTAATACGACTCACTATAGGaagcgactcggttcctttttN2 cDNA1 . 1F58A4 . 8 ( tbg-1 ) AATTAACCCTCACTAAAGGctcaagccttctggaaatcgTAATACGACTCACTATAGGccatgctcttcagcaacgN2 cDNA1 . 1F26E4 . 8 ( tba-1 ) AATTAACCCTCACTAAAGGccgatactggaaacggaagaTAATACGACTCACTATAGGtggtgtaacttggacggtcaN2 cDNA1 . 5 For localization analysis in Figure 5D , E and cell division analysis in Figure 2—figure supplement 1 , dsRNA was delivered by injecting L4 hermaphrodites . For all other RNAi experiments , dsRNA was delivered by soaking L4 hermaphrodites for 24 hr at 20°C . After recovery from injection or soaking , worms were incubated at 20°C for 18–54 hr before different experiments . For brood size counting and embryonic lethality assays , worms were singled 24 hr post recovery and removed from the plates at 48 hr post recovery . The number of hatched larvae and unhatched embryos were counted 1 day later . For germline imaging , injected worms were incubated at 20°C for 48–54 hr before imaging . For larval lethality assays , embryos were obtained by bleaching adult worms with freshly mixed 20% household bleach and 0 . 5 N NaOH for 10 min . Embryos were then rinsed twice in M9 and rotated in M9 at room temperature ( ∼23°C ) overnight to allow hatching . On the following day , synchronized , starved L1 worms were recovered on food . Phenotypes were quantified 72 hr post recovery . For the permeability assay ( Moribe et al . , 2004 ) , synchronized worms were rinsed with M9 in a depression slide well and transferred into 1 μg/ml HOECHST 33258 ( Sigma–Aldrich , St . Louis , MO ) in M9 . After 15 min , worms were rinsed twice in M9 and anesthetized in M9 with 1 mg/ml Tricaine ( ethyl 3-aminobenzoate methanesulfonate salt , Sigma–Aldrich; A5040-25G ) and 0 . 1 mg/ml TMHC ( tetramisole hydrochloride , Sigma–Aldrich; T1512-10G ) for 30 min . Worms were mounted onto a 2% agarose pad and imaged on a Nikon eclipse E800 microscope ( see ‘Light microscopy’ section ) . For brood size and embryonic lethality assays , L4 hermaphrodites were incubated at 20°C and singled 24 hr later . After another 24 hr , the adult worms were removed from the plates . Embryos were allowed to develop for 20–30 hr and hatched and unhatched ( embryonic lethal ) worms were counted the following day . For nuclear migration assays , worms were maintained at 20°C . Healthy L1 worms were partially anesthetized in 20 mM NaN3 , mounted on a 2% agarose pad , and imaged using DIC optics on an inverted Zeiss Axio Observer Z1 microscope system ( see ‘Light microscopy’ section ) and the number of nuclei in the dorsal cord was counted . Images and videos in Figures 2B–E , 3A , B , D , 4A , C , D , 6A–C , 7A , Figure 2—figure supplement 2 , Figure 2—figure supplement 4 , Figure 2—figure supplement 6 , Figure 3—figure supplement 1 , Figure 4—figure supplement 1A ( control and noca-1 ( RNAi ) ) , Videos 2 , 3 , 5 , 6 were acquired using an inverted Zeiss Axio Observer Z1 system with a Yokogawa spinning-disk confocal head ( CSU-X1 ) , a 63× 1 . 40 NA Plan Apochromat lens ( Zeiss , Oberkochen , Germany ) , and a Hamamatsu ORCA-ER camera ( Model C4742-95-12ERG , Hamamatsu photonics , Shizuoka , Japan ) . Images in Figures 1G , 4B , E , 5A–5C , Figure 1—figure supplement 5A , Figure 2—figure supplement 1A , Figure 4—figure supplement 1A ( noca-1Δ ) , Figure 4—figure supplement 2B and Figure 5—figure supplement 1A were acquired using the same system with an EMCCD camera ( QuantEM:512SC , Photometrics , Tucson , AZ ) . Imaging parameters were controlled using AxioVision software ( Zeiss ) . Images in Figure 5D–F , 6D were acquired using a Nikon TE2000-E inverted microscope with a Yokogawa spinning-disk confocal head ( CSU-10 ) , a 60× 1 . 40 NA Plan Apochromat lens ( Nikon , Tokyo , Japan ) and an EMCCD camera ( iXon DV887ECS-BV , Andor Technology , Belfast , United Kingdom ) . Imaging parameters were controlled using the Andor iQ2 software . Images in Figure 1—figure supplement 5B was acquired using the same system as above except that a 100× 1 . 40 NA Plan Apochromat lens ( Nikon ) was used . Images in Figure 2F and Figure 2—figure supplement 5 were acquired using a Nikon eclipse E800 microscope with a 60× 1 . 40 NA Plan Apo lens ( Nikon ) and a Hamamatsu ORCA-ER camera ( Model C4742-95-12ERG , Hamamatsu photonics ) . Imaging parameters were controlled using the Metamorph software ( Molecular Devices , Sunnyvale , CA ) . Images and videos in Figures 1D , 3E , Videos 1 , 4 were acquired using the DinoEye eyepiece camera ( AM7023B , Dino-Lite , Hsinchu , Taiwan ) mounted on a Nikon SMZ800 dissection scope . Imaging parameters were controlled using the DinoXcope software ( Dino-Lite ) . EB1 comet counting was performed using imageJ software ( FIJI ) in a macro-aided semi-automatic fashion . Tiff images were first subject to ‘Gaussian blur’ ( Sigma = 1 pixel ) , ‘subtract background’ ( rolling ball based , r = 20 pixels ) and ‘threshold’ ( manually adjusted parameters to retain most comets ) . Effective area ( excluding nuclear area ) was obtained by subjecting images to ‘threshold’ so that only the cytoplasmic region was highlighted . EB1 comet density was calculated by dividing the number of EB1 comets by the effective area . Quantifications of GIP-2::GFP intensity were performed automatically using an imageJ ( FIJI ) macro script . For measuring the intensity of GIP-2::GFP puncta in the epidermis , tiff images were first subjected to ‘Gaussian blur’ ( Sigma = 10 pixels ) and ‘setThreshold’ ( lowThreshold = 0 , highThreshold = 205 ) to get the total imaged area ( Image Area ) . The images were then subjected to ‘Gaussian blur’ ( Sigma = 1 pixel ) and ‘Find Maxima . . . ’ ( noise = 2 ) to pinpoint all GFP puncta . GFP puncta were enclosed for signal measurements by expanding the selection by 3 pixels on all sides to generate 7-pixel diameter circular regions centered over the 1-pixel-sized selections; area [in] and mean intensity [in] were measured for these regions . Background measurements were generated for each region by expanding the 7-pixel diameter circles by 3 more pixels on each side to generate 13-pixel diameter circular regions centered over the same 1-pixel-sized selections and measuring Area [out] and Mean Intensity [out] . The average background intensity was calculated as Background Intensity = ( Area [out] × Mean Intensity [out] − Area [in] × Mean Intensity [in] ) / ( Area [out] − Area [in] ) . The total intensity of the signal was calculated as Total Intensity = Area [in] × ( Mean Intensity [in] − Background Intensity ) . The normalized intensity is the Total Intensity divided by the Image Area . For measuring the membrane and centrosomal GIP-2::GFP intensity in the germline , the Image Area was obtained using exactly the same strategy as in the epidermis . The tiff images were subjected to ‘Gaussian blur’ ( sigma = 1 pixel ) and ‘setThreshold’ ( lowThreshold = 0 , highThreshold = 211 ) to narrow down the selection to membranes and centrosomes . To get all signals , the selection was expanded by adding 6 pixels on all sides and measuring Area [in] and Mean Intensity [in] in this region . The background was calculated by expanding the selected area by 6 more pixels on all sides to measure Area [out] and Mean Intensity [out] . The Background Intensity , Total Intensity , and normalized intensity were then calculated as described above for the epidermis . To generate the NOCA-1 and PTRN-1 antibodies , the regions encoding amino acids 569–717 of NOCA-1h and 910–1110 of PTRN-1a were amplified from an N2 cDNA library using the oligos listed in Table 3 and cloned into pGEX6P-1 . GST fusions were purified from bacteria and outsourced for injection into rabbits ( Covance , Princeton , NJ ) . NOCA-1 antibodies were affinity purified using the same antigen after cleavage of the GST tag as previously described ( Desai et al . , 2003 ) . PTRN-1 antibodies were affinity purified using a GFP-PTRN-1-6×His fusion purified from baculovirus-infected insect cells ( see ‘Protein purification’ section ) . 10 . 7554/eLife . 08649 . 034Table 3 . Oligos used in antibody productionDOI: http://dx . doi . org/10 . 7554/eLife . 08649 . 034TargetOligonucleotide 1Oligonucleotide 2TemplateNOCA-1ttgaattcCTCcgattgcaagaaatgattgaattcTTAgagttcttcaactgctcgN2 cDNAPTRN-1aagttctgttccaggggcccAAGGAGCTCGGTGCTGAGagtcgacccgggaattcttaGTTATTCTTATGAGCCGGAGTTCN2 cDNA For the Western blot in Figure 1B right panel , Figure 4E and Figure 1—figure supplement 2 , 20–50 control , noca-1 ( RNAi ) or noca-1h ( RNAi ) worms were transferred into a pre-weighed Eppendorf tube containing 1 ml of M9 + 0 . 1% Triton X-100 and washed three times . After the last wash , excess buffer was removed until the net weight of worms and buffer was proportional to the number of worms ( 1 mg per worm ) and 1/3 vol of 4× sample buffer was added . Worms were then placed in a sonicating water bath at ∼70°C for 10 min and boiled for 3 min . For all other worm Western blots , a mixed population of worms growing at 20°C were collected and washed three times in M9 + 0 . 1% Triton X-100 in an Eppendorf tube . After the last wash , ∼100 µl buffer was left , and then 50 µl of 4× sample buffer and 100 µl of glass beads ( Sigma–Aldrich; G8772 ) were added . Worms were vortexed for 5 min , boiled for 3 min , and then vortexed and boiled again . Samples were separated on an SDS-PAGE gel , transferred to a nitrocellulose membrane , probed with 1 µg/ml anti-NOCA-1 , anti-PTRN-1 or anti-α-tubulin ( mouse monoclonal DM1-α; Sigma–Aldrich , St . Louis , MO; T9026 ) , and detected with an HRP-conjugated secondary antibody ( rabbit or mouse; GE Healthcare , Little Chalfont , United Kingdom ) . The C . elegans extract was made as previously described ( Zanin et al . , 2011 ) . Briefly , ∼1 g of frozen worms from a large-scale liquid culture were weighed and resuspended in 1 . 5× vol of worm lysis buffer ( 50 mM Hepes-KOH pH 7 . 6 , 1 mM MgCl2 , 1 mM EGTA , 75 mM KCl , 0 . 5 mM DTT , 1 µg/ml Pepstatin A , 1 tablet of Roche cOmplete EDTA-free protease inhibitor per 50 ml ) . The suspension was sonicated to obtain a crude extract . The crude extract was centrifuged at 20 , 000×g in a TLA100 . 3 rotor ( Beckman , Pasadena , CA ) for 10 min at 2°C , and the supernatant was re-centrifuged at 50 , 000×g using a TLA100 . 3 rotor ( Beckman ) for 20 min at 2°C . The supernatant after the second centrifugation was used for the experiment . To determine whether NOCA-1 and PTRN-1 co-pellet with taxol-stabilized microtubules , a procedure modified from Kellogg et al . ( 1989 ) was used . For each experimental condition , 200 µl of worm extract prepared as above was supplemented with 0 . 5 mM DTT , 1 mM GTP , and 0 . 4 µl of DMSO ( solvent control ) , 0 . 5 mg/ml nocodazole ( no microtubule control ) or 10 mM taxol ( stabilized microtubule ) . The samples were warmed to room temperature ( ∼23°C ) for 10 min to allow microtubule polymerization , incubated on ice for 15 min and then pelleted through a glycerol cushion ( worm lysis buffer with 40% glycerol ) by centrifugation at 48 , 000×g in a TLA120 . 2 rotor ( Beckman ) for 30 min at 2°C . The sample/cushion interface was washed three times with worm lysis buffer . The pellet was resuspended in 200 µl of 1× sample buffer , and 12 µl of each sample were separated on an SDS-PAGE gel for either Coomassie staining or Western blots . For microtubule anchoring and gliding assays , DmPatronin , PTRN-1 , NOCA-1LICR+NHD , NOCA-1NHD , and MBP-NOCA-1NHD were PCR-amplified from a plasmid or N2 cDNA , and then cloned into the pFL vector ( Berger et al . , 2004 ) downstream of the p10 viral promoter with GFP and His tags . Plasmids were transformed into DH10EMBacY to generate bacmids , which were transfected into Sf9 cells to produce baculovirus . High Five or Sf9 cells were infected at 1–2 × 106 cells/ml using the baculoviruses ( 1:50 or 1:100 dilutions for High Five and 1:10 dilutions for Sf9 cells ) and cultured for 48 hr at 27°C ( High Five cells ) or 24 . 5°C ( Sf9 cells ) before being collected . Expression of GFP-tagged protein was monitored using a fluorescence dissection scope . For purifications of DmPatronin , PTRN-1 , NOCA-1LICR+NHD , and NOCA-1NHD with GFP and 6×His tags , baculovirus-infected High Five cells from 150 ml culture were lysed by sonication in 50 ml lysis buffer ( 50 mM Hepes pH 7 . 6 , 500 mM NaCl , 10 mM imidazole , 10% sucrose , 1 mM DTT , 0 . 1% Tween-20 , 1 µg/ml pepstatin A , 1 tablet of Roche cOmplete EDTA-free protease inhibitor ) . The crude extract was spun at 40 , 000 rpm in a Ti 45 rotor ( Beckman ) for 30 min at 4°C , and the soluble fraction was incubated with 1-ml nickel beads for 1 hr at 4°C . The beads were then washed three times with 50 ml wash buffer ( 25 mM Hepes pH 7 . 6 , 500 mM NaCl , 25 mM imidazole , 10% sucrose , 1 mM DTT , 0 . 1% Tween-20 ) and eluted with 1 ml fractions of elution buffer ( 25 mM Hepes pH 7 . 6 , 500 mM NaCl , 250 mM imidazole , 10% sucrose , 1 mM DTT , 0 . 1% Tween-20 ) . Elutions were either used for flow-cell assays directly or were stored at −80°C after snap-freezing of 50–100 µl aliquots in liquid nitrogen . For purifications of MBP::NOCA-1NHD::GFP-6×His and MBP::GFP::6×His , baculovirus-infected Sf9 cells from 150 ml culture were lysed by sonication in 50 ml lysis buffer ( 50 mM Hepes pH 7 . 6 , 500 mM NaCl , 10% glycerol , 1 mM DTT , 0 . 1% Tween-20 , 1 µg/ml pepstatin A , 1 tablet of Roche cOmplete EDTA-free protease inhibitor ) . The crude extract was spun at 40 , 000 rpm in a Ti 45 rotor ( Beckman ) for 30 min at 4°C , and the soluble fraction was incubated with 1 ml nickel beads for 1 hr at 4°C . The beads were then washed three times with 50 ml wash buffer ( 25 mM Hepes pH 7 . 6 , 500 mM NaCl , 10% glycerol , 1 mM DTT , 0 . 1% Tween-20 ) and eluted with 1 ml fractions of elution buffer ( 25 mM Hepes pH 7 . 6 , 500 mM NaCl , 10% glycerol , 1 mM DTT , 0 . 1% Tween-20 , 10 mM maltose ) . Elutions were either used for flow-cell assays directly or dialyzed into the dialysis buffer ( 25 mM Hepes pH 7 . 6 , 500 mM NaCl , 10% glycerol , 1 mM DTT ) and then used for flow-cell assays . The kinesin motor domain ( K560; Woehlke et al . , 1997 ) with or without GFP tag was expressed in Escherichia coli cells ( Rosetta or BL21 ) induced at OD600 0 . 6–0 . 8 and cultured at 13°C overnight . Bacteria from 1 . 5 l culture were lysed in 50 ml lysis buffer ( 50 mM Hepes-K pH 7 . 6 , 500 mM KCl , 10 mM imidazole , 10% Glycerol , 1 mM DTT , 1 mM ATP , 1 mM MgCl2 , 1 µg/ml pepstatin A , 1 tablet of Roche cOmplete EDTA-free protease inhibitor ) . The crude extract was spun at 40 , 000 rpm in a Ti 45 rotor ( Beckman ) and the soluble fraction was incubated with 1 ml nickel beads for 1 hr at 4°C . The beads were then washed three times with 50 ml wash buffer ( 50 mM Hepes-K pH 7 . 6 , 500 mM KCl , 25 mM imidazole , 10% Glycerol , 1 mM DTT , 1 mM ATP , 1 mM MgCl2 ) and eluted with 1 ml fractions of elution buffer ( 50 mM Hepes-K pH 7 . 6 , 500 mM KCl , 250 mM imidazole , 10% Glycerol , 1 mM DTT , 1 mM ATP , 1 mM MgCl2 ) . Elutions were used for flow-cell assays either directly or stored at −80°C after snap-frozen in 50–100 µl aliquots in liquid nitrogen . Coverslips were cleaned by sonication for 10 min in 5 M KOH dissolved in pure ethanol followed by 10 min of sonication in clean water , 2× rinse with water and 1× rinse with ethanol . After being dried in 37°C incubator for overnight , the coverslips were used to make flow cells using microscope slides ( Gold Seal; Thermo Scientific , Waltham , MA ) and double-sided tape ( Scotch , St . Paul , MN ) . To make rhodamine-labeled GMPCPP microtubules , labeled and unlabeled bovine tubulin were clarified by centrifugation at 90 , 000 rpm using a TLA120 . 2 rotor ( Beckman ) for 5 min at 2°C . Then the concentrations of labeled and unlabeled tubulins were measured . An elongation mix was prepared by mixing labeled and unlabeled bovine tubulin at a 1:20 ratio and 10 µM total concentration in BRB80 ( 80 mM Pipes-KOH pH 6 . 8 , 1 mM MgCl2 , 1 mM EGTA ) supplemented with 1 mM DTT and 0 . 5 mM GMPCPP ( NU-405; Jena Bioscience , Jena , Germany ) . The mix was snap frozen in liquid nitrogen at 5 µl and stored at −80°C . The stock of labeled microtubules were made by thawing an elongation mix aliquot , diluting with 5 µl of BRB80 plus 1 mM DTT and incubating in a 37°C water bath for 30–60 min . For microtubule anchoring assays , 10 µg/ml of llama GFP nanobody diluted in Tris-KAc buffer ( 50 mM Tris-HCl 8 . 0 , 150 mM KAc , 10% Glycerol , 1 mM DTT ) was introduced into the flow cell and incubated for 5 min . Then the coverslip was blocked by flowing in 1 mg/ml casein diluted in Tris-KAc buffer and incubating for 5 min . 5 nM of the GFP fusion to be tested diluted in Tris-KAc buffer were flowed in and incubated for another 5 min . Finally , 0 . 1 µM of rhodamine-labeled GMPCPP microtubules diluted in microtubule buffer with an oxygen scavenger mix ( 1×BRB80 , 1 mM DTT , 1 mg/ml casein , 0 . 8 mg/ml glucose , 0 . 04 mg/ml glucose oxidase , 0 . 016 mg/ml catalase ) was flowed in before imaging . For kinesin gliding assays , the kinesin motor domain K560 ( most concentrated fraction after His purification ) was introduced into the flow cell and incubated for 5 min . The coverslip was blocked with the Gliding Buffer ( 1×BRB80 , 1 mg/ml casein , 100 mM KCl , 0 . 1% Tween-20 , 10% sucrose , 1 mM DTT , 1 mM ATP ) and 0 . 1 µM of rhodamine-labeled GMPCPP microtubules diluted in Gliding Buffer was flowed in and incubated for 5 min . Finally , samples containing ∼200 pM of GFP-DmPatronin or ∼300 pM of GFP-PTRN-1 diluted in the Gliding Buffer , 60 nM of MBP-NOCA-1NHD-GFP or MBP-GFP diluted in BRB80 with 1 mg/ml Casein or 1 µM of MBP-NOCA-1NHD-GFP diluted in H100 ( 25 mM Hepes-NaOH pH 7 . 6 , 100 mM NaCl and 1 mM DTT ) with 1 mg/ml Casein ( all supplemented with the oxygen scavenger mix: 0 . 8 mg/ml glucose , 0 . 04 mg/ml glucose oxidase , 0 . 016 mg/ml catalase ) were flowed in before imaging . To make taxol stabilized microtubules , 2 mg/ml bovine tubulin in BRB80 with 1 mM DTT and 1 mM GTP was clarified by centrifugation at 90 k rpm using a TLA120 . 2 rotor ( Beckman ) for 5 min at 2°C . The clarified tubulin was incubated at 37°C for 2 min , and then 2 µM , 20 µM and 200 µM taxol was added stepwise . Each taxol addition was followed by 10 min of incubation at 37°C . Polymerized microtubules were then pelleted through a 40% glycerol in BRB80 cushion in a pre-warmed TLA120 . 2 rotor at 80 k rpm for 20 min at 25°C . The pellet was resuspended in BRB80 with 200 µM taxol , and the concentration was determined by the absorbance at 280 nm . For microtubule co-sedimentation assay , reaction mixes of 1 µl of 10 mM taxol in DMSO , 5 µl of 60 µM microtubules or microtubule resuspension buffer , 74 µl of H100 buffer ( 25 mM Hepes-NaOH pH 7 . 6 , 100 mM NaCl , and 1 mM DTT ) and 20 μl test protein in dialysis buffer ( 25 mM Hepes-NaOH pH 7 . 6 , 100 mM NaCl , 1 mM DTT and 10% Glycerol ) or dialysis buffer alone were incubated at room temperature for 5 min . 90 µl of each reaction mix was layered onto 100 µl of cushion ( 40% glycerol in 50 mM Hepes-KOH pH 7 . 6 , 75 mM KCl , 1 mM MgCl2 , 1 mM EGTA , 1 mM DTT , and 3 µM taxol ) in a tiny centrifuge tube and spun at 100 k rpm for 10 min at 25°C using a TLA100 rotor ( Beckman ) . 30 µl of supernatant sample was taken from the top of each tube , and 10 µl of 4× sample buffer was added in each sample . The cushion–layer interface was subsequently washed for three times using the H100 buffer before all supernatant was removed . The pellet was then re-suspended in 120 µl of 1× sample buffer . Supernatant and pellet samples were separated on an 8% poly-acrylamide gel for Coomassie blue staining .	Microtubules are hollow , rigid filaments that are found in the cells of animals and other eukaryotes . These filaments are built from smaller building blocks called tubulin heterodimers; and in dividing animal cells , they mainly emerge from structures called centrosomes . When a cell is dividing , arrays of microtubules that originate from centrosomes help assemble the spindle-like structure that segregates the chromosomes . Many non-dividing or specialized cells—including neurons , skin cells and muscle fibers—assemble other arrays of microtubules that do not emerge from centrosomes , but nevertheless perform a variety of structural , mechanical and transport-based roles . Compared to the centrosomal arrays , much less is known about how these non-centrosomal microtubules are assembled . A vertebrate protein called ‘ninein’ had previously been shown to be involved in anchoring microtubules at centrosomes . Ninein can change its localization from centrosomes to the cell surface in mammalian skin cells , suggesting that it might also have a role in assembling the peripheral microtubule arrays that are found in these cells . Now , Wang et al . have identified a protein from worms called NOCA-1 , which contains a region similar to the part of ninein that was previously shown to be needed to anchor microtubules at centrosomes . The experiments show that NOCA-1 guides the assembly of non-centrosomal microtubule arrays in multiple tissues in C . elegans worms . This includes in the outer layer of the worm's larvae , which is similar to mammalian skin . The results also highlight that NOCA-1 performs many of the same roles as a member of the Patronin family of proteins called PTRN-1 , which interacts with the ‘minus’ end of a microtubule to prevent the microtubule from breaking apart . Wang et al . also found that NOCA-1 works with another protein called γ-tubulin , which helps new microtubules to form and also interacts with microtubule minus ends . In contrast , PTRN-1 works independently of γ-tubulin . This suggests that NOCA-1 works together with γ-tubulin to protect new microtubule ends or promote their assembly , a role similar to what has been proposed for Patronin family proteins . Overall , Wang et al . 's results highlight the importance of ninein-related proteins in the assembly of non-centrosomal microtubule arrays and suggest overlapping roles for the ninein and Patronin families of proteins .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2015	NOCA-1 functions with γ-tubulin and in parallel to Patronin to assemble non-centrosomal microtubule arrays in C. elegans
Stem cell homeostasis in plant shoot meristems requires tight coordination between stem cell proliferation and cell differentiation . In Arabidopsis , stem cells express the secreted dodecapeptide CLAVATA3 ( CLV3 ) , which signals through the leucine-rich repeat ( LRR ) -receptor kinase CLAVATA1 ( CLV1 ) and related CLV1-family members to downregulate expression of the homeodomain transcription factor WUSCHEL ( WUS ) . WUS protein moves from cells below the stem cell domain to the meristem tip and promotes stem cell identity , together with CLV3 expression , generating a negative feedback loop . How stem cell activity in the meristem centre is coordinated with organ initiation and cell differentiation at the periphery is unknown . We show here that the CLE40 gene , encoding a secreted peptide closely related to CLV3 , is expressed in the SAM in differentiating cells in a pattern complementary to that of CLV3 . CLE40 promotes WUS expression via BAM1 , a CLV1-family receptor , and CLE40 expression is in turn repressed in a WUS-dependent manner . Together , CLE40-BAM1-WUS establish a second negative feedback loop . We propose that stem cell homeostasis is achieved through two intertwined pathways that adjust WUS activity and incorporate information on the size of the stem cell domain , via CLV3-CLV1 , and on cell differentiation via CLE40-BAM1 . In angiosperms , the stem cell domain in shoot meristem is controlled by the directional interplay of two adjacent groups of cells . These are the central zone ( CZ ) at the tip of the dome-shaped meristem , comprising slowly dividing stem cells , and the underlying cells of the organizing centre ( OC ) . Upon stem cell division , daughter cells are displaced laterally into the peripheral zone ( PZ ) , where they can enter differentiation pathways ( Fletcher et al . , 1999; Hall and Watt , 1989; Reddy et al . , 2004; Schnablová et al . , 2020; Stahl and Simon , 2005; Steeves and Sussex , 1989 ) . Cells in the OC express the homeodomain transcription factor WUSCHEL ( WUS ) , which moves through plasmodesmata to CZ cells to maintain stem cell fate and promote expression of the secreted signalling peptide CLAVATA3 ( CLV3 ) ( Brand et al . , 2000; Daum et al . , 2014; Müller et al . , 2006; Schoof et al . , 2000; Yadav et al . , 2011 ) . Perception of CLV3 by plasma membrane-localized receptors in the OC cells triggers a signal transduction cascade and downregulates WUS activity , thus establishing a negative feedback loop ( Mayer et al . , 1998; Ogawa et al . , 2008; Yadav et al . , 2011 ) . Mutants of CLV3 or its receptors ( see below ) fail to confine WUS expression and cause stem cell proliferation , while WUS mutants cannot maintain an active stem cell population ( Brand et al . , 2002; Clark et al . , 1993; Clark et al . , 1995; Endrizzi et al . , 1996; Laux et al . , 1996; Schoof et al . , 2000 ) . WUS function in the OC is negatively regulated by HAM transcription factors , and only WUS protein that moves upwards to the stem cell zone , which lacks HAM expression , can activate CLV3 expression ( Han et al . , 2020; Zhou et al . , 2018 ) . The CLV3-WUS interaction can serve to maintain the relative sizes of the CZ and OC , and thereby meristem growth along the apical-basal axis . However , cell loss from the PZ due to production of lateral organs requires a compensatory size increase of the stem cell domain . The CLV3 signalling pathway , which acts along the apical-basal axis of the meristem , has been widely studied in several plant species and shown to be crucial for stem cell homeostasis in shoot and floral meristems ( Somssich et al . , 2016 ) . The CLV3 peptide is perceived by a leucine-rich repeat ( LRR ) receptor kinase , CLAVATA1 ( CLV1 ) , which interacts with coreceptors of the CLAVATA3 INSENSITIVE RECEPTOR KINASES ( CIK ) 1–4 family ( Clark et al . , 1997; Cui et al . , 2018 ) . CLV1 activation involves autophoshorylation , interaction with membrane-associated and cytosolic kinases and phosphatases ( Blümke et al . , 2021; Defalco et al . , 2021 ) . Furthermore , heterotrimeric G-proteins and MAPKs have been implicated in this signal transduction cascade in maize and Arabidopsis ( Betsuyaku et al . , 2011; Bommert et al . , 2013; Ishida et al . , 2014; Lee et al . , 2019 ) . Besides CLV1 , several other receptors contribute to WUS regulation , among them RECEPTOR-LIKE PROTEIN KINASE2 ( RPK2 ) , the CLAVATA2-CORYNE heteromer ( CLV2-CRN ) and BARELY ANY MERISTEM1-3 ( BAM1-3 ) ( Bleckmann et al . , 2010; Deyoung and Clark , 2008; Hord et al . , 2006; Jeong et al . , 1999; Kinoshita et al . , 2010; Müller et al . , 2008 ) . The BAM receptors share high-sequence similarity with CLV1 and perform diverse functions throughout plant development . Double mutants of BAM1 and BAM2 maintain smaller shoot and floral meristems , thus displaying the opposite phenotype to mutants of CLV1 ( DeYoung et al . , 2006; Deyoung and Clark , 2008; Hord et al . , 2006 ) . Interestingly , ectopic expression experiments showed that CLV1 and BAM1 can perform similar functions in stem cell control ( Nimchuk et al . , 2015 ) . In addition , one study showed that CLV3 could interact with CLV1 and BAM1 in cell extracts ( Shinohara and Matsubayashi , 2015 ) , although another in vitro study did not detect BAM1-CLV3 interaction at physiological levels of CLV3 ( Crook et al . , 2020 ) . Furthermore , CLV1 was shown to act as a negative regulator of BAM1 expression , which was interpreted as a genetic buffering system , whereby a loss of CLV1 is compensated by upregulation of BAM1 in the meristem centre ( Nimchuk , 2017; Nimchuk et al . , 2015 ) . Comparable genetic compensation models for CLE peptide signalling in stem cell homeostasis were established for other species , such as tomato and maize ( Rodriguez-Leal et al . , 2019 ) . Maintaining the overall architecture of the shoot apical meristem during the entire life cycle of the plant requires replenishment of differentiating stem cell descendants in the PZ , indicating that cell division rates and cell fate changes in both regions are closely connected ( Stahl and Simon , 2005 ) . Overall meristem size is restricted by the ERECTA-family signalling pathway , which is activated by EPIDERMAL PATTERNING FACTOR ( EPF ) -LIKE ( EPFL ) ligands from the meristem periphery and confines both CLV3 and WUS expression ( Mandel et al . , 2014; Shpak , 2013; Shpak et al . , 2004; Torii et al . , 1996; Zhang et al . , 2021 ) . In the land plant lineage , the shoot meristems of bryophytes such as the moss Physcomitrium patens appear less complex than those of angiosperms and carry only a single apical stem cell which ensures organ initiation by continuous asymmetric cell divisions ( de Keijzer et al . , 2021; Harrison et al . , 2009 ) . Broadly expressed CLE peptides were here found to restrict stem cell identity and act in division plane control ( Whitewoods et al . , 2018 ) . Proliferation of the apical notch cell in the liverwort Marchantia polymorpha is promoted by MpCLE2 peptide which acts from outside the stem cell domain via the receptor MpCLV1 , while cell proliferation is confined by MpCLE1 peptide through a different receptor ( Hata and Kyozuka , 2021; Hirakawa et al . , 2019; Hirakawa et al . , 2020; Takahashi et al . , 2021 ) . Thus , antagonistic control of stem cell activities through diverse CLE peptides is conserved between distantly related land plants . In the grasses , several CLEs were found to control the stem cell domain . In maize , ZmCLE7 is expressed from the meristem tip , while ZmFCP1 is expressed in the meristem periphery and its centre . Both peptides restrict stem cell fate via independent receptor signalling pathways ( Liu et al . , 2021; Rodriguez-Leal et al . , 2019 ) . In rice , overexpression of the CLE peptides OsFCP1 and OsFCP2 downregulates the homeobox gene OSH1 and arrests meristem function ( Ohmori et al . , 2013; Suzaki et al . , 2008 ) . Common for rice and maize , CLE peptide signalling restricts stem cell activities in the shoot meristem , but a stem cell-promoting pathway was not identified so far . Importantly , how stem cell activities in the CZ and OC are coordinated to regulate organ initiation and cell differentiation in the PZ , which is crucial to maintain an active meristem , is not yet known . In maize , the CLV3-related peptide ZmFCP1 was suggested to be expressed in primordia and convey a repressive signal on the stem cell domain ( Je et al . , 2016 ) . In Arabidopsis , the most closely related peptide to CLV3 is CLE40 , which was shown to act in the root meristem to restrict columella stem cell fate and regulate the expression of the WUS paralog WOX5 ( Berckmans et al . , 2020; Hobe et al . , 2003; Pallakies and Simon , 2014; Stahl et al . , 2013; Stahl and Simon , 2010 ) . Endogenous functions of CLE40 in the SAM have not previously been described , although overexpression of CLE40 causes shoot stem cell termination , while CLE40 expression from the CLV3 promoter fully complements the shoot and floral meristem defects of clv3 mutants ( Hobe et al . , 2003 ) . We therefore hypothesized that CLE40 could act in a CLV3-related pathway in shoot stem cell control . Here , we show that the expression level of WUS in the OC is subject to feedback regulation from the PZ , which is mediated by the secreted peptide CLE40 . In the shoot meristem , CLE40 is expressed in a complementary pattern to CLV3 and excluded from the CZ and OC . In cle40 loss-of-function mutants , WUS expression is reduced , and shoot meristems remain small and flat , indicating that CLE40 signalling is required to maintain WUS expression in the OC . Ectopic expression of WUS represses CLE40 expression , while in wus loss-of-function mutants CLE40 is expressed in the meristem centre , indicating that CLE40 , in contrast to CLV3 , is subject to negative feedback regulation by WUS . CLE40 likely acts as an autocrine signal that is perceived by BAM1 in a domain flanking the OC . Based on our findings , we propose a new model for the regulation of the stem cell domain in the shoot meristem in which signals and information from both the CZ and the PZ are integrated through two interconnected negative feedback loops that sculpt the dome-shaped shoot meristems of angiosperms . Previous studies showed that CLE40 expression from the CLV3 promoter can fully complement a clv3-2 mutant , indicating that CLE40 can substitute CLV3 function in the shoot meristem to control stem cell homeostasis , if expressed from the stem cell domain . Furthermore , while all other CLE genes in Arabidopsis lack introns , the CLE40 and CLV3 genes carry two introns at very similar positions ( Hobe et al . , 2003 ) , indicating close evolutionary relatedness . Phylogenetic analysis revealed that CLV3 and CLE40 locate in the same cluster together with CLV3 orthologues from rice , maize and tomato ( Goad et al . , 2017 ) . The amino acid sequences of CLV3 and CLE40 differ at 4 out of 13 positions ( Figure 1A ) . Mutations in CLE40 were previously found to affect distal stem cell maintenance in the root meristem , revealing that a CLV3-related signalling pathway also operates in the root stem cell niche . To uncover a potential role of CLE40 in shoot development , we analysed seedling and flower development , and inflorescence meristem ( IFM ) sizes of the wild-type Col-0 , and clv3-9 and cle40-2 loss-of-function mutants . At 4 weeks after germination ( WAG ) , leaves of clv3-9 mutants remained shorter than those of Col-0 or cle40-2 ( Figure 1—figure supplement 1 ) . After floral induction , the inflorescences of clv3-9 mutants were compact with many more flowers than the wild type , while cle40-2 mutant inflorescences appeared smaller than the control ( B–D ) . To first investigate effects on meristem development in detail , longitudinal sections through the IFM at 6 WAG were obtained by confocal microscopy and meristem areas were analysed ( Figure 1B–E ) . In clv3-9 mutants , meristem areas increased to ~450% of wild-type ( Col-0 ) levels , while shoot meristems from four independent cle40 mutant alleles in a Col-0 background ( cle40-2 , cle40-cr1 , cle40-cr2 , cle40-cr3 ) reached only up to 65% of wild type ( Figure 1E , Figure 1—figure supplement 2C; Yamaguchi et al . , 2017 ) . Next , we used carpel number as a rough proxy for flower meristem ( FM ) size , which was 2 ± 0 . 0 ( N = 290 ) in Col-0 and cle40-2 ( N = 290 ) but 3 . 7 ± 0 . 4 ( N = 340 ) in clv3-9 ( Figure 1—figure supplement 3 ) . Hence , we concluded that CLE40 mainly promotes IFM growth , whereas CLV3 serves to restrict both IFM and FM sizes . We next analysed the precise CLE40 expression pattern using a transcriptional reporter line , CLE40:Venus-H2B , which showed the same expression pattern as a previously described reporter ( Stahl et al . , 2009; Figure 2—figure supplement 1 ) . We first concentrated on the IFMs and FMs . CLE40 is expressed in IFMs and FMs , starting at P5 to P6 onwards ( Figure 2A–C ) . We found stronger expression in the PZ than in the CZ , and no expression in young primordia . Using MorphoGraphX software , we extracted the fluorescence signal originating from the outermost cell layer ( L1 ) of the IFM and noted downregulation of CLE40 expression in the centre of the meristem ( Figure 2B ) . Longitudinal optical sections through the IFM showed that CLE40 is not expressed in the CZ , and only occasionally in the OC region ( Figure 2C , Figure 2—figure supplement 2A-E' ) . Expression of CLE40 changed dynamically during development: expression was concentrated in the IFM , but lacking at sites of primordia initiation ( P0 to P4/5 , Figure 2C ) . In older primordia from P5/6 onwards , CLE40 expression is detectable from the centre of the young FM and expands towards the FM periphery . In the FMs , CLE40 is lacking in young sepal primordia ( P6 ) , but starts to be expressed on the adaxial sides of petals at P7 ( Figure 2A , P1–P7 ) . To compare the CLE40 pattern with that of CLV3 , we introgressed a CLV3:NLS-3xmCherry transcriptional reporter into the CLE40:Venus-H2B background . CLV3 and CLE40 are expressed in almost mutually exclusive domains of the IFM , with CLV3 in the CZ surrounded by CLE40 expressing cells ( Figure 2D–F’’ , Figure 2—figure supplement 2 ) . In the deeper region of the IFM , where the OC is located , both CLV3 and CLE40 are not expressed ( Figure 2F , Figure 2—figure supplement 2 ) . We noted that CLE40 is downregulated where WUS is expressed , or where WUS protein localizes , such as the OC and CZ . Furthermore , CLE40 is also lacking in very early flower primordia and incipient organs . To further analyse the regulation of CLE40 expression , we introduced the CLE40 transcriptional reporter into the clv3-9 mutant background ( Figure 3A and B , Figure 3—figure supplement 1 ) . In clv3-9 mutants , WUS is no longer repressed by the CLV signalling pathway , and the CZ of the meristem increases in size as described previously ( Clark et al . , 1995 ) . In the clv3-9 mutant meristems , both CLV3 and WUS promoter activity is now found in an expanded domain ( Figure 3—figure supplement 1 ) . CLE40 is not expressed in the tip and centre of the IFM but is rather confined to the peripheral domain , where neither CLV3 nor WUS are expressed ( Figure 3B’ , Figure 3—figure supplement 1B' ) . To further explore the expression dynamics of CLE40 in connection with regulation of stem cell fate and WUS , we misexpressed WUS from the CLV3 promoter and introgressed it into plants carrying the CLE40:Venus-H2B construct . Since WUS activates the CLV3 promoter , CLV3:WUS misexpression triggers a positive feedback loop . This results in a continuous enlargement of the CZ ( Brand et al . , 2002 ) . Young seedlings carrying the CLV3:WUS transgene at 10 days after germination ( DAG ) displayed a drastically enlarged SAM compared to wild-type seedlings of the same age ( Figure 3C–D’ ) . Wild-type seedlings at this stage express CLE40 in older leaf primordia and deeper regions of the vegetative SAM ( Figure 3E–E’ ) . The CLV3:WUS transgenic seedlings do not initiate lateral organs from the expanded meristem , and CLE40 expression is confined to the cotyledons ( Figure 3F–F’ ) . CLE40 is also lacking in the deeper regions of the vegetative SAM ( Figure 3F’ ) . Thus . we conclude that either WUS itself or a WUS-dependent regulatory pathway represses CLE40 gene expression . We next determined if CLE40 repression in the CZ can be alleviated in mutants with reduced WUS activity . Since wus loss-of-function mutants fail to maintain an active CZ and shoot meristem , we used the hypomorphic wus-7 allele ( Graf et al . , 2010; Ma et al . , 2019 ) . wus-7 mutants are developmentally delayed . Furthermore , wus-7 mutants generate an IFM , but the FMs give rise to sterile flowers that lack inner organs ( Figure 3—figure supplement 2 ) . We introgressed the CLE40 reporter into wus-7 and found that at 5 WAG all wus-7 mutants expressed CLE40 in both the CZ and the OC of the IFM ( Figure 3G–H’ , Figure 3—figure supplement 2 ) . Similar to wild type , CLE40 is only weakly expressed in the young primordia of wus-7 . Therefore , we conclude that a WUS-dependent pathway downregulates CLE40 in the centre of the IFM during normal development . Given that CLV1 and BAM1 perform partially redundant functions to perceive CLV3 in shoot and floral meristems , we asked if these receptors also contribute in a CLE40 signalling pathway . We therefore generated the translational reporter lines CLV1:CLV1-GFP and BAM1:BAM1-GFP , and analysed their expression patterns in detail . Both reporter lines rescued the shoot phenotype in a clv1-101 and bam1-3;clv1-20 mutant background , respectively . The CLV1 reporter line additionally showed the same expression pattern in the shoot compared to a previously published reporter line ( Nimchuk et al . , 2011; Figure 4—figure supplement 1 , Figure 4—figure supplement 2 , Figure 4—figure supplement 3 , Figure 5—figure supplement 1 ) . We observed dynamic changes of CLV1 expression during the different stages of flower primordia initiation . CLV1:CLV1-GFP is continuously expressed in deeper regions of the IFM comprising the OC , and in the meristem periphery where new FMs are initiating ( Figure 4A , Figure 4—figure supplement 2 , Figure 4—figure supplement 3 ) . CLV1 is expressed strongly in cells of the L1 and L2 of incipient organ primordia ( P-1 , P0 ) , and only in L2 at P1 . P2 and P3 show only very faint expression in the L1 , but in stages from P4 to P6 , CLV1 expression expands from the L3 into the L2 and L1 ( Figure 4 , P1–P6 , Figure 4—figure supplement 2 , Figure 4—figure supplement 3 ) . The translational BAM1:BAM1-GFP reporter is expressed in the IFM , the FMs and in floral organs ( Figure 5A , Figure 5—figure supplement 2A–C’ , Figure 5—figure supplement 3A–E’ ) . In the IFM , expression is found throughout the L1 layer of the meristem , and , at an elevated level , in L2 and L3 cells of the PZ , but not in the meristem centre around the OC , where CLV1 expression is detected ( Figure 5B and C , compare to Figure 4C ) . BAM1 is less expressed in the deeper regions of primordia from P6 onwards ( Figure 5C , Figure 5—figure supplement 3A’–E’ ) . BAM1 transcription was reported to be upregulated in the meristem centre in the absence of CLV3 or CLV1 signalling ( Nimchuk , 2017 ) . Using our translational BAM1 reporter in the clv1-20 mutant background , we quantified and thereby confirmed that BAM1 is now expressed in the meristem centre , similar to the pattern of CLV1 in the wild type , and that BAM1 is upregulated in the L1 of the meristem . Importantly , in a clv1-20 background BAM1 is absent in the peripheral region of the IFM and the L2 ( Figure 5D–F , Figure 4—figure supplement 3F–J’ , Figure 5—figure supplement 4 ) . In longitudinal and transversal optical sections through the IFM , we found that complementarity of CLE40 and CLV3 is reflected in the complementary expression patterns of BAM1 and CLV1 ( Figure 5—figure supplement 5 ) . Therefore , we conclude that expression patterns of CLV1 and BAM1 are mostly complementary in the meristem itself and during primordia development . When comparing CLE40 and BAM1 expression patterns , we found a strong overlap in the PZ of the meristem , during incipient primordia formation , in older primordia and in L3 cells surrounding the OC ( Figure 5—figure supplement 5A’ and B’ , Figure 5—figure supplement 6 ) . Similarly , CLV3 and CLV1 are confined to the CZ and OC , respectively . To analyse if CLE40-dependent signalling requires CLV1 or BAM1 , we measured the sizes of IFMs in the respective single and double mutants ( Figure 6 ) . While cle40-2 mutant IFMs reached 65% of the wild-type size , clv1-101 plants develop IFMs that were 140% wild-type size , whereas bam1-3;clv1-101 double mutant meristems reached 450% wild-type size , similar to those of clv3-9 mutants . This supports the notion that BAM1 can partially compensate for CLV1 function in the CLV3 signalling pathway when expressed in the meristem centre ( Figure 5F; Nimchuk et al . , 2015 ) . The relationship between CLV1 and BAM1 is not symmetrical since CLV1 is expressed in a wild-typic pattern in bam1-3 mutants ( Figure 8—figure supplement 4 ) . Meristem sizes of bam1-3 mutants reached 70% of the wild type , and double mutants of cle40-2;bam1-3 did not differ significantly . However , double mutants of cle40-2;clv1-101 developed like the clv1-101 single mutant , indicating an epistatic relationship . Importantly , both clv1-101 and bam1-3 mutants lack BAM1 function in the meristem periphery ( Figure 5F ) , where also CLE40 is highly expressed , which could explain the observed epistatic relationships of cle40-2 with both clv1-101 and bam1-3 . Similar genetic relationships for CLV3 , CLE40 , CLV1 and BAM1 were noticed when analysing carpel number as a proxy for FM sizes . We also noted that generation of larger IFMs and FMs in different mutants was negatively correlated with leaf size , which we cannot explain so far ( Figure 1—figure supplement 1 ) . In the root meristem , we found that the BAM1 receptor , but not CLV1 , is required for CLE40-dependent root meristem development , suggesting that CLE40 can act through BAM1 ( Figure 6—figure supplement 1 ) . For the shoot , we hypothesize that CLE40 signals from the meristem periphery via BAM1 to promote meristem growth . Next , we aimed to determine if the commonalities between cle40-2 and bam1-3 mutants extend beyond their effects on meristem size . We next analysed the number of WUS-expressing cells in wild type and mutant meristems using a WUS:NLS-GFP transcriptional reporter . Compared to wild type , the WUS expression domain was laterally strongly expanded in both clv3-9 and clv1-101 . Interestingly , WUS signal extended also into the L1 layer of clv1-101 , albeit in a patchy pattern ( Figure 7A–C’ and F , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 ) . Also noteworthy is that BAM1 was expressed at a higher level in the L1 layer of clv1 mutants . cle40-2 mutants showed a reduction in the number of WUS-expressing cells down to ~50% wild-type levels ( Figure 7D–D’ and F , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 ) . Importantly , WUS remained expressed in the centre of the meristem , but was found in a narrow domain . In bam1-3 mutants , the WUS domain was similarly reduced as in cle40-2 , and WUS expression focussed in the meristem centre ( Figure 7E , E’ and F , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 ) . In contrast , both clv3-9 and clv1-101 mutants express WUS in a laterally expanded domain ( Figure 7B’ and C’ , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 ) . To integrate our finding that CLE40 expression is repressed by WUS activity with the observation that WUS , in turn , is promoted by CLE40 signalling , we hypothesize that the CLE40-BAM1-WUS interaction establishes a new negative feedback loop . The CLE40-BAM1-WUS negative feedback loop acts in the meristem periphery , while the CLV3-CLV1-WUS negative feedback loop acts in the meristem centre along the apical-basal axis . Both pathways act in parallel during development to regulate the size of the WUS expression domain in the meristem , possibly by perceiving input signals from two different regions , the CZ and the PZ , of the meristem . We then asked how the two signalling pathways converge on the regulation of WUS expression , control meristem growth and development . So far , we showed that both CLV3-CLV1 and CLE40-BAM1 signalling control meristem size , but in an antagonistic manner . However , we noticed that the different mutations in peptides and receptors affected distinct aspects of meristem shape . We therefore analysed meristem shape by measuring meristem height ( the apical-basal axis ) at its centre , and meristem diameter ( the radial axis ) at the base in longitudinal sections . The ratio of height to width then gives a shape parameter ‘σ’ ( from the Greek word σχήμα = shape ) . In young IFMs at 4–5 WAG , when inflorescence stems were approximately 5–8 cm long , meristems of cle40-2 and bam1-3 mutants were slightly reduced in width , and strongly reduced in height , resulting in reduced σ in comparison to Col-0 ( Figure 7G , Figure 7—figure supplement 3 ) . Meristems of clv1-101 and clv3-9 mutants were similar in width to wild type , but strongly increased in height , giving high σ values ( Figure 7 , Figure 7—figure supplement 3 ) . This indicates that CLV3-CLV1 signalling mostly restricts meristem growth along the apical-basal axis , while CLE40-BAM1 signalling promotes meristem growth along both axes . Our data expand the current model of shoot meristem homeostasis by taking into account that stem cells are lost from the OC during organ initiation in the PZ ( Figure 8 ) . CLV3 signals from the CZ via CLV1 in the meristem centre to confine WUS expression to the OC . The diffusion of WUS protein along the apical-basal axis towards the meristem tip establishes the CZ and activates CLV3 expression as a feedback signal . During plant growth , rapid cell division activity and organ initiation requires the replenishment of PZ cells from the CZ , which can be mediated by increased WUS activity . We now propose that the PZ generates CLE40 as a short-range or autocrine signal that acts through BAM1 in the meristem periphery . Since BAM1 and WUS expression does not overlap , we postulate the generation of a diffusible factor that relies on CLE40-BAM1 and acts from the PZ to promote WUS expression . WUS , in turn , represses CLE40 expression from the OC , thus establishing a second negative feedback regulation . Together , the two intertwined pathways serve to adjust WUS activity in the OC and incorporate information on the actual size of the stem cell domain , via CLV3-CLV1 , and the growth requirements from the PZ via CLE40-BAM1 . Shoot meristems are the centres of growth and organ production throughout the life of a plant . Meristems fulfil two main tasks , which are the maintenance of a non-differentiating stem cell pool , and the assignment of stem cell daughters to lateral organ primordia and differentiation pathways ( Hall and Watt , 1989 ) . Shoot meristem homeostasis requires extensive communication between the CZ , the OC and the PZ . The discovery of CLV3 as a signalling peptide , which is secreted exclusively from stem cells in the CZ , and its interaction with WUS in a negative feedback loop was fundamental to our understanding of such communication pathways ( Fletcher , 2020 ) . Here , we analysed the function of CLE40 in shoot development of Arabidopsis and found that WUS expression in the OC is under positive control from the PZ due to the activity of a CLE40-BAM1 signalling pathway . IFM size is reduced in cle40 mutants , indicating that CLE40 signalling promotes meristem size . Importantly , CLE40 is expressed in the PZ , late-stage FMs and differentiating organs . A common denominator for the complex and dynamic expression pattern is that CLE40 expression is confined to meristematic tissues , but not in organ founder sites or in regions with high WUS activity , such as the OC and the CZ . Both misexpression of WUS in the CLV3 domain ( Figure 3F ) , studies of clv3 mutants with expanded stem cell domains ( Figure 3B , Figure 3—figure supplement 1 ) and analysis of wus mutants ( Figure 3 , Figure 3—figure supplement 2 ) underpinned the notion that CLE40 expression , in contrast to CLV3 , is negatively controlled in a WUS-dependent manner . Furthermore , we found that the number of WUS-expressing cells in cle40 mutant IFMs is strongly reduced , indicating that CLE40 exerts its positive effects on IFM size by expanding the WUS expression domain . So far , the antagonistic effects of Arabidopsis CLV3 and CLE40 on meristem size can only be compared to the antagonistic functions of MpCLE1 and MpCLE2 on the gametophytic meristems of M . polymorpha , which signal through two distinct receptors , MpTDR and MpCLV1 , respectively ( Hata and Kyozuka , 2021 ) . By the complementation of clv3 mutants through expression of CLE40 from the CLV3 promoter , it was shown previously that CLE40 and CLV3 are able to activate the same downstream receptors ( Hobe et al . , 2003 ) . Our detailed analysis of candidate receptor expression patterns showed that CLV3 and CLV1 are expressed in partially overlapping domains in the meristem centre , while CLE40 and BAM1 are confined to the meristem periphery . Like cle40 mutants , bam1 mutant IFMs are smaller and maintain a smaller WUS expression domain , supporting the notion that CLE40 and BAM1 comprise a signalling unit that increases meristem size by promoting WUS expression . The antagonistic functions of the CLV3-CLV1 and CLE40-BAM1 pathways in the regulation of WUS are reflected in their complementary expression patterns . There is cross-regulation between these two signalling pathways at two levels: ( 1 ) WUS has been previously shown to promote CLV3 levels in the CZ , and we here show that WUS represses ( directly or indirectly ) CLE40 expression in the OC and in the CZ ( Figure 3B , Figure 3—figure supplement 1 ) and ( 2 ) CLV1 represses BAM1 expression in the OC , and thereby restricts BAM1 to the meristem periphery ( Figures 5 and 6 ) . In clv1 mutants , BAM1 shifts from the meristem periphery to the OC , and the WUS domain laterally expands in the meristem centre ( Figures 5F and 7B’ ) . Furthermore , BAM1 expression increases also in the L1 , which could cause the observed irregular expression of WUS in the outermost cell layer of clv1 mutants . The role of BAM1 in the OC is not entirely clear: despite the high-sequence similarity between CLV1 and BAM1 , the expression of BAM1 in the OC is not sufficient to compensate for the loss of CLV1 ( Figure 5D–F , Nimchuk et al . , 2015 ) . In the OC , BAM1 appears to restrict WUS expression to some extent since clv1;bam1 double mutants reveal a drastically expanded IFM ( Deyoung and Clark , 2008 ) . However , it is possible that BAM1 in the absence of CLV1 executes a dual function: to repress WUS in response to CLV3 in the OC as a substitute for CLV1 and simultaneously to promote WUS expression in the L1 in response to CLE40 . The expression domains of CLE40 and its receptor BAM1 largely coincide , suggesting that CLE40 acts as an autocrine signal . Similarly , protophloem sieve element differentiation in roots is inhibited by CLE45 , which acts as an autocrine signal via BAM3 ( Kang and Hardtke , 2016 ) . Since WUS is not expressed in the same cells as BAM1 , we have to postulate a non-cell-autonomous signal X that is generated in the PZ due to CLE40-BAM1 signalling and diffuses towards the meristem centre to promote WUS expression ( Hohm et al . , 2010 ) . As a result , CLE40-BAM1 signalling from the PZ will provide the necessary feedback signal that stimulates stem cell activity and thereby serves to replenish cells in the meristem for the initiation of new organs . The CLV3-CLV1 signalling pathway then adopts the role of a necessary feedback signal that avoids an excessive stem cell production . The two intertwined , antagonistically acting signalling pathways that we described here allow us to better understand the regulation of shoot meristem growth , development and shape . The previous model , which focussed mainly on the interaction of the CZ and the OC via the CLV3-CLV1-WUS negative feedback regulation , lacked any direct regulatory contribution from the PZ . EPFL peptides were shown to be expressed in the periphery and to restrict both CLV3 and WUS expression via ER ( Zhang et al . , 2021 ) . However , EPFL peptide expression is not reported to be feedback regulated from the OC or CZ , and the main function of the EPFL-ER pathway is therefore to restrict overall meristem size ( Zhang et al . , 2021 ) . The second negative feedback loop controlled by CLE40 , which we uncovered here , enables the meristem to fine-tune stem cell activities in response to fluctuating requirements for new cells during organ initiation . Due to the combined activities of CLV3 and CLE40 , the OC ( with WUS as a key player ) can now record and compute information from both , the CZ and PZ . Weaker CLV3 signalling , indicating a reduction in the size of the CZ , induces preferential growth of the meristem along the apical-basal axis ( increasing σ ) , while weaker CLE40 signals , reporting a smaller PZ , would decrease σ and flatten meristem shape . It will be intriguing to investigate if different levels of CLV3 and CLE40 also contribute to the shape changes that are observed during early vegetative development or upon floral transition in Arabidopsis . Many shoot-expressed CLE peptides are encoded in the genomes of maize , rice and barley , which could act analogously to CLV3 and CLE40 of Arabidopsis . It is tempting to speculate that in grasses a CLE40-like , stem cell-promoting signalling pathway is more active than a CLV3-like , stem cell-restricting pathway . This could contribute to the typical shape of cereal SAMs , which are , compared to the dome-shaped SAM of dicotyledonous plants , extended along the apical-basal axis . All wild-type Arabidopsis thaliana ( L . ) Heynh . plants used in this study are ecotype Columbia-0 ( Col-0 ) , except for wus-7 mutants which are in Landsberg erecta ( L . er . ) background . Details about A . thaliana plants carrying mutations in the following alleles – bam1-3 , cle40-2 , cle40-cr1 , cle40-cr2 , cle40-cr3 , clv1-101 , clv3-9 and wus-7 – are described in Table 1 . All mutants are in Col-0 background and are assumed to be null mutants , except for wus-7 mutants . cle40 mutants ( cle40-2 , cle40-cr1 , cle40-cr2 , cle40-cr3 ) have either a stop codon , a T-DNA insertion or deletion in or before the crucial CLE box domain . clv3-9 mutants were generated in 2003 by the lab of R . Simon . clv3-9 mutants were created by EMS , resulting in a W62STOP mutation before the critical CLE domain region . bam1-3 and clv1-101 mutants have been described as null mutants before ( DeYoung et al . , 2006; Kinoshita et al . , 2010 ) , while clv1-20 is a weak allele which contains a insertion within the 5′-UTR of CLV1 and results in a reduced mRNA level ( Durbak and Tax , 2011 ) . wus-7 is a weak allele and mutants were described in previous publications ( Graf et al . , 2010 ) . Double mutants were obtained by crossing the single mutant plants until both mutations were proven to be homozygous for both alleles . Genotyping of the plants was performed either by PCR or dCAPS method with the primers and restrictions enzymes listed in Table 2 . Before sowing , seeds were either sterilized for 10 min in an ethanol solution ( 80% v/v ethanol , 1 . 3% w/v sodium hypochloride , 0 . 02% w/v SDS ) or for 1 hr in a desiccator in a chloric gas atmosphere ( 50 mL of 13% w/v sodium hypochlorite with 1 mL 37% HCl ) . Afterwards , seeds were stratified for 48 hr at 4°C in darkness . Seeds on soil were then cultivated in phytochambers under long day ( LD ) conditions ( 16 hr light/8 hr dark ) at 21°C . For selection of seeds or imaging of vegetative meristems , seeds were sowed on ½ Murashige & Skoog ( MS ) media ( 1% w/v sucrose , 0 . 22% w/v MS salts + B5 vitamins , 0 . 05% w/v MES , 12 g/L plant agar , adjusted to pH 5 . 7 with KOH ) in squared Petri dishes . Seeds in Petri dishes were kept in phytocabinets under continuous light conditions at 21°C and 60% humidity . The CLE40 ( CLE40:Venus-H2B ) reporter line was cloned using the Gateway method ( Curtis and Grossniklaus , 2003 ) . The vector CLE40:Venus-H2B carries a 2291 bp long DNA fragment extending 5´ from the translational start codon of CLE40 that drives the expression of a Venus-H2B fusion protein . The DNA fragment was amplified via PCR using the oligonucleotides proCLE40_F and proCLE40_R ( Table 5 ) . As PCR template , wild-type Col-0 DNA was used . The fragment was inserted in pENTR-D-TOPO via directional TOPO-cloning . The insert was then transferred into a modified plant transformation vector pMDC99 containing the Venus-H2B sequence ( Curtis and Grossniklaus , 2003 ) . CLV1 ( CLV1:CLV1-GFP ) , BAM1 ( BAM1:BAM1-GFP ) and CLV3 ( CLV3:NLS-3xmCherry ) reporter lines were cloned using the GreenGate method ( Lampropoulos et al . , 2013 ) . Entry and destination plasmids are listed in Table 3 and Table 4 . Promoter and coding sequences were PCR amplified from genomic Col-0 DNA which was extracted from rosette leaves of Col-0 plants . Primers used for amplification of promoters and coding sequences can be found in Table 5 with the specific overhangs used for the GreenGate cloning system . Coding sequences were amplified without the stop codon to allow transcription of fluorophores at the C-terminus . BsaI restriction sites were removed by site-directed mutagenesis using the ‘QuikChange II Kit’ following the manufacturer’s instructions ( Agilent Technologies ) . Plasmid DNA amplification was performed by heat-shock transformation into Escherichia coli DH5α cells ( 10 min on ice , 1 min at 42°C , 1 min on ice , 1 hr shaking at 37°C ) , which were subsequently plated on selective LB medium ( 1% w/v tryptone , 0 . 5% w/v yeast extract , 0 . 5% w/v NaCl ) and cultivated overnight at 37°C . All entry and destination plasmids were validated by restriction digest and Sanger sequencing . Generation of stable A . thaliana lines was done by using the floral dip method ( Clough and Bent , 1998 ) . Translational CLV1 ( CLV1:CLV1-GFP ) and transcriptional CLV3 ( CLV3:NLS-3xmCherry ) reporter carry the BASTA plant resistance cassette . T1 seeds were sown on soil and sprayed with Basta ( 120 mg/mL ) at 5 and 10 DAG . Seeds of ~10 independent Basta-resistant lines were harvested . The transcriptional CLE40 ( CLE40:Venus-HB ) reporter carries the hygromycin plant resistance cassette . T1 seeds were sown on ½ MS media containing 15 μg/mL hygromycin . The translational BAM1 ( BAM1:BAM1-GFP ) reporter line carries a D-Alanin resistance cassette and T1 seeds were sown on ½ MS media containing 3–4 mM D-Alanin . Only viable plants ( ~10 T1 lines ) were selected for the T2 generation . T2 seeds were then selected on ½ MS media supplied with either 15 μg/mL hygromycin , 3–4 mM D-Alanin or 10 µg/mL of DL-phosphinothricin ( PPT ) as a BASTA alternative . Only plants from lines showing about ~75% viability were kept and cultivated under normal plant conditions ( 21°C , LD ) . Last , T3 seeds were plated on ½ MS media supplied with 3–4 mM D-Alanin or PPT again and plant lines showing 100% viability were kept as homozygous lines . The CLE40:Venus-H2B , CLV3:NLS-3xmCherry and CLV1:CLV1-GFP constructs were transformed into Col-0 wild-type plants , and stable T3 lines were generated . Plants carrying the CLE40:Venus-H2B reporter were crossed into homozygous clv3-9 or heterozygous wus-7 mutants . Homozygous clv3-9 mutants were detected by its obvious phenotype and were bred into a stable F3 generation . Homozygous wus-7 mutants were identified by phenotype and DNA genotype . Seeds were kept in the F2 generation since homozygous wus-7 plants do not develop seeds . The CLE40:Venus-H2B reporter line was crossed with the CLV3:NLS-3xmCherry reporter line and was brought into a stable F3 generation . To generate the CLE40:Venus-H2B//CLV3:WUS line , plants carrying the CLE40:Venus-H2B line were transformed with the CLV3:WUS construct . T1 seeds were sown on 10 µg/mL of DL-phosphinothricin ( PPT ) and the viable seedlings were imaged . Plants carrying the CLV1:CLV1-GFP construct were crossed into bam1-3 , cle40-2 , clv3-9 and clv1-101 mutants until a homozygous mutant background was reached . BAM1:BAM1-GFP lines were transformed into bam1-3 mutants and subsequently crossed into the clv1-20 mutant background which rescued the extremely fasciated meristem phenotype of bam1-3;clv1-20 double mutants ( Figure 5D–F ) . BAM1:BAM1-GFP//bam1-3 plants were also crossed into cle40-2 and clv3-9 mutants until a homozygous mutant background was achieved . The CLE40:CLE40-GFP line was previously described in Stahl et al . , 2009 . The WUS:NLS-GFP;CLV3:NLS-mCherry reporter line was a gift from the Lohmann lab and was crossed into clv3-9 , cle40-2 , clv1-101 and bam1-3 mutants until a stable homozygous F3 generation was reached respectively . Detailed information of all used A . thaliana lines can be found in Table 6 . To image IFMs in vivo , plants were grown under LD ( 16 hr light/8 hr dark ) conditions and inflorescences were cut off at 5 or 6 WAG . Inflorescences were stuck on double-sided adhesive tape on an objective slide and dissected until only the meristem and primordia from P0 to maximum P10 were visible . Next , inflorescences were stained with either propidium iodide ( PI 5 mM ) or 4′ , 6-diamidin-2-phenylindol ( DAPI 1 µg/mL ) for 2–5 min . Inflorescences were then washed three times with water and subsequently covered with water and a cover slide and placed under the microscope . Imaging was performed with a Zeiss LSM780 or LSM880 using a W Plan-Apochromat 40×/1 . 2 objective . Laser excitation , emission detection range and detector information for fluorophores and staining can be found in Table 7 . All IFMs were imaged from the top taking XY images along the Z axis , resulting in a Z-stack through the inflorescence . The vegetative meristems were imaged as described for IFMs . Live imaging of the reporter lines in A . thaliana plants was performed by dissecting primary inflorescences ( except for clv3-9 mutants ) at 5 WAG under LD conditions . For imaging of the reporter lines in the mutant backgrounds of clv3-9 , secondary IFMs were dissected since the primary meristems are highly fasciated . Vegetative meristems were cultivated in continuous light conditions at 21°C on ½ MS media plates and were imaged at 10 DAG . For each reporter line , at least three independent experiments were performed and at least five IFMs were imaged . For meristem measurements ( area size , width and height ) , primary and secondary IFMs of wild-type ( Col-0 ) and mutant plants ( cle40-2 , cle40-cr1-3 , bam1-3 , cle40-2;bam1-3 , clv1-101 ) were dissected at 6 WAG under LD conditions . For clv3-9 and clv1-101;bam1-3 , only secondary IFMs were imaged and analysed due to the highly fasciated primary meristems . Longitudinal optical sections of the Z-stacks were performed through the middle of the meristem starting in the centre of primordia P5 and ending in the centre of primordia P4 . Based on the longitudinal optical sections ( XZ ) , meristem height and area size were measured as indicated in Figure 6 . IFM sizes from Figure 1E are also used in Figure 6E for Col-0 , cle40-2 and clv3-9 plants . Same procedure was used to count the cells expressing WUS in different mutant backgrounds ( Figure 7A–E ) . Longitudinal optical sections of IFMs at 5 WAG were performed from P4 to P5 , and only nuclei within the meristem area were counted and plotted . For analyses of carpel numbers , the oldest 10–15 siliques per plant at 5 WAG were used . Each carpel was counted as 1 , independent of its size . N number depicts number of siliques . Leaf measurements were performed at 4 WAG , and four leaves of each plant were measured and plotted . Data was obtained from at least three independent experiments . Effects of CLE40 peptide treatment on root growth were analysed by cultivating seedlings ( Col-0 , clv1-101 , bam1-3 and bam1-3;clv1-101 ) on ½ MS agar plates ( squared ) supplied with or without synthetic CLE40p at indicated concentrations . The plates were kept upright in continuous light at 21°C and 60% relative humidity . Root growth was measured at 11 DAG by scanning plates and analysed using ImageJ to measure root lengths . For each genotype and condition , 20–48 single roots were measured . Graphs and statistical analyses were done with Prism v . 8 . For visualization of images , the open-source software ImageJ v 1 . 53c ( Schneider et al . , 2012 ) was used . All images were adjusted in ‘Brightness and Contrast’ . IFMs in Figure 7 were imaged with identical microscopy settings ( except for clv3-9 mutants ) and were all changed in ‘Brightness and Contrast’ with the same parameters to ensure comparability . clv3-9 mutants were imaged with a higher laser power since meristems are highly fasciated . MIPs were created by using the ‘Z-Projection’ function and longitudinal optical sections were performed with the ‘Reslice…’ function , resulting in the XZ view of the image . Meristem width , height and area size were measured with the ‘Straight line’ for width and height and the ‘Polygon selection’ for area size . The shape parameter σ was calculated by the quotient of height and width from each IFM . For L1 visualization , the open-source software MorphoGraphX ( https://www . mpipz . mpg . de/MorphoGraphX/ ) was used that was developed by Richard Smith . 2½ D images were created by following the steps in the MorphoGraphX manual ( Barbier de Reuille et al . , 2015 ) . After both channels ( PI and fluorophore signal ) were projected to the created mesh , both images were merged using ImageJ v 1 . 53c . For all statistical analyses , GraphPad Prism v8 . 0 . 0 . 224 was used . Statistical groups were assigned after calculating p-values by ANOVA and Tukey’s or Dunnett’s multiple comparison test ( differential grouping from p≤0 . 01 ) as indicated under each figure . Same letters indicate no statistical differences . Intensity plot profiles were measured with the ‘plot profile’ function in Fiji and plotted in GraphPad Prism . Each intensity profile was normalized . For each genotype , nine meristems were analysed and the mean of all nine meristems with its corresponding error bars ( standard deviation ) was plotted ( Figure 5—figure supplement 3 ) . Imaris software was used to detect WUS-expressing cells within the entire IFMs ( MIP ) of different mutant backgrounds . The ‘spot detection’ function in Imaris was used with the same algorithm for Col-0 , cle40-2 , bam1-3 genotypes ( [Algorithm] Enable Region Of Interest = false; Enable Region Growing = false; Enable Tracking = false; [Source Channel]; Source Channel Index = 2; Estimated; Diameter = 3 . 00 um; Background Subtraction = true; [Classify Spots] "Quality" above 3000 ) . Due to the highly fasciated meristems in clv1-101 and clv3-9 mutants , the threshold for ‘Quality’ for WUS-expressing cells was set to 1000 . All plasmid maps and cloning strategies were created and planned using the software VectorNTI .	Plants are sessile lifeforms that have evolved many ways to overcome this challenge . For example , they can quickly adapt to their environment , and they can grow new organs , such as leaves and flowers , throughout their lifetime . Stem cells are important precursor cells in plants ( and animals ) that can divide and specialize into other types of cells to help regrow leaves and flowers . A region in the plant called meristem , which can be found in the roots and shoots , continuously produces new organs in the peripheral zone of the meristem by maintaining a small group of stem cells in the central zone of the meristem . This is regulated by a signalling pathway called CLV and a molecule produced by the stem cells in the central zone , called CLV3 . Together , they keep a protein called WUS ( found in the deeper meristem known as the organizing zone ) at low levels . WUS , in turn , increases the production of stem cells that generate CLV3 . However , so far it was unclear how the number of stem cells is coordinated with the rate of organ production in the peripheral zone . To find out more , Schlegel et al . studied cells in the shoot meristems from the thale cress Arabidopsis thaliana . The researchers found that cells in the peripheral zone produce a molecule called CLE40 , which is similar to CLV3 . Unlike CLV3 , however , CLE40 boosts the levels of WUS , thereby increasing the number of stem cells . In return , WUS reduces the production of CLE40 in the central zone and the organizing centre . This system allows meristems to adapt to growing at different speeds . These results help reveal how the activity of plant meristems is regulated to enable plants to grow new structures throughout their life . Together , CLV3 and CLE40 signalling in meristems regulate stem cells to maintain a small population that is able to respond to changing growth rates . This understanding of stem cell control could be further developed to improve the productivity of crops .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "developmental", "biology" ]	2021	Control of Arabidopsis shoot stem cell homeostasis by two antagonistic CLE peptide signalling pathways
The poles of the mitotic spindle contain one old and one young centrosome . In asymmetric stem cell divisions , the age of centrosomes affects their behaviour and their probability to remain in the stem cell . In contrast , in symmetric divisions , old and young centrosomes are thought to behave equally . This hypothesis is , however , untested . In this study , we show in symmetrically dividing human cells that kinetochore–microtubules associated to old centrosomes are more stable than those associated to young centrosomes , and that this difference favours the accumulation of premature end-on attachments that delay the alignment of polar chromosomes at old centrosomes . This differential microtubule stability depends on cenexin , a protein enriched on old centrosomes . It persists throughout mitosis , biasing chromosome segregation in anaphase by causing daughter cells with old centrosomes to retain non-disjoint chromosomes 85% of the time . We conclude that centrosome age imposes via cenexin a functional asymmetry on all mitotic spindles . The bipolar spindle has a symmetric appearance; nevertheless it contains two centrosomes of different ages , as every centrosome is duplicated once during the cell cycle , resulting in the presence of an old and young centrosome at mitotic onset ( Nigg and Stearns , 2011 ) . In asymmetric stem cell divisions centrosome age differentially affects their capacity to nucleate microtubules and their positioning with respect to the polarity and cell division axis ( Yamashita et al . , 2007; Wang et al . , 2009; Januschke et al . , 2013 ) . Stem cells stereotypically inherit the centrosome , which nucleates more microtubules , which in most cases is the old centrosome ( except in fly neuroblast divisions , where stem cells retain the young active centrosome; Januschke et al . , 2011 , Conduit and Raff , 2010 ) . Old centrosomes also co-segregate with the ciliary membrane in stem cell divisions , allowing daughter stem cells to form a primary cilium earlier than the differentiating daughter cells ( Paridaen et al . , 2013 ) . In symmetric divisions , the old and young centrosomes can be differentiated at the ultra-structural level , and in terms of their microtubule-anchoring capacity during interphase ( Rieder and Borisy , 1982; Piel et al . , 2000 ) . The oldest centriole within the old centrosome contains distal and subdistal appendages: the first are necessary for centrioles to become basal bodies that can contact the plasma membrane ( Graser et al . , 2007; Hoyer-Fender , 2010 ) , while the latter are involved in the organization of the interphase microtubule network , due to the presence of ninein , a key microtubule-anchoring protein ( Mogensen et al . , 2000 ) . Both structures require the presence of cenexin , the oldest known marker for old centrosomes and appendages ( Lange and Gull , 1995; Ishikawa et al . , 2005 ) . Importantly , all these structural differences disappear progressively as cells enter mitosis; therefore , it is assumed that the old and the young centrosomes behave indistinguishably in symmetrically dividing cells , resulting in a symmetric bipolar spindle . This hypothesis has , however , never been directly tested at the functional level . Here , we tested whether centrosome age affects cell division in symmetrically dividing human cells , focusing on the ability of centrosomes to organize the alignment and segregation of sister-chromatids into two daughter cells . The first key task of the mitotic spindle is to bind to chromosomes via kinetochores and align them onto the metaphase plate ( Kops et al . , 2010 ) . To distinguish between old and young centrosomes , we used untransformed hTert-RPE1 and transformed HeLa cell lines expressing eGFP-centrin1 , a centriolar protein whose abundance correlates with centriole age , or an anti-cenexin antibody , a marker for old centrosomes ( Figure 1A , B , Kuo et al . , 2011; Lange and Gull , 1995 ) . In the vast majority of the cases both markers were enriched at the same centriole pair , indicating a robust recognition of the old centrosomes ( data not shown ) . To investigate whether half-spindles associated with old or new centrosomes align chromosomes with the same efficiency , we analyzed late prometaphase cells that contained few unaligned chromosomes . We found that 61 . 23% of the unaligned chromosomes were in the vicinity of old centrosomes in Hela-eGFP-centrin1 cells as opposed to 50% expected for an unbiased distribution , suggesting a difference in the efficiency of chromosome alignment ( Figure 1C , D , statistical tests for the chromosome alignment assays throughout the study are shown in Table 1 ) . As such unaligned chromosomes were rare , we also treated cells with 10 ng/ml nocodazole , a condition that moderately stabilizes microtubules ( Vasquez et al . , 1997 ) , and that delays chromosome alignment , leading to 3–6 unaligned chromosomes per cell ( Figure 1C ) . Unaligned chromosomes were again preferentially found in the vicinity of the old centrosomes in HeLa eGFP-centrin1 ( 63 . 9% ) and hTert-RPE1-eGFP-centrin1 cells ( 71 . 8% ) , confirming the bias in chromosome alignment ( Figure 1E ) . We found the same bias ( 67 . 8% ) in wild-type nocodazole-treated hTert-RPE1 cells stained with cenexin , excluding any effect due to eGFP-centrin1 expression ( Figure 1E ) . We conclude that the half-spindles associated to the old centrosomes accumulate more unaligned chromosomes or that unaligned chromosomes align less efficiently when bound to microtubules emanating from the old centrosomes . 10 . 7554/eLife . 07909 . 003Figure 1 . Centrosome age affects chromosome alignment . ( A ) HeLa-eGFP-centrin1 ( green ) cell stained for cenexin ( red , old centrosome marker ) and DAPI ( blue , DNA ) . One spindle pole contains the old centriole ( brightest centrin1 signal and cenexin positive ) and an accompanying daughter centriole ( dim signal ) , which together form the old centrosome . The other spindle pole contains the young centriole ( intermediate centrin1 signal ) , which is also accompanied by a daughter centriole and which together form the young centrosome . Scale bar in all panels = 5 μm . ( B ) Amounts of eGFP-centrin1 on the old , young and daughter centrioles in HeLa-eGFP-centrin1 cells determined from 3 independent experiments in 140 cells . ( C ) Untreated HeLa-eGFP-centrin1 cell ( upper panel ) and hTert RPE-eGFP-centrin1 cell treated with 10 ng/ml nocodazole ( lower panel ) stained for CENP-A ( kinetochore marker ) and DAPI . Yellow arrowheads indicate unaligned chromosomes; white arrowheads old centrosomes . ( D and E ) Proportion of unaligned chromosomes at old centrosomes in HeLa-eGFP-centrin1 cells ( D ) , and in RPE1 cells stained for cenexin , RPE1-eGFP-centrin1 cells and HeLa-eGFP-centrin1 cells treated with 10 ng/ml nocodazole ( E ) . For experiment and cell numbers , and p-values see Table 1 . For results of individual experiments see Figure 1—source data 1 . Error bars indicate s . e . m . * indicates p ≤ 0 . 05 in Binomial test compared to random distribution , *** indicates p ≤ 0 . 01 in Binomial test compared to random distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 00310 . 7554/eLife . 07909 . 004Figure 1—source data 1 . Values of individual experiments of graphs shown in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 00410 . 7554/eLife . 07909 . 005Table 1 . Percentage of unaligned chromosomes at old centrosomesDOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 005ConditionN0 of repeatsN0 of cellsN0 of chromosomes% Chromosomes at old centrosomes2-tailed Binomial test pHeLa-eGFP-centrin1 DMSO3339361 . 230 . 037hTert-RPE1 10 ng/ml nocodazole316122767 . 808 . 08e-8hTert-RPE1-eGFP-centrin1 10 ng/ml nocodazole712729571 . 811 . 0e-12HeLa-eGFP-centrin1 10 ng/ml nocodazole35753263 . 911 . 42e-10hTert-RPE1 Eg5 inhibition recovery35315674 . 368 . 9e-10hTert-RPE1 nocodazole recovery35916461 . 580 . 00373HeLa-eGFP-CENP-A/eGFP-centrin1 10 ng/ml nocodazole314294658 . 038 . 68e-07HeLa-eGFP-CENP-A/eGFP-centrin1 Eg5 inhibition recovery56830661 . 110 . 000121hTert-RPE1-eGFP-centrin1 siCtrl 10 ng/ml nocodazole39220668 . 451 . 26e-07hTert-RPE1-eGFP-centrin1 siNinein 10 ng/ml nocodazole47716966 . 860 . 0000138hTert-RPE1-eGFP-centrin1 CENP-E inhibitor36539349 . 87n . shTert-RPE1-eGFP-centrin1 CENP-E inhibitor 10 ng/ml nocodazole35945952 . 29n . shTert RPE-eGFP-centrin1 5 nM Taxol35010543 . 81n . shTert RPE-eGFP-centrin1 siDsn1 10 ng/ml nocodazole34216243 . 82n . shTert RPE-eGFP-centrin1 siNnf1 10 ng/ml nocodazole5214652 . 17n . sNon-significant . The bias in unaligned chromosomes could reflect faster kinetics in the initial capture of sister-kinetochore pairs by old centrosomes , for example , because they are closer to kinetochores at nuclear envelope breakdown or because they mature—that is , acquire a high , mitotic microtubule-nucleating capacity—earlier . Alternatively , the bias could reflect a permanent difference between the two centrosomes to capture or to align chromosomes . To test whether at mitotic onset old centrosomes capture kinetochores faster because they are closer , we compared the distances between kinetochores and old and young centrosomes at mitotic onset , but found no difference ( Figure 2A ) . We next forced hTert-RPE1 or HeLa cells to enter mitosis with monopolar spindles by treating them with monastrol , a reversible inhibitor of Eg5 , the kinesin that separates centrosomes ( Mayer et al . , 1999 ) . A monastrol washout led to bipolar spindles with few unaligned chromosomes , 74 . 4% of which were adjacent to old centrosomes , indicating that the bias is independent of the initial centrosome position ( Figure 2B , C; and Figure 2—figure supplement 1 ) . To test whether old centrosomes capture more kinetochores because they mature earlier , we treated cells with high doses of nocodazole ( 1 μg/ml ) , allowing them to enter mitosis without microtubules and to fully mature the two centrosomes ( Khodjakov and Rieder , 1999 ) , before washing out nocodazole for 1 hr: 62 . 6% of the unaligned chromosomes were adjacent to old centrosomes , indicating that the alignment bias reflects a permanent difference between the centrosomes that is independent of the initial conditions at mitotic onset ( Figure 2B , C ) . The two washout experiments also confirmed that this bias does not require low nocodazole concentrations , since in both cases , cells were released in nocodazole-free medium . 10 . 7554/eLife . 07909 . 006Figure 2 . The asymmetric distribution of unaligned chromosomes depends on CENP-E . ( A ) An illustrative example of a HeLa-eGFP-centrin1/mCherry-CENP-A cell where the distances from centrosomes to kinetochores were measured 30 s before nuclear envelope breakdown ( left top ) , assay to calculate the distances between kinetochores and centrosomes ( left bottom ) , and distribution ( right ) of the measured distances . Values are determined from 24 cells and 1434 kinetochores in 6 independent experiments . White arrowheads indicate old centrosomes in all panels . Scale bars in all panels = 5 μm . ( B ) Proportion of unaligned chromosomes at the old centrosomes in hTert-RPE1-eGFP-centrin1 cells after indicated treatments . Error bars indicate s . e . m . * indicates p ≤ 0 . 01 in Binomial test . ( C ) hTert-RPE1-eGFP-centrin1 cells stained for CENP-A and DAPI after indicated treatments . ( D ) Differences in the intensity of the microtubule asters in a re-nucleation assay at old and young centrosomes as shown in E , calculated in 32–49 cells in 3 independent experiments . Columns indicate the median , errors bars the 99% CI . Precise methodology is shown in Figure 2—figure supplement 2 . ( E ) HeLa-eGFP-centrin1 or hTert-RPE1-eGFP-centrin1 cells stained for α-tubulin after a microtubule re-nucleation assay . ( F ) hTert-RPE1-eGFP-centrin1 cells treated with 10 ng/ml nocodazole and/or CENP-E inhibitor , and stained for CENP-A . Yellow arrowheads indicate unaligned kinetochores in the proximity of the young centrosome ( G ) Proportion of unaligned chromosomes at old centrosomes in hTert-RPE1-eGFP-centrin1 cells treated with 10 ng/ml nocodazole and/or CENP-E inhibitor . Error bars indicate s . e . m . * indicates p ≤ 0 . 01 in Binomial test . For results of all individual experiments see Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 00610 . 7554/eLife . 07909 . 007Figure 2—source data 1 . Values of individual experiments of graphs shown in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 00710 . 7554/eLife . 07909 . 008Figure 2—figure supplement 1 . Monastrol wash-out does not change the alignment bias in HeLa cells . Proportion of unaligned chromosomes at the old centrosome in HeLa-eGFP-centrin1 cells after monastrol washout . Errors bars indicate s . e . m . * indicates p ≤ 0 . 01 in Binomial test . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 00810 . 7554/eLife . 07909 . 009Figure 2—figure supplement 2 . Methodology to compare microtubule re-nucleation at old and new pole . To calculate in hTert-RPE1-eGFP-centrin1 cells the relative differences in microtubule nucleation , cells were stained after a release from an ice-cold treatment with a-tubulin antibodies ( microtubules ) and DAPI ( chromosomes ) . The intensity of the centrosomal aster ( minus the background ) at old and young centrosomes was measured and the relative differences calculated with the indicated formula . Scale bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 009 In asymmetric cell divisions the old and young centrosomes have different capacities to nucleate microtubules , providing a key clue for centrosome positioning and inheritance ( Wang et al . , 2009; Januschke et al . , 2013; Lerit and Rusan , 2013 ) . If microtubule nucleation from the two centrosomes also differed in symmetric cell division , this might allow one centrosome to capture more kinetochores . However , a microtubule re-nucleation assay revealed no difference in microtubule nucleation capacity between the two centrosomes in HeLa and RPE1 cells ( Figure 2D , E ) , suggesting that the centrosomal microtubule nucleation capacity did not cause the biased distribution of unaligned chromosomes . To study if chromosome alignment process per se is asymmetric , we inhibited the Centromere-associated Protein E ( CENP-E ) , the kinetochore-bound kinesin that aligns polar chromosomes by transporting them along existing spindle microtubules ( Wood et al . , 1997; Kapoor et al . , 2006; Barisic et al . , 2014 ) . Partial CENP-E inhibition , yielding few polar chromosomes , abolished the bias in the distribution of unaligned chromosomes in the absence or presence of 10 ng/ml nocodazole , ( Figure 2F , G , 49 . 87% and 52 . 29% respectively ) , indicating that the bias depends on CENP-E , and that chromosome alignment itself is biased by centrosome age . The CENP-E-dependent alignment bias could be due to an asymmetric abundance of CENP-E; however , the levels of CENP-E on unaligned chromosomes associated to old or young centrosomes were equal ( Figure 3A , B and Figure 3—figure supplement 1A ) . Alternatively , since CENP-E favours lateral kinetochore–microtubule attachments to transport unaligned chromosomes towards the metaphase plate ( Kapoor et al . , 2006 ) , we reasoned that a difference in the types of kinetochore–microtubule attachments might bias the alignment of unaligned chromosomes: specifically end-on attachments might delay CENP-E driven chromosome alignment , by creating a poleward drag . Indeed , chromosomes that are not captured by microtubules emanating from both poles , bind laterally to microtubules from the closest pole , and are first driven to this pole in a dynein-dependent manner , before CENP-E aligns them on the metaphase plate ( Barisic et al . , 2014 ) . During these movements , kinetochores can in some cases form end-on monotelic or syntelic attachments . These non-bipolar end-on attachments are normally destabilized in an Aurora-B-dependent manner ( Hauf et al . , 2003 ) , favouring the formation of lateral kinetochore–microtubule attachments . However , if end-on attachments were to be more stable at one centrosome , they would delay this conversion and create a drag on the CENP-E driven alignment . To test this hypothesis , we visualized by 3D-high-resolution microscopy kinetochore–microtubule attachments of individual , single kinetochores on unaligned chromosomes in cells treated with 10 ng/ml nocodazole and fixed with glutaraldehyde . At old centrosomes nearly three times more individual kinetochores had end-on attachments ( 13 . 0% vs 4 . 7% at the young centrosome , p = 0 . 003 in paired t-test ) and fewer lateral attachments ( 83 . 6% vs 89 . 1% at the young centrosomes; p = 0 . 0007 in paired t-test; overall p = 0 . 0024 in 2way-ANOVA-test; Figure 3C , D and Figure 3—figure supplement 2 ) ; the number of unattached kinetochores was higher at young centrosomes , even tough this difference was statistically not significant ( p = 0 . 06 in paired t-test; Figure 3—figure supplement 2 ) . This implied an overall higher stability of kinetochore–microtubules at old centrosomes . To confirm this result , we quantified the levels of tubulin acetylation on individual kinetochore–fibres of sister-kinetochore pairs aligned on the metaphase plate , as tubulin acetylation preferentially accumulates on stable microtubules ( Webster and Borisy , 1989 ) . This analysis revealed higher level of acetylation on kinetochore–microtubules associated with old centrosomes ( median difference of 22% in tubulin acetylation at the plus ends of microtubules attached to sister-kinetochores vs 2% in CREST levels between sister-kinetochores , p < 0 . 0001 in Wilcoxon Signed Rank Test; Figure 3E , F and Figure 3—figure supplement 1C ) . In contrast , when we measured the levels of detyrosinated tubulin , a modification that has been linked to preferential CENP-E motor activity ( Barisic et al . , 2015 ) , we found no difference ( median difference of 0 . 3%; Figure 3F and Figure 3—figure supplement 1D ) . To also functionally confirm the difference in microtubule stability , metaphase cells were treated for 10 min with 0 . 5 mM Ca2+ , a condition that gradually destabilizes microtubules , before fixing them with glutaraldehyde and staining for kinetochores and microtubules . While such a treatment did not reveal strong overall differences in the two half-spindles ( Figure 3G ) , a detailed analysis of kinetochore–microtubule attachments of aligned sister-kinetochores revealed that kinetochores oriented towards young centrosomes were significantly more likely to have lost their attachment , than those oriented towards the old centrosome ( 56 . 8% vs 43 . 2%; p = 0 . 014 in paired t-test; Figure 3G , H ) . Together , these data indicated that the kinetochore–microtubules emanating from old centrosomes are more stable . 10 . 7554/eLife . 07909 . 010Figure 3 . Kinetochore–microtubules bound to old centrosomes are more stable . ( A ) hTert-RPE1-eGFP-centrin1 cells treated with 10 ng/ml nocodazole and stained for CENP-E and DAPI . White arrowheads indicate old centrosomes in all panels . Scale bars in all panels = 5 μm . ( B ) Differences in the abundance of CENP-E and CREST ( kinetochore marker ) at kinetochores bound to old and young centrosomes , calculated from 27 cells in 3 independent experiments . Columns indicate the median; error bars the 99% CI . ( C ) Immunofluorescence image of a HeLa-eGFP-centrin1/eGFP-CENP-A ( green ) cells treated with 10 ng/ml nocodazole , fixed with glutaraldehyde , and stained for α-tubulin ( red ) and DAPI ( blue ) . Single kinetochores in every unaligned sister-kinetochore pair were classified as end-on attached , laterally attached or unattached . Inset 1 on the left shows an illustrative example of a kinetochore pair with one unattached and one end-on attached kinetochore; inset 2 on the right shows an illustrative example with 2 laterally attached kinetochores . ( D ) Quantification of individual end-on attached kinetochores at old and young centrosomes in HeLa-eGFP-centrin1/eGFP-CENP-A cells treated with 10 ng/ml nocodazole , 5 nM taxol and the indicated siRNAs and 10 ng/ml nocodazole . Percentages are based on 3 independent experiments with 29–50 cells . Error bars indicate s . e . m; * indicates p ≤ 0 . 01 in paired t-test . ( E ) hTertRPE1-eGFP-centrin1 cells stained with anti-acetylated tubulin ( red ) and CREST ( green ) antibodies . Shown are total projections ( upper panels ) or maximum-intensity projections of 5–10 planes around the focal plane of interest ( lower panels ) . White arrowheads indicate kinetochore–microtubules with stronger acetylation , yellow with weaker acetylation . Note that the white arrows are on the side of the old centrosome . ( F ) Differences in the abundance of acetylated tubulin on k-fibres of sister-kinetochores , and detyrosinated tubulin on the two spindle halves in hTertRPE1-eGFP-centrin1 cells , based on 3 independent experiments and 32–33 cells . Methodology is explained in Figure 3—figure supplement 2 . Columns indicate the meadian , error bars the 99% CI . ( G ) HeLa-eGFP-centrin1/eGFP-CENP-A cells treated with 0 . 5 mM Ca2+ for 10 min stained for α-tubulin ( red ) and DAPI ( blue ) . Shown are total projections ( upper panels ) or maximum-intensity projections of 5–10 planes around the focal plane of interest ( lower panels ) . White arrowheads in zoom-ins indicate end-on attached kinetochores and yellow arrow the unattached kinetochore . ( H ) Percentage of unattached kinetochores oriented towards old or young poles based on 3 independent experiments and 33 cells . ( I ) hTert-RPE1-eGFP-centrin1 cells stained with CENP-A antibodies ( red ) and DAPI ( blue ) after treatment with 5 nM taxol or the indicated siRNAs and 10 ng/ml nocodazole . White arrowheads indicate old centrosome . ( J ) Proportion of unaligned chromosomes at old centrosome in hTert-RPE1-eGFP-centrin1 cells treated with 5 nM taxol or with 10 ng/ml nocodazole after the indicated siRNA treatment . Error bars indicate s . e . m; * indicates p ≤ 0 . 01 in Binomial test . ( K ) Differences in the abundance of cenexin , phospho-Aurora-A , and Plk1 at old and young centrosomes in HeLa and hTert-RPE1-eGFP-centrin1 cells , based on 3 independent experiments and 41–113 cells . Methodology is explained in Figure 3—figure supplement 2 . Columns indicate the median , error bars the 99% CI . ( L ) Proportion of unaligned chromosomes at old centrosome in HeLa-eGFP-centrin1 cells treated with Aurora-A or Plk1 inhibitors . Error bars indicate s . e . m; *** indicates p ≤ 0 . 01 in Binomial test . For results of all individual experiments see Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01010 . 7554/eLife . 07909 . 011Figure 3—source data 1 . Values of individual experiments of graphs shown in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01110 . 7554/eLife . 07909 . 012Figure 3—figure supplement 1 . Methodology to compare kinetochore- and centrosome-associated protein intensities at old and young spindle poles . ( A ) To calculate in hTert-RPE1-eGFP-centrin1 cells the relative differences in the abundance of CENP-E on unaligned kinetochores , cells were stained with CENP-E antibodies and DAPI ( chromosomes ) . The intensity of CENP-E on unaligned kinetochores ( minus the background ) at old and young centrosomes was measured and the relative differences calculated with the indicated formula . Scale bars in all panels = 5 μm . ( B ) Methodology to calculate the relative differences in phospho-Aurora-A , Plk1 , or cenexin in hTert-RPE1-eGFP-centrin1 cells as displayed in Figure 3K . In this example , cells were stained with phospho-Aurora-A antibodies , before determining its abundance on old and young centrosomes . The relative difference was calculated with the indicated formula . ( C ) Methodology to calculate the relative differences in acetylated tubulin ( ac-tubulin ) in hTert-RPE1-eGFP-centrin1 cells as displayed in Figure 3E , F . In this example cells were stained with acetylated tubulin antibodies , before determining its abundance on the kinetochore–microtubules associated with old and young centrosomes . The relative difference was calculated with the indicated formula . ( D ) Methodology to calculate the relative differences in detyrosinated tubulin ( dt-tubulin ) in hTert-RPE1-eGFP-centrin1 cells as displayed in Figure 3F . Cells were stained with detyrosinated tubulin antibodies , before determining its abundance on the half-spindles associated with old and young centrosomes . Same area around the old and the young centrosome , containing centrioles , was excluded from the measurement ( green rectangles ) . The relative difference was calculated with the indicated formula . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01210 . 7554/eLife . 07909 . 013Figure 3—figure supplement 2 . Quantification of the proportion of laterally and unattached kinetochores . Proportion of laterally attached ( left ) or unattached kinetochores ( right ) at old and young centrosomes in HeLa-eGFP-centrin1/eGFP-CENP-A cells treated with 10 ng/ml nocodazole , as determined in immunofluorescence pictures shown in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01310 . 7554/eLife . 07909 . 014Figure 3—figure supplement 3 . Quantification of kinetochore–microtubule stability in cold treated cells . The bar graph indicates the percentage of HeLa-eGFP-centrin1/eGFP-CENP-A cells treated with the indicated siRNAs or taxol displaying intact kinetochore–microtubules ( Category 1 ) , destabilized kinetochore–microtubules ( Category 2 ) or no kinetochore–microtubules ( Category 3 ) , after 20 min of cold treatment on ice . Values are based on measurements in 50–60 cells in 3 independent experiments . Scale bar = 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01410 . 7554/eLife . 07909 . 015Figure 3—figure supplement 4 . Validation of siRNA treatments . ( A–E ) hTert-RPE1-eGFP-centrin1 cells treated with Dsn1 ( A ) , Nnf1 ( B ) , MCAK ( C ) , Ninein ( D ) or Cenexin ( E ) siRNA and stained with indicated antibodies . Note that ninein levels were quantified in interphase , as ninein is barely visible at centrosomes in mitotic cells . Scale bar in all panels = 5 μm . Errors bars indicate s . e . m . * indicates p ≤ 0 . 05 in paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 015 To test if the difference in kinetochore–microtubule stability is at the origin of the alignment bias , we depleted the kinetochore proteins Dsn1 or Nnf1 ( both Mis12 complex ) , or treated the cells with a low dose of taxol ( 5 nM ) . These conditions strongly destabilize kinetochore–microtubules ( Dsn1 and Nnf1 depletion; ( Kline et al . , 2006 ) , or hyperstabilize kinetochore–microtubules ( taxol; Figure 3—figure supplement 3 ) . Either treatment abolished the bias in alignment and equalized the number of end-on attached unaligned kinetochores ( Figure 3D , I , J ) . In contrast , knock-down of the Mitotic Centromere-Associated Kinesin ( MCAK ) , a microtubule depolymerase that is required for destabilization of erroneous kinetochore–microtubule attachments ( Knowlton et al . , 2006 ) , whose depletion leads to mild microtubule stabilization in metaphase , but not in prometaphase ( ( Bakhoum et al . , 2009 ) and Figure 3—figure supplement 3 ) , did not change the bias in chromosome alignment ( Figure 3I , J ) . This suggested that a massive stabilization or destabilization of all kinetochore–microtubules equilibrates the difference in kinetochore–microtubule stability and chromosome alignment , but that a mild stabilization does not change this bias . We conclude that the difference in kinetochore–microtubule stability biases chromosome alignment . Which factors at centrosomes could generate an age-dependent difference in kinetochore–microtubule stability causing a bias in chromosome alignment ? We first considered two centrosomal kinases , Aurora-A and Plk1 , which can both affect kinetochore–microtubule stability ( Liu et al . , 2012; Bakhoum et al . , 2014 ) . We compared by quantitative immunofluorescence the levels of Plk1 or the activity of Aurora-A ( with an antibody that is specific for active Aurora-A ) at old and new centrosomes , to reveal a potential asymmetry in kinase levels/activity . While Plk1 was symmetrically distributed , we found a modest increase of active Aurora-A on old centrosomes in HeLa cells ( Figure 3K and Figure 3—figure supplement 1B ) . This difference was , however , not present in RPE1 cells ( Figure 3K ) ; moreover inhibition of Aurora-A or Plk1 did not abolish the bias in chromosome alignment , indicating that it does not depend on these two kinases ( Figure 3L ) . In a second step , we investigated the possible involvement of ninein , as it is essential for cell fate determination in asymmetric cell divisions of neuronal progenitors and preferentially localizes to old centrosomes in asymmetric cell division ( Wang et al . , 2009 ) , and of cenexin itself , the classical marker for old centrosomes ( Figure 3K—note that ninein levels could not be compared on old and young mitotic centrosomes , as it is only present at very low levels ( Logarinho et al . , 2012 ) ) . While ninein depletion had no effect on chromosome alignment , cenexin depletion randomized the distribution of unaligned chromosomes ( Figure 3I , J ) . Furthermore , it also equalized the percentage of end-on attached kinetochores at unaligned chromosomes , indicating that cenexin affects kinetochore–microtubule stability ( Figure 3D ) . We conclude that old centrosomes stabilize kinetochore–microtubules in a cenexin-dependent manner . The ultimate function of the mitotic spindle is to accurately segregate sister chromatids . If kinetochore–microtubules bound to the old centrosomes were more stable , we predicted that this should affect the fate of chromosomes that fail to fully disjoin in anaphase; chromosome non-disjunction is a frequent cause of chromosome mis-segregation in cancer cells , that can be caused by various defects , such as stretches of unreplicated DNA , telomere fusions , or chromosome entanglements ( Aguilera and García-Muse , 2013 ) . To monitor the fate of such chromosome non-disjunction , we monitored by live-cell imaging HeLa-eGFP-centrin1/mCherry-CENP-A cells released for synchronization purpose from a monastrol arrest . As previously reported , this procedure produced a number of single lagging , most likely merotelic , kinetochores , whose exact fate could not be tracked . However , in addition in roughly 2–5% of anaphases , we observed the presence of two lagging kinetochores that moved in synchrony between the two daughter DNA masses , but were separated by several microns , suggesting a sister-kinetochore pair on a non-disjoint chromosome ( Figure 4A and Video 1 ) . This assumption was confirmed by high-resolution immunofluorescence imaging , as such kinetochore pairs were invariably connected by a DNA thread ( see representative images in Figure 4B ) . In those instances where both sister-kinetochores segregated to the same daughter cell , we found a strong bias in chromosome mis-segregation as 18 out of the 21 analyzed kinetochore pairs co-segregated with the old centrosome ( Figure 4C; Video 1; number of experiments , cells and statistical tests for all chromosome mis-segregation events are in Table 2 ) . This suggested that non-disjoint chromosomes are preferentially pulled towards the old centrosomes , possibly due to a higher stability of the kinetochore–microtubules emanating from the old centrosomes , which in a tug-of-war would favour a destabilization and release of the kinetochore–microtubules bound to the young pole . To test this hypothesis , we treated cells with Nnf1 and Cenexin siRNAs , which had abolished the bias in chromosome alignment and the asymmetry in the percentage of end-on attached kinetochores . In both cases , the bias in chromosome mis-segregation was abolished ( Figure 4C; Video 2 and 3 ) ; in contrast when we depleted MCAK , which did not abolish the bias in chromosome alignment , chromosome mis-segregation was still biased ( Figure 4C ) . We conclude that the difference in the stability of kinetochore–microtubules bound to the old or the young centrosome persists in anaphase , and that this difference causes non-disjoint chromosomes to co-segregate with old centrosomes . 10 . 7554/eLife . 07909 . 016Figure 4 . Non-disjoined chromosomes co-segregate with old centrosomes . ( A ) Time lapse images of a HeLa-eGFP-centrin1/mCherry-CENP-A cell with a non-disjoined sister-kinetochore pair in anaphase . White arrowhead indicates the old centrosome , yellow arrowheads the non-disjoined sister-kinetochore pair . Scale bar = 10 μm . ( B ) Illustrative example of a HeLa-eGFP-centrin1/eGFP-CENP-A cell in anaphase stained for α-tubulin with a non-disjoined chromosome . Insets highlight the non-disjoined chromosomes . Scale bar = 5 μm . ( C ) Proportion of non-disjoined chromosomes that co-segregate with the old centrosomes in HeLa-eGFP-centrin1/mCherry-CENP-A cells treated with the indicated siRNAs . For statistics and number of experiments , see Table 2 . ( D ) Proposed model of how old and new centrosomes differentially affect chromosome alignment and chromosome segregation via kinetochore–microtubule stability . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01610 . 7554/eLife . 07909 . 017Video 1 . HeLa cell expressing mCherry-CENP-A ( kinetochore marker in red ) and eGFP-centrin1 ( centrosome age marker in green ) after a monastrol washout . Note that the mis-segregating chromosome moves towards the brighter , old centrosome . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01710 . 7554/eLife . 07909 . 018Table 2 . Percentage of mis-segregating chromosomes that co-segregate with the old centrosomesDOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 018ConditionN0 of repeatsNo of cellsNo of chromosomesNo of chromosomes to the old centrosome2-tailed binomial test pHeLa-eGFP-centrin1/ mCherry-CENP-A112121180 . 0015HeLa-eGFP-centrin1/ mCherry-CENP-A siMCAK38880 . 0078HeLa-eGFP-centrin1/ mCherry-CENP-A siNnf163030100 . 0990HeLa-eGFP-centrin1/ mCherry-CENP-A siCenexin61616100 . 454510 . 7554/eLife . 07909 . 019Video 2 . HeLa cell expressing mCherry-CENP-A ( kinetochore marker in red ) and eGFP-centrin1 ( centrosome age marker in green ) depleted of Nnf1 , after a monastrol washout . Note that the mis-segregating chromosome moves towards the dimmer , young centrosome . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 01910 . 7554/eLife . 07909 . 020Video 3 . HeLa cell expressing mCherry-CENP-A ( kinetochore marker in red ) and eGFP-centrin1 ( centrosome age marker in green ) depleted of Cenexin , after a monastrol washout . Note that the mis-segregating chromosome moves towards the dimmer , young centrosome . DOI: http://dx . doi . org/10 . 7554/eLife . 07909 . 020 Our results demonstrate that even in symmetrically dividing cells the two half-spindles behave in an asymmetric manner , and that centrosome age imposes a functional asymmetry on all mitotic spindles ( see model Figure 4D ) . We find that this asymmetry reflects a differential stability of kinetochore–microtubules that depends on the presence of cenexin at old centrosomes , indicating that its presence influences the relative stability of kinetochore–microtubules . Previous studies demonstrated that knock-out of cenexin does not impair mitotic progression; it contributes , however , to the stability of the centrosome-bound microtubules during interphase , consistent with our findings that cenexin affects mitotic microtubules ( Ishikawa et al . , 2005; Tateishi et al . , 2013 ) . Cenexin is known since a long time as a marker for old centrioles ( Lange and Gull , 1995 ) , yet the molecular mechanisms by which it affects microtubules are unclear . It is required for the formation of distal and sub-distal appendages on the oldest centriole , two structures that are essential for the formation of basal bodies during ciliogenesis ( Ishikawa et al . , 2005 ) . These structures consist of more then ten different centrosomal proteins , including CEP164 , CEP123 , CEP83 , SCLT1 and FBF1 at the distal appendages and ninein , centriolin , ε-tubulin , trichoplein and CEP170 at sub-distal appendages ( Mogensen et al . , 2000; Chang et al . , 2003; Gromley et al . , 2003; Guarguaglini et al . , 2005; Graser et al . , 2007; Ibi et al . , 2011; Sillibourne et al . , 2013; Tanos et al . , 2013; Tateishi et al . , 2013 ) . Moreover Plk1 has been shown to bind Odf2 at centrosomes ( Soung et al . , 2009 ) . Future work will thus have to evaluate whether the ability to stabilize microtubules is a direct function of cenexin , or a more general function of centriolar appendages . We speculate that cenexin or some of its associated centrosomal proteins might control microtubule stability via three potential mechanisms: first , they could directly interact with microtubule plus-ends , as has been seen for γ-tubulin ( Bouissou et al . , 2009 ) ; second , they could affect microtubule plus-end dynamics by controlling the dynamics of the minus-end , a type of regulation that has been seen in the context of poleward microtubule flux ( Maddox et al . , 2003; Ganem et al . , 2005; Matos et al . , 2009 ) ; third they could act via centrosomal protein kinases , as it has been recently shown as a proof-of-principle that centrosomal-bound Aurora-A has the ability to regulate kinetochore–microtubule attachments ( Chmátal et al . , 2015; Ye et al . , 2015 ) . The asymmetry in kinetochore–microtubule stability persists throughout mitosis and directs the fate of non-disjoint chromosomes , which co-segregate mostly with the old centrosome . At present it is unclear whether this asymmetry serves a direct purpose during mitosis , or if it a consequence of a centrosome asymmetry that is required for a non-mitotic function , but which cells have to deal with during each cell division . Possible non-mitotic purposes of this asymmetry include the necessity to only have one centriole capable of generating the basal body of a cilium , or the requirement to only have one centriole capable of anchoring interphase microtubules ( Piel et al . , 2000; Nigg and Raff , 2009 ) . A possible mitotic function for an asymmetric behaviour of centrosomes is that it might protect stem cells that inherit the old centrosome , from losing non-disjoint chromosomes . This would provide a selective advantage , as haplo-insufficiency is much more frequent than triplo-lethality at the level of single genes in animal cells ( Lindsley et al . , 1972; Torres et al . , 2007 ) . We speculate that the asymmetric distribution of non-disjoint chromosomes might , however , also favour the rapid acquisition of new traits by co-occurrence of chromosome gains in cancer-stem cells . Even though gain/loss of non-disjoint chromosomes is only one of the causes of chromosomal instability in cancer cells , chromosome gain of non-disjoint chromosomes would not be random , but a favoured behaviour in cancer stem cells . Such a potential bias should thus be considered when modelling chromosomal instability in aneuploid cancer cell populations . Consistent with this hypothesis we note that the analysis of large human cancer samples reveals that whole chromosome gains ( or losses ) co-occur at much higher frequencies than combined chromosome gains and losses ( Ozery-Flato et al . , 2011 ) . This phenomenon was so far thought to be the result of an evolutionary pressure; we propose that an asymmetric chromosome mis-segregation in cancer stem cells might provide a direct mechanistic explanation for this behaviour . Moreover , it could suggest a more general asymmetry in chromosomal instability , pointing to the need to determine whether other forms of chromosome mis-segregation , such as gain/loss of merotelic chromosomes , depend on centrosome age or not . HeLa , hTert-RPE1 , and hTert-RPE1-eGFP-centrin1 cells ( kind gift of A . Khodjakov ) were grown in Dulbecco's modified medium ( DMEM ) supplemented with 10% FCS , 100 U/ml penicillin , 100 mg/ml streptomycin , at 37°C with CO2 in a humidified incubator . HeLa-eGFP-centrin1 cells ( kind gift of S . Doxsey ) were further maintained in 500 μg/ml G418 . HeLa-eGFP-centrin1/mCherry-CENP-A cells were generated by stably transfecting eGFP-centrin1 in HeLa-mCherry-CENP-A cells ( kind gift of A . McAinsh , U . of Warwick ) ; as HeLa-eGFP-centrin1/CENP-A-GFP , they were further supplemented with 500 μg/ml puromycin and 500 μg/ml G418 . Live-cell imaging experiments were performed at 37°C in Lab-Tek II ( Thermofisher , Switzerland ) and Ibidi IV ( Ibidi , Switzerland ) chambers in L-15 medium supplemented with 10% FCS . To enrich for unaligned chromosomes , mitotic cells were removed by shake-off and the remaining cells treated with 10 ng/ml nocodazole for 2 hr . For nocodazole and monastrol washout experiments , cells were treated with either 1 μg/ml nocodazole or 100 nM monastrol for 4 hr ( Sigma , Switzerland ) , washed twice with fresh medium and left to recover for 1 hr . Aurora-A was inhibited for 2 hr with 100 nM MLN8237 ( Selleckchem . com , Switzerland ) , Plk1 for 2 hr with 100 nM BI2536 ( Axon Lab AG , Switzerland ) , and CENP-E for 2 hr with 5 nM GSK-923295 ( Chem Express , Switzerland ) . To stabilize microtubules , cells were treated with 5 nM Taxol ( Sigma , Switzerland ) for 2hr . To monitor anaphase cells , cells were released from a monastrol arrest and followed by live cell imaging . The following SiRNA oligonucleotides ( Invitrogen and Thermofisher , Switzerland ) were used: siControl ( scrambled ) ( Mchedlishvili et al . , 2012 ) , siNinein ( Logarinho et al . , 2012 ) , siDsn1 ( Kline et al . , 2006 ) , siNnf1 ( McAinsh et al . , 2006 ) , siMCAK ( Ganem and Compton , 2004 ) , siCenexin ( OnTarget smart pool , L-017319-01-0005 , Thermofisher ) ; they were transfected using RNAi Max Lipofectamine ( Invitrogen ) and validated by immunofluorescence microscopy ( Figure 3—figure supplement 4 ) . Cells were fixed with methanol at −20°C for 6 min , or with 20 nM Pipes ( pH 6 . 8 ) , 10 mM EGTA , 1 mM MgCl2 , 0 . 2% Triton X-100 , 4% formaldehyde for 7 min at room temperature . For the microtubule nucleation assay , cells were incubated on ice for 1 hr before release in 37°C medium for 15 s for RPE-eGFP-centrin1 and 30 s for HeLa-eGFP-centrin1 cells . To image the attachment state of unaligned kinetochores , cells were rinsed with cytoskeleton buffer ( 10 mM MES , 150 mM NaCl , 5 mM MgCl2 , 5 mM glucose ) prior and after fixation with 3% formaldehyde , 0 . 1% Triton X-100 , and 0 . 05% glutaraldehyde for 10 min at room temperature . To analyze tubulin acetylation and detyrosination , cells were fixed with 20 nM Pipes ( pH 6 . 8 ) , 10 mM EGTA , 1 mM MgCl2 , 0 . 2% Triton X-100 , 4% formaldehyde for 7 min at room temperature . To image the Calcium stability of kinetochore–microtubules , cells were treated with 0 . 5 nM CaCl2 dissolved in warm DMEM for 10 min at room temperature , rinsed with cytoskeleton buffer ( 10 mM MES , 150 mM NaCl , 5 mM MgCl2 , 5 mM glucose ) prior and after fixation with 3% formaldehyde , 0 . 1% Triton X-100 and 0 . 05% glutaraldehyde for 10 min at room temperature . Three-dimensional image stacks of mitotic cells were acquired in 0 . 1- or 0 . 2-μm steps using 100x and 60x NA 1 . 4 objectives on an Olympus DeltaVision microscope ( GE Healthcare , Switzerland ) equipped with DAPI/FITC/TRITC/CY5 filter set ( Chroma , Bellow Falls , VT ) and a CoolSNAP HQ camera ( Roper Scientific , Tuscon USA ) . 3D image stacks were deconvolved with SoftWorx ( GE Healthcare ) and analyzed with SoftWorx , Imaris ( Bitplane , Switzerland ) or ImageJ . For the nucleation assay , deconvolved total projections were analyzed as shown in Figure 1B . For the attachment status of unaligned kinetochores , single kinetochores were analyzed in 3D reconstruction of several z-stacks , and classified as shown in Figure 3C . To analyze Calcium stability , individual kinetochores were displayed in single z-planes and classified as shown in Figure 3E . The difference in microtubule nucleation capacity at old and young centrosomes was calculated as shown in Figure 2—figure supplement 2 . The differences in protein levels at centrosomes or unaligned kinetochores at old and young centrosomes were calculated as shown in Figure 3—figure supplement 1 . Images were mounted as figures using Adobe Illustrator . Primary antibodies used were mouse anti-CENP-A ( 1:2000 , Abcam , United Kingdom ) , mouse anti-α-tubulin ( 1:10′000 , Sigma ) , mouse anti-acetylated tubulin ( 1:1000; Sigma ) , rabbit anti-detyrosinated tubulin ( 1:1000 , Merck-Millipore , Switzerland ) , rabbit anti-α-tubulin ( 1:500 , Abcam ) , human CREST ( 1:400 , Antibodies Inc , Davis USA ) , rabbit anti-phosphoT288-Aurora-A ( 1:1000 , Cell Signalling , Danvers USA ) , rabbit anti-Plk1 ( 1:1000 , Abcam ) , rabbit anti-Cenexin ( 1:1000 , Abcam ) , rabbit anti-Ninein ( 1:500 , Abcam ) , rabbit anti-Nnf1 ( 1:1000 , McAinsh et al . , 2006 ) , rabbit anti-Dsn1 ( 1:2000 , kind gift of Iain Cheeseman Kline et al . , 2006 ) , rabbit anti-MCAK ( 1:1000 , Amaro et al . , 2010 ) , and rabbit anti-CENP-E ( 1:1000 , Meraldi et al . , 2004 ) . Cross-adsorbed secondary antibodies were used ( Invitrogen ) . To visually monitor the fate of non-disjoint chromosomes , Hela-eGFP-centrin1/mCherry-CENP-A cells were recorded every 2 or 4 min in 26 × 0 . 7-μm steps using a 60 × 1 . 4 NA objective on an Olympus DeltaVision microscope equipped with a GFP/mRFP filter set ( Chroma ) and a CoolSNAP HQ camera . To distinguish eGFP-centrin1 intensities in both experiments , reference images were taken at the end of the experiment , as three-dimensional stacks of 40 × 0 . 3-μm steps and a high-exposure times using the same objective and camera . Time-lapse movies were visualized in Imaris ( Bitplane ) . To calculate distances , three-dimensional positions of the old and the young centriole and of all the kinetochores were detected using Imaris ( Bitplane ) and distances calculated with a custom MatLab function ( see source code 1 ) . To check for biased distribution of polar chromosomes , Binomial probability test with expected probability success on a single trial of 0 . 5 was used . To calculate 2-ANOVAs , medians and median confidence intervals , and run t-tests PRISM ( GraphPad , La Jolla , CA ) were used . Graphs were plotted in Excel ( Microsoft , Redmond , WA ) and PRISM and mounted in Adobe Illustrator ( Adobe , Mountain View , CA ) .	Cells are able to copy their DNA and then divide to make two daughter cells that each have a complete set of genetic material . In animal cells , the DNA is arranged within structures called chromosomes and groups of proteins called centrosomes control the process that separates the chromosome copies as the cell divides . Each cell starts off with one centrosome , but before it divides , this centrosome is copied so that the cell now has two centrosomes at opposite ends of the cell , one old and one new . Filaments called microtubules assemble from the centrosomes and attach to the chromosomes . The microtubules first align all the chromosomes in the middle of the cell before pulling them towards the centrosomes as the cell divides . Some cells divide such that the two daughter cells are destined to take on different roles , for example , a stem cell may divide to produce one stem cell and one skin cell . The end of the dividing cell that will become the stem cell contains the older centrosome , while the half that forms the skin cell will receive the younger centrosome . Other cells in the body may divide to form daughter cells that have the same fate , known as symmetrical division . In these cases , it is thought that there is no difference between the behaviour of the old and young centrosomes , but this idea has never been tested . Here , Gasic et al . studied symmetrical division of human cells using fluorescent tags that made it possible to tell the centrosomes apart . The experiments show that the microtubules that assemble from the older centrosome bind the chromosome more tightly than those that form from the younger centrosome . This delays the alignment of the chromosomes that are connected to the old centrosome , as this process requires a flexible attachment . Moreover , in case the two chromosome copies fail to separate properly as cells divide , the older centrosome is more likely to receive both chromosome copies at the expense of the other centrosome . A protein called cenexin is present at higher levels around older centrosomes than around younger ones and is responsible for this effect . Gasic et al . 's findings show that the age of the centrosomes leads to asymmetry in all cell divisions , even those that produce cells that are destined to have the same role in an organism . The next challenge will be to understand whether this asymmetry has any consequences for cells , in particular cancer cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2015	Centrosome age regulates kinetochore–microtubule stability and biases chromosome mis-segregation
The gene desert upstream of the MYC oncogene on chromosome 8q24 contains susceptibility loci for several major forms of human cancer . The region shows high conservation between human and mouse and contains multiple MYC enhancers that are activated in tumor cells . However , the role of this region in normal development has not been addressed . Here we show that a 538 kb deletion of the entire MYC upstream super-enhancer region in mice results in 50% to 80% decrease in Myc expression in multiple tissues . The mice are viable and show no overt phenotype . However , they are resistant to tumorigenesis , and most normal cells isolated from them grow slowly in culture . These results reveal that only cells whose MYC activity is increased by serum or oncogenic driver mutations depend on the 8q24 super-enhancer region , and indicate that targeting the activity of this element is a promising strategy of cancer chemoprevention and therapy . Deregulated expression of the MYC oncogene is associated with many cancer types ( Reviewed in Albihn et al . , 2010; Dang , 2012; Evan , 2012 ) . MYC acts primarily as a transcriptional activator that increases expression of many genes required for RNA and protein synthesis above the level that is required in resting cells . In cancer cells , aberrantly elevated levels of MYC drive global amplification of transcription rates , providing the cells with necessary resources for rapid proliferation ( see , for example Brown et al . , 2008; van Riggelen et al . , 2010; Ji et al . , 2011; Lin et al . , 2012; Sabò et al . , 2014; Walz et al . , 2014 ) . Transcription of the MYC gene is regulated by a diverse array of regulatory elements located both upstream and downstream of the MYC transcription start site ( TSS ) . Variants in the MYC upstream region contribute to inherited susceptibility to most major forms of human cancer , and account for a very large number of cancer cases at the population level ( Amundadottir et al . , 2006; Gudmundsson et al . , 2007; Yeager et al . , 2007; Al Olama et al . , 2009; Yeager et al . , 2009 ) . For example , the polymorphism rs6983267 linked to colorectal ( Tomlinson et al . , 2007 ) and prostate ( Yeager et al . , 2007 ) cancers contributes more to cancer morbidity and mortality than any other known inherited variant or mutation , including the inherited mutations in classic tumor suppressors such as RB , TP53 and APC . Through computational and experimental analyses , we and others have shown that the risk allele G of rs6983267 creates a strong binding site for the colorectal-cancer associated transcription factor Tcf7l2 ( Pomerantz et al . , 2009; Tuupanen et al . , 2009 ) . This binding site is located within the Myc-335 enhancer element that is dispensable for mouse viability , but required for efficient Tcf7l2-driven intestinal tumorigenesis ( Sur et al . , 2012b ) . More recently , another enhancer element , located 1 . 47 Mb downstream of Myc was shown to be required for formation of acute lymphoblastic leukemia ( ALL ) in mice ( Herranz et al . , 2014 ) . However , in contrast to the Myc-335 element , this element is also required for normal T-cell development . Thus , the mechanism by which individual Myc enhancer elements contribute to normal development and tumorigenesis is still unclear . Several studies have shown that the 8q24 region contains a large number of additional enhancer elements ( see , for example [Hallikas et al . , 2006; Ahmadiyeh et al . , 2010; Yan et al . , 2013; Yao et al . , 2014] ) and super-enhancers that are active in many different types of human cancer ( Hnisz et al . , 2013; Lovén et al . , 2013; Zhou et al . , 2015 ) . The MYC-associated super-enhancers are activated during the process of tumorigenesis ( Hnisz et al . , 2013 ) , and downregulation of super-enhancer activity leads to selective inhibition of MYC expression ( Lovén et al . , 2013 ) . Thus , MYC-associated super-enhancer activity is required for tumorigenesis , but the role of these elements in normal tissue morphogenesis and homeostasis has been unclear . To address this problem , we have in this work generated multiple mouse strains deficient of regulatory elements upstream of the Myc promoter . Since this region contains multiple tumor type and tissue -specific enhancers and super-enhancers , for the sake of clarity we refer to the deleted region here as the `super-enhancer region´ . By analysis of the mice , we found that the entire super-enhancer region conferring multi-cancer susceptibility contributes to MYC expression in vivo , yet is not required for mouse embryonic development and viability . However , this region is required for the growth of normal cells in culture and cancer cells in vivo . As cultured cells are exposed to serum , which is a signal of tissue damage , this finding suggests that tumor cells and cells responding to damage signals share regulatory mechanisms that are dispensable for normal physiological growth control . To dissect functional significance of the 8q24 region during normal development , we generated series of Myc alleles in mice using homologous recombination in ES cells . These include the Myc-335 enhancer deletion allele we have described previously ( Sur et al . , 2012b ) , and deletions of two additional conserved enhancer elements , Myc-196 and Myc-540 , both of which are active in mouse intestine and colorectal cancer cells . In addition , we generated a point mutation that inactivates a conserved CCCTC-Binding factor ( CTCF ) site 2 kb upstream of the Myc TSS . This site has previously been reported to be required for MYC expression ( Gombert and Krumm , 2009 ) , and to have insulator activity ( Gombert et al . , 2003 ) ( Figure 1a ) . Each allele contained loxP site ( s ) in the same orientation to allow conditional knockouts of the enhancers , and to facilitate generation of large deletions and duplications by interallelic recombination ( Wu et al . , 2007 ) . All alleles were bred to homozygosity , and resulted in generation of viable mice . Expression of Myc in the colon of Myc-196−/− and Myc-540−/− mice was not markedly altered , suggesting that these elements have little effect on regulation of Myc in the intestine under normal laboratory conditions ( Figure 1b ) . Myc expression level was also normal in Myc-CTCFmut/mut mouse colon despite loss of CTCF and cohesin ( Rad21 ) binding to the region proximal to the Myc promoter ( Figure 1c ) . 10 . 7554/eLife . 23382 . 003Figure 1 . Cancer susceptibility region upstream of Myc contains several conserved enhancer elements that are dispensable for normal mouse development and MYC expression . ( a ) Comparison of Myc locus between human and mouse . The susceptibility regions for prostate cancer ( PrCa ) , chronic lymphocytic leukemia ( CLL ) , breast cancer ( BrCa ) , colorectal cancer ( CRC ) and bladder cancer ( BlCa ) are marked . Red vertical lines mark the location of the Tcf7l2-binding CRC Myc enhancers in the two species . The lower panel shows the regional conservation probability predicted by PhastCons ( hg19 assembly , UCSC ) with non-overlapping sliding windows for the whole region and each enhancer locus with a size of 500 bp and 10 bp , respectively . ( b ) Deletion of Myc-196 and Myc-540 enhancer elements does not affect Myc expression in the colon as determined by qPCR analysis ( Myc-196−/− n = 2 , Myc-540−/− n = 3 and wild-type n = 5 ) . See Figure 1—source data 1 for details . ( c ) Mutation of the Myc-CTCF site causes loss of CTCF and Rad21 binding at the Myc locus ( top panel ) . Binding of CTCF and Rad21 at a control Actb locus is not affected . Red and black arrowheads denote binding sites at Myc and Actb loci , respectively; green: Myc-CTCFmut/mut , blue: wild-type . The gene body for Myc and Actb is shown below the respective panels . The qPCR analysis reveals that despite loss of CTCF/cohesin binding , the expression of Myc mRNA is not altered in the colon ( for qPCR , Myc-CTCFmut/mut n = 4 , wild-type n = 3 ) . See Figure 1—source data 1 for details . Error bars denote one standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 00310 . 7554/eLife . 23382 . 004Figure 1—source data 1 . Myc transcript levels in wild-type and mutant mice in Figure 1b-c . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 004 As the individual mutations and deletions had limited effect , we next decided to generate two large deletions in the Myc locus using interallelic recombination between the Myc-CTCFmut loxP site and the loxP sites at Myc-335− and Myc-540− , yielding deletions of 365 kb ( GRCm38/mm10 chr15:61618287–61983375 ) and 538 kb ( chr15:61445326–61983375 ) , respectively ( Figure 2a ) . The resulting alleles , Myc△2-367 and Myc△2–540 , were then segregated out from the corresponding duplications , and bred to homozygosity . Given the very large regions that were deleted ( Figure 2b ) , we expected to see a strong phenotype . However , no overt phenotype was identified in the Myc△2-367/△2-367 mice . The mice were born at the expected mendelian ratio , and both males and females were viable and fertile . Analysis of Myc expression , however , revealed a strong decrease in Myc expression in the colon and ileum of the mice ( not shown ) . 10 . 7554/eLife . 23382 . 005Figure 2 . Deletion of the 8q24 super-enhancer region is well tolerated during normal development and homeostasis . ( a ) Schematic representation of the 365 kb and 538 kb deletions . ( b ) Myc△2–540/△2–540 deletion removes a region containing several active enhancer elements upstream of Myc as shown by ChIP-seq analysis of histone H3 lysine 27 acetylation ( H3K27ac ) and lysine four trimethylation ( H3K4me3 ) . The deletion also removes several Tcf7l2 ChIP-seq peaks . Signal from Myc△2–540/△2–540 and wild-type mice are shown in green and blue , respectively . Red arrowheads and horizontal lines mark the different enhancer positions . ( c ) Haematoxylin/ Eosin stained sections of spleen , bladder , prostate , colon ( Bar = 100 µm ) and Carmine Alum stained whole mounts of mammary glands , Bars = 3 mm , 100 µm ( inset ) showing normal development and homeostasis of different organs in Myc△2–540/△2–540 mice . ( d ) Myc△2–540/△2–540 mice have a reduced number of B-cells compared to the wild-types . Left panel: FACS plots of a representative Myc△2–540/△2–540 and wild-type mouse spleen showing B-cell ( B ) population . Right panel: Scatter dot plot of total number of B cells in the spleen and bone marrow of wild-type ( squares ) , n = 5 and Myc△2–540/△2–540 ( filled circles ) , n = 5 . Each point represents individual mouse . Line represents the median . See Figure 2—source data 1 for details . The number of CD4+ and CD8+ T-cells is not affected by the deletion ( see Figure 2—figure supplement 1 and appendix 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 00510 . 7554/eLife . 23382 . 006Figure 2—source data 1 . B cell numbers in the wild-type and MycΔ2-540/Δ2-540 mice in Figure 2d . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 00610 . 7554/eLife . 23382 . 007Figure 2—figure supplement 1 . The loss of the Myc super-enhancer region results in a decrease in the number of B-cells , but no major defects in hematopoiesis . ( a ) Super-enhancer region deletion results in a lower number of total cells in the spleen ( average number in wild-type = 150 . 1×106 and Myc△2–540/△2–540 = 117 . 2×106 ) which manifests as a specific decrease of B-cell numbers both in the spleen and peripheral blood without an affect on the CD4+ and CD8+ T-cell population . Line marks the median . See Figure 2—figure supplement 1—source data 1 for details . ( b ) ChIP-seq analysis of histone marks H3K4me2 ( red ) , H3K27ac ( green ) and RNA-seq ( grey: signals from plus and minus strand are shown ) shows the presence of B-cell specific enhancer both 3´and 5´ side of the Myc ORF . One of these enhancers ( black arrow , mature B-cells panel ) is located at the 5´boundary of the deleted region ( black solid bar ) . The 2–540 deletion will bring this element very close to the Myc TSS , potentially explaining the increased Myc expression in the spleen . The different enhancer regions used in this study are marked by red arrowhead and the 5´ boundary of the super-enhancer region is marked by dashed line . The increase in MYC levels in the spleen and decrease in B-cell number could potentially be explained by a direct effect of MYC in inducing apoptosis of B-cells ( Hoffman and Liebermann , 2008 ) . However , further studies are necessary for dissection of the role of specific Myc enhancers during hematopoiesis . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 00710 . 7554/eLife . 23382 . 008Figure 2—figure supplement 1—source data 1 . B and T-cell populations in the wild-type and MycΔ2-540/Δ2-540 mice in Figure 2—figure supplement 1a . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 008 The larger deletion , Myc△2–540 , could also be bred to homozygosity , and both males and females were viable . Given that the entire Myc regulatory region spans more than 2 Mb of DNA and is located on both sides of the Myc coding region ( Rosenbloom et al . , 2013; Sloan et al . , 2016 ) , the deletion is not expected to be equivalent to deletion of the Myc gene itself . Still , the viability of the mice is striking , since the region deleted contains regions linked to risk for myeloma , chronic lymphocytic leukemia and pancreatic , thyroid , bladder , prostate , breast , and colon cancers ( Chung and Chanock , 2011; Sahasrabudhe et al . , 2015; Mitchell et al . , 2016; Zhang et al . , 2016 ) . To characterize the mice further , we analyzed histology and MYC expression in the tissues where these tumors originate from . This analysis revealed normal morphology of mammary gland , spleen , bladder , prostate and colon in Myc△2–540/△2–540 mice ( Figure 2c ) . Although the Myc△2–540/△2–540 mice exhibited a normal phenotype , Myc expression was altered in a tissue-specific manner in these mice . This is expected since this region contains individual tissue specific regulatory elements . The expression of Myc was strongly decreased in colon , small intestine and prostate of these mice ( Figure 3a and not shown ) . Immunohistochemical analysis of MYC expression in intestine revealed strong decrease of nuclear staining , and loss of MYC expression from the transit amplifying cell compartment . However , expression of MYC was still detected at the base of the crypt in the region where the intestinal stem cells are known to reside ( Figure 3b ) . These results are consistent with the role of the deleted region in tumorigenesis of colon and prostate . To analyze the effect of decreased MYC expression on the proliferation in the transit amplifying compartment , we performed immunohistochemistry ( IHC ) for the proliferation marker Ki-67 . Both the wild-type and Myc△2–540/△2–540 had similar proliferative activity in the intestinal crypts ( Figure 3b ) . 10 . 7554/eLife . 23382 . 009Figure 3 . Tissue-specific effect of Myc△2–540/△2–540 deletion on MYC expression . ( a ) qPCR data showing the percentage of Myc expression in Myc△2–540/△2–540 relative to the wild-type in colon ( Co ) n = 4 , prostate ( Pr ) n = 2 , bladder ( Bl ) n = 5 , spleen ( Sp ) n = 4 and mammary gland ( Ma ) n = 3 . Red line marks the expression level ( 100% ) in wild-type mice . Error bars indicate one standard deviation . See Figure 3—source data 1 for details . ( b ) Immunohistochemistry shows reduced expression of MYC ( n = 3 for each genotype ) protein in intestinal crypts of Myc△2–540/△2–540 mice without any significant effect on proliferation as indicated by Ki-67 ( n = 2 for each genotype ) immunostaining , Bar = 10 µm . Brown: IHC staining , Blue: Haematoxylin staining . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 00910 . 7554/eLife . 23382 . 010Figure 3—source data 1 . Myc transcript levels in MycΔ2-540/Δ2-540 mice relative to the wild-types in Figure 3a . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 010 In contrast to colon and prostate , Myc expression was not markedly affected in the bladder , and was elevated in the spleen ( Figure 3a ) . To analyze the cellular composition of the spleen , we performed flow cytometric analysis of markers for hematopoietic stem cells and lymphoid lineage cells . Myc△2–540/△2–540 mice had a near normal hematopoietic compartment ( Figure 2d ) . The only observed difference was a small reduction of B cells in the Myc△2–540/△2–540 mice compared to the wild-type mice both in the spleen and the bone marrow . In contrast to the decrease in B-cells , the T cell numbers were not affected by the deletion ( Figure 2—figure supplement 1a ) . This finding is consistent with the published data that regulatory elements controlling T-cell development and T-cell acute lymphoblastic leukemia are located 1 . 47 Mb downstream of the Myc ORF ( Herranz et al . , 2014 ) . To identify regulatory elements that could explain the effect in B-cells , we performed ChIP-seq analysis of chromatin from LSK-Flt3neg hematopoietic stem cells and mature B-cells isolated from wild-type mice . This analysis identified two B-cell specific regulatory elements . The Myc 2–540 deletion results in loss of one of the elements , and moves the other element very close to the Myc TSS ( Figure 2—figure supplement 1b ) . Although the exact regulatory mechanism is not clear and requires further study , the above data is consistent with a role of the super-enhancer region in development of chronic lymphocytic leukemia , which is primarily a B-cell malignancy . However , the decrease in B-cell number does not affect viability , and the Myc△2–540/△2–540 mice are healthy and do not display an immune-deficient phenotype under normal ‘clean’ mouse housing conditions in the absence of known pathogenic microorganisms . To compare the role of the 8q24 super-enhancer region in growth of cells in vivo and in cell culture , we isolated fibroblasts from the skin of adult Myc△2–540/△2–540 and wild-type mice . Based on presence of active histone marks , and undermethylation of focal elements , the super-enhancer region is active in fibroblasts from both humans and mice ( Figure 4a and Figure 4—figure supplement 1 ) . However , the resident fibroblasts in the skin of Myc△2–540/△2–540 mice appeared normal as judged by Vimentin expression ( Figure 4b ) . Ki-67 staining ( IHC ) of skin sections showed comparable proliferation levels in wild-type and Myc△2–540/△2–540 mice ( Figure 4b ) . In contrast , most lines of fibroblasts ( 6 out of 7 ) isolated from Myc△2–540/△2–540 mice grew slower in culture compared to fibroblasts isolated from wild-type mice ( Figure 4c; p-value=0 . 0256 , Mann-Whitney one tailed test ) . 10 . 7554/eLife . 23382 . 011Figure 4 . Myc△2–540/△2–540 deletion results in a proliferation defect of adult skin fibroblasts cultured in vitro . ( a ) The super-enhancer region deleted in the Myc△2–540/△2–540 has under methylated DNA as determined through bisulfite sequencing of the wild-type fibroblasts grown in culture . H3K27ac ChIP-seq shows the presence of active enhancer marks within this region in the wild-type fibroblasts whereas the Myc△2–540/△2–540 fibroblasts show a complete absence of the super-enhancer region . The Myc super-enhancer region is also active in human fibroblasts ( see Figure 4—figure supplement 1 ) . ( b ) Normal morphology and proliferation of resident fibroblasts in the mouse skin as determined by Vimentin and Ki-67 IHC staining respectively in both the wild-type ( n = 3 ) and Myc△2–540/△2–540 mice ( n = 3 ) , Bar = 50 µm . Brown: IHC staining , Blue: Haematoxylin staining ( c ) Representative phase contrast images of wild-type and Myc△2–540/△2–540 primary fibroblasts showing growth defect of Myc△2–540/△2–540 fibroblasts in culture . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 01110 . 7554/eLife . 23382 . 012Figure 4—figure supplement 1 . The Myc super-enhancer region is also active in human fibroblasts . ChIP-seq analysis of H3 lysine 27 acetylation ( H3K27ac ) , cohesion ( SMC1A ) and CTCF binding shows several active enhancer marks in the super-enhancer region upstream of the MYC gene in human fibroblasts . Red arrows mark the positions of the conserved enhancer elements used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 012 To understand the mechanism by which the deletion of the 8q24 super-enhancer region has a differential effect on growth during normal tissue homeostasis and growth under culture conditions , we subjected both the mouse tissues and cultured cells to RNA-seq analysis . Analysis of mouse tissues confirmed the changes in Myc expression observed by qPCR ( Figure 5a and Figure 5—figure supplement 1 ) . Surprisingly , despite more than 80% decrease of Myc expression in the colon , very few genes were downregulated in the tissues , and none of the significantly altered genes were known MYC targets ( Supplementary file 1 ) . These results suggest that expression of canonical MYC target genes is not sensitive to decreases in MYC protein level during normal tissue homeostasis . In contrast to the in vivo situation , where Myc is downregulated but key target genes are not affected , in cultured Myc△2–540/△2–540 fibroblasts that grew slowly in culture , the downregulation of Myc lead to a loss of expression of key target genes that drive cell growth and division . Upstream regulator analysis performed using Ingenuity Pathway Analysis revealed that the highest-ranked potential regulator for the identified gene set was MYC ( Figure 5b ) . 10 . 7554/eLife . 23382 . 013Figure 5 . Differential effect of Myc△2–540/△2–540 deletion on MYC target gene expression . ( a ) Scatter plot comparing the average Fragments per kilobase of exons per million fragments mapped ( FPKM ) values of gene transcripts in colon and spleen of wild-type ( n = 4 ) and Myc△2–540/△2–540 ( n = 4 ) mice . Genes showing significant ( q < 0 . 05 ) differential expression are marked in red ( Myc ) or green ( other genes ) . For median FPKM values of gene transcripts see Figure 5—figure supplement 1 ( b ) Upstream regulator analysis of RNA-seq data shows that the highest ranked potential regulator affected in the slow growing Myc△2–540/△2–540 fibroblasts is MYC . The activation z-scores are to infer the activation states of predicted upstream regulators . The overlap p-values were calculated from all the regulator-targeted differential expression genes using Fisher’s Exact Test . Two independent Myc△2–540/△2–540 fibroblasts lines were analysed to confirm the downregulation of Myc . Ingenuity pathway analysis performed on one of these is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 01310 . 7554/eLife . 23382 . 014Figure 5—figure supplement 1 . Scatter plot comparing the median of FPKM values of gene transcripts in colon of wild-type ( n = 4 ) and Myc△2–540/△2–540 ( n = 4 ) mice . Genes showing significant ( q < 0 . 05 ) differential expression are marked in red ( Myc ) or green ( other genes ) . The plot was generated using ggplot2 ( version 2 . 2 . 1 , RRID:SCR_014601 ) . The median FPKM values of gene transcripts were plotted for the colon data since one of the wild-type samples for unknown reason had higher amounts of ribosomal structural protein transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 014 Measured by FPKM values , the cultured wild-type fibroblasts had higher Myc mRNA levels than normal tissues , whereas the cultured null fibroblasts had Myc levels that were comparable to or lower than those of normal wild-type tissues . The elevated Myc levels in cultured cells are caused by serum stimulation , as Myc mRNA levels are low in serum-starved fibroblasts and strongly induced by serum ( Ref . [Dean et al . , 1986] and our unpublished data ) . These results indicate that the 8q24 super-enhancer region is dispensable for normal tissue homeostasis under conditions where MYC activity is relatively low . However , the region is required for induction of MYC activity to levels that are high enough to drive the expression of MYC target genes above their basal levels during pathological growth . We have shown earlier that deletion of a 1 . 7 kb Myc-335 enhancer sequence located at the 8q24 super-enhancer region is required for intestinal tumorigenesis in mice ( Sur et al . , 2012b ) . As the super-enhancer region deleted in Myc△2–540/△2–540 mice carries risk also for other cancer types , including breast and bladder cancer , we tested the susceptibility of the Myc△2–540/△2–540 mice to carcinogen induced bladder and mammary tumorigenesis . The Myc△2–540/△2–540 mice were not resistant to N-Butyl-N ( 4-hydroxybutyl ) nitrosamine ( BBN ) induced bladder tumors . Both wild-type ( n = 8 ) and Myc△2–540/△2–540 ( n = 8 ) mice developed urothelial changes ranging from hyperplasia to high grade invasive urothelial carcinoma after 5 months of BBN treatment . In contrast , comparison of median tumor-free survival times of wild-type and Myc△2–540/△2–540 mice exposed to mammary-tumor inducing dimethylbenz[a]anthracene/ medroxypregesterone ( DMBA/MPA ) regimen revealed that the Myc△2–540/△2–540 mice were partially resistant to mammary tumorigenesis ( Figure 6a ) . The median tumor-free survival time for the wild-type and Myc△2–540/△2–540 mice was 88 and >120 days , respectively . Although we cannot pinpoint the specific regions that contribute to breast tumorigenesis by analysis of the Myc△2–540/△2–540 mice , our work is consistent with the presence of a breast cancer susceptibility locus in humans at a region syntenic to the deletion . The region is distinct from the colon cancer susceptibility locus that harbors Myc-335 . 10 . 7554/eLife . 23382 . 015Figure 6 . Myc −2 to −540 kb genomic region is required for the growth of cancers in vivo and cancer cells in vitro . ( a ) Tumor-free survival plots showing resistance of Myc△2–540/△2–540 mice to development of DMBA/MPA induced mammary tumors . p-value=0 . 0002 ( Mantel-Cox Log-rank test ) . See Figure 6—source data 1 for details . ( b ) The Myc −2 to −540 kb deletion results in fewer polyps than the Myc-335 deletion alone . p-value=0 . 00019 ( Students T-test , 2-tailed ) . Apcmin mice were of 4 months of age ( n = 5 ) and Apcmin; Myc△2–540/△2–540 mice were 6 months old ( n = 5 ) at the time of analysis . Filled circles correspond to individual mice and red color denotes the median . See Figure 6—source data 1 for details . Bar equals 5 mm . ( c ) Crispr-Cas9 mediated deletion of region corresponding to Myc△2–540/△2–540 in human GP5d colon cancer cells , results in a loss of the edited cells over time . Top panel shows the active enhancer elements in GP5d cells within this region as determined by ChIP-seq analysis of histone H3 lysine 27 acetylation ( H3K27ac ) . The sites of sgRNAs ( black lines ) and genotyping primers ( blue arrows ) used are indicated ( not to scale ) . Red arrows mark the enhancer regions used in this study . Bottom panel shows the PCR-genotyping of the MYC locus and the control IGH locus showing the specific loss of the cells with the edited MYC locus over time . GAPDH was used as internal control . The right panel in each set shows absence of any deletion in the non-transfected cells ( day 2 ) . 100 bp ladder DNA molecular weight marker is shown ( M ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 01510 . 7554/eLife . 23382 . 016Figure 6—source data 1 . Survival time and intestinal polyp numbers for mice in Figure 6a-b . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 016 To determine whether additional elements outside of the Myc-335 region are playing a role in tumorigenesis , we crossed the Myc△2–540/△2–540 mice with the Apcmin mouse that is susceptible to intestinal tumors . The Myc△2–540/△2–540 mice had fewer polyps than the Myc-335−/− mice in the Apcmin background . In this study the wild-type mice had on an average 56 polyps at around 4 months of age ( n = 5 ) when they had to be euthanized for ethical reasons similar to what we reported previously . The Apcmin; Myc△2–540/△2–540 looked healthy and had on an average 2 . 4 polyps even at 6 months of age ( n = 5 ) compared to an average of 14 . 33 polyps reported for the Apcmin;Myc-335−/− null mice at 4 months of age ( Figure 6b ) . Together with our earlier findings , these results indicate that loss of the 8q24 super-enhancer region makes mice resistant to both genetically and chemically induced tumors . We further tested the requirement of this region for the proliferation of cancer cell lines in cultures . We found that the corresponding region ( hg19: chr8:128226490–128746456 ) was also required for GP5d colon cancer cell growth , as indicated by progressive loss of cells bearing a CRISPR/Cas9 induced deletion of the region during co-culture with unedited cells in the population ( Figure 6c ) . The region around the MYC gene carries inherited risk towards multiple major forms of cancer . On aggregate , this region contributes more to inherited cancer than any other locus in the human genome . The risk alleles for different cancer types are located in multiple distinct linkage disequilibrium blocks , indicating that different variants contribute to different cancer types . Several of these regions containing risk variants have been implicated in regulation of MYC expression ( Hallikas et al . , 2006; Sur et al . , 2012b; Herranz et al . , 2014; Uslu et al . , 2014 ) , suggesting that a large number of enhancers within this region can drive tumorigenesis . Some of the identified elements have also been shown to have roles in normal development ( Herranz et al . , 2014; Uslu et al . , 2014 ) . To study the role of the 8q24 region more systematically , we have in this work deleted several individual enhancer elements , and also analyzed the effect of larger deletions on normal development and carcinogenesis in mice . Our analysis of mice lacking a 538 kb region upstream of the Myc gene suggests that enhancer elements within this region cooperatively enhance Myc expression . Deletion of individual enhancers in this region has only a weak ( Sur et al . , 2012b ) or no effect on Myc expression in the mouse intestine in contrast to the deletion of the entire super-enhancer region , which leads to severe decrease in Myc expression in multiple tissues . MYC deficient mouse embryos die due to placental defect at E9 . 5 . The embryos are also smaller in size than wild-type embryos ( Davis et al . , 1993 ) . However , when Myc is deleted only in the epiblast , the embryos grow normally and survive until E11 . 5 , when they die due to defects in hematopoiesis ( Dubois et al . , 2008 ) . None of these defects are observed in mice homozygous for the deletion of super-enhancer region . The 8q24 super-enhancer region is thus dispensable for MYC function both in the placenta and during early hematopoiesis . In our mouse colony , the super-enhancer region deficient mice also do not display the size or weight differences reported for Myc heterozygous mice that have a 50% reduction in MYC activity ( Trumpp et al . , 2001 ) . These results indicate that tissue-specific enhancers that reside outside of the deleted regions drive sufficient MYC expression in the tissues that contribute to the phenotypes observed in Myc+/− and Myc−/− mice . Consistently with this , several hematopoietic enhancers have been identified in the region 3' of the MYC ORF ( Hnisz et al . , 2013; Shi et al . , 2013; Herranz et al . , 2014 ) . Myc heterozygous mice also display increased longevity and enhanced healthspan ( Hofmann et al . , 2015 ) . Although the deletion of the super-enhancer region that contains tissue-specific enhancers regulating MYC expression is not equivalent to a heterozygous deletion of the Myc gene in the whole body , the Myc△2–540/△2–540 mice could be an interesting model for identification of the tissues that contribute to the longevity phenotype . Despite decreased levels of MYC in multiple adult tissues , the mice lacking the super-enhancer region are viable , fertile and display normal tissue morphology in all the tissues we investigated . They display no overt phenotype and do not have marked defects in cell proliferation . The mice are , however , resistant to intestinal tumorigenesis , and DMBA-induced mammary tumors , indicating that this region is important for tumorigenesis also in mice . Our data thus shows that despite the central role of this region in tumorigenesis ( Sur et al . , 2012b; Lovén et al . , 2013 ) , it is dispensable for normal tissue development and homeostasis under laboratory conditions . Whereas this result may appear very surprising , it is consistent with the original identification of this region using genome-wide association studies ( GWAS ) . GWAS has a high power to identify common variants , and most variants that are common have only a limited effect on physiological functions . This is because a variant that has strong positive or negative effect is rapidly fixed or lost , respectively . Thus , GWAS are specifically biased to find variants that have a relatively large effect on disease , but a small effect on fitness . Most genes in mammals do not have haploinsufficient phenotypes . Such buffering could be due to mechanisms that maintain constant expression level irrespective of gene dose . However , a simpler buffering mechanism involves either expressing a gene at a very low level where it has no effect , or at a high level where it can contribute its functions even if its expression level is decreased due to transcriptional noise or loss-of-function of one allele . A similar two state mechanism where physiological transcription factor ( TF ) activity levels in the relevant cell types are either too low to drive any target genes ( off state ) , or high enough to activate all important targets ( on state ) could also mechanistically explain why most heterozygous null mutations of TF genes have no apparent phenotype . Our analysis of the role of MYC in normal colon is consistent with such a simple buffering model ( Figure 7 ) . However , it should be noted that this buffering mechanism does not operate in all tissues and under all conditions . For example , Myc gene dose has effects on mouse size , longevity and hematopoiesis ( Davis et al . , 1993; Trumpp et al . , 2001; Dubois et al . , 2008; Hofmann et al . , 2015 ) . In addition , the level of expression of the Myc gene has quantitative effects on cell proliferation under pathological conditions such as activation of T-cells ( Heinzel et al . , 2017 ) . These results indicate that in some situations , MYC is expressed at a level where cell growth responds linearly to small changes in MYC levels ( Figure 7 , middle panel ) . However , the lack of an overt phenotype in our model under normal physiological conditions in the absence of infection or tissue damage suggests that growth during normal tissue homeostasis in at least some adult tissues does not linearly respond to changes in MYC levels . The lack of an overt phenotype should not , however , be taken to mean that the mice have no phenotype at all . As the super-enhancer region contains several highly conserved DNA segments , and affects cell growth in culture , we expect that it will also affect responses to tissue damage or some other perturbation that we have not investigated here . Therefore , further studies are needed to determine the role of the super-enhancer region in various chronic and acute models of injury and infection . 10 . 7554/eLife . 23382 . 017Figure 7 . Model showing the activity of the Myc super-enhancer region during normal homeostasis ( left ) and cancer ( right ) . During normal tissue homeostasis ( left ) , Myc enhancers are not strongly active , and MYC activity is relatively low . The MYC expression level is insufficient to recruit enough MAX proteins to MYC/MAX heterodimers to drive strong induction of the MYC target genes , which instead remain under the control of basal transcription factors ( BTF ) . Under conditions of normal rapid proliferation as seen during embryonic development or during pathological insults in the adult , MYC is expressed at intermediate levels to elicit response from targets with high affinity binding sites ( red ) . In cancer cells or cells grown in culture ( right ) , upstream regulators such as Tcf7l2 and β-catenin activate the Myc super-enhancers , driving high levels of MYC expression . This leads to the formation of MYC/MAX heterodimers that strongly activate transcription of MYC target genes driving cancer cell growth . The high levels of MYC are also sufficient to induce target genes that harbor low affinity MYC binding sites ( grey ) . The model is consistent with the model of Lorenzin et al ( Lorenzin et al . , 2016 ) who showed genes differ in their response to MYC levels due to differences in the MYC affinity of their promoters . Given that Myc super-enhancer region is tumor-specific , and induction of the MYC target genes are not required for normal homeostasis , it provides a promising target for antineoplastic therapies . DOI: http://dx . doi . org/10 . 7554/eLife . 23382 . 017 Based on our data and the earlier literature we propose that under normal physiological conditions in the intestine , the Myc gene regulatory system is in the off state , and a basal level of expression of the MYC target genes is maintained by a MYC-independent mechanism . The target genes are thus only sensitive to an increase in MYC levels . Consistently , an 80% decrease of Myc mRNA expression does not lead to a proliferation defect , or major changes in expression of known MYC target genes . In contrast , in tumors the system is locked to an on state , where MYC targets are driven to a maximal level by MYC , and the targets are now only sensitive to a decrease in MYC activity ( Figure 7 ) . The requirement of MYC in tumor cells appears absolute . In transgenic animal models , overexpression of MYC leads to deregulated proliferation and tumor development in multiple tissues ( Felsher and Bishop , 1999; Pelengaris et al . , 1999; D'Cruz et al . , 2001; Jain et al . , 2002; Shachaf et al . , 2004 ) . Furthermore , inhibition of MYC almost invariably causes growth arrest of cancer cells both in culture and in vivo ( Soucek et al . , 2002 , 2004; Hart et al . , 2014 ) . Despite the importance of MYC for cancer growth , it appears that the role of MYC in controlling growth during adult tissue homeostasis is limited . In the adult tissues , MYC is expressed in rapidly proliferating compartments of the body like the intestinal crypts and skin . Deletion of Myc in these compartments does not result in prominent proliferation defects ( Wilson et al . , 2004; Baena et al . , 2005; Bettess et al . , 2005; Muncan et al . , 2006 ) . Although there is still controversy regarding MYC requirement for the intestinal homeostasis , in the skin MYC is dispensable under normal adult proliferation and homeostasis in vivo ( Oskarsson et al . , 2006 ) . It is however required for Ras mediated tumorigenesis and growth of fibroblasts and keratinocytes in vitro ( Mateyak et al . , 1997; Oskarsson et al . , 2006 ) . Taken together , these results suggest that MYC is required for pathological proliferation , but is less important and in many cases dispensable for normal homeostasis of tissues in the adult . Our results are consistent with these observations . Prior to our study it was not clear whether the MYC dependence of cancer cells in vivo and normal cells in culture is due to shared regulatory mechanisms . Our results have uncovered striking mechanistic similarities between growth of normal cells in culture , and growth of cancer cells in vivo by showing that MYC expression depend on the same genetic elements in cultured normal cells and in cancer cells . The similarity between tumor cells and cultured normal cells also suggest that many potential drugs that block cancer cell growth may have been inadvertently discarded due to their negative effects on growth of normal cells in culture , even when they might not have affected normal tissue homeostasis in vivo . Our results show that the MYC super-enhancer region that carries multi-cancer susceptibility in humans contributes to the formation of multiple tumor types also in mice . Despite its role in tumor formation , it is dispensable for normal development and homeostasis . Loss of the super-enhancer region leads to low MYC expression , but the lowered expression does not translate to changes in expression of MYC target genes in the intestine . Thus , the MYC/MAX/MNT system ( Grandori et al . , 2000 ) that drives cell growth and proliferation is robustly set to an off state during normal homeostasis , whereas in cancer , the system is locked to a pathological on state . This also explains how physiological growth control can be robust to small perturbations and transcriptional noise . Taken together , our results reveal an important difference between the transcriptional states of normal and cancer cells , and suggest that therapeutic interventions that decrease the activity of the Myc super-enhancer region would be well tolerated . We generated cKO Myc-196 and cKO Myc-540 strains with loxP sites flanking the regions chr15:61445326–61447611 and chr15:61789274–61791107 , respectively ( Taconic ) . These mice were crossed to EIIa-cre mouse strain ( Jackson Laboratory ) to generate mice with enhancer deletions . Myc-CTCFmut mouse strain was generated by mutating the CTCF-binding site at chr15:61983375–61983647 TGGCCAGTAGAGGGCAC to TGGAACGTCTTGAATGC . In order to generate large deletions at the Myc locus ( Myc△2-367 and Myc△2–540 ) Myc-367− and Myc-540− were crossed to Myc-CTCFmut that were also heterozygous for the Rosa26-Cre ( Taconic ) . The Myc-540− , Myc-196− and Myc-CTCFmut carry one lox P site at the respective loci ( chr15:61445326 , chr15:61618287 and chr15:61983375 ) . The loxP site on chr15:61983375 is located immediately 5' of the mutant CTCF binding site . We obtained compound heterozygotes carrying the chr15:61445326 or the chr15:61618287 loxP site together with the loxP site on chr15:61983375 and the Rosa26-Cre . The compound heterozygotes were screened by PCR for the interallelic recombination and the resultant deletion and duplication of the intervening sequence . Mice mosaic for the deletion and duplication were backcrossed to the C57Bl/6 mice in order to segregate the chromosomes carrying the deletion . The F1 heterozygotes were intercrossed to generate mice with homozygous large deletions . Myc-335 strain has been previously described ( Sur et al . , 2012b ) . All mice used in the study were on a C57Bl/6 genetic background . All mouse experiments were conducted in accordance with the local ethical guidelines , after approval of the protocols by the ethics committee of the Board of Agriculture , Experimental Animal Authority , Stockholm South , Sweden ( Dnr S50/13 , S11/15 and S16/15 ) . The sequences of the different primer pairs used for genotypings are given in Supplementary file 2 . Inguinal mammary glands were removed from 8 week old virgin females and spread on glass slides . These were fixed for 4 hr in Carnoy’s fixative and subsequently stained O/N with Carmine Alum . The whole mounts were rinsed and dehydrated through increasing series of ethanol and cleared in xylene before mounting with the pertex mounting medium . qPCR was performed as described previously ( Sur et al . , 2012b ) . Essentially , total RNA was isolated from whole tissue by homogenizing in RNA Bee reagent ( ambios AMS Biotechnology ) followed by RNA isolation using Qiagen’s RNA MinElute kit according to manufacturers' protocols . 0 . 5–1 µg of total RNA was reverse transcribed using high capacity reverse transcription kit in a 20 μl reaction ( Applied Biosystems ) . Quantitative PCR in triplicates was performed using the SYBR select master mix ( Applied Biosystems ) on the LightCycler 480 instrument ( Roche ) . For normalization , mouse β-actin transcripts were used as internal controls . Following primer pairs were used for quantitative PCR analysis . Myc-Fw: 5'-GGGGCTTTGCCTCCGAGCCT-3' , Myc-Rev: 5'-TGAGGGGCATC GTCGTGGCT-3' , β-actin-Fw: 5'CTGTCGAGTCGCGTCCACCCG-3' , β-actin-Rev: 5'-CATGCCGGAGCCGTTGTCGAC-3' . NEBNext Ultra Directional RNA library Prep kit ( NEB ) was used for preparing the samples for RNA-seq together with the NEBNext Poly ( A ) mRNA magnetic isolation module ( NEB ) according to manufacturers protocol . In the case of tissues 1–2 µg and for cultured fibroblasts 200 ng of total RNA was used as starting material . For library preparation , adapters and index primers from NEBNext Multiplex Oligos for Illumina kit were used . The RNA-seq library was sequenced on a HiSeq2000 ( Illumina ) . Sequencing reads were mapped to the mouse reference genome ( NCBI37/mm9 ) using Tophat2 ( version 2 . 0 . 13; RRID:SCR_013035 ) ( Kim et al . , 2013 ) . Cuffdiff ( version 2 . 2 . 1; RRID:SCR_001647 ) was used for differential gene expression analysis and for graphical representation , CummeRbund package ( version 2 . 8 . 2; RRID:SCR_014568 ) ( Trapnell et al . , 2012 ) was used . The upstream regulator analysis was performed on all the significant differentially expressed genes ( Cuffdiff q-value <0 . 05 ) using QIAGEN’s Ingenuity Pathway Analysis ( IPA , QIAGEN Redwood city , www . qiagen . com/ingenuity; version 24718999 , updated 2015-09-14; RRID:SCR_008653 ) . ChIP-seq was performed as described in ( Sur et al . , 2012b; Yan et al . , 2013 ) with the following modifications: Adult 8–10 week old mice were euthanized and colon was removed , rinsed with cold PBS and cut into fine pieces . Tissue was crosslinked with 1 . 5% formaldehyde and cultured cells were crosslinked with 1% formaldehyde for 10 min at room temperature and quenched with 0 . 33M Glycine . Sequences were mapped to the mouse reference genome ( NCBI37/mm9 ) and human reference genome ( hg19 ) using Burrows-Wheeler Alignment tool ( bwa ) ( version 0 . 6 . 2 ) ( Li and Durbin , 2009 ) with default parameters . All antibodies used in ChIP-seq experiments were ChIP-grade . In each experiment a non-specific IgG was used as control . Following antibodies were used for ChIP-seq experiments: rabbit anti-H3 lysine 27 acetylation ( H3K27ac ) ( abcam , ab4729: RRID:AB_2118291 ) , mouse anti-H3 lysine four trimethylation ( H3K4me3 ) ( abcam , ab1012; RRID:AB_442796 ) , rabbit anti-Rad21 ( Santa Cruz , sc-98784; RRID:AB_2238151 ) , goat anti-CTCF ( Santa Cruz , sc-15914X; RRID:AB_2086899 ) , rabbit anti-SMC1A ( Bethyl Laboratories , A300-055A; RRID:AB_2192467 ) , rabbit IgG ( Santa Cruz , sc-2027; RRID:AB_737197 ) , mouse IgG ( Santa Cruz , sc-2025; RRID:AB_737182 ) , goat IgG ( Santa Cruz , sc-2028; RRID:AB_737167 ) . ChIPseq data for Tcf7l2 was used from ENA accession number PRJEB3354 ( Sur et al . , 2012a ) and for GP5d cells from ENA accession number PRJEB1429 ( Yan et al . , 2013a ) . For visualization , ChIP-seq read depth data were average smoothed across windows of 10 pixels ( H3K27ac and H3K4me3 ) or five pixels ( Tcf7l2 ) in UCSC Genome Browser; RRID:SCR_005780 or alternatively visualized in Integrative Genomics Viewer ( IGV , version 2 . 3; RRID:SCR_011793 ) . Genomic DNA was isolated using Qiagen’s Blood & Tissue Genomic DNA extraction kit . Around 1 µg of wild-type and 250 ng of Myc△2–540/△2-540 null fibroblast genomic DNA was sonicated to 300 bp fragments using Covaris S220 sonicator . Subsequent to end polishing and A base addition , cytosine methylated paired end adapters ( Integrated DNA technologies ) were ligated to the DNA fragments . The adapter sequences are as follows 5'-P-GATCGGAAGAGCGGTTCAGCAGGAATGCCGAG 5'-ACACTCTTTCCCTACACGACGCTCTTCCGATCT After adapter ligation 300–600 bp fragments were size-selected on a 2% agarose gel . Bisulfite-conversion was carried out using ZYMO EZ DNA Methylation-Gold kit ( cat . no . D5005 ) . PCR amplification with 12 and 18 cycles was carried out to prepare libraries from the wild-type and Myc△2–540/△2–540 null mouse fibroblasts , respectively . The primer pair used for PCR amplification was as follows Five micron paraffin embedded tissue sections were processed for immuno-histochemistry as previously described ( Sur et al . , 2012b ) . Rabbit polyclonal anti-Myc ( Santa Cruz , sc-764; RRID:AB_631276 ) ( 1:500 ) , Rabbit monoclonal anti Ki-67 ( abcam , ab16667; RRID:AB_302459 ) ( 1:200 ) , Goat polyclonal anti-Vimentin ( Santa Cruz , sc-7557; RRID:AB_793998 ) ( 1:500 ) , biotinylated goat anti-Rabbit IgG ( Vector Laboratories , BA1000; RRID:AB_2313606 ) and biotinylated rabbit anti-Goat IgG ( Vector Laboratories , BA5000; RRID:AB_2336126 ) ( 1:350 ) antibodies were used . For flow cytometry , single cell suspensions of spleen and bone-marrow and cells from peripheral blood were stained with Fc-block ( CD16/CD32 clone 93 , Biolegend , 101302 , RRID:AB_312801 ) and subsequently with CD19 ( clone 1D3 , BD Biosciences , RRID:AB_11154223 ) , TER119 ( clone TER119 , Biolegend 116210 , RRID:AB_313711 ) , CD3ε ( clone 145–2 C11 , Biolegend 100308 , RRID:AB_312673 ) , NK1 . 1 ( clone PK136 , Biolegend , 108716 , RRID:AB_493590 ) , GR1/LY6G ( clone RB6-8C5 , Biolegend , 108410 , RRID:AB_313375 ) , CD4 ( clone RM4-5 , BD Biosciences , 563747 ) and CD8a ( clone 53–6 . 7 , BD Biosciences , 563332 ) . Dead cells were visualized using Propidium iodide . Samples were analyzed using a BD LSRFortessa instrument . Fibroblasts were isolated from adult mice by dissecting the skin to ~1 mm3 pieces , and allowing the pieces to adhere to cell culture plates , followed by addition of DMEM medium supplemented with 10% FCS and antibiotics . The fibroblasts were allowed to migrate out from the explants , after which the cells were collected by trypsinization and passaged in the same media for 1–3 passages . For growth assays , 2 × 103 cells were plated per well in 96 well plates . Cells were trypsinized and counted using hemocytometer at respective time points . CRISPR-Cas9 mediated deletion of MYC super-enhancer region on chromosome 8q24 ( GRCh37/hg19 chr8: 128226403–128746490 ) and Immunoglobulin Heavy ( IGH ) gene locus on chromosome 14q32 . 33 ( GRCh37/hg19 chr14: 106527004–107035452 ) were carried out in GP5d ( Sigma , 95090715; RRID:CVCL_1235 , confirmed by STR profiling at ECACC ) colon cancer cell line stably expressing Cas9 protein . A lentiviral plasmid containing Cas9 fused via a self-cleaving 2A peptide to a blasticidin resistance gene , was packaged into lentiviral particles using the packaging plasmids psPAX2 ( a gift from Didier Trono , Addgene plasmid # 12260 , RRID:SCR_002037 ) and pCMV-VSV-G ( a gift from Robert Weinberg ( Addgene plasmid # 8454 , RRID:SCR_002037 ) . The virus was used to transduce GP5d colon cancer cells . 48 hr after transduction , GP5d cells expressing Cas9 ( GP5d-Cas9 ) were selected in 5 µg/ml Blasticidin ( Thermo Fisher Scientific Inc . , Cat . no . A1113903 ) . The single guide RNA ( sgRNA’s ) were designed ( http://www . broadinstitute . org/rnai/public/analysis-tools/sgrna-design ) to span the entire MYC super-enhancer region and IGH locus ( Figure 6 ) , respectively ( Eurofins MWG Operon ) . sgRNAs were cloned into an sgRNA Cloning Vector ( Addgene Plasmid #41824 , RRID:SCR_002037 ) using Gibson assembly master mix ( NEBuilder HiFi DNA assembly Master Mix , Cat no . E2621S ) . GP5d-Cas9 ( 2 × 106 ) cells were transfected ( using FuGENE HD Transfection Reagent , Cat . no E2312 ) with 10 µg of eight pooled equimolar sgRNA constructs . Post transfection half of the cultured cells were collected for PCR genotyping , while the other half was re-plated for culturing . Cells were collected at day 2 , 4 and subsequently every fourth day till day 32 . DNA from cells was extracted ( using DNeasy Blood & Tissue Kit , Qiagen Cat . no . 69506 ) and genotyped with 300 ng of DNA at following conditions - Initial denaturation of 95°C for 5 min; denaturation of 98°C for 15 s , annealing at 60°C for 30 s , extension at 72°C for 30 s ( 30 cycles for MYC super-enhancer region and 35 cycles for IGH gene locus deletion genotyping ) ; final extension at 72°C , 5 min . Each experiment was done in triplicate . The sequences of the different guide RNAs and primer pairs used for PCR genotyping of the deletions are given in Supplementary file 2 . GP5d cells were cultured in DMEM supplemented with 10% FBS and antibiotics . The cell line was mycoplasma free .	Our cells each contain close to 20 , 000 genes , which provide the instructions needed to build our bodies and keep us alive . At any one time in the life of the cell , only some of these genes are active . The activity of each gene is constantly regulated to allow the cell to respond to changes in its environment . Enhancers are sections of DNA , outside of the genes , that act as molecular switches controlling the activity of genes . A gene can have many such enhancers; each enhancer is linked to a particular set of signals and having multiple enhancers allows the same gene to be activated by different signals in different tissues in the body . Changes to enhancers can have serious consequences . By altering the activity of genes , an enhancer can have widespread effects on the health and behavior of a cell , including transforming it from healthy to cancerous . The small differences in enhancers also make some people more susceptible to cancers than others . If we can identify enhancers whose activity is commonly altered in cancers , it could be possible to target them through treatment . Yet , it is not clear whether targeting enhancers in this way could be effectively used to treat cancer without damaging healthy cells . Now , Dave , Sur et al . have examined a large enhancer region with known links to several different cancers – including prostate , breast and colon cancers – to uncover whether it also plays a critical role in healthy cells and if it could be safely targeted for treatment . The region has multiple enhancers for a cancer-linked gene called MYC and is implicated in many cancer-associated deaths every year . This particular enhancer region is found in both humans and mice , which share many genes in common . Using genetic engineering , Dave , Sur et al . removed this enhancer region from the genetic information of a group of mice . The experiment showed that mice without the enhancer region were completely healthy . Also , when tested for cancer development , these mice were much less susceptible to several major types of cancer . This investigation reveals that it may be possible to create drugs to shut down or inhibit certain enhancers to prevent or treat cancer without damaging healthy cells . However , this is currently just one example in mice under laboratory conditions . Further research is needed to determine if a similar approach can be developed to treat patients in the clinic .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cancer", "biology" ]	2017	Mice deficient of Myc super-enhancer region reveal differential control mechanism between normal and pathological growth
The mechanosensing ability of lymphocytes regulates their activation in response to antigen stimulation , but the underlying mechanism remains unexplored . Here , we report that B cell mechanosensing-governed activation requires BCR signaling molecules . PMA-induced activation of PKCβ can bypass the Btk and PLC-γ2 signaling molecules that are usually required for B cells to discriminate substrate stiffness . Instead , PKCβ-dependent activation of FAK is required , leading to FAK-mediated potentiation of B cell spreading and adhesion responses . FAK inactivation or deficiency impaired B cell discrimination of substrate stiffness . Conversely , adhesion molecules greatly enhanced this capability of B cells . Lastly , B cells derived from rheumatoid arthritis ( RA ) patients exhibited an altered BCR response to substrate stiffness in comparison with healthy controls . These results provide a molecular explanation of how initiation of B cell activation discriminates substrate stiffness through a PKCβ-mediated FAK activation dependent manner . Cells interact with an intricate mechanical extracellular microenvironment by sensing physical signals and using surface receptors to convert them into chemical signals and , subsequently , cellular responses ( Sun et al . , 2012; Liu et al . , 2015 ) . Lymphocytes – the cells of the adaptive immune system – include B and T cells , both of which use surface expressed antigen receptors to sense external antigens . The B cell receptor ( BCR ) is composed of a membrane-bound entity of immunoglobulin ( Ig ) and an Igα-Igβ heterodimer in a 1:1 stoichiometry ( Pierce and Liu , 2010 ) . The engagement of the BCR and antigen leads to initiation of B cell activation . It is known that antigens exhibit great diversity and B cell activation is remarkably sensitive to the diversity of antigen properties , including antigen density ( Liu et al . , 2010a; Fleire et al . , 2006 ) , antigen affinity ( Liu et al . , 2010a; Fleire et al . , 2006 ) , antigen valency ( Bachmann et al . , 1993; Liu and Chen , 2005; Liu et al . , 2004 ) , and Brownian mobility of the antigen ( Wan and Liu , 2012 ) . Previous studies have suggested the extraordinary capability of B cells to sense the chemical and physical features of the antigen , although the presenting forms of antigen are far more complicated under physiological conditions . Recent studies have also suggested that the antigens encountered by B cells in vivo are presented on substrates with various stiffness ( also referred as rigidity or elasticity ) features ( Bachmann and Jennings , 2010 ) . Stiffness describes the extent that an object resists deformations in response to an applied force , which is quantified by Young's modulus in units of Pascal ( Pa or N/m2 or m−1·kg·s−2 ) ( Discher et al . , 2005 ) . For example , antigen molecules presented by a viral capsid exhibit a high degree of stiffness ( 45 , 000–1 , 000 , 000 kPa ) ( Mateu , 2012 ) , while the same antigen molecules expressed on the membrane of a virus-infected host cell show a medium level of stiffness ( 0 . 01–1000 kPa ) ( Nemir and West , 2010 ) , and if released from the virus into the plasma display a particularly low degree of stiffness ( several Pa ) ( Araujo et al . , 2012 ) . B cells acquire extracellular matrix ( ECM ) -associated antigens in a B cell receptor ( BCR ) and contact-dependent manner ( Ciechomska et al . , 2014 ) , in the presence of a considerable range of stiffness of ECM in the tissue , from 0 . 012 kPa to 20 kPa ( Bao and Suresh , 2003; Paszek et al . , 2005; Engler et al . , 2004; Nemir and West , 2010 ) . Recent studies have shown that the degree of stiffness of the substrates presenting the antigens efficiently regulates the activation of both B and T cells ( Wan et al . , 2013; Bashour et al . , 2014; O'Connor et al . , 2012; Judokusumo et al . , 2012; Zeng et al . , 2015; Wan et al . , 2015; Saitakis et al . , 2017 ) ; however , the underlying molecular mechanism remains unexplored . The ECM-associated microenvironment is an abundant source of antigens ( Tesniere et al . , 2008; Schaefer , 2010; Knight , 2015 ) . The capability of B cells to sense the stiffness features of antigen-presenting surfaces may be related to the pathological activation of auto-reactive B cells in autoimmune disease patients . For example , in patients with rheumatoid arthritis ( RA ) , reduced cartilage stiffness leads to auto-antigen-presenting B cells , which subsequently causes the production of auto-antibodies ( Mauri and Ehrenstein , 2007 ) . Clinical reports strongly suggest that the altered stiffness properties of the ECM are linked to aberrant activation of auto-antigen reactive B cells . However , it is unknown whether the primary B cells in RA patients maintain their capability to discriminate the stiffness of the substrates presenting antigens , and if they do , how such capabilities differ from those of primary B cells derived from healthy individuals . Canonically , it is well established that the cells are capable of sensing the mechanical signals from the microenvironment by the conventional mechanosensor that is , integrin . The major types of integrin molecules expressed by B cells are leukocyte function-associated antigen-1 ( LFA-1 ) ( αLβ2 ) and very late antigen-4 ( VLA-4 ) ( α4β1 ) ( Arana et al . , 2008a ) . The ligands of LFA-1 and VLA-4 are adhesion molecules that is , intercellular adhesion molecules ( ICAM-1/2 ) and vascular cell adhesion molecule-1 ( VCAM-1 ) , respectively ( Arana et al . , 2008a ) . Previous studies investigating the responses of both B and T cells upon encountering antigens that were tethered to stiff and soft substrates generally did not use adhesion molecules in their experimental systems ( Wan et al . , 2013; Bashour et al . , 2014; O'Connor et al . , 2012; Judokusumo et al . , 2012; Zeng et al . , 2015 ) . Early studies suggested that lymphocytes discern substrate stiffness independently of integrin signaling , but the details are unclear . More intriguingly , it is also unclear if or how the ability of lymphocytes to discriminate substrate stiffness is regulated by direct interaction between the integrin and adhesion molecule . Here , we addressed all three of the aforementioned questions through a combination of molecular imaging with genetic and pharmacological approaches by examining initiation of B cell activation on antigen-presenting substrates with stiff or soft features . The two most commonly used substrates for in vitro mechanosensing studies are polyacrylamide ( PA ) and polydimethylsiloxane ( PDMS ) . Both were used in this study to fabricate the soft and stiff substrates at contrasting ranges of 2 . 6 versus 22 . 1 kPa for PA ( Wan et al . , 2013; Judokusumo et al . , 2012 ) and 20 versus 1100 kPa for PDMS ( O'Connor et al . , 2012 ) . The results revealed that BCR signaling competent B cells can discriminate substrate stiffness through accumulating/polarizing more BCR molecules into the B cell immunological synapse ( IS ) when encountering stiff rather than soft substrates . In contrast , B cells deficient for each of the early BCR signaling molecules including tyrosine-protein kinase ( Lyn ) , spleen tyrosine kinase ( Syk ) , phospholipase Cγ2 ( PLCγ2 ) , Bruton’s tyrosine kinase ( Btk ) , B-cell linker protein ( BLNK ) , and protein kinase C ( PKCβ ) , lose this discrimination capability . Moreover , we found that PKCβ functions downstream of Btk and PLCγ2 , as PMA-induced activation of PKCβ can bypass the requirements of Btk and PLCγ2 for B cell discrimination of substrate stiffness . Mechanistically , we excluded PKCβ-mediated NF-κB activation in the capability of B cells to discriminate substrate stiffness . Instead , we provide evidence that PKCβ-dependent activation of FAK is required for B cells to discriminate substrate stiffness , whereby FAK potentiates B cell spreading and adhesion responses . As supporting evidence for this model , we observed that a pharmaceutical inhibitor targeting FAK , the key downstream molecule in integrin signaling , drastically impaired the capability of B cells to discriminate substrate stiffness . FAK-deficient B cells also lost this capacity to discriminate substrate stiffness , which was rescued by exogenous expression of FAK-WT but not the inactivated mutant FAK-Y926F . Further data showed that the ‘outside-in’ activation of integrin by the adhesion molecules , ICAM-1 or VCAM-1 , greatly enhanced the B cell’s capability to discriminate substrate stiffness . Lastly , the capability of B cells to discriminate substrate stiffness could be readily recapitulated in human primary PBMC B cells . It was striking to observe that B cells from RA patients exhibited a disordered and weakened capability to discriminate substrate stiffness in comparison with the healthy controls . Our data dissect the molecular mechanisms whereby B cells discriminate substrate stiffness during B cell activation , and also explicitly determine the contribution of adhesion molecules to such an event . The conclusion that B cells discriminate substrate stiffness through a PKCβ-mediated FAK activation-dependent manner in the initiation of B cell activation improves our understanding of the sophisticated mechanosensing capability of B lymphocytes and provides a potential explanation for the dysregulated activation of auto-reactive B cells in RA patients . To dissect the molecular mechanism of substrate stiffness discrimination by B cells , we used a library of DT40 B cells deficient for specific signaling molecules through a gene-targeted knock out ( KO ) technique ( Kurosaki et al . , 2010; Kurosaki , 1999 ) . The KO library of DT40 B cells was established by Kurosaki and his colleagues to investigate the function of specific signaling molecules in BCR signal transduction ( Kurosaki et al . , 2010; Kurosaki , 1999 ) . To determine whether DT40 wild-type ( DT40-WT ) B cells can discriminate between stiff and soft PDMS substrates during initiation of B cell activation , we used mouse anti-chicken IgM antibodies that were pre-tethered to stiff or soft PDMS substrates as a surrogate antigen to activate DT40 B cells . As antigen density can drastically influence B cell activation ( Liu et al . , 2010a; Fleire et al . , 2006 ) , we first examined the distribution and density of the surrogate antigen on both stiff and soft PDMS using Alexa Fluor 647-conjugated mouse IgM antibody ( clone M4 ) and anti-chicken IgM as surrogate antigens . We measured an even distribution of the antigens on both types of PDMS substrates through confocal fluorescence microscopy ( Figure 1A , B ) . We confirmed that there was comparable accessibility to the non-fluorophore conjugated surrogate antigens on both of the substrates using the DyLight 649-conjugated Fab anti-mouse IgM antibody ( Figure 1C , D ) . To avoid inadvertently altered antigen accessibility for B cells when varying the stiffness of PDMS substrates , we examined the accessibility of the antigens to DT40-WT B cells . We set equal amounts of DT40-WT B cells on either stiff or soft PDMS substrates that were pre-coated with the same density of surrogate antigens . We counted the number of tethered DT40-WT cells under the microscope and found that DT40-WT B cells can equally tether to both stiff and soft PDMS substrates ( Figure 1E–G ) . Thus , we established an experimental system constituting stiff and soft PDMS substrates that presented the same density of antigen with equal accessibility to DT40 B cells . 10 . 7554/eLife . 23060 . 003Figure 1 . Surrogate antigens tethered to stiff or soft PDMS substrate show similar surface density and accessibility for B cells . ( A ) Distribution of Alexa 647-conjugated mouse IgM monoclonal antibody ( clone M4 ) anti-chicken IgM as a surrogate antigen on the surface of PDMS substrates was equal and highly uniform as shown by confocal fluorescence microscopy . Scale bar is 3 µm . ( B ) Quantification of antigen density on both the surfaces of stiff and soft PDMS . ( C ) Representative confocal fluorescence microscope images showing the equal and highly uniform accessibility of the surrogate antigen on both the substrates as probed by the DyLight 649-conjugated Fab anti-mouse IgM antibody . Scale bar is 3 µm . ( D ) Quantification of antibody accessibility on both the surfaces of stiff and soft PDMS . Shown are mean ± SEM from one representative of three independent experiments . Two-tailed t tests were performed for statistical comparisons . ( E ) Representative images of the adhesion of DT40 B cells on the surface of either stiff or soft PDMS substrates before and after wash with 10 ml of PBS-1% FBS washing buffer . Scale bar is 50 µm . ( F , G ) Statistical quantification of the percentage of DT40 B cells adhered to stiff or soft substrates with or without tethered antigens . Adhesion rate is used for quantification as detailed in Materials and methods . The results were obtained using two different washing speeds of 0 . 5 ( F ) or 1 ml/sec ( G ) for a total amount of 10 ml of PBS-1% FBS washing buffer . Bar represents mean ± SEM from one representative of two independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 003 Next , we compared the capability of DT40-WT B cells to discriminate substrate stiffness during their activation initiation by quantifying the amount of BCRs that accumulated at the contact interface between B cells and the antigen-presenting surfaces on either soft or stiff substrates ( Figure 2A , B ) . BCRs are evenly distributed in quiescent B cells , and the strength of the initiation of B cell activation can be measured by the level of polarization of the BCRs to the site of contact with the antigen-presenting surfaces in activated B cells ( Liu et al . , 2010b , 2010c , 2012; Seeley-Fallen et al . , 2014; Liu et al . , 2013; Arana et al . , 2008b; Carrasco and Batista , 2007; Lin et al . , 2008; Treanor et al . , 2011; Weber et al . , 2008; Depoil et al . , 2008; Fleire et al . , 2006 ) . To quantify the amount of accumulated BCRs , we used the mean fluorescence intensity ( MFI ) of BCRs within the B cell contact interface rather than the total fluorescent intensity ( TFI ) value , as the latter will increase in response to B cell spreading during B cell activation , which increases the size of the contact interface . Thus , it is not possible to distinguish whether the increase of TFI results from polarization of BCRs to the B cell contact interface or from an increase in the size of the contact interface . In contrast , the value of MFI is resilient to such changes as MFI is defined by a value of TFI / size of the contact interface , equal to the density of BCRs within the contact interface , a change that can only be introduced by the enrichment of BCRs . Indeed , the results showed a much higher BCR MFI in B cells that were placed on stiff substrates compared with B cells on soft substrates ( Figure 2B ) . To better compare the efficiency of the accumulation of BCRs at the B cell’s contact interface with either stiff or soft PDMS substrates , we defined a ratio index as the BCR MFI of each cell on the stiff substrate divided by the averaged BCR MFI value of all cells on the soft substrate . A ratio larger than 1 would indicate that B cells can accumulate more BCRs when on a stiff substrate versus a soft substrate , and a higher ratio value would indicate better discrimination capability . Another advantage of using such a ratio is to enable multi-grouped comparisons , which are problematic for absolute MFI values because of the presence of inter-sample and inter-batch variations . Using this approach with DT40-WT B cells , we found the ratio of the MFI of BCR on stiff/soft PDMS substrates was larger than 1 . 5 , suggesting that stiff substrates induced the accumulation of significantly more BCRs into the B cell IS compared with soft substrates ( Figure 2B ) . Thus , DT40-WT B cells could clearly discriminate between stiff and soft PDMS substrates ( Figure 2A , B ) . Similar results were acquired in the same experimental system using PA substrates ( Figure 2C , D ) . These results validate the utility of using DT40 B cells in this PDMS or PA based experimental system for dissecting the molecule mechanisms underlying the capability of B cells to discriminate substrate stiffness during the initiation of B cell activation . 10 . 7554/eLife . 23060 . 004Figure 2 . DT40-WT B cells exhibit excellent capability to discriminate substrate stiffness . ( A ) Representative confocal images of DT40 B cells showing the contact interface with the antigens tethered on either stiff or soft PDMS substrates . Scale bar is 3 µm . ( B ) Synaptic accumulation of BCRs on either stiff or soft substrates and a ratio figure showing the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft PDMS substrates . ( C ) Representative confocal images of DT40 B cells showing the contact interface with the antigens tethered on either stiff or soft PA substrates . Scale bar is 5 µm . ( D ) Synaptic accumulation of BCRs on either stiff or soft substrates and a ratio figure showing the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft PA substrates . Bar represents mean ± SEM from one representative of two independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 004 We first performed experiments in DT40 B cells by knocking out the key molecules in BCR signaling , including Lyn ( Takata et al . , 1994 ) , Syk ( Takata et al . , 1994 ) , PLCγ2 ( Watanabe et al . , 2001 ) , Btk ( Hashimoto et al . , 1999 ) , BLNK ( Ishiai et al . , 1999 ) , and PKCβ ( Shinohara et al . , 2005 ) . Lyn , Syk , and PLCγ2 are well characterized signaling molecules that function as an initiator complex to trigger the BCR signaling cascade , leading to subsequent assembly of a multi-protein complex of adaptors and various signaling molecules proximal to the plasma membrane , known as the signalosome ( Kurosaki , 1999; Kurosaki et al . , 2010 ) . To reduce batch to batch variation , each of the three types of DT40 KO was respectively assayed in parallel with the corresponding DT40-WT B cells . In each case , DT40-WT B cells showed a clear capability to discriminate substrate stiffness with a ratio of approximately 1 . 5 . However , the DT40-Lyn-KO , DT40-Syk-KO , and DT40-PLCγ2-KO B cells only showed a ratio of 1 . 0 , suggesting that similar amounts of BCRs accumulated at the contact interface of B cells and the PDMS substrates , regardless of stiffness ( Figure 3A–C ) . Similar results were obtained using PA substrates ( Figure 3—figure supplement 1A–C ) . As the synaptic accumulation of BCRs in these three types of KO B cells was largely BCR signaling-independent , these results indicate that only BCR signaling-dependent enrichment of BCRs to the B cell IS is regulated by the stiffness feature of the substrates . 10 . 7554/eLife . 23060 . 005Figure 3 . Lyn , Syk , and PLCγ2 are required for B cells to discriminate substrate stiffness . ( A–C ) Statistical comparison for the substrate stiffness discrimination capability of B cells from the following three groups: ( A ) DT40-WT , DT40-Lyn-KO , and DT40-Lyn-Rescue cells; ( B ) DT40-WT , DT40-Syk-KO , and DT40-Syk-Rescue cells; ( C ) DT40-WT , DT40-PLCγ2-KO , and DT40-PLCγ2-Rescue cells . The calculation of ratio of BCR MFI is defined as the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft substrates . ( D–F ) B1-8 primary B cell pre-treatment with DMSO as a control ( NC ) versus Lyn inhibitor PP2 ( D ) , Syk inhibitor Piceatannol ( E ) , and PLCγ2 inhibitor U73122 ( E ) . ( G–I ) CH27 B cell pre-treatment with DMSO as a control ( NC ) versus Lyn inhibitor PP2 ( G ) , Syk inhibitor Piceatannol ( H ) and PLCγ2 inhibitor U73122 ( I ) . In ( A ) to ( I ) , bar represents mean ± SEM from at least 20 cells in one representative of three independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 00510 . 7554/eLife . 23060 . 006Figure 3—figure supplement 1 . Lyn , Syk , PLCγ2 , Btk , BLNK , or PKCβ molecules are required for B cells to discriminate between stiff and soft substrates . ( A–C ) Ratio data on PA substrate showing the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft substrates in a comparison of the following groups: ( A ) DT40-WT , DT40-Lyn-KO , ( B ) DT40-WT , DT40-Syk-KO , and ( C ) DT40-WT , DT40-PLCγ2-KO . In ( A ) - ( C ) , bar represents mean ± SEM from at least 25 cells in one representative of at least two independent experiments . Two-tailed t tests were performed for statistical comparisons . ( D–F ) Inhibitor treatment of signaling molecules ( Lyn , Syk , PLCγ2 ) primarily blocked the synaptic accumulation of BCRs on stiff substrates , while the changes of BCR MFI on soft substrates were very mild in the comparison of inhibitor treated versus DMSO control B cells . In ( D ) - ( F ) , bar represents mean ± SD from at least 25 cells in one representative of at least two independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 006 To verify that these observations were dependent on the knocked out molecules rather than subordinate effects during the construction of these three types of KO B cells , we cloned the cDNA of Lyn , Syk , and PLCγ2 molecules by RT-PCR from the mRNA of DT40-WT B cells ( Table 1 ) and performed rescue experiments ( Figure 3A–C ) . Importantly , exogenous overexpression of these molecules rescued the capability of B cells to discriminate substrate stiffness , suggesting that the observed phenotype is indeed mediated through these signaling molecules . To further validate these observations in other types of B cells including primary B cells , we used pharmaceutical inhibitors specifically targeting each of these three signaling molecules . Pharmaceutical inhibitor PP2 , Piceatannol , and U73122 were used to block the function of Lyn , Syk , and PLCγ2 molecules , respectively , as detailed in the Materials and methods section . Hapten antigen 4-hydroxy-3-nitrophenyl acetyl ( NP ) -specific B1-8 primary B cells from B1-8 transgenic mice ( Hauser et al . , 2007 ) or B lymphoma cells , CH27 that were pre-treated with each of these inhibitors lost the capacity to discriminate between the different degrees of stiffness of the substrate-presenting antigens ( Figure 3D–I ) . We next examined whether this loss of ability to discriminate substrate stiffness resulted from changes in the BCR MFI ( as a parameter indicating BCR microclustering ) on a stiff or soft substrate surface . We achieved this by comparing the BCR MFI of the inhibitor pre-treated versus DMSO-control pre-treated primary B cells ( Figure 3—figure supplement 1D–F ) on the surface of the substrates with the same Young's modulus ( stiffness ) . The results demonstrated that inhibition of proximal signaling molecules , Lyn , Syk , or PLCγ2 blocked the synaptic accumulation of BCRs specifically on stiff substrates . In comparison , we found only mild differences in BCR MFI of primary B cells pre-treated with inhibitor versus DMSO-control on soft substrates ( Figure 3—figure supplement 1D–F ) . Thus , BCR signaling initiating molecules , Lyn , Syk , and PLCγ2 are required for B cells to discriminate between stiff and soft substrates during the initiation of B cell activation . 10 . 7554/eLife . 23060 . 007Table 1 . Primer sequences used to amplify cDNADOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 007Gene nameUpstream primerDownstream primer Lynatgggatgtataaaatcaaaaaggactatggctgctgttgatattgcc Sykatggcttccaacatggccaaccaatcaccctttacagcattatcatcaaggcatt PLCγ2atgcctcgaaagagtgtagattatgttaagagtagaatttgctgttactg Btkatggccagcatcatcctgtcacggctcttcgtctg PKCβgcctaccccaagtccatgtcttggtcatgagccctttg FAKggcagcagcttaccttgatccggcctggactggctgatcatt BLNKgcggccgcaccggtctgcagctagctggacaagctgaataagataactgctcaccatcgaagcagctcctcctcctgaaaccttcacagcatatttcagtc We further assessed the contribution of downstream BCR signaling molecules in B cell discrimination of substrate stiffness by examining DT40 B cells deficient for Btk , BLNK , or PKCβ . We first observed that DT40-Btk-KO , DT40-BLNK-KO , and DT40-PKCβ-KO B cells encountering antigens presented on soft substrates showed comparable accumulation of BCRs into the B cell IS as DT40-WT B cells ( Figure 4A–C ) . These results are consistent with an earlier report showing that DT40-Btk-KO , DT40-BLNK-KO , and DT40-PKCβ-KO B cells accumulated 81% , 98% , and 82% , respectively , of the membrane-bound antigens on fluid planar lipid bilayer membranes to the B cell IS compared with DT40-WT B cells ( Weber et al . , 2008 ) . However , when quantifying the substrate stiffness discrimination capability of these KO B cells on substrates with different stiffness features , it was striking to observe that each of these three B cell lines completely lost the capability to discriminate substrate stiffness ( Figure 4D–F ) . Similar results were obtained using the PA gel system ( Figure 4—figure supplement 1A–C ) . Exogenous supply of Btk , BLNK , or PKCβ rescued the capability of B cells to discriminate substrate stiffness ( Figure 4D–F ) . Furthermore , the deficiency of each of these three proximal signaling molecules blocked the synaptic accumulation of BCRs specifically on stiff substrates , while there were only minor changes in BCR MFI on soft substrates ( Figure 4—figure supplement 1D–F ) . As Btk , BLNK , and PKCβ are all well-established signaling molecules in BCR signal transduction , our data also demonstrated that the initiation of B cell activation is sensitive to substrate stiffness in a BCR signaling-dependent manner . 10 . 7554/eLife . 23060 . 008Figure 4 . Genetic ablation of Btk , BLNK , or PKCβ blunts the ability of B cells to discriminate substrate stiffness . ( A–C ) B cells showed comparable capability to accumulate BCRs into B cell IS in the following groups ( A ) DT40-WT and DT40-Btk-KO; ( B ) DT40-WT and DT40-BLNK-KO; and ( C ) DT40-WT , DT40-PKCβ-KO . ( D–F ) Statistical comparison for the substrate stiffness discrimination capability of B cells from the following three groups: ( D ) DT40-WT , DT40-Btk-KO , and DT40-Btk-Rescue cells; ( E ) DT40-WT , DT40-BLNK-KO , and DT40-BLNK-Rescue cells; ( F ) DT40-WT , DT40-PKCβ-KO , and DT40-PKCβ-Rescue cells . In ( A ) - ( F ) , bar represents mean ± SEM from at least 27 cells in one representative of three independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 00810 . 7554/eLife . 23060 . 009Figure 4—figure supplement 1 . Genetic ablation of Btk , BLNK , or PKCβ blunts the ability of B cells to discriminate substrate stiffness . ( A–C ) Ratio data on PA substrate showing the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft substrates in a comparison of the following groups: ( A ) DT40-WT , DT40-Btk-KO , ( B ) DT40-WT , DT40-BLNK-KO; and ( C ) DT40-WT , DT40-PKCβ-KO . In ( A ) - ( C ) , bar represents mean ± SEM from at least 25 cells in one representative of at least two independent experiments . Two-tailed t tests were performed for statistical comparisons . ( D–F ) Deficiency of proximal signaling molecules primarily blocked the synaptic accumulation of BCRs on stiff substrates , while the changes of BCR MFI on soft substrates were very mild in the comparison of KO versus WT B cells . In ( D ) - ( F ) , bar represents mean ± SD from at least 25 cells in one representative of at least two independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 009 In the sequential initiation and transduction of the membrane proximal BCR signaling cascade , antigen stimulation of BCR leads to phosphorylation of the immunoreceptor tyrosine-based activation motif ( ITAM ) in the cytoplasmic domains of BCR components Igα/Igβ by Lyn kinase . The phosphorylated ITAMs provide a docking site for Syk kinase , which results in rapid auto-phosphorylation of Syk within its linker regions ( Saouaf et al . , 1994 ) , which subsequently provides the additional docking sites for PLCγ2 ( Weber et al . , 2008 ) and Btk ( Smith et al . , 1994 ) . Btk actively promotes the phosphorylation and activation of PLCγ2 . Activated PLCγ2 potently hydrolyzes phosphatidylinositol 4 , 5-biphosphate to generate inositol 1 , 4 , 5-trisphosphate ( IP3 ) and diacylglycerol ( DAG ) . DAG recruits PKCβ to the membrane proximal BCR signalosome . PMA , a phorbol ester , has been widely used as a DAG analog to directly induce activation of PKC . Thus , we tested whether addition of PMA could rescue the inability of DT40-BTK-KO , DT40-PLCγ2-KO , and DT40-PKCβ-KO B cells to discriminate substrate stiffness . We chose these three types of B cell lines considering that DT40-BTK-KO and DT40-PLCγ2-KO B cells lack the ability to produce DAG upon BCR activation while maintaining normal expression of PKCβ . In contrast , DT40-PKCβ-KO can normally produce DAG upon BCR activation , but lacks PKCβ . In our experimental system , we pre-treated either WT or KO DT40 B cells with different concentrations of PMA , 5 , 20 , and 50 ng/ml , following a published protocol ( Quann et al . , 2009 ) , and then examined the substrate stiffness discrimination capability of B cells under each of these conditions . DT40-WT B cells showed good substrate stiffness discrimination capability ( ratio = 1 . 5 ) in the absence of PMA as expected , and this capability was largely unaffected in the presence of PMA ( Figure 5A–D ) . In contrast , PMA rescued the ability of DT40-PLCγ2-KO , DT40-BTK-KO , and DT40-BLNK-KO B cells to discriminate substrate stiffness in a dose-dependent manner . This suggests that PMA-mediated activation of PKCβ that bypasses early BCR signaling molecules is sufficient to restore the substrate stiffness discrimination capability of B cells ( Figure 5A–C ) . In contrast , PMA-treated DT40-PKCβ-KO B cells failed to exhibit such discrimination capability ( Figure 5D ) . A much higher concentration of PMA ( 100 ng/ml ) still failed to rescue the defect in DT40-PKCβ-KO B cells ( Figure 5E ) . DT40-WT B cells pre-treated with a PKCβ inhibitor and PMA also failed to discriminate stiffness ( Figure 5F ) . Together , these results suggest that Btk and PLCγ2 function upstream of PKCβ , demonstrating the key role of PKCβ in mediating the capability of B cells to discriminate substrate stiffness . 10 . 7554/eLife . 23060 . 010Figure 5 . PMA-induced activation of PKCβ can bypass the requirements of Btk and PLCγ2 for B cells to discriminate substrate stiffness . ( A–D ) Statistical comparison for the substrate stiffness discrimination capability of PMA pre-treated B cells in the following groups: ( A ) DT40-WT and DT40-PLCγ2-KO; ( B ) DT40-WT and DT40-Btk-KO; ( C ) DT40-WT and DT40-BLNK-KO; and ( D ) DT40-WT and DT40- PKCβ-KO . ( E ) DT40-WT versus DT40-PKCβ-KO B cells that were pre-treated with high concentration of 100 ng/ml PMA . ( F ) DT40 WT B cells treated with PMA plus DMSO as a control or PMA plus PKCβ inhibitor . In ( A ) - ( F ) , bar represents mean ± SEM from at least 21 cells in one representative of three independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 010 Next , we investigated the requirement for BCR signaling molecules by B cells to discriminate substrate stiffness . PKCβ is the most downstream signaling molecule of all those investigated in this report , including Lyn , Syk , PLCγ2 , Btk , BLNK , and PKCβ . A canonical function of PKCβ in B cell activation is to directly phosphorylate the downstream signaling molecule Carma-1 at Ser-559 , Ser-644 , and Ser-652 , which induces the association of Carma1 with Bcl10 and Malt1 . The formation of Carma1 , Bcl10 , and Malt1 ( CBM ) signaling complexes leads to the activation of NF-κB ( Su et al . , 2002; Shinohara et al . , 2005 ) . However , the activation of NF-κB leads to a downstream gene expression signature in B cells at a late time point , usually hours after the initial BCR and antigen recognition , whereas substrate stiffness discrimination in B cells occurs within a few minutes of interaction between the BCR and antigen . Therefore , we speculated that PKCβ-mediated NF-κB activation would not contribute to the capabilities of B cells to discriminate substrate stiffness . Indeed , we found that Carma1-KO DT-40 B cells effectively maintained their discrimination capability ( Figure 6A ) . 10 . 7554/eLife . 23060 . 011Figure 6 . PKCβ-dependent FAK activation accounts for B cells to discriminate substrate stiffness . ( A ) Statistical comparison for the substrate stiffness discrimination capability of DT40-WT versus DT40-Carma-1-KO B cells . ( B–D ) Statistical comparison for the substrate stiffness discrimination capability of DT40-WT ( B ) , CH27 ( C ) , or B1-8 primary ( D ) B cells that were pre-treated with either DMSO as a control ( NC ) or FAK inhibitor PF573-228 ( FAK inhibitor ) . ( E ) Statistical comparison for the substrate stiffness discrimination capability of DT40-WT , DT40-FAK-KO , and DT40-FAK-Rescue B cells . ( F ) Statistical comparison for the substrate stiffness discrimination capability of DT40-FAK-Rescue and FAK-Y926F Mutant B cells . ( G ) The representative confocal images of B1-8 primary B cells showing the spatial co-distribution of BCR and pFAK ( Tyr 925 ) molecules within the B cell immunological synapse . Scale bar is 4 µm . ( H ) Statistical comparison of the MFI of pFAK molecules within the B cell immunological synapse of DT40-WT B cells that were placed on the PDMS substrates presenting antigen or lacking antigen . ( I ) Statistical comparison of the MFI of pFAK molecules within the B cell immunological synapse of DT40-WT B cells that were placed on the antigen-presenting surfaces of either stiff or soft PDMS substrates . ( J ) Statistical comparison for the substrate stiffness discrimination capability of DT40-WT , DT40-PKCβ-KO , and DT40-PKCβ-Rescue B cells . ( K ) Statistical comparison of the MFI of pFAK molecules within the B cell immunological synapse of DT40-WT , DT40-PKCβ-KO , and DT40-PKCβ-Rescue B cells in response to antigen stimulation . In ( A ) – ( F ) and ( H ) – ( K ) , bar represents mean ± SEM from at least 30 cells in one representative of three independent experiments . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 01110 . 7554/eLife . 23060 . 012Figure 6—figure supplement 1 . PKCβ-dependent FAK activation accounts for B cells to discriminate substrate stiffness . ( A ) FAK ( PTK2 ) gene knock-out through CRISPR/Cas9 technique . ( B ) Western blot confirmed the FAK-KO efficiency compared with that of the WT . ( C ) DT40-WT and FAK-KO B cells showed equal BCR expression levels . ( D ) Sequence homology comparison showed the similarity of DT40-FAK-Tyr926 with Mouse and Human FAK-Tyr925 . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 012 The initiation of B cell activation is dependent on adhesion of B cells to the surface of the substrate presenting the antigen . We thus assessed the potential requirement of PKCβ in the discrimination capability of B cells through its function in regulating B cell adhesion responses . Our hypothesis was based on published studies showing that PKC family members play key roles in mediating the adhesion responses of many types of mammalian cells through ‘inside-out’ activation of integrin molecules ( Haller et al . , 1998; Besson et al . , 2002; Disatnik et al . , 2002; Buensuceso et al . , 2005 ) . Downstream of integrin activation , focal adhesion kinase ( FAK ) , a member of the nonreceptor protein-tyrosine kinase family , is a key player in mechanosensing-mediated cell adhesion responses ( Yu et al . , 2012; Bashour et al . , 2014; Slack-Davis et al . , 2007 ) . Moreover , it has also been reported that FAK is required for the chemoattractant-induced migration and adhesion response in B cells ( Tse et al . , 2012 ) . Thus , we first examined the dependence of FAK activation on the substrate stiffness discrimination capability of B cells . To test this hypothesis , we used the FAK specific inhibitor , PF573-228 ( Slack-Davis et al . , 2007 ) , to inactivate FAK . Strikingly , PF573-228 pre-treated DT40 B cells , CH27 , and B1-8 primary B cells were completely blunted in their capability to discriminate substrate stiffness ( Figure 6B–D ) . To confirm this , we knocked out FAK in DT40 B cells ( DT40-FAK-KO ) using CRISPR/Cas9 gene editing technology ( Figure 6—figure supplement 1A–C ) and found that DT40-FAK-KO B cells lost the capability to discriminate substrate stiffness during their activation . This was rescued by the exogenous expression of chicken FAK-WT ( Figure 6E ) . The phosphorylation of FAK at Tyr-925 is considered to be a critical step in FAK activation-mediated migration and adhesion responses ( Deramaudt et al . , 2011 ) . As expected , exogenous expression of a chicken FAK-Y926F mutant ( chicken Tyr 926 has sequence homology with Tyr 925 in mouse and human FAK ) ( Figure 6—figure supplement 1D ) in DT40-FAK-KO B cells failed to rescue the discrimination capability of B cells ( Figure 6F ) . Thus , FAK activation is required for the B cells to discriminate substrate stiffness . The importance of the phosphorylation of FAK at Tyr-925 ( pFAK ) in FAK activation-mediated migration and adhesion responses ( Deramaudt et al . , 2011 ) led us to examine the spatial colocalization of BCR and pFAK molecules within the IS of mouse primary B cells upon activation ( Figure 6G ) . As predicted , BCRs primarily aggregated at the central region of the B cell IS , while pFAK accumulated at the peripheral region , consistent with the report that integrin molecules are mainly located at the peripheral region of the B cell IS ( Tolar et al . , 2009 ) . Because we did not use adhesion molecules in our experimental system , we speculated that phosphorylation of FAK would result from activation of BCR signaling pathway molecules upon antigen stimulation . Indeed , as further supporting evidence , we observed no obvious upregulation of pFAK in B cells that were placed on substrates without antigen coating , while very strong pFAK signaling was evident only in the antigen-activated B cells ( Figure 6H ) . Strikingly , a significantly stronger pFAK signal was observed in B cells that were placed on stiff substrates than in B cells on soft substrates ( Figure 6I ) . Using the method presented above for quantifying the BCR MFI ratio , we also calculated the pFAK MFI ratio index by dividing the pFAK MFI of each cell on a stiff substrate by the averaged value of the pFAK MFI of all cells on the soft substrate ( Figure 6J ) . The obtained ratio value of 2 . 5 for MFI of pFAK on stiff/soft PDMS substrates indicated that stiff substrates induced accumulation of significantly more pFAK molecules into the B cell IS compared with soft substrates ( Figure 6J ) . Together , these experiments suggest that the activation of FAK could be induced by antigen-binding-induced BCR signaling in B cells in the absence of adhesion molecules , and more importantly , that FAK activation is required for the B cells to discriminate substrate stiffness . Based on all these findings , we speculated that activation of FAK by the BCR signaling molecule PKCβ is important for B cells to discriminate substrate stiffness . To test this , we examined the activation of FAK in DT40-WT and DT40-PKCβ-KO B cells . As expected , the activation of FAK was drastically impaired in DT40-PKCβ-KO cells compared with DT40-WT B cells . Exogenous PKCβ rescued the capability of B cells to accumulate pFAK in the B cell IS in response to BCR and antigen recognition ( Figure 6K ) . As mentioned above , further quantification using the pFAK MFI ratio index demonstrated that DT40-WT B cells exhibited a ratio value of 2 . 5; in marked contrast , DT40-PKCβ-KO B cells had a ratio of only 1 , suggesting that similar amounts of pFAK molecules accumulated at the contact interface of B cells on both stiff and soft PDMS substrates ( Figure 6J ) . Furthermore , DT40-PKCβ-KO B cells expressing exogenous PKCβ accumulated more pFAK molecules on stiff substrates than on soft substrates ( Figure 6J ) . As a consequence , exogenous expression of PKCβ rescued the capability of B cells to discriminate stiffness of the substrates ( Figure 6J ) . These results further confirmed the requirement of PKCβ-dependent FAK activation for B cells to effectively discern the stiffness of the substrate . We next investigated how PKCβ-dependent FAK activation accounts for the capability of B cells to discriminate the stiffness of antigen-presenting substrates . We hypothesized that stronger FAK activation in B cells that were placed on stiff substrates leads to better B cell spreading and adhesion responses , both of which are known to be essential for efficient B cell activation . To assess this , we first performed correlation analyses of the pFAK MFI value with each of the following three parameters in the same B cell: ( 1 ) the size of the B cell contact interface ( or the size of the B cell IS ) ; ( 2 ) the strength of B cell adhesion , and ( 3 ) the BCR MFI value . Quantifications of the size of the B cell contact interface and the BCR MFI within the contact area were introduced above . The strength of B cell adhesion was quantified by interference reflection microscopy ( IRM ) following our published protocol ( Xu et al . , 2015 ) . As shown in the representative image in Figure 7A , we quantified the MFI of IRM images by subcellular level analysis through ImageJ Software; the darker the region in the IRM image , the smaller the MFI of IRM , and thus the stronger the B cell adhesion , and vice versa ( Figure 7A ) . We found that the pFAK MFI value was strongly correlated with each of these three parameters ( Figure 7B–D ) . As B cells that were placed on stiff substrates always displayed a much higher pFAK MFI value , as mentioned above , it was not unexpected that these B cells also presented a much higher BCR MFI value and thus a higher BCR MFI ratio index than B cells that were placed on soft substrates ( Figure 7E ) . These correlation analyses demonstrated that the enhanced FAK activation in B cells on stiff substrates leads to better B cell spreading and adhesion toward the antigen-presenting surfaces . This suggests that PKCβ-dependent FAK activation accounts for the B cell discrimination capability by potentiating B cell spreading and adhesion responses . 10 . 7554/eLife . 23060 . 013Figure 7 . PKCβ-dependent FAK activation accounts for B cell discrimination capability by potentiating B cell spreading and adhesion responses . ( A ) Representative confocal images showing the adhesion strength of B cells on the basis of IRM . In both IRM and pFAK images , two representative region of interests ( ROIs , ( a and b ) demonstrated the calculation of the IRM and pFAK MFI within the same ROI . Scale bar is 4 µm . ( B–D ) Correlation analysis of the pFAK MFI with the size of spreading area ( B ) , the adhesion strength on the basis of IRM MFI ( C ) , or the BCR MFI ( D ) . ( E ) Correlation analysis of the ratio of the pFAK MFI to the ratio of BCR MFI . Data in B , D , E were analyzed based on the contact area of a single cell , thus one dot represents one cell; while data in ( C ) were analyzed based in a ROI within a cell’s contact area as demonstrated in ( A ) , thus one dot represents one ROI . In ( B ) – ( E ) , inserted correlation function was the linear regression analysis; data are one representative of at least two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 013 In the previous experiments , although only antigens and not adhesion molecules were present on the substrate surfaces , B cells were still able to detect substrate stiffness during their activation . This suggests that this capability of B cells is independent of the interaction between adhesion and integrin molecules . To test this directly , we used EDTA to block the function of integrins that may be activated by the presence of ECM molecules in the serum and/or the blocking reagent in the incubation buffer . EDTA-treated B cells lost the capability to discriminate substrate stiffness ( Figure 8A ) . Although EDTA can interfere with the function of integrin , it is also a potent calcium chelator with a much broader impact on the activation and function of B cells , including BCR-antigen binding-mediated calcium influx . Thus , to confirm this conclusion we used splenic primary B cells from CD11a ( ITGAL ) knockout ( KO ) mice , which lack the lymphocyte function-associated antigen 1 ( LFA-1 ) ( Figure 8B ) . As the adhesion and activation of lymphocytes requires integrin and adhesion molecules in vivo under physiological conditions ( Arana et al . , 2008a ) , it could follow that the capability of B cells to discern substrate stiffness is also regulated by direct activation of integrins with external adhesion molecules . B cells mainly express two types of integrins , LFA-1 and VLA-4 , that bind to their respective adhesion molecules , ICAM-1 and VCAM-1 ( Arana et al . , 2008a ) . Thus , we compared in parallel how NP-specific B1-8 primary B cells discriminate between stiff and soft PDMS substrates that were coated with antigens alone or coated with both antigens and adhesion molecules . The results showed that the addition of either ICAM-1 or VCAM-1 significantly enhanced the B cell’s capacity to discriminate between stiff and soft substrates ( Figure 8C–D , G ) . Similar results were also acquired using mouse CH27 B cells ( Figure 8E–G ) . Thus , during the initiation of B cell activation , integrin signaling plays an important role in maintaining the substrate stiffness discrimination capability of B cells , which is greatly enhanced by the outside-in activation of integrin by adhesion molecules . 10 . 7554/eLife . 23060 . 014Figure 8 . Adhesion molecules enhance B cell’s capability to discriminate between stiff and soft substrates . ( A ) Blocking the integrin with EDTA reduces the ratio of BCR MFI of CH27 B cells . ( B ) LFA-1 KO primary B cells lost the substrate stiffness discrimination compared with the WT B cells . ( C–F ) Adhesion molecules , ICAM-1 and VCAM-1 , enhanced the B cell’s capability to discriminate between stiff and soft substrates as shown in B1-8 primary B cells ( C , D ) or CH27 B cells ( E , F ) . ( G ) Representative confocal images showing the synaptic accumulation of BCRs from either B1-8 Primary B cells or CH27 B cells that were placed on antigen-presenting substrates with the additional condition of lacking ( NC ) or presenting adhesion molecules . Scale bar is 3 µm . Bar represents mean ± SEM from one representative of three independent experiments . Data were from at least 20 cells . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 014 Lastly , we examined the physiological and pathological relevance of our finding that PKCβ-dependent FAK activation accounts for B cell discrimination of substrate stiffness . A very recent study showed that auto-reactive human primary B cells from RA patients can efficiently acquire the auto-antigen , aggrecan , in a BCR- and adhesion-dependent manner ( Ciechomska et al . , 2014 ) . Thus , we compared the activation of primary B cells placed on either stiff or soft PDMS substrates from either healthy controls or RA patients . To reduce inter-sample and inter-batch variations , we chose nine age- and gender-matched pairs of healthy controls and RA patients . In each batch , we only compared one pair of samples , a healthy individual versus an RA patient . We pre-labeled PBMC B cells from the paired samples using Alexa Fluor 647-conjugated Fab fragment anti-human IgM constant region antibodies and placed these cells on either stiff or soft PDMS substrates presenting anti-human Igκ and anti-human Igλ antibodies , which functioned as the surrogate antigens . The cells were in contact with the antigen-coated PDMS substrates for 15 min before image acquisition . The accumulation of BCR ( or pFAK ) at the contact site between the B cell and the antigen-presenting substrates was quantified , and the BCR ( or pFAK ) MFI ratio index values were calculated as above . The results showed that all the BCR ( or pFAK ) MFI ratio index values were larger than 1 , indicating that all the human primary B cells from both healthy controls and RA patients exhibited substrate discrimination capabilities ( Figure 9A–D ) . However , careful examination of the BCR MFI ratio index values of the paired samples in the same batch of experiments revealed the presence of different efficiencies in terms of the capability of B cells to discriminate substrate stiffness: B cells from RA patients exhibited a much weaker discrimination efficiency as indicated by the much lower BCR ( or pFAK ) MFI ratio index , in comparison with B cells from the paired healthy controls ( Figure 9A , B ) . We observed this phenomenon in six out of the nine paired samples that we examined ( Figure 9—figure supplement 1A–F ) . In two pairs , we observed a comparable B cell discrimination capability in both the healthy control and RA patient ( Figure 9—figure supplement 1G–H ) . In only one pair , B cells from the healthy control showed weaker discrimination capability than those from the RA patient ( Figure 9—figure supplement 1I ) . It should be noted that the pFAK MFI ratio index value was more in line with this conclusion than the BCR index value , as indicated by much larger differences when comparing the ratio index of healthy controls versus RA patients within the paired samples ( Figure 9C , D ) . 10 . 7554/eLife . 23060 . 015Figure 9 . RA patient B cells exhibited disordered capability to discriminate substrate stiffness . ( A , B ) Paired comparison of healthy control and RA patient B cells on the basis of ratio of either BCR MFI ( A ) or pFAK MFI ( B ) . ( C , D ) Paired comparison of the ratio of BCR MFI and pFAK MFI on the basis of either healthy control or RA patient B cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 01510 . 7554/eLife . 23060 . 016Figure 9—figure supplement 1 . RA patient B cells exhibited disordered capability to discriminate substrate stiffness . ( A–I ) Comparison of original ratio data of BCR or pFAK MFI between first six pairs of the healthy control and RA patient B cells in which patients showed weak discrimination capability of substrate stiffness ( A–F ) ; two pairs showed comparable substrate stiffness discrimination between healthy control and RA patient B cells ( G–H ) ; while one pair showed a higher level of substrate stiffness discrimination in RA patient B cells than healthy control B cells ( I ) . Bar represents mean ± SEM from at least 20 cells . Two-tailed t tests were performed for statistical comparisons . ( J ) Representative confocal Images showing the colocalization of BCR and F-actin or the colocalization between pFAK and F-actin on stiff or soft PDMS substrate . Scale bar is 3 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 01610 . 7554/eLife . 23060 . 017Figure 9—figure supplement 2 . RA patient B cells exhibited disordered capability to discriminate substrate stiffness . ( A-1 ) Comparison of the BCR MFI of B cells from the paired healthy controls versus RA patients on either stiff or soft substrates . The results show that B cells from healthy controls preferentially enhanced the synaptic accumulation of BCR microclusters on stiff substrates , while RA patient B cells exhibited a different preference of mainly enhancing the BCR accumulation on soft substrates . Bar represents mean ± SD from at least 20 cells . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 01710 . 7554/eLife . 23060 . 018Figure 9—figure supplement 3 . RA patient B cells exhibited disordered capability to discriminate substrate stiffness . ( A-I ) Comparison of the pFAK MFI of B cells from the paired healthy controls versus RA patients on either stiff or soft substrates . The results showed that B cells from healthy controls preferentially enhanced pFAK MFI on stiff substrates , while RA patient B cells exhibited a different preference of mainly enhancing pFAK MFI on soft substrates . Bar represents mean ± SD from at least 20 cells . Two-tailed t tests were performed for statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 23060 . 018 Furthermore , we examined whether alteration in the ability of RA patient B cells to discriminate substrate stiffness of the antigen-presenting surface resulted from changes in BCR MFI ( as a parameter indicating BCR microclustering ) on a stiff or a soft substrate surface . We did so by comparing the BCR MFI of B cells from the paired healthy controls versus the RA patients on either stiff or soft substrates ( Figure 9—figure supplement 2A–I ) . The results showed that B cells from healthy controls preferentially enhanced the synaptic accumulation of BCR microclusters on stiff substrates . In direct contrast , RA patient B cells preferentially enhanced BCR accumulation on soft substrates ( Figure 9—figure supplement 2A–I ) . We also examined the original MFI value of pFAK and obtained similar results ( Figure 9—figure supplement 3A–I ) . All of these outcomes further support our model that PKCβ-mediated FAK activation accounts for B cell discrimination of substrate stiffness . In conclusion , the capability of B cells to discriminate substrate stiffness features can be readily recapitulated in human primary PBMC B cells , and more importantly , RA patient B cells exhibited a disordered capability of discriminating substrate stiffness in comparison with healthy controls . We discuss how these findings may help explain the dysregulated activation of auto-reactive B cells in RA patients ( see Discussion section below ) . Initiation of B cell activation has been shown to be sensitive to antigen density ( Liu et al . , 2010a; Fleire et al . , 2006 ) , antigen affinity ( Liu et al . , 2010a; Fleire et al . , 2006 ) , antigen valency ( Bachmann et al . , 1993; Liu and Chen , 2005; Liu et al . , 2004 ) , Brownian mobility of antigen ( Wan and Liu , 2012 ) , and the stiffness feature of the substrates tethering the antigens ( Wan et al . , 2013; Zeng et al . , 2015 ) . Remarkably , this suggests that B cells use mechanosensing capability to sense the biochemical and biophysical properties of the antigens and their presenting substrates . Recent studies suggested that antigens encountered by B cells in vivo are actually presented on substrates with diverse stiffness ( Bachmann and Jennings , 2010 ) . The stiffness feature of the substrates presenting the antigen can greatly influence the initiation of B cell activation by regulating the efficiency of the accumulation of BCR and antigen molecule into the B cell immunological synapse ( Wan et al . , 2013; Zeng et al . , 2015; Wan et al . , 2015 ) . Here we investigated the underlying molecular mechanisms used by B cells to discriminate substrate stiffness . We found that synaptic recruitment of BCRs is significantly enhanced on activation by antigens on stiff substrates compared with antigens on soft substrates . Our experimental system was very similar to that used by Kam and colleagues , in which only IgG anti-CD3 and anti-CD28 surrogate antigens were tethered to PA substrates ( Judokusumo et al . , 2012 ) , and to the system used by Milone and colleagues in which only IgG anti-CD3 and anti-CD28 surrogate antigens were tethered to PDMS substrates ( O'Connor et al . , 2012 ) . In these systems , no adhesive ligands were used , but , in both cases , T cells similarly showed strong mechanosensing abilities . These T cell studies , in addition to the B cell studies presented in this report , show that mechanosensing by lymphocyte cells may not solely function through direct interaction between ICAM-1 and the well-characterized mechanosensor LFA-1 ( Chen et al . , 2010 , 2012 ) . However , direct ICAM-1 and LFA-1 interactions may still be required to maintain and regulate the mechanosensing ability of lymphocyte cells as other studies have shown that they can fine-tune the activation of both B and T cells ( Carrasco et al . , 2004; Arana et al . , 2008a; McLeod et al . , 2004; Spaargaren et al . , 2003; Arana et al . , 2008b ) . Considering this , in this study , we show that the presence of adhesion molecules ( ICAM-1 or VCAM-1 ) greatly enhanced the B cells’ capability to discriminate between the degrees of stiffness . Interestingly , recent studies indicate that germinal center B cells ( GCBs ) recognize antigens on antigen-presenting cells through a specialized immune synapse architecture that is distinct from that of mature naïve B cells ( Nowosad et al . , 2016 ) . Moreover , it has also been reported that specialized antigen-presenting cells , follicular dendritic cells , are usually stiff cells that can promote efficient antigen extraction and stringent affinity discrimination of GCBs , while regular dendritic cells that are mainly responsible for the antigen presentation for mature naïve B cells are mostly soft cells ( Spillane and Tolar , 2017 ) . Based on these published studies and the data in this report , we propose that adhesion molecules on antigen-presenting cells in germinal centers play an important role in enhancing the activation of GCBs . Indeed , this is supported by a recent study showing that integrin-–ligand interaction within the germinal center B cells and the antigen-presenting FDCs are important for the responses in GC ( Wang et al . , 2014 ) . In this report , we investigated the molecular nature of the mechanosensor machinery used by B cells to distinguish substrate stiffness in the absence of adhesion molecule-triggered integrin activation . We used a library of chicken DT40 B cell lines deficient for specific signaling molecules ( Kurosaki et al . , 2010; Kurosaki , 1999 ) , including Lyn , Syk , PLCγ2 , Btk , BLNK , PKCβ , and Carma-1 , to dissect the underlining molecular mechanism . The first striking observation was that only BCR signaling-dependent , not BCR signaling-independent , accumulation of BCRs into the B cell IS is subjected to strict regulation by the mechanosensing capability of B cells . Published studies and the data in this report show that BCR signaling-independent accumulation of BCRs and/or antigens was likely induced by the passive trapping of BCRs and antigens into the B cell IS , as both types of molecules exhibit free Brownian thermal diffusion before recognition ( Weber et al . , 2008; Liu et al . , 2010a , 2010b; Tolar et al . , 2009 ) . These published studies also indicate that both signaling-independent and signaling-dependent mechanisms account for accumulation of BCRs into the BCR microclusters . These two steps are not mutually exclusive , but are instead closely related in a sequential and synergistic way to maximize efficient BCR clustering . Specifically , signaling-independent BCR clustering initiates the earliest signaling on antigen and BCR recognition , which further enhances BCR clustering in a signaling-dependent manner . Our data in this report show that the initial signaling-independent accumulation of BCRs into the BCR microclusters is not sensitive to the stiffness features of the substrates , whereas the subsequent signaling-dependent accumulation of BCRs into the BCR microclusters is . Based on all these studies , it is reasonable to conclude that thermal diffusion-mediated and BCR signaling-independent passive trapping of BCR molecules is not sensitive to substrate stiffness as the Brownian diffusion of BCRs on the plasma membrane would not be affected by the stiffness of the substrates . In this report , major efforts were made to address how BCR signaling transduction enables B cells to discriminate substrate stiffness during initiation of B cell activation . Mechanistically , the results indicate that PKCβ functions downstream of Btk and PLCγ2 , as PMA-induced activation of PKCβ can bypass the requirements of Btk and PLCγ2 for discriminating substrate stiffness , demonstrating the vital importance of PKCβ in mediating the capability of B cells to discriminate substrate stiffness . We excluded an involvement for PKCβ-mediated NF-κB activation by showing that Carma1-KO B cells maintained their discrimination capability . This result makes sense as the activation of NF-κB leads to alterations in gene expression profiles in B cells at a later time point ( on the scale of hours ) , while the substrate stiffness discrimination capability of B cells kicks in minutes after the initial BCR and antigen interaction . Then , the next question becomes which molecular mechanisms determine this dependence of PKCβ and other upstream BCR signaling molecules for the substrate stiffness discrimination of B cells ? Some interesting insight came from a study showing that membrane-proximal BCR signaling molecules including Btk , PLCγ2 , BLNK , and PKCβ are essential for the inside-out activation of integrin molecules ( Arana et al . , 2008a , 2008b; Spaargaren et al . , 2003 ) . Taking this into account with our findings that an outside-in activating signal of integrin significantly enhanced the capability of B cells to discriminate substrate stiffness , we validated our hypothesis that PKCβ-dependent FAK activation accounts for the substrate stiffness discrimination capability of B cells . FAK plays a key role in the activation of integrin signaling pathways by providing a key docking site for Src family kinases , which in turn phosphorylate downstream signaling molecules ( Yu et al . , 2012; Bashour et al . , 2014; Slack-Davis et al . , 2007 ) . Specifically , early biochemical studies demonstrated that FAK phosphorylation at Tyr-925 regulates cross-talk between focal adhesion turnover and cell protrusion in embryonic fibroblasts ( Deramaudt et al . , 2011 ) . Indeed , FAK inactivation completely blunted the capability of the B cells to discriminate substrate stiffness , which could be rescued by the exogenous expression of FAK-WT but not the inactivated mutant at Tyr-925 . Mechanistically , we found that FAK activation is triggered by BCR engagement in a PKCβ-dependent manner . These findings are consistent with earlier studies showing that integrin-induced FAK phosphorylation can be blocked by inhibiting PKC , and that the PKC activator , PMA , drastically enhanced phosphorylation , and thus activation , of FAK ( Disatnik and Rando , 1999 ) . Further colocalization analyses in this report indicated that pFAK mainly localized to the peripheral region of the B cell IS , and highly colocalized with F-actin , strongly suggesting that FAK activation supported B cell spreading and adhesion responses ( Figure 9—figure supplement 1J ) . Indeed , additional analyses demonstrated that PKCβ-dependent FAK activation accounts for B cell discrimination of stiffness by potentiating B cell spreading and adhesion responses . These findings in B cells are consistent with an earlier study showing that the spreading and adhesion responses of muscle cells are regulated by typical PKC-mediated FAK activation ( Disatnik and Rando , 1999 ) . A similar observation was also reported in the case of snail defense cells from Lymnaea stagnalis ( Disatnik and Rando , 1999; Walker et al . , 2010 ) . What is the physiological significance of the core finding in this report that PKCβ-dependent FAK activation accounts for B cell discrimination against substrate stiffness ? We propose two obvious applications: Firstly , we propose that B cells have evolved to maintain a high efficiency to discriminate substrate stiffness features because non-self-antigens presented by viral capsids usually exhibit a high degree of stiffness ( 45 , 000–1 , 000 , 000 kPa ) ( Mateu , 2012 ) . In contrast , self-antigens presented by the plasma membrane usually show a low level of stiffness ( 0 . 01–1000 kPa ) ( Nemir and West , 2010 ) and soluble self-antigens in humoral microenvironment usually display a particularly low degree of stiffness ( several Pa ) ( Araujo et al . , 2012 ) . We believe that our findings here provide a new explanation for the widely accepted observations in vaccine research and administration that viral like particle ( VLP ) antigens are more potent than soluble antigens to induce antibody responses ( Bachmann and Jennings , 2010; Bachmann et al . , 1997 , 1993 ) . Secondly , this report also suggests that RA patient B cells exhibit a disordered capability to discriminate substrate stiffness . B cells from RA patients exhibited weaker discrimination capability than B cells from healthy controls . Specifically , B cells from healthy controls preferentially enhanced the synaptic accumulation of BCR microclusters on stiff substrates . In marked contrast , B cells from RA patients exhibited a different preference of enhancing the accumulation on soft substrates . As B cells from RA patients are known to exhibit hyper BCR signaling ( Nakken et al . , 2011b; Oligino and Dalrymple , 2003; Szodoray et al . , 2006 ) , it is our speculation that the enhanced activation of RA patient B cells on soft substrates results from signaling-dependent accumulation of BCRs into the BCR microclusters , as discussed above . These findings have obvious clinical relevance as there is a diverse range in stiffness features of the antigen-presenting substrates as well as changes in substrate stiffness at the physiological level versus pathological conditions which are associated with disease ( Knight , 2015 ) . It is reported that auto-reactive human primary B cells from RA patients can efficiently acquire the auto-antigen aggrecan , in a BCR and adhesion-dependent manner ( Ciechomska et al . , 2014 ) and RA patients usually exhibit highly enriched B cells in the synovia ( Nakken et al . , 2011a ) . It is expected that the altered stiffness properties of the ECM , which displays the auto-antigens , can drive the auto-reactive B cells to break the anergy state and instead undergo aberrant activation . Indeed , our finding is well supported by a report showing that reduced cartilage stiffness renders B cells into auto-antigen presenters in RA patients , which subsequently causes production of auto-antibodies ( Mauri and Ehrenstein , 2007 ) . As the ECM-associated microenvironment provides an abundant source of antigens , it has been proposed that any change in ECM stiffness could be identified as a threatening signal by the immune system and thus trigger a response from immune cells ( Tesniere et al . , 2008; Schaefer , 2010; Knight , 2015 ) . Taking our findings into consideration , it is an intriguing hypothesis that the capability of B cells to sense the stiffness features of antigen-presenting surfaces may be potentially related to detection of that danger by the immune system ( Knight , 2015 ) . Conclusively , all these data shed light on the precise molecular mechanism of how B cells discriminate substrate stiffness in a PKCβ- and FAK-dependent manner during initiation of B cell activation , improving our understanding of the sophisticated mechanosensing capability of B cells . We also propose that the mechanosensing and mechanotransducing abilities of immune cells deserve further investigation as these studies could enhance our understanding of immune cell activation on antigen recognition and may help build better vaccines to ultimately cure autoimmune diseases . CH27 B cell line ( RRID:CVCL_7178 , Source: mouse lymphoma ) was gifted by Dr . Susan K . Pierce ( NIAID-NIH ) that was originally purchased from ATCC ( USA ) . Similarly , B1-8 specific primary B cells were negatively selected from IgH B1-8/B1-8 Igκ −/− transgenic mice as described previously ( Liu et al . , 2010a ) . All the chicken DT40 B cells , including DT40-WT ( RRID:CVCL_J437 ) , DT40-LYN KO ( RRID:CVCL_1T41 ) , DT40-SYK KO ( RRID:CVCL_1T43 ) , DT40-PLCγ2 ( RRID:CVCL_1T47 ) , DT40-BTK ( RRID:CVCL_1T45 ) , DT40-BLNK KO ( RRID:CVCL_1T38 ) and DT40-PKCβ KO ( RRID:CVCL_1T42 ) were gifts for laboratory scientific studies from Dr . Tomohiro Kurosaki ( RIKEN , Japan ) . CD11a ( ITGAL ) KO mice were gifted by Dr . Yan Shi ( Tsinghua University , China ) . Mice B cell lines including CH27 and mouse primary naïve B cells were cultured in RPMI-1640 medium supplemented with 10% FBS , 50 μM β-mercaptoethanol ( Sigma-Aldrich ) , and penicillin/streptomycin antibiotics ( Invitrogen ) . DT40 chicken B cell lines including WT and KO used in this study were maintained at 37°C in the same medium as above , but 1% chicken serum was added into the above mentioned culture medium . 293 T cells were purchased from Cell bank ( Chinese academy of sciences , Shanghai ) . 293 T cells were maintained in the DMEM culture medium supplemented with 10% FBS , and penicillin/streptomycin antibiotics ( Invitrogen ) . All cell lines used in this study were negative for mycoplasma contamination test using a PCR detection method . Mouse anti-chicken IgM antibody ( clone M1 or M4 ) ( cat# 8310–01 or cat# 8300–08 ) and Goat anti-mouse Igκ kappa light chain , Goat anti-human kappa and lambda light chain antibodies were purchased from Southern Biotech . DyLight 649-conjugated Fab anti-mouse IgM constant region antibodies , Alexa Fluor 647 conjugated Fab fragment of IgM constant region antibodies anti-human , Goat F ( ab ) 2 anti-mouse IgM + IgG ( H+L ) ( Lot# 119316 ) and Cy5-conjugated Fab Goat anti-mouse IgM were purchased from Jackson ImmunoResearch . Phospho-FAK ( Try 925 ) ( RRID: AB_10831810 ) was purchased from Cell Signaling . Labeling mouse anti-chicken IgM antibody ( clone M1 ) with Alexa Fluor-647 and digesting the Fab fragment of mouse anti-chicken IgM antibody ( clone M1 ) were done following our published protocol ( Liu et al . , 2010b , 2010a ) . NP8-BSA was obtained from Bioresearch technology . Rabbit polyclonal anti-mouse FAK antibodies ( Cat# AMO0672 ) were purchased from ThermoFisher . Chicken Lyn , Syk and PLCγ2 , Btk and PKCβ were cloned from DT-40 cDNA and constructed by fusing GFP with the C-terminal , then incorporated into pEGFP vectors . Chicken FAK was cloned from DT40-WT cDNA and constructed by fusing mCherry with the N-terminal , then incorporated into pHAGE vector . A PCR-based mutagenesis strategy was used to construct pHAGE-mCherry-FAK-Y926F plasmid through Gibson Assembly . pSpCas9-2a-GFP was a gift from Dr . Feng Zhang ( MIT , Cambridge ) . FAK was knocked out in DT40 B cells using the CRISPR/Cas9 technique . Guide RNA was designed using the website http://crispr . mit . edu . Two target sites on the N-terminal and C-terminal of the gene were used to promote the KO efficiency , and the sequences of those are AACCTTTAGGACTCGCTCCA and GGCTGGTCATGACGTACTGC . The DT40 B cells transiently expressing pSpCas9-2a-GFP by electroporation were sorted and cultured in a 96-well plate . The KO cells were detected by PCR and western-blot . For DT40 B cell electroporation , the Buffer T and B-009 program of Amaxa Nucleofector was used according to the protocol from Lonza . For the retroviral transduction , 293 T cells were transfected with pHAGE and packaging plasmids by the calcium phosphate method . After 48 hr at 37°C ecotropic viral supernatants were collected and added to B cell culture medium in the presence of 5 μg/mL polybrene . Positive cells were sorted by flow cytometry . Cell sorting was performed using the FACSAria III Cell Sorter ( BD ) , by following the BD protocols . 2×Lysis buffer containing Tris-Hcl pH 7 . 4 50 mM , NaCl 150 mM , EDTA 20 mM and NP40 4% was used to lyse 2 × 106 DT40 WT and FAK-KO B cells . Protein was separated by 10% Bis-Tris PAGE ( Life technologies ) , and then transferred to a polyvinylidene fluoride ( PVDF ) membrane . FAK was probed with primary antibody , Rabbit pAb anti-mouse FAK ( RRID:AB_1500093 ) , and then appropriate HRP-conjugated secondary antibodies ( Dako ) . We used 5 mM concentration of EDTA in PBS containing 1% FBS . We pre-treated 2 × 106 CH27 B cells with EDTA-containing buffer for 20 min at 37°C , washed the cells twice with 1X PBS and then proceeded to the next step . A total of nine RA patients , matching the criteria of RA disease according to the American College of Rheumatology , were enrolled . For the control group healthy volunteers were recruited . This study was approved by the committee of ethics at Beijing People’s Hospital of Peking University . Each healthy volunteer and patient submitted their informed consent . A total of 8 ml of peripheral blood was acquired from each person . PBMC were isolated from the healthy and RA patient samples using Ficoll-Paque plus density separation , and were frozen at −80 under liquid nitrogen prior to use . Before use for imaging experiments , human PBMS cells were cultured in a RPMI-1640 medium supplemented with 10% FBS , 50 μM β-mercaptoethanol ( Sigma-Aldrich ) , and penicillin/streptomycin antibiotics ( Invitrogen ) for 2–3 hr . Cells were stimulated with Goat anti-human kappa and lambda light chain antibodies . For staining the BCRs of PBMS , Alexa 647 conjugated Fab fragment of IgM constant region anti-human antibodies were used . To prepare antigen-presenting polyacrylamide gel ( PA ) substrates , glass coverslips were treated with NaOH , 3-aminopropyltrimethoxysilane , and glutaraldehyde in a step-by-step manner . After washing extensively in distilled H2O , the glass coverslips were ready to support the polyacrylamide gels . The rigidity of the polyacrylamide gel substrate was controlled using different amounts of bisacrylamide cross-linker while keeping the total acrylamide concentration constant at 10% ( w/v ) . In our report , polyacrylamide gels with different Young’s modulus were produced using bisacrylamide concentrations of 0 . 8% and 0 . 05% ( w/v ) . Antigens were tethered to the surface of polyacrylamide gel substrates following methods described previously ( Judokusumo et al . , 2012 ) . In our report , streptavidin-conjugated acrylamide was polymerized into the polyacrylamide gel for the purpose of tethering the biotinylated F ( ab’ ) two anti-IgM antibody as surrogate antigen , which was generally incubated with the substrate at 37°C for 30 min at a concentration of 30 μg/ml . After extensive washing , the polyacrylamide gel substrate tethering specific antigens were ready for use ( Wan et al . , 2013 ) . Poly-dimethylsiloxane ( PDMS ) substrates were prepared following a standard protocol from our previously published studies ( Zeng et al . , 2015 ) . In brief , we prepared the PDMS substrates by mixing dimethylsiloxane monomer ( Dow Corning Sylgard 184 ) with a cross-linking agent , according to the manufacturer's instructions . The ratio of cross-linking agent to base polymer was 1:5 ( Stiff ) or 1:50 ( Soft ) to prepare PDMS substrates with different stiffness features . After defoaming in a mixer and degassing under vacuum , PDMS elastomers were cured at 60°C for 4 hr on glass coverslips or in 24-well-cell culture plates . PDMS forms a planar surface with highly hydrophobic features that can easily tether proteins through adsorption . To coat antigens which can activate B cells , PDMS substrates were coated with 5 µg/ml NP8-BSA or 5 µg/ml anti-BCR surrogate antigens in PBS as surrogate antigen overnight at 4°C . After washing with PBS , the PDMS elastomers were blocked with 5% BSA in PBS at 37°C for 1 hr , followed by thorough washing before downstream experiments ( Zeng et al . , 2015 ) Alexa Fluor 647-conjugated mouse anti-chicken IgM antibody ( clone M4 ) was used on PDMS surfaces to examine the tethered concentration of antigen , then imaged by confocal fluorescence microscopy ( Zeiss , LSM710 ) . To determine the antigen accessibility towards antibody , mouse anti-chicken IgM antibody ( clone M4 ) was incubated on soft and stiff PDMS substrate as surrogate antigen overnight at 4°C . After washing with PBS , it was blocked with 5% BSA for 1 hr at 37°C and then washed thoroughly . 100 nM Alexa DyLight 649-conjugated Fab anti-mouse IgM constant region antibody was added and incubated for 1 hr at 37°C , and then images were acquired after washing thoroughly . The mean fluorescence intensity ( MFI ) was analyzed by Image J ( NIH , U . S . ) . For cell accessibility , DT40 B cells were loaded on both the surfaces for 10 min at 37°C . Adhered B cells were imaged under a microscope before washing or after washing with 10 ml PBS-1% FBS at the speed of 0 . 5 or 1 ml/s . Images were acquired and quantified for the number of B cells adhered before washing and the number left on the PDMS surface after washing . The rate of adhesion was quantified according to the following equation: Adhesion Rate = the number of cells in a defined area on the gel surface after washing/the number of cells on the same area of the gel surface before washing Recombinant mouse ICAM-1/Fc and VCAM-1/Fc ligands were purchased from Sino Biological Inc and stored at −80°C as per the instructions of the manufacturer . After overnight incubation of surrogate antigen on PDMS at 4°C , ICAM-1 or VCAM-1 ( 2 μg/ml ) were added and incubated for 2 hr at 37°C . After washing , the chambers were blocked with 5% BSA for 30 min at 37°C and washed thoroughly before use . Then , pre-stained cells were loaded into the chamber and incubated for 10 min at 37°C . Cells were fixed with 4% paraformaldehyde at room temperature for 15 min , then washed carefully with PBS and imaged under TIRFM . Cells were pre-treated with the inhibitors under different conditions following protocols in published studies or manufacturer’s instructions as detailed below . Piceatannol ( CALBiochem ) was used at 50 nM working concentration at RT for 15 min ( Liu et al . , 2010a ) . PP2 was used at 20 µM ( Sigma-Aldrich ) ( Liu et al . , 2010a ) , U73122 at 5 µM ( EMD Millipore ) ( Mao et al . , 2006 ) , PF573-228 at 1 µM ( Sigma-Aldrich ) ( Slack-Davis et al . , 2007 ) , and bisindolylemaleimide at 3 . 5 µM ( Sigma-Aldrich ) ( Zhou et al . , 1999 ) . DT40 WT and KOs were treated with different concentrations of PMA ( 5 , 20 , 50 , and 100 ng/ml ) for 1 hr at RT ( Sigma-Aldrich ) ( Quann et al . , 2009 ) . After the inhibitor treatment as described above , the cells were washed three times with PBS and were ready for further processing . Surrogate antigens were tethered on the PDMS substrates and B cells were labeled with 200 nM of the DyLight 649-conjugated Fab anti-mouse IgM constant region antibodies or Alexa 647 conjugated Fab fragment of anti-human IgM constant region antibodies . After washing twice carefully , cells were stimulated with the surrogate antigens for 15–20 min at 37°C and then were fixed with the 4% paraformaldehyde for 30 min at RT . Cells were washed with the PBS slowly and then treated with the 0 . 2% Triton-X 100 in PBS for 20 min at RT . Donkey nonspecific IgG ( Jackson Immunoresearch Laboratory ) was employed for 1 hr at 37°C as a blocking reagent . Phospho-FAK ( Try 925 ) ( Cell Signaling ) antibody was used as the primary antibody for 1 hr at 37°C . After careful washing , Alexa 555 Donkey anti-rabbit IgG antibody was used as a secondary antibody for 45 min at RT . After careful washing , cells were imaged under a confocal microscope . For all of the respective experiments , B cells were obtained from chicken cell line DT40 , mice cell line CH27 or mouse primary naïve B cells , or human PBMC cells , and stained with Alexa Fluor 647-conjugated mouse Fab anti-chicken IgM ( clone M1 ) , Alexa Fluor 647-conjugated Goat Fab anti-mouse IgM specific for Fc5μ , or Alexa 647 conjugated Fab fragment of IgM constant region antibodies anti-human , respectively . After washing twice they were placed on either the soft or stiff PDMS elastomer surfaces with tethered antigens for 10 or 15 min at 37°C , 5% CO2 depending on the experiment . The cells were then fixed with 4% paraformaldehyde at room temperature for a minimum of 10 min . An Olympus IX-81 Microscope was used to acquire images supported by the Port of TIRF . EMCCD electron-multiplying camera ANDOR iXon + DU897D , 100X Olympus 1 . 49 NA objective . Lenses with lasers of 488 nm , 561 nm , and 647 nm were used ( Sapphire lasers , Coherent ) . The time of exposure was 100 ms until it was indicated specifically . Metamorph software was used to control the acquisition ( MDS Analytical Technologies ) . The images on PA gel and the IRM images were acquired using scanning laser confocal Olympus FLUOVIEW FV1000 with a 60x oil objective lens . Images were analyzed by Image J ( NIH , U . S . ) software as our previous studies reported ( Wan and Liu , 2012; Liu et al . , 2010c , 2010b , 2010a ) . Mean fluorescence intensity ( MFI ) value of BCRs , was calculated by Image J software based on mean pixel intensity . Briefly , regions of interest ( ROIs ) were marked in the images that were already subtracted for background . The MFI values were calculated as the ratio of integrated fluorescence intensity of a ROI to the total area , as our previous studies reported ( Wan and Liu , 2012; Liu et al . , 2010c , 2010b , 2010a ) . We calculated the ratio by dividing the BCR MFI of each cell on stiff substrates to the averaged value of the BCR MFI of all the cells on soft substrates . We used that value to quantify the substrate stiffness discrimination capability of B cells . The ratio was obtained from the following equation: Ratio = BCRs MFI of each cell on stiff substrates / the averaged value of the BCR MFI of all the cells on soft substrates	The human immune system protects the body from harmful bacteria , viruses and other microbes . Immune cells called B cells use proteins called B cell receptors on their surface to identify these invaders . When the B cell receptors detect molecules called antigens on the surface of the microbes , they produce signals that activate the B cell and enable it to combat the infection . Previous research has found that B cells react differently to antigens depending on the stiffness of the surface to which an antigen is attached . Now , Shaheen , Wan , Li et al . have attached antigens to artificial surfaces that were either stiff or soft and examined how B cells responded to them . Some of the B cells were modified to lack particular molecules that are important for B cell receptor signaling . The results of the experiments suggest that two signaling molecules – called protein kinase C beta ( PKCβ ) and focal adhesion kinase ( FAK ) – enable B cells to distinguish between the stiffness of different surfaces . PKCβ activates FAK , which causes the B cell to spread onto the surface and stick to it . However , B cells that had an inactive version of FAK – or lacked the protein entirely – did not efficiently spread onto the surfaces and were less able to discriminate between stiff and soft surfaces . In autoimmune diseases such as rheumatoid arthritis , B cells are overactive and attack the body’s own cells . Shaheen et al . found that the B cells of people with rheumatoid arthritis are less able to distinguish between stiff and soft surfaces than normal cells . Further research that investigates how to change the ability of a B cell to detect stiffness could therefore help researchers to develop treatments or vaccines for rheumatoid arthritis and other autoimmune conditions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "immunology", "and", "inflammation" ]	2017	Substrate stiffness governs the initiation of B cell activation by the concerted signaling of PKCβ and focal adhesion kinase
Proper development of the CNS axon-glia unit requires bi-directional communication between axons and oligodendrocytes ( OLs ) . We show that the signaling lipid phosphatidylinositol-3 , 5-bisphosphate [PI ( 3 , 5 ) P2] is required in neurons and in OLs for normal CNS myelination . In mice , mutations of Fig4 , Pikfyve or Vac14 , encoding key components of the PI ( 3 , 5 ) P2 biosynthetic complex , each lead to impaired OL maturation , severe CNS hypomyelination and delayed propagation of compound action potentials . Primary OLs deficient in Fig4 accumulate large LAMP1+ and Rab7+ vesicular structures and exhibit reduced membrane sheet expansion . PI ( 3 , 5 ) P2 deficiency leads to accumulation of myelin-associated glycoprotein ( MAG ) in LAMP1+perinuclear vesicles that fail to migrate to the nascent myelin sheet . Live-cell imaging of OLs after genetic or pharmacological inhibition of PI ( 3 , 5 ) P2 synthesis revealed impaired trafficking of plasma membrane-derived MAG through the endolysosomal system in primary cells and brain tissue . Collectively , our studies identify PI ( 3 , 5 ) P2 as a key regulator of myelin membrane trafficking and myelinogenesis . In the vertebrate CNS , the majority of long axons are myelinated . Myelin greatly increases the conduction velocity of action potentials and provides metabolic support for axons . Bidirectional axo-glial signaling is critical for nervous system myelination and fiber stability ( Nave and Trapp , 2008; Simons and Lyons , 2013 ) . Myelin development is regulated by oligodendrocyte ( OL ) intrinsic mechanisms ( Zuchero and Barres , 2013 ) , astrocyte secreted factors ( Ishibashi et al . , 2006 ) , neuronal electrical activity ( Barres and Raff , 1993; Ishibashi et al . , 2006 ) and axon derived chemical signals ( Coman et al . , 2005; Ohno et al . , 2009; Winters et al . , 2011 , Yao et al . , 2014 ) . Disorders associated with defective CNS white matter range from multiple sclerosis and inherited leukodystrophies to psychiatric disorders ( Fields , 2008; Makinodan et al . , 2012; Perlman and Mar , 2012 ) . FIG4 is an evolutionarily conserved lipid phosphatase that removes the 5’ phosphate group from phosphatidylinositol ( 3 , 5 ) bisphosphate [PI ( 3 , 5 ) P2] to produce PI ( 3 ) P . Together with its antagonistic kinase PIKFYVE and the scaffold protein VAC14 , FIG4 forms an enzyme complex that regulates the interconversion of PI ( 3 ) P and PI ( 3 , 5 ) P2 on membranes of the late endosomal/ lysosomal ( LE/Lys ) compartment ( Jin et al . , 2008; McCartney et al . , 2014 ) . In addition to its 5’-phosphatase activity , Fig4 is required to stabilize the enzyme complex . PI ( 3 , 5 ) P2 directly regulates the lysosomal cation channels TRPML1 , TPC1 and TPC2 ( Dong et al . , 2010; Wang et al . , 2012; 2014 ) . Reduced activity of these lysosomal channels and the resulting osmotic enlargement of the LE/Lys may underlie vacuolization in Fig4 null cells ( Lenk and Meisler , 2014 ) . Consistent with this model , overexpression of TRPML1 in Vac14 and Fig4 mutant cells appears to rescue vacuolization ( Dong et al . , 2010; Zou et al . , 2015 ) . In Drosophila , loss of TRPML1 generates a muscle vacuolization phenotype reminiscent of FIG4 deficiency ( Bharadwaj et al . , 2016 ) . FIG4 deficiency is particularly harmful for neural cells with elaborate morphologies , including projection neurons and myelinating glia . Mutations of human FIG4 result in neurological disorders including Charcot-Marie-Tooth type 4J , a severe form of peripheral neuropathy ( Chow et al . , 2007; Nicholson et al . , 2011 ) , polymicrogyria with epilepsy ( Baulac et al . , 2014 ) , and Yunis-Varon syndrome ( Campeau et al . , 2013 ) . Mice null for Fig4 exhibit severe tremor , brain region-specific spongiform degeneration , hypomyelination , and juvenile lethality ( Chow et al . , 2007; Ferguson et al . , 2009; Winters et al . , 2011 ) . We previously demonstrated that a Fig4 transgene driven by the neuron-specific enolase ( NSE ) promoter rescued juvenile lethality and neurodegeneration in global Fig4 null mice , and that these phenotypes were not rescued by an astrocyte-specific Fig4 transgene ( Ferguson et al . , 2012 ) . The neuron-specific transgene also rescued conduction in peripheral nerves ( Ferguson et al . , 2012 ) and structural defects in CNS myelination ( Winters et al . , 2011 ) . Conversely , inactivation of Fig4 specifically in neurons resulted in region-specific neurodegeneration ( Ferguson et al . , 2012 ) . The cellular and molecular mechanisms relating loss of Fig4 to hypomyelination are poorly understood . To further characterize the requirement of PI ( 3 , 5 ) P2 for CNS myelination , we manipulated individual components of the PI ( 3 , 5 ) P2 biosynthetic complex . Pikfyve and Vac14 global null mice die prematurely , before the onset of CNS myelination ( Zhang et al . , 2007; Ikonomov et al . , 2011 ) . To circumvent this limitation , we employed a combination of conditional null alleles and hypomorphic alleles in the mouse . Our study shows that multiple strategies to perturb the FIG4/PIKFYVE/VAC14 enzyme complex , and by extension the lipid product PI ( 3 , 5 ) P2 , result in the common endpoints of arrested OL differentiation , impaired myelin protein trafficking through the LE/Lys compartment , and severe CNS hypomyelination . We demonstrate that these defects in myelin biogenesis are functionally relevant and result in faulty conduction of electrical impulses . In the early postnatal brain , Fig4 is broadly expressed and enriched in oligodendrocyte progenitor cells ( OPCs ) and newly formed OLs ( NFOs ) ( Zhang et al . , 2014 ) . Mice in which exon 4 of the Fig4 gene is flanked by loxP sites ( Ferguson et al . , 2012 ) were used to generate Fig4-/flox , SynCre and Fig4-/flox , Olig2Cre mice deficient for Fig4 in neurons or OLs , respectively . Myelin development in these conditional mutants , as well as the Fig4 global mutant ( Fig4-/- ) and control mice ( Fig4+/+ and Fig4flox/+ ) , was analyzed by Fluoromyelin Green labeling ( Figure 1 ) . In control brains , the corpus callosum and internal capsule were prominently labeled ( Figure 1A and A’ ) . Staining of these structures was weaker in Fig4-/flox , SynCre brains and further reduced in Fig4-/flox , Olig2Cre and Fig4-/- brains ( Figure 1B-D’ ) . For a quantitative comparison of the myelination defects , whole brain membranes were prepared from P21 pups and analyzed by immunoblotting with antibodies specific for the myelin markers myelin-associated glycoprotein ( MAG ) , 2’ , 3’-cyclic-nucleotide 3’-phosphodiesterase ( CNPase ) , proteolipid protein ( PLP ) , and myelin basic protein ( MBP ) ( Figure 1E ) . Compared to Fig4+/+ membranes , a significant reduction in myelin proteins was evident in Fig4-/- mice , Fig4-/flox , SynCre mice and Fig4-/flox , Olig2Cre mice ( Figure 1F -I ) . The finding that the neuronal marker classIII β-tubulin is not significantly decreased in any of these mice indicates that the decrease in CNS myelin is not secondary to neuronal loss . While the Olig2 promoter is highly active in the OL lineage , activity has also been reported in astrocytes and a subset of neurons ( Dessaud et al . , 2007; Zhang et al . , 2014 ) . To independently assess the role of Fig4 in the OL lineage , we generated Fig4-/flox , PdgfraCreER mice that permit tamoxifen inducible gene ablation . At postnatal-days ( P ) 5 and 6 , before the onset of CNS myelination , Fig4-/flox , PdgfraCreER pups were injected with 4-hydroxytamoxifen and brains were analyzed at P20-P21 . Inducible ablation of Fig4 in the OL-linage resulted in reduced expression of the myelin proteins CNPase , MAG , and MBP , as assessed by Western blot analysis ( Figure 1—figure supplement 1A–B’ ) as well as myelin loss in forebrain structures and cerebellar white matter ( Figure 1—figure supplement 1C–D’ ) . Fewer Plp1+ OLs were present in optic nerve sections of Fig4-/flox , PdgfraCreER mice ( Figure 1—figure supplement 1E and E’ ) . Together , these studies indicate that proper CNS myelination is dependent upon OL cell-autonomous ( intrinsic ) functions of Fig4 , in addition to non-OL-autonomous ( extrinsic ) functions of Fig4 provided by neurons . 10 . 7554/eLife . 13023 . 003Figure 1 . Conditional ablation of Fig4 in neurons or OLs leads to CNS hypomyelination . ( A-D ) Coronal sections of juvenile ( P21-30 ) mouse forebrain stained with FluoroMyelin Green . ( A ) Fig4 control mice ( harboring at least one Fig4 WT allele ) , ( B ) Fig4 germline null mice ( Fig4-/- ) , ( C ) Fig4-/flox , SynCre mice and ( D ) Fig4-flox , Olig2Cre mice . Thinning of the corpus callosum and internal capsule ( white arrowheads ) is observed in Fig4-/- , Fig4-/flox , SynCre , and Fig4-flox , Olig2Cre mice . ( A’-D’ ) Higher magnification images of the corpus callosum . Scale bar ( A-D ) , 1 mm and ( A’-D’ ) , 400 µm . ( E ) Representative Western blots of P21 brain membranes prepared from Fig4+/+ ( WT ) , Fig4-/- , Fig4-/flox , SynCre and Fig4-/flox , Olig2Cre mice probed with antibodies specific for the myelin proteins MAG , CNPase , PLP , and MBP . To control for protein loading , membranes were probed for the neuronal marker class III β-tubulin ( βIII Tub ) . ( F-I ) Quantification of Western blot signals for MAG , MBP , CNPase , and PLP in Fig4+/+ ( black bars ) , Fig4-/- ( purple bars ) , Fig4-/flox , SynCre ( light blue bars ) , and Fig4-flox , Olig2Cre ( red bars ) brain membranes . Quantification of myelin protein signals is normalized to βIII Tub . Relative protein intensities compared to WT brain are shown as mean value ± SEM . For each of the four genotypes , three independent membrane preparations were carried out . One-way ANOVA with multiple comparisons , Dunnett posthoc test; p<0 . 01 , p<0 . 001 and **p<0 . 0001 . An independent strategy for OL-specific Fig4 deletion results in a similar phenotype as shown in Figure 1—figure supplement 1 . Histochemical staining of brain , spinal cord and dorsal root ganglion tissue sections of Fig4 conditional knock-out mice , as well as Kaplan-Meier plots for Fig4-/flox , SynCre and Fig4-flox , Olig2Cre mice are shown in Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 00310 . 7554/eLife . 13023 . 004Figure 1—figure supplement 1 . CNS hypomyelination in Fig4-/flox , PdgfrαCreER mice . Representative Western blots of ( A ) P21 forebrain and ( B ) P21 hindbrain ( cerebellum/brainstem ) lysates prepared from Fig4 control littermate mice ( Fig4+/flox , PdgfrαCre-ER ) and Fig4-/flox , PdgfrαCreER mutant mice , probed with antibodies specific for the myelin proteins MAG , CNPase and MBP . To control for protein loading , blots were probed for the neuronal marker class III β-tubulin ( βIII Tub ) . ( A’ and B’ ) Quantification of Western blot signals for MAG , CNPase and MBP in ( A’ ) forebrain and ( B’ ) cerebellum/brainstem lysates . Relative protein intensities compared to control tissue are shown as mean value ± SEM . Six pairs of control littermate and Fig4-/flox , PdgfrαCreER mice were analyzed and quantified . Unpaired Student’s t-test , p=0 . 0323 ( A' , MAG ) , *p=0 . 0006 ( A' , CNPase ) , p=0 . 0096 ( A’ , MBP ) , p=0 . 027 ( B' , MAG ) , p=0 . 038 ( B’ , CNPase ) , **p=0 . 0004 ( B’ , MBP ) . ( C and C’ ) Sagittal sections of P21 forebrain of control littermate ( Fig4+/flox , PdgfrαCreER ) and Fig4-/flox , PdgfrαCreER mutant mice probed for Mbp mRNA expression . ( D and D’ ) Sagittal sections of P21 cerebellum of control littermate ( Fig4+/flox , PdgfrαCreER ) and Fig4-/flox , PdgfrαCreER mutant mice probed for Mbp mRNA expression . ( E and E’ ) Longitudinal optic nerve sections of P21 littermate control and Fig4-/flox , PdgfraCreER mice probed for Plp1 mRNA expression . Scale bar ( C-D’ ) , 500 μm and ( E and E’ ) , 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 00410 . 7554/eLife . 13023 . 005Figure 1—figure supplement 2 . Loss of Fig4 in the OL-lineage or neurons differentially affects spongiform degeneration and lifespan . ( A-D’’’ ) Hematoxylin/eosin stained tissue sections of P30 mouse neocortex , cerebellum , dorsal root ganglion ( DRG ) and spinal cord ventral horn . Tissue sections of mice with the following genotypes are shown: ( A , B , C , D ) control mice ( Fig4flox/- ) , ( A’ , B’ , C’ , D’ ) Fig4 germline null mice ( Fig4-/- ) , ( A” , B” , C” , D” ) Fig4-/flox , SynCre conditional mutants and ( A’’’ , B’’’ , C’’’ , D’’’ ) Fig4-/flox , Olig2Cre conditional mutants . Most notable are the large vacuolar ( sponge-like ) structures in different regions of the Fig4-/- nervous system , including ( A’ ) deep layers of the neocortex , ( B’ ) deep cerebellar nuclei , ( C’ ) DRGs and ( D’ ) ventral horn of the spinal cord . ( A”-D” ) A milder but similar phenotype is observed in Fig4-/flox , SynCre mice . ( A’’’ ) In the Fig4-/flox , Olig2Cre neocortex small vacuoles are observed in all layers of the neocortex . ( B’’’ and C’’’ ) In Fig4-/flox , Olig2Cre mice deep cerebellar nuclei and DRGs look largely normal . ( D’’’ ) The large vacuoles in the spinal cord ventral horn of Fig4-/flox , Olig2Cre mice likely represent motoneurons , as the Olig2 promoter is known to drive cre expression in motoneurons and the OL-linage . ( E ) The Hb9-cre driver line was used for conditional deletion of Fig4 specifically in motoneurons . Toluidine blue stained section of Fig4-/flox , Hb9Cre ventral horn shows multiple large vacuolar structures within the gray and white matter of the spinal cord . Examples of vacuolar structures are labeled with asterisks . Apparently normal motoneuron profiles are indicated by arrows . ( F ) Electron micrograph of Fig4-/flox , Hb9Cre ventral horn with large vacuolar structures labeled by asterisks . Vacuolar structures are mostly devoid of electron-dense material and found in axons surrounded by thin myelin sheaths ( arrows ) . Vacuoles cause peripheral displacement of axoplasm and mitochondria . The arrowhead points to a dystrophic axon with accumulation of numerous smaller vesicles . ( G ) Viability of Fig4 conditional mutants . Kaplan-Meier plot shows an average life-span of 6 months for Fig4-/flox , SynCre mice ( n = 15 ) , while Fig4-/flox , Olig2Cre mice ( n = 5 ) are viable for 12–14 months ( the oldest mice currently in our colony ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 005 As previously described , Fig4-/flox , SynCre mice exhibit impaired movement and region-specific vacuolization and neurodegeneration ( Figure 1—figure supplement 2A” , B” , C” , D” ) ( Ferguson et al . , 2012 ) . In contrast , Fig4-/flox , Olig2Cre mice exhibit very mild vacuolization in brain ( Figure 1—figure supplement 2A’’’ , B’’’ , C’’’ , D’’’ ) . Consistent with the known expression of the Olig2 promoter in motor neurons ( Mizuguchi et al . , 2001 ) ventral spinal cord of Fig4-/flox , Olig2Cre mice shows extensive vacuolization ( Figure 1—figure supplement 2D’’’ ) , similar to Fig4-/flox , Mnx1Cre ( otherwise referred to as Fig4-/flox , Hb9Cre ) mice ( Figure 1—figure supplement 2E ) ( Vaccari et al . , 2015 ) . Analysis of Fig4-/flox , Hb9Cre spinal cord identified enlarged vacuoles within motoneuron axons , greatly extending their diameter and pushing the axoplasm into a thin peripheral rim near the plasma membrane ( Figure 1—figure supplement 2F ) . In contrast to the movement disability and reduced survival of Fig4-/flox , SynCre mice , ( Ferguson et al . , 2012 ) the movement of Fig4-/flox , Olig2Cre mice is normal and no premature death was observed , with the oldest now surviving beyond 14 months of age ( Figure 1—figure supplement 2G ) . There are no obvious defects in mobility of littermate controls and Fig4-/flox , Olig2Cre conditional mutant mice at P23 , as demonstrated in the Videos 1 and 2 . 10 . 7554/eLife . 13023 . 006Video 1 . Normal locomotion of juvenile Fig4+/flox , Olig2Cre mice . A representative video of a control Fig4+/flox , Olig2Cre mouse at P23 . N = 10DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 00610 . 7554/eLife . 13023 . 007Video 2 . Normal locomotion of juvenile Fig4-/flox , Olig2Cre mice . A representative video of a Fig4-/flox , Olig2Cre conditional mutant mouse at P23 shows no obvious pathology in locomotion . N = 10DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 007 Analysis of P21 retina revealed the presence of numerous vacuoles in the inner retina of Fig4-/flox , SynCre mice but no defects in overall morphology or stratification ( Figure 2A’ ) . No vacuoles were detected in the Fig4-/flox , Olig2Cre retina ( Figure 2A’’ ) . For ultrastructural analysis , optic nerves of Fig4 conditional knock-out mice were processed for transmission electron microscopy ( TEM ) . In P21 Fig4 control mice ( retaining at least one intact allele of Fig4 ) , the fraction of myelinated axons in the optic nerve is 79± 2% . In optic nerves of Fig4-/flox , SynCre mice , only 9± 3% of axons are myelinated and in Fig4-/flox , Olig2Cre mice only 12± 1% of axons are myelinated ( Figure 2B–B’’ and D ) . To assess myelin health , we determined the g-ratio ( the ratio of the inner axonal diameter to the total fiber diameter ) of myelinated axons in the optic nerve of Fig4 control and conditional mutants . Compared to control mice , a small but significant increase in g-ratio was observed in Fig4-/flox , SynCre and Fig4-/flox , Olig2Cre mice , an indication of myelin thinning ( Figure 2E ) . To determine whether the optic nerve hypomyelination at P21 reflects a transient delay in myelin development , rather than a lasting defect , we repeated the analysis with adult mice . Similar to P21 optic nerves , ultrastructural analysis of both types of adult optic nerves revealed profound hypomyelination ( Figure 2C–C’’ ) . At P60-75 , 92± 2% of axons are myelinated in Fig4 control nerves . This is reduced to 16± 4% in Fig4-/flox , SynCre mice and 12± 2% in Fig4-/flox , Olig2Cre mice ( Figure 2F ) . It is noteworthy that conditional ablation of Fig4 either in neurons or OLs leads to preferential absence of myelin sheaths on small and intermediate caliber axons , while many large caliber axons undergo myelination ( Figure 2B’ , B” , C’ and C” ) . 10 . 7554/eLife . 13023 . 008Figure 2 . Conditional ablation of Fig4 in neurons or in OLs leads to severe dysmyelination of the optic nerve . ( A-A’’ ) Sagittal sections of juvenile ( P21 ) mouse retina embedded in epoxy resin and stained with toluidine blue . ( A ) Fig4 control mice , harboring at least one Fig4 WT allele , ( A’ ) Fig4-/flox , SynCre mice and ( A’’ ) Fig4-/flox , Olig2Cre mice . Scale bar , 100 µm . ( B-B’’ ) Representative TEM images of optic nerve cross sections of P21 ( B ) Fig4 control , ( B’ ) Fig4-/flox , SynCre and ( B’’ ) Fig4-/flox , Olig2Cre mice . ( C-C’’ ) Representative TEM images of optic nerve cross sections of adult ( P60-75 ) mice . ( C ) Fig4 control , ( C’ ) Fig4-/flox , SynCre and ( C’’ ) Fig4-/flox , Olig2Cre mice . Black arrows in C’ indicate the presence of dystrophic axons . Scale bar ( B-C’’ ) = 1 μm . ( D ) Quantification of percentage of myelinated fibers in the optic nerve at P21 and P60-75 . At P21 , Fig4 controls ( n = 3 mice , 3 nerves ) ; Fig4-/flox , SynCre ( n = 2 mice , 3 nerves ) and Fig4-/flox , Olig2Cre ( n = 3 mice , 3 nerves ) . ( E ) Quantification of myelinated fiber g-ratios in the optic nerve at P21 , n = 3 animals , 3 nerves for all groups . ( F ) Quantification of myelinated fibers in the optic nerve at P60-P75 . Fig4 control ( n = 4 mice , 4 nerves ) , Fig4-/flox , SynCre ( n = 4 mice , 4 nerves ) ; Fig4-/flox , Olig2Cre ( n = 3 mice , 4 nerves ) . Results are shown as mean value ± SEM , one-way ANOVA with multiple comparisons , Tukey posthoc test; n . s . p>0 . 05 , p=0 . 0211 , p=0 . 0055 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 008 Few axons in the optic nerve of adult Fig4-/flox , SynCre mice showed signs of degeneration ( Figure 2C’ ) . No evidence for axonal degeneration was observed in Fig4-/flox , Olig2Cre optic nerves . CNS hypomyelination in Fig4-/flox , Olig2Cre mice was still present at P150 , the oldest time point examined by TEM ( data not shown ) . Thus , the optic nerve hypomyelination observed at P21 is not transient in nature but persists into adulthood . We conclude that selective ablation of Fig4 either in neurons or in the OL lineage leads to profound CNS dysmyelination . To determine whether the morphological defects in CNS myelin of Fig4 conditional mutants result in functional deficits , we performed electrophysiological recordings . We measured the conduction velocity and amplitude of compound action potentials ( CAPs ) in optic nerves acutely isolated from P21 mice . Global deletion of Fig4 ( Fig4-/- ) results in a dramatic reduction in a population of fast conducting fibers and a corresponding increase in the proportion of slowly conducting fibers ( Figure 3A , B , E ) ( Winters et al . , 2011 ) . The average velocity of the largest peak in Fig4 control nerves carrying at least one intact allele of Fig4 is 1 . 9 ± 0 . 1 m/s but in Fig4-/- nerves this is reduced to 0 . 7 ± 0 . 2 m/s . A similar CAP redistribution was observed in optic nerves prepared from Fig4-/flox , SynCre mice ( 0 . 7 ± 0 . 1 m/s ) and Fig4-/flox , Olig2Cre mice ( 0 . 6 ± 0 . 03 m/s ) ( Figure 3C , D , E ) . Thus , consistent with biochemical and morphological analyses ( Figures 1 and 2 ) , loss of Fig4 in neurons or in the OL-lineage results in slowed nerve conduction . 10 . 7554/eLife . 13023 . 009Figure 3 . Conditional ablation of Fig4 in neurons or OLs leads to impaired conduction of electrical impulses in the optic nerve . Compound action potential ( CAP ) recordings from acutely isolated optic nerves of P21 mice . ( A ) Representative CAP traces recorded from Fig4 control mice , harboring at least one Fig4 WT allele ( n = 14 nerves ) , ( B ) Fig4-/- mice ( n = 5 nerves ) , ( C ) Fig4-/flox , SynCre mice ( n = 11 nerves ) and ( D ) Fig4-/flox , Olig2Cre mice ( n = 9 nerves ) . For each graph , the arrow indicates the largest amplitude peak , as identified by Gaussian fit . ( E ) Quantification of average conduction velocity of largest amplitude peaks identified in A-D . Results are shown as mean value ± SEM , one-way ANOVA with multiple comparisons , Dunnett posthoc , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 009 To assess the cellular basis of the CNS hypomyelination phenotype , we stained optic nerve cross sections from Fig4 conditional mutants for markers in the OL lineage . Compared to Fig4 control optic nerves , the diameter of nerves from P21 Fig4-/flox , SynCre and Fig4-/flox , Olig2Cre mice were each reduced by 20% . The density of NG2+ progenitor cells in optic nerve tissue sections is comparable among the three genotypes ( Figure 4A–A’’ and D ) . The density of Olig2+ cells , a marker that labels immature and mature OLs , is reduced , as is labeling of Plp1 , a mature OL marker ( Figure 4B-B’’ , C–C’’ , E and F ) . These studies indicate that OPCs are present at normal density and tissue distribution in the Fig4 conditional null optic nerves , but they fail to generate the normal population of mature myelin-forming OLs . 10 . 7554/eLife . 13023 . 010Figure 4 . Conditional ablation of Fig4 in neurons or OLs results in a decrease of mature OLs . ( A , B , C ) Optic nerve cross sections from P21 Fig4 control mice , harboring at least one Fig4 WT allele , ( A’ , B’ , C’ ) Fig4-/flox , SynCre mice and ( A’’ , B’’ , C’’ ) Fig4-/flox , Olig2Cre mice were stained with anti-NG2 , anti-Olig2 or probed for Plp1 mRNA expression . Scale bar = 100 µm . ( D-F ) Quantification of labeled cells in optic nerve cross sections normalized to area in arbitrary units ( A . U . ) . ( D ) The density of NG2+ cells in Fig4 control ( n = 4 mice ) , Fig4-/flox , SynCre ( n = 3 mice ) and Fig4-/flox , Olig2Cre ( n = 3 mice ) optic nerves is not significantly ( n . s . ) different . ( E ) Quantification of the density of Olig2+ cells in Fig4 control ( n = 6 mice ) , Fig4-/flox , SynCre ( n = 3 mice ) and Fig4-/flox , Olig2Cre ( n = 4 mice ) optic nerves . ( F ) Quantification of the density of Plp1+ cells in Fig4 control ( n = 8 mice ) , Fig4-/flox , SynCre ( n = 4 mice ) and Fig4-/flox , Olig2Cre ( n = 4 mice ) optic nerves . Results are shown as mean value ± SEM , one-way ANOVA with multiple comparisons , Dunnett’s posthoc test . p=0 . 001 , *p=0 . 0002 and p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 010 For a more detailed analysis of the OL lineage , we isolated primary OPCs from P6-P14 Fig4 pups by anti-PDGFRα immunopanning ( Emery and Dugas , 2013 ) . Yields of OPCs per brain did not differ between control and Fig4-deficient mice ( data not shown ) . OPCs were cultured for two days in vitro ( DIV2 ) under proliferating conditions , fixed and analyzed by double-immunofluorescence staining of Ki67 and PDGFRα . The density of Ki67+/PDGFRα+ cells in Fig4+/+ and Fig4-/- cultures is very similar ( Figure 5—figure supplement 1A–B ) . After culture under standard differentiation conditions for 4 days , absence of PDGF and presence of triiodothyronine ( T3 ) , OPCs isolated from Fig4+/+ ( control ) or Fig4-/- pups both acquire a highly arborized morphology and positive staining for OL markers . The density of NG2+ cells and CNPase+ cells , normalized to Hoechst 33 , 342 dye+ nuclei , is comparable among wildtype and Fig4-deficient cultures ( Figure 5A–B’ , and C ) . However , the fraction of cells expressing the more mature OL markers MAG and MBP was significantly reduced in Fig4-/- cultures ( Figure 5A–B’ , and C ) . A more detailed categorization of post-mitotic OLs , based on actin and MBP double-labeling , revealed a significantly decreased number of Fig4-deficient OLs that matured to a stage with lamellar MBP+ membrane sheets ( Zuchero et al . , 2015 ) ( Figure 5D–E ) . The reduced number of mature OLs in Fig4-/- cultures was not caused by increased cell death ( Figure 5—figure supplement 1C–E ) . For a quantitative assessment of protein expression in primary OLs from Fig4+/+ and Fig4-/- brains , DIV 3 cultures were lysed and analyzed by capillary Western blotting ( Figure 5—figure supplement 2A–C ) . FIG4 is clearly detected in Fig4+/+ OL lysates but not in Fig4-/- OL lysates . In Fig4-/- lysates MAG is significantly reduced . Collectively , these data suggest that the initial programs of OL maturation progress normally in the absence of Fig4 while later stages of OL-differentiation , including lamellar membrane expansion , are Fig4-dependent . 10 . 7554/eLife . 13023 . 011Figure 5 . Fig4-deficient OLs show impaired differentiation and membrane expansion in vitro . Representative images of Fig4 control ( Fig4+/+ or Fig4+/- ) and Fig4-/- primary OLs after 4 days in differentiation medium , fixed and stained for the OL-lineage markers ( A and A’ ) NG2 and MAG; ( B and B’ ) CNPase and MBP . Scale bar in A-B’ , 200 µm . ( C ) Quantification of NG2 , CNPase , MAG , and MBP/CNPase labeled cells in Fig4 control ( n = 3 ) and Fig4-/- ( n = 3 ) cultures normalized to Hoechst 33342 dye labeled cells . The ratio of immunolabeled cells over Hoechst+ cells in Fig4 control cultures was set at 1 . Results are shown as mean value ± SEM , multiple t-test analysis with Holm-Sidak method . p=0 . 0075 ( MAG ) , p=0 . 012 ( MBP ) . ( D and D’ ) Confocal images of MBP+ and Actin Red 555+ OLs in Fig4 control and Fig4-/- cultures . Nuclei were labeled with TO-PRO-3 , scale bar = 20 µm . ( E ) Quantification of the fraction of “arborized” ( actin rich , no MBP ) , “partial” ( partial actin disassembly , onset of MBP expansion ) , and “ring + lamellar” ( full MBP expansion , actin disassembly ) in Fig4 control cultures ( n = 4 ) and Fig4-/- ( n = 4 ) cultures . Results are shown as mean value ± SEM , multiple t-test analysis with Holm-Sidak method . p=0 . 0008 ( “partial” ) , p=0 . 009 ( “ring + lamellar” ) . The effects of Fig4 deletion on OPC proliferation and OL survival are shown in Figure 5—figure supplement 1 . Quantitative Western blot analysis of myelin proteins in primary OL lysates is shown in Figure 5—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01110 . 7554/eLife . 13023 . 012Figure 5—figure supplement 1 . Loss of Fig4-/- in primary OLs does not affect cell proliferation or cell death . ( A-A’ ) Representative images of control ( Fig4+/+ or Fig4+/- ) and Fig4-/- OPCs cultured for 2 days under proliferative conditions , fixed and stained with anti-PDGFRα ( green ) and Ki67 ( red ) . Hoechst 33342 dye was included for nuclear staining of all cells . Scale bar = 200 µm . ( B ) Quantification of PDGFRα and Ki67 double-labeled cells . The number of double stained cells in Fig4 control cultures was set at 1 and is comparable to Fig4-/- cultures ( n = 4 experiments per genotype ) . Results are shown as mean value ± SEM , unpaired Student’s t-test . ( C-C’ ) Representative images of OLs after 4 days in T3 containing differentiation medium . Cultures were fixed and stained with calcein-AM ( green , living cells ) and ethidium homodimer ( red , dead cells ) . Scale bar = 200 µm . Quantification of live cells after 4 days ( D ) and 5 days ( E ) in differentiation medium revealed no differences among the two genotypes . Fig4control cultures ( n = 4 ) and Fig4-/- cultures ( n = 4 ) . Results are shown as mean value ± SEM , unpaired Students t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01210 . 7554/eLife . 13023 . 013Figure 5—figure supplement 2 . Capillary Western analysis of primary OL lysates . ( A ) Representative capillary immunoassay data of Fig4 control and Fig4-/- OPC/OLs are shown in Simple Western lane view . ( B ) Representative chemiluminescence signal intensity graphs and protein molecular weight of individual proteins . Fig4 control and Fig4-/- OPC/OLs lysates are shown as black and pink lines respectively . Specific peaks corresponding to each protein target are marked . ( C ) Quantification of protein of Erk1-normalized peak area of each protein target . Three independent experiments were used for quantification . Results are shown as mean value ± SD . p<0 . 05; *p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 013 Together with the kinase PIKFYVE and the scaffolding protein VAC14 , FIG4 forms a biosynthetic complex necessary for acute interconversion of PI ( 3 ) and PI ( 3 , 5 ) P2 . The complex is located on the cytosolic surface of vesicles trafficking through the LE/Lys compartment ( McCartney et al . , 2014 ) . As an independent test of the effect of perturbation of the FIG4/PIKFYVE/VAC14 enzyme complex on CNS myelination , we generated Pikfyveflox/flox , Olig2cre mice predicted to be more severely deficient in PI ( 3 , 5 ) P2 than the FIG4 and VAC14 mutants . Consistent with this expectation , the phenotype of the Pikfyve mutant mice is much more severe , with a significant tremor ( Videos 3 and 4 ) and death at 2 weeks of age ( n = 16 pups ) . FluoroMyelin Green staining of P13 brain tissue revealed profound hypomyelination of the corpus callosum , internal capsule and cerebellar white matter of Pikfyveflox/flox , Olig2cre pups ( Figure 6A–A’ ) . In situ hybridization of Plp1 revealed a virtual absence of mature OLs in the Pikfyveflox/flox , Olig2cre brain , including structures in the forebrain and cerebellar white matter ( Figure 6B–D’ ) . Toluidine blue staining of P13 optic nerve sections revealed many fibers with clearly visible myelin profiles in Pikfyve positive control mice and a striking absence of myelin profiles in Pikfyveflox/flox;Olig2cre conditional mutants ( Figure 6—Supplement 1B–D ) . Moreover , deficiency of Pikfyve in OLs results in a pronounced accumulation of large perinuclear vesicles in the optic nerve ( Figure 6—figure supplement 1B , D ) . Defects in differentiation of Pikfyve-/- OL cultures are more pronounced than in Fig4-/-OL cultures . Deficiency of Pikfyve reduces OPC proliferation ( Figure 6E–E' and G ) and results in a 95 ± 1% reduction in cells that progress to the MBP+ stage , compared with wildtype cells ( Figure 6F–F’ and H ) . In addition to Fig4 and Pikfyve mutants , we also examined myelinogenesis in the well-characterized recessive Vac14 mouse mutant L156R ( Vac14L156R ) ( Jin et al . , 2008 ) . The L156R missense mutation impairs the interaction of VAC14 with PIKFYVE , but not with FIG4 ( Figure 7A ) . Similar to Fig4-/- mice , Vac14L156R/L156R mice exhibit ~50% reduction in PI ( 3 , 5 ) P2 . Immunoblots of brain membranes prepared from Vac14L156R/L156R mice showed significantly reduced levels of the myelin markers MAG , CNPase , and MBP ( Figure 7B–E ) . The electrical properties of optic nerve from Vac14L156R homozygous mice were also impaired , with a significant increase in the population of slowly conducting fibers ( Figure 7F–H ) . Consistent with this observation , toluidine blue staining of optic nerve sections of adult wild-type mice revealed many myelinated fibers but optic nerves of adult Vac14L156R/L156R mice showed few myelinated fibers ( Figure 7—figure supplement 1A–D ) . Thus , independent genetic disruptions of the FIG4/PIKFYVE/VAC14 enzyme complex all result in severe hypomyelination and a PI ( 3 , 5 ) P2 dosage-dependent decline in CNS white matter development . 10 . 7554/eLife . 13023 . 014Figure 6 . Conditional deletion of Pikfyve in OLs results in profound CNS hypomyelination . ( A-D’ ) Sagittal sections of P13 mouse brains . ( A ) Pikfyve control ( Pikfyveflox/+ or Pikfyveflox/flox; n = 3 ) mice and ( A’ ) Pikfyve conditional null ( Pikfyveflox/flox , Olig2Cre; n = 3 ) mice stained with FluoroMyelin Green . In Pikfyveflox/flox , Olig2Cre , no myelin staining was observed , Scale bar , 1 mm . ( B-D’ ) in situ hybridization for Plp1 shows virtual absence of mature OLs in P13 Pikfyveflox/flox , Olig2Cre brain tissue , including ( B and B’ ) internal capsule and corpus callosum , ( C’ and C’ ) hippocampus and corpus callosum and ( D and D’ ) cerebellar white matter . Scale bar ( B-D’ ) = 500 µm . ( E-H ) Cultures of primary OPCs/OLs isolated from Pikfyve control and Pikfyveflox/flox , Olig2cre mouse pups . ( E , E’ ) At DIV2 , cells were fixed and stained with anti-PDGFRα , anti-Ki67and Hoechst 33342 dye . ( F , F’ ) After 3 days in differentiation medium , supplemented with T3 , cells were fixed and stained with anti-MBP and Hoechst 33342 . ( G ) Quantification of proliferating OPCs revealed a Pikfyve-dependent reduction in Ki67+/PDGFRα+ double-labeled cells ( n = 3 experiments per genotype ) . ( H ) Quantification of MBP+ OLs normalized to Hoechst+ cells shows a highly significant decrease in the number of MBP+ OLs in Pikfyveflox/flox , Olig2cre cultures ( n = 3 experiments per genotype ) . Unpaired Student’s t-test; mean value ± SEM . p=0 . 011 and **p<0 . 0001 . Toluidine blue labeling of epoxy resin embedded optics nerves of Pikfyve control and Pikfyveflox/flox , Olig2cre conditional mutant mice is shown in Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01410 . 7554/eLife . 13023 . 015Figure 6—figure supplement 1 . Optic nerve axons are not myelinated in Pikfyveflox/flox , Olig2Cre mice . Semi-thin sections of P14 optic nerves in ( A , B ) longitudinal and ( C , D ) cross sectional view stained with toluidine blue . In control optic nerve ( Pikfyveflox/+ , Olig2Cre ) many myelinated fibers are observed ( n =3 pups ) . In Pikfyveflox/flox , Olig2Cre conditional mutants ( n = 3 pups ) , there is a striking absence of myelinated axons in the optic nerve . Arrows in photomicrographs B and D point to presumptive oligodendrocytes laden with large vacuoles . Scale bar = 15 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01510 . 7554/eLife . 13023 . 016Figure 7 . Homozygosity for VAC14L156R leads to CNS hypomyelination and impaired conduction of compound action potentials . ( A ) Schematic of the PIKfyve/Vac14/Fig4 enzyme complex and its phosphoinositide products PI ( 3 ) P and PI ( 3 , 5 ) P2 . The red asterisk in VAC14 indicates the L156R point mutation that perturbs the interaction with PIKfyve , but not with Fig4 . ( B ) Western blot analysis of brain membranes prepared from adult ( P90-120 ) WT and VAC14L156R/VAC14L156R littermate mice revealed a reduction in the myelin markers MAG , CNPase , and MBP . Anti-class III β-tubulin ( βIII-Tub ) , a neuronal marker , is shown as a loading control . ( C-E ) Quantification of protein bands detected by Western blotting , shows a significant decrease in MAG , CNPase , and MBP in VAC14 mutant brain tissue ( n = 3 independent blots per genotype ) . Unpaired Student’s t-test; mean value ± SEM . p<0 . 001 , p=0 . 0015 and p=0 . 0238 . ( F and G ) Representative CAP traces recorded from acutely isolated optic nerves of WT and VAC14L156T homozygous mice . ( H ) Quantification of average conduction velocity ( CV ) of largest amplitude peaks identified in F and G . Results are shown as mean value ± SEM , unpaired Student’s t-test , p=0 . 0063 . WT n = 6 nerves , 3 mice and for Vac14L156R mutants n = 6 nerves , 3 mice . Toluidine blue staining of epoxy resin embedded optic nerve sections from VAC14L156R/VAC14L156R mice is shown in Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01610 . 7554/eLife . 13023 . 017Figure 7—figure supplement 1 . Severe optic nerve hypomyelination in VAC14L156R/L156R mice . Semi-thin sections of P21 optic nerves in ( A , B ) longitudinal and ( C , D ) cross sectional view stained with toluidine blue . In wildtype ( WT ) optic nerve sections many myelinated fibers are observed ( n = 3 pups ) . In marked contrast , very few axons are myelinated in the VAC14L156R/L156R optic nerves ( n = 2 ) . Scale bar = 15 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01710 . 7554/eLife . 13023 . 018Video 3 . Normal locomotion of juvenile Pikfyveflox/+ control mice . A representative video of a control Pikfyveflox/+ mouse at P13 . N = 16DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 01810 . 7554/eLife . 13023 . 019Video 4 . Severe tremor in juvenile Pikfyveflox/flox , Olig2Cre mice . A representative video of a Pikfyveflox/flox , Olig2Cre conditional mutant mouse at P13 reveals a severe tremor phenotype . N = 16DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 019 The FIG4/PIKFYVE/VAC14 biosynthetic complex regulates intracellular PI ( 3 , 5 ) P2 and thereby influences membrane trafficking through the endo-lysosomal system . DIV2 primary OPC cultures established from Fig4 control ( Fig4+/+ or Fig4-/+ ) and Fig4-/- mice were fixed and subjected to anti-LAMP1 and anti-PDGFRα double-immunofluorescence labeling . The majority of Fig4-/- OPCs showed normal-sized lysosomes with a diameter of < 1 µm , while a fewcells ( < 20% ) exhibited enlarged LAMP1+ vesicles ( Figure 8—figure supplement 1A–B” ) . Upon OL differentiation , an increase in size and number of perinuclear LAMP1+ vesicles is observed in Fig4-/-cultures . The enlarged perinuclear LAMP1+ structures are prominently labeled with anti-MAG ( Figure 8—figure supplement 1C–D” ) . In a parallel approach , Fig4-/- OLs were transfected with Rab7-YFP , a reporter for LE . Enlarged perinuclear vacuoles in Fig4-/- OLs are positive for Rab7-YFP ( Figure 8—figure supplement 2A–A’ ) . Live imaging of primary OLs revealed that the majority of enlarged perinuclear vacuoles in Fig4-/- OLs are stable for several days . However , vacuole size varies and live imaging revealed that some vacuoles appear and disappear over a period of 12 hr ( Video 5 ) . Collectively , these studies demonstrate that in Fig4-/- OLs , myelin building blocks that are normally trafficked through the LE/Lys are present in abnormal , enlarged vesicles the majority of which is stable for several days . 10 . 7554/eLife . 13023 . 020Video 5 . Vacuoles in Fig4-/- OLs appear and disappear within hours . Time-lapse live cell analysis of Fig4-/- primary OPC/OLs imaged with an IncuCyte ZOOM microscope . Phase contrast images were taken every 2 hr over a time interval of 60 hr . The majority of Fig4-/- cells contain large perinuclear vacuoles . Some of these vacuoles appear and disappear within hours ( n = 3 ) . Scale bar = 60 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 020 In developing OLs , myelin proteins such as MAG and PLP transiently accumulate on the plasma membrane ( PM ) at the cell soma , prior to undergoing endocytosis and LE/Lys dependent transport to the myelin sheet ( Winterstein et al . , 2008 ) . To monitor trafficking of MAG , we used antibody tagging in live OL cultures . In wildtype OLs , anti-MAG-Alexa488 binds to MAG on the PM surface , undergoes endocytosis and is targeted to LAMP1+ vesicles in the LE/Lys compartment ( Figure 8—figure supplement 2B–B” ) . In these wildtype cultures , anti-MAG+ vesicles are small , with a median volume of 0 . 3 ± 0 . 06 µm3 , and partially overlap with LysoTracker+ vesicles ( Figure 8A–A” ) . In contrast , in Fig4-/- OLs , anti-MAG-Alexa488 is endocytosed and accumulates in LAMP1+ perinuclear vacuoles with greatly enlarged size ( ≥5 µm3 , mean volume 94 ± 41 µm3 ) and also in smaller MAG+/LAMP1+ vesicles with a median volume of 0 . 7 ± 0 . 25 µm3 . The average size of all vesicles in Fig4-/- OLs is 1 . 65 ± 0 . 32 µm3 ( Figure 8B–B” and C , Figure 8—figure supplement 2C–C” ) . This suggests that independent of Fig4 genotype , MAG is transported to the PM and is rapidly endocytosed . In Fig4-/- OLs , large MAG+/LAMP1+ vesicles rarely overlap with LysoTracker staining ( Figure 8B-8B” ) , suggesting that large vesicles may exhibit reduced acidification . As an independent approach to assess whether perturbation of PI ( 3 , 5 ) P2 synthesis causes accumulation of MAG in large perinuclear vacuoles , wildtype OL cultures were treated with 1 µM apilimod , a potent inhibitor of PIKfyve ( Cai et al . , 2013 ) . Treatment with apilimod for 90–120 min leads to the formation of large perinuclear vacuoles laden with MAG ( Figure 8D–D” ) , similar to those in Fig4-/- OLs . To evaluate the specificity of the anti-MAG-Alexa488 antibody , experiments were repeated with primary OLs isolated from Mag-/- pups ( Pan et al . , 2005 ) . Bath application of anti-MAG-Alexa488 to Mag-/- OLs treated with vehicle or apilimod did not result in immunostaining , demonstrating that the antibody is specific for MAG ( Figure 8—figure supplement 3A–D” ) . The myelin protein MOG has a different endocytotic fate from MAG , trafficking through recycling endosomes ( RE ) but not the lysosomal compartment ( Winterstein et al . , 2008 ) . Simultaneous antibody labeling of cell surface MAG and MOG in live OLs confirmed distinct endocytotic trafficking routes in both Fig4 control and Fig4-/- cultures . Importantly , in Fig4-/- OLs , MOG was not present in the enlarged vacuoles that are typically laden with MAG ( Figure 8—figure supplement 4A–B” ) . This suggests that the defect in Fig4-/- OLs in trafficking of myelin building blocks from the PM is specific for trafficking through the LE/Lys compartment and does not affect trafficking through the RE . 10 . 7554/eLife . 13023 . 021Figure 8 . In Fig4-/- OLs , MAG accumulates in large perinuclear vacuoles . Confocal images of live OLs acutely labeled with bath applied anti-MAG-Alexa488 ( green ) and LysoTracker Deep Red . ( A-A’’ ) Fig4 control ( Fig4+/+ or Fig4+/- ) OLs incubated with anti-MAG-Alexa488 and LysoTracker , single channel and merged images are shown . ( B-B” ) Fig4-/- OLs incubated with anti-MAG-Alexa488 and Lysotracker shows accumulation of MAG in large perinuclear vacuoles ( arrows ) , single channel and merged images are shown . Of note , large perinucler MAG+ vacuoles do not stain with LysoTracker . ( C ) Scatter plot depicting the volume of anti-MAG-Alexa488+ particles in live Fig4 control and Fig4-/- OLs . Each dot represents an individual vesicle ( n = 4 experiments , 9 cells per genotype ) . Mean volumes ± SEM are shown . ( D-D” ) Wildtype OLs were incubated with anti-MAG-Alexa488 and LysoTracker and then acutely treated with the PIKfyve inhibitor apilimod . MAG accumulates in large perinuclear vacuoles , the majority of which does not stain with LysoTracker ( n = 4 for Fig4 controls and n = 4 for Fig4-/-cultures ) . For apilimod treatment , n = 3 independent cultures . Maximum projection confocal z-stack images are shown , scale bar = 10 μm . Further characterization of enlarged perinuclear vacuoles in Fig4-/- OL cultures , specificity control for the anti-MAG-Alexa488 antibody and distinct trafficking routes of MAG and MOG are shown in Figure 8—figure supplement 1–3 and 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 02110 . 7554/eLife . 13023 . 022Figure 8—figure supplement 1 . Fig4-/- OLs show enlarged perinuclear vacuoles that stain positive for LAMP1 . Confocal images of ( A-A” ) Fig4 control ( Fig4+/+ or Fig4+/- ) and ( B-B” ) Fig4-/-OPCs cultured for two days in the presence of PDGF , fixed and double-stained with anti-LAMP1 and anti-PDGFRα antibodies . TO-PRO-3 dye was added to stain nuclei . Few OPCs ( <20% ) in Fig4-/- cultures showed enlarged LAMP1+ vesicles ( white arrows ) . ( C-D’’ ) Confocal images of ( C-C” ) Fig4 control and ( D-D” ) Fig4-/- OLs after 4 days in T3 containing differentiation medium . Cultures were fixed and double-stained with anti-LAMP1 and anti-MAG antibodies . TO-PRO-3 dye was added to stain nuclei . In Fig4-/- cultures , the majority of OLs ( >65% ) showed multiple large perinuclear vacuoles that were double-positive for LAMP1and MAG ( white arrows ) . Observations were made in 4 independent experiments per culture condition . Scale bar , A-D” = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 02210 . 7554/eLife . 13023 . 023Figure 8—figure supplement 2 . In Fig4-/- OLs , PM derived MAG is transported to enlarged vesicles in the LE/Lys compartment . Representative confocal images of ( A ) Fig4 control OLs and ( A’ ) Fig4-/- OLs transfected with a Rab7-YFP expression construct . Large perinuclear Rab7-YFP+ vesicular structures are observed in Fig4-/- OLs ( arrows ) . Scale bar = 20 μm . Confocal images of ( B-B” ) Fig4 control ( Fig4+/+ or Fig4+/- ) and ( C-C” ) Fig4-/- OL cultures transfected with a LAMP1-mCherry expression construct and incubated in bath applied anti-MAG-Alexa488 antibody . ( B” ) In Fig4 control cultures MAG is localized to LAMP1+ vesicles with a diameter of less than 1 µm . ( C” ) In Fig4-/-cultures , MAG labeling is frequently observed in enlarged perinuclear and LAMP1+ vesicles ( arrows ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 02310 . 7554/eLife . 13023 . 024Figure 8—figure supplement 3 . Specificity control for anti-MAG-Alexa488 antibody . Live-cell imaging of primary OLs prepared from Mag+/+ and Mag-/-pups following bath application of anti-MAG-Alexa488 ( green ) and LysoTracker Deep Red . Representative confocal Z-stack images are shown . ( A-A” ) Anti-MAG-488 labeling of intracellular vesicles is robust in Mag+/+ OLs . ( B-B” ) No signal is detected in parallel processed Mag-/- OLs . Independent of MAG genotype , prominent LysoTracker staining is observed . ( C-D’’ ) To rule out the possibility that large vacuoles are non-specifically labeled by anti-MAG-Alexa488 , Mag+/+ and Mag-/-OL cultures were treated with the PIKfyve kinase inhibitor apilimod . Apilimod leads to accumulation of enlarged perinuclear vacuoles in Mag+/+and Mag-/-cultures . ( C-C’’ ) In Mag+/+ cultures , vacuoles are strongly labeled with anti-MAG-Alexa488 . ( D-D’’ ) In Mag-/- cultures , no labeling with anti-MAG-Alexa488 was observed . Scale bar in A-D” = 7 . 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 02410 . 7554/eLife . 13023 . 025Figure 8—figure supplement 4 . Live imaging of primary OLs reveals distinct trafficking routes for PM-derived MAG and MOG . Confocal images of ( A-A” ) Fig4 control and ( B-B” ) Fig4-/- OLs simultaneously incubated with anti-MAG-Alexa488 and anti-MOG-Alexa555 antibodies . Independent of Fig4 genotype , there is little overlap among MAG+ ( green ) and MOG+ ( red ) structures . In Fig4-/- OLs , enlarged MAG+ vesicular structures ( arrows ) are MOG- . Scale bar A-B’ = 7 . 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 025 The perinuclear location and large size of MAG+ vacuoles suggests that their mobility may be compromised , potentially leading to impaired trafficking of MAG and other myelin building blocks transported via the LE/Lys route . To explore this possibility , we assessed movement of MAG+ vesicles in live OLs using time-lapse imaging ( Figure 9A–B’ ) . Small vesicles labeled with anti-MAG-Alexa488 are observed in Fig4+/+ and Fig4-/- primary OLs , with average volumes of 0 . 3 µm3 and 0 . 7 µm3 , respectively . The average velocity of these 'normal-sized' vesicles is comparable in Fig4+/+ and Fig4-/- cells: 0 . 09 ± 0 . 01 µm/s and 0 . 07 ± 0 . 01 µm/s , respectively ( Figure 9C ) . The large MAG+ vesicles in the Fig4-/- OLs with an average volume of 94 ± 41 µm3 are more stationery , with an average velocity of 0 . 033 ± 0 . 005 µm/s ( Figure 9C ) , and they fail to reach the nascent myelin sheet . These data suggest that trafficking of MAG and other LE/Lys dependent myelin building blocks is impaired in the Fig4-/- OLs . Collectively , these studies indicate that PI ( 3 , 5 ) P2 is critical for myelin protein trafficking through the LE/Lys compartment in developing OLs . 10 . 7554/eLife . 13023 . 026Figure 9 . In Fig4-/- OLs , vesicular trafficking through the LE/Lys compartment is defective . Representative confocal images of live , anti-MAG-Alex488 labeled ( A ) Fig4 control OLs and ( A’ ) Fig4-/- OLs . Time-lapse imaging was used to track movement of MAG+ vesicles . ( B ) Using Imaris software , MAG+ vesicles were labeled with pink spheres and vesicular movement was tracked ( yellow lines ) in Fig4 control cultures . ( B’ ) Imaris software was used to track movement of large vesicles ( white color ) and small vesicles ( purple color ) in Fig4-/- OLs: tracks of individual vesicles are shown . ( C ) Quantification of mean velocity of MAG+ vesicles in Fig4 control OLs and Fig4-/- OLs . In Fig4-/- OLs , movement of small vesicles ( 0 . 7 µm3 ) and large vesicles ( 94 µm3 ) was assessed separately . The velocity is shown as mean value ± SEM . N = 4 independent experiments and a total of 9 cells per genotype were analyzed . One-way ANOVA with Dunnett posthoc , *p= 0 . 001 . ( n . s . = not significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 026 Inter-cellular communication is critical for proper development of the axo-glial unit . To extend the studies of myelin protein trafficking to a system that contains intact axo-glial units , we prepared acute forebrain slices from P10-P14 mice and kept them in oxygenated artificial cerebrospinal fluid . Trafficking of MAG was monitored by bath application of mouse anti-MAG-Alexa555 for 2 hr at 32°C . To distinguish between endocytosed MAG and PM localized MAG , brain slices were fixed and incubated with a secondary anti-mouse-Alexa488 conjugated antibody under non-permeabilizing conditions . Endocytosed MAG containing vesicles were prominently found in OL perinuclear regions and along cellular processes that form the myelin internode ( Figure 10A–A” ) . Only a small fraction of MAG is labeled with both antibodies , and thus localized to the PM on the cell surface ( Figure 10A–A” ) . To visualize cells in the OL lineage , we repeated MAG trafficking studies with brain slices from the ROSA-LacZ/EGFP , Olig2Cre reporter mouse . Vesicular MAG labeling was abundant in EGFP+ cells , indicating that endocytosis of PM localized MAG does occur in cells of the OL-lineage and vesicular labeling is not the result of nonspecific antibody uptake by microglia or other cell types ( Figure 10—figure supplement 1A–C ) . To control for antibody specificity , brain slices from Mag-/- mice were processed in parallel and revealed no significant labeling ( Figure 10—figure supplement 1D–F ) . Thus , acute brain slices provide an opportunity to study myelin protein trafficking in live tissue . To assess whether PI ( 3 , 5 ) P2 is required for endocytosis and trafficking of PM derived MAG in live brain tissue , the experiment was repeated with forebrain slices prepared from Pikfyveflox/flox , Olig2Cre pups . Strikingly , in the absence of PI ( 3 , 5 ) P2 , MAG+ labeling was restricted to abnormal perinuclear accumulations , and trafficking to cell processes was virtually absent ( Figure 10B’–B” ) . The data demonstrate that in brain slices , as well as cultured cells , PI ( 3 , 5 ) P2 is required for proper membrane trafficking from the PM through the LE/Lys compartment . 10 . 7554/eLife . 13023 . 027Figure 10 . Impaired trafficking of MAG in Pikfyveflox/flox , Olig2Cre brain slices . Confocal images of acute brain slices in oxygenated ACSF treated with bath-applied anti-MAG-Alexa555 antibody , fixed and stained with anti-mouse-Alexa488 secondary antibody to distinguish between endocytosed MAG ( red ) and PM localized MAG ( green ) . ( A ) OLs in the striatum of Pikfyve control mice ( P13 ) show punctate MAG labeling in the cell soma ( arrows ) and along processes that form internodes . Only few MAG+ structures are also stained with anti-mouse-Alexa488 , and thus , localized on the PM . ( B-B” ) Limited perinuclear MAG labeling is observed in the Pikfyveflox/flox , Olig2Cre striatum . Many MAG+ structures are labeled red and green , and thus localized to the PM , however intracellular MAG is observed in some cells . Scale bar = 20 µm . Small inset shows a 3D view of the two cells labeled with arrows ( B-B” ) . MAG+ vesicles ( red ) only partially overlap with PM localized MAG ( green ) . Alexa488+ isosurface transparency is adjusted to 50% to demonstrate intracellular Alexa555+ ( red ) and LiveNuc 647+ ( blue ) structures . Scale bar = 10 µm . To directly demonstrate that anti-MAG antibody labeled cells are OLs , parallel experiments were carried out with brain slices of LacZ/EGFP , Olig2Cre reporter mice . Anti-MAG antibody specificity is demonstrated with Mag-/-slices in Figure 10—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 02710 . 7554/eLife . 13023 . 028Figure 10—figure supplement 1 . Anti-MAG labeling of EGFP+ OLs and specificity control for anti-MAG antibody in acute brain slices . ( A-C ) The LacZ/EGFP , Olig2Cre reporter mouse was used for genetic labeling of cells in the OL lineage . In acute brain slices , punctate anti-MAG-Alexa555 labeling is observed in the soma and processes of EGFP+ OLs in the developing neocortex of P14 mice ( arrow ) . For nuclear staining slices were incubated with NucRed 647 . Insert: Isosurface rendering of the MAG+/EGFP+ cell labeled with the arrow . EGFP+ isosurface transparency is increased to 50% to demonstrate intracellular Alexa555+ ( red ) and LiveNuc 647+ ( blue ) structures . Scale bar = 10 µm . ( D-F ) Parallel processed brain slices from Mag-/-pups labeled live with mouse anti-MAG-Alexa555 antibody , fixed and incubated with anti-mouse Alexa488 , show no staining above background . Scale bar = 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13023 . 028 Immunohistological studies of Fig4-/flox , Olig2Cre optic nerves and experiments with Fig4-/- primary OLs did not detect a significant change in OPC density or reduction in viability . OPCs deficient for Fig4 progress and differentiate normally to the stage of newly formed OLs ( NFOs ) , a postmitotic cell type characterized as PDGFRα- , GalC+ , MOG- ( Zhang et al . , 2014 ) . However , differentiation of NFOs into mature OLs is PI ( 3 , 5 ) P2-dosage dependent . The arrest of OL differentiation becomes more severe as PI ( 3 , 5 ) P2 levels are reduced to ~50% of wildtype levels , in Fig4 and VAC14 mice , or completely depleted in Pikfyve mutant mice . OL maturation is highly regulated , and can be attenuated or blocked by perturbation of numerous signaling pathways and transcriptional programs ( Emery et al . , 2009; Bercury and Macklin , 2015; Marinelli et al . , 2016 ) . The fate of immature OLs that fail to progress to the mature stage remains unclear . However , these cells are likely to be short-lived and destined to die . The number of activated caspase-3+ cells in the OL lineage of Fig4-/- mice is not significantly increased ( Winters et al . , 2011 ) , suggesting that immature OLs either do not die in large numbers or die in a caspase-independent manner . Additional studies are needed to determine exactly at which stage of OL lineage progression PI ( 3 , 5 ) P2 deficiency impairs differentiation and how PI ( 3 , 5 ) P2 regulates progression to a mature myelin producing cell . Like epithelial cells , OLs are polarized , with the myelin sheath resembling the apical membrane domain and the membrane near the OL cell body the basolateral membrane domain ( Salzer , 2003; Maier et al . , 2008; Masaki , 2012 ) . Myelin-producing OLs synthesize and transport large quantities of myelin building blocks ( lipids and proteins ) in order to segmentally ensheath multiple axons . Myelinogenesis also requires membrane sorting and trafficking to specific subdomains of the nascent myelin membrane sheath . Indeed , the final destination of myelin proteins may vary between compact myelin ( e . g . PLP ) , peri-axonal loops ( MAG ) or abaxonal loops ( MOG ) of non-compact myelin ( Arroyo and Scherer , 2000 , Salzer , 2003; White and Kramer-Albers , 2014 ) . As in other polarized cells , OL proteins may be targeted through direct transport pathways from the Golgi to their final destination ( Salzer , 2003 ) . Alternative strategies are also employed to target key myelin constituents to their final destination . The mRNA for MBP , encoding a protein important for axon wrapping and myelin compaction , is packaged into RNA-granules and transported to distal sites within OL processes for regulated translation ( Müller et al . , 2013 ) . MAG , PLP , and MOG are synthesized in the endoplasmatic reticulum and transported through the Golgi network to the PM near the OL cell body ( analogous to the basolateral domain ) as an intermediate target . From there MOG is targeted to the recycling endosome ( RE ) while MAG and PLP are targeted to the LE/Lys for delivery to the myelin sheath ( analogous to the apical membrane domain ) ( Simons and Trajkovic , 2006; Maier et al . , 2008; Winterstein et al . , 2008 ) . LAMP1 is a marker for LE/Lys and we show that MAG is targeted to LAMP1+ vesicles in both Fig4+/+ and Fig4-/- OLs . A key feature of the MAG/LAMP1 double-labeled vesicles in Fig4-/- mutant OLs is their greatly enlarged size and perinuclear position . The average velocity of these vesicles is significantly reduced , suggesting impaired membrane trafficking through the LE/Lys compartment . Trafficking defects in Fig4-/- OLs are confined to the LE/Lys compartment as trafficking of MOG through RE occurs apparently normal , independent of Fig4 genotype . The severe CNS hypomyelination phenotype in Fig4-/flox , Olig2Cre mice is likely not the result of impaired MAG trafficking alone , but rather the result of mistrafficking of numerous myelin building blocks normally migrating through the LE/Lys compartment . For example , cholesterol ( in part bound to PLP ) and glycosphingolipids are endocytosed from the PM and stored in LE/Lys vesicles ( Trajkovic et al . , 2006 , Winterstein et al . , 2008 ) . During OL maturation , neuronal signals trigger a profound redistribution of PLP-containing membrane domains; endocytosis is reduced and PLP together with cholesterol and glycosphingolipids is moved from the LE/Lys to the PM ( Trajkovic et al . , 2006 ) . In humans , impaired trafficking of PLP due to mutation or altered dosage of the Plp1 gene , causes Pelizaeus-Merzbacher disease ( PMD ) and Spastic Paraplegia Type 2 ( SPG2 ) , developmental disorders with severe neurological impairment ( Inoue , 2005 ) . Overexpression of PLP in mice leads to accumulation of the protein in autophagic vesicles and LE/Lys , leading to reduction of other myelin proteins such MBP , MAG , and MOG ( Karim et al . , 2007 ) . As in Fig4-/- mice , PMD results in reduced number of OLs and CNS dysmyelination . Failure of lysosomal trafficking or function is thus a common underlying mechanism for a growing number of hereditary disorders that cause CNS dysmyelination , including PMD , Niemann-Pick type C disease , and several lysosomal storage diseases ( Folkerth , 1999; Yaghootfam et al . , 2005; Prolo et al . , 2009; Schweitzer et al . , 2009; Faust et al . , 2010; Grishchuk et al . , 2014 ) . Different phosphoinositides exhibit unique distribution to intracellular membrane compartments and have been implicated as key regulators of membrane sorting and targeted vesicular trafficking ( Mayinger , 2012 ) . PI ( 3 , 5 ) P2 , for example , decorates vesicles in the LE/Lys compartment and serves as a docking site for cytosolic proteins ( Mayinger , 2012 ) . PIP binding proteins frequently interact with small GTPases belonging to the Rab or Arf families , establishing a combinatorial code that defines membrane identity ( Behnia and Munro , 2005; Stenmark , 2009; Jean and Kiger , 2012; Mayinger , 2012; Egami et al . , 2014 ) . The phosphorylation status of PIPs and the activation state of small GTPases can be rapidly modified , providing an identification code that is both unique and dynamic , two prerequisites for targeted membrane transport . In HeLa cells , for example , the lysosomal membrane is characterized by the presence of PI ( 3 , 5 ) P2 and the small GTPases Rab7 and Arf-like ( Bucci et al . , 2000; Hofmann and Munro , 2006 ) . In fibroblasts cultured from Fig4-/- or VAC14L156R/L156R mice , PI ( 3 , 5 ) P2 levels are reduced by ~50% leading to formation of greatly enlarged LAMP1+ vacuoles ( Chow et al . , 2007; Jin et al . , 2008; Zou et al . , 2015 ) . In Fig4-/- OLs , Rab7-YFP localizes to large perinuclear vacuoles ( Figure 8—figure supplement 2 ) . In HeLa cells , overexpression of constitutively active Rab7 leads to formation of large LAMP1+ and LAMP2+ vacuoles ( Bucci et al . , 2000 ) . A direct interaction of VAC14 with the Rab7 GTPase activating protein ( GAP ) TBC1D15 has recently been described in HeLa cells ( Schulze et al . , 2014 ) . This suggests the existence of a large protein complex that controls the interconversion of PI ( 3 ) P and PI ( 3 , 5 ) P2 and the activity of select Rab GTPases , an emerging theme for directed membrane trafficking ( Jean et al . , 2015 ) . Rab GTPases constitute a large protein family whose members are localized to distinct intracellular membrane microdomains to coordinate vesicle trafficking ( Stenmark , 2009; Hutagalung and Novick , 2011 ) . The GTPase Rab3A is expressed in OLs and has been shown to participate in membrane trafficking and myelination ( Schardt et al . , 2009 ) . As discussed above , transport of myelin membrane components , including PLP , cholesterol and MAG , involves membrane sorting and trafficking through the LE/Lys compartment prior to insertion into the nascent myelin sheath ( White and Kramer-Albers , 2014 ) . Thus , interference with PI ( 3 , 5 ) P2 synthesis , turnover , or binding partners that define LE/Lys membrane identity results in impaired cargo delivery of key myelin membrane components required for membrane expansion and sheath formation . The severe hypomyelination phenotype in Fig4-/flox , SynCre mice suggests that Fig4-dependent neuronal signals are necessary for proper CNS myelination . When coupled with our previous finding that transgenic Fig4 directed by the NSE promoter on a Fig4-/-background ( Fig4-/- , NSE-Fig4 ) rescues the myelination defect ( Winters et al . , 2011; Ferguson et al . , 2012 ) , this suggests that normal levels of Fig4 in neurons is necessary for CNS myelination and that neuronal overexpression of recombinant Fig4 on a global Fig4-/- background is sufficient to drive CNS myelination . Multiple lines of evidence have demonstrated that neuron-derived signals regulate OL maturation and axon myelination ( Coman et al . , 2005; Trajkovic et al . , 2006; Ohno et al . , 2009; Winters et al . , 2011; Yu and Lieberman , 2013; Yao et al . , 2014 ) . We speculate that neuronal Fig4 regulates LE/Lys-dependent transport and axonal presentation of a 'pro-myelination' signal ( s ) necessary for OL differentiation and CNS axon myelination and that transgenic overexpression of Fig4 in neurons ( NSE-Fig4 ) leads to an elevated production of 'pro-myelination' signals ( s ) sufficient to rescue the deficiency of Fig4 in the OL lineage of the Fig4-/- , NSE-Fig4 transgenic mice . Alternatively , neuronal Fig4 may accelerate the loss of 'anti-myelination' signal ( s ) on the axonal surface , e . g . , through endocytosis . Inter-cellular communication may occur through paracrine action of secreted molecules or shedding vesicles . Exosomes are extracellular vesicles produced by many cells that facilitate transport and exchange of proteins , mRNAs and regulatory RNAs with important functions in cellular processes including myelination ( Frühbeis et al . , 2012; Pusic and Kraig , 2014 ) . Because Fig4 plays an important role in membrane trafficking through the LE/Lys system , it is possible that protein secretion or the content and abundance of exosomes may be altered in the mutant mice . Two independent approaches to delete Fig4 in the OL lineage ( Olig2Cre and PdgfrαCreER ) revealed that Fig4 is required in the OL lineage for proper CNS myelination . These data were corroborated by in vitro studies with primary OLs . Taken together , our observations suggest that endogenous levels of Fig4 gene expression in both neurons and OLs are necessary for normal CNS myelination . Technical limitations in the specificity of transgene promoters may affect the interpretation of these experiments . For neuron-specific loss-of-function we employed female SynapsinCre/+ mice driven by a synapsin-1 gene ( SYN1 ) promoter fragment ( Rempe et al . , 2006 ) , and for neuron-specific gain-of-function studies we used a 4 . 6 kb NSE promoter fragment ( Winters et al . , 2011; Ferguson et al . , 2012 ) . While these are commonly used strategies , it is recognized that in the developing mouse the NSE ( ENO2 ) and SYN1 promoters may have some leakiness that results in transient expression in non-neuronal cells including glia . A low level of expression of the endogenous SYN1 and ENO2 genes in OPCs/OLs has been reported ( Zhang et al . , 2014 ) , but it is not clear whether this expression is retained by the promoter fragments that were used to drive transgene expression . Independent of these technical limitations , we provide multiple lines of evidence that genetic manipulations that compromise PI ( 3 , 5 ) P2 synthesis profoundly impact OL differentiation and CNS myelination . Acutely prepared brain slices are viable for several hours when maintained in oxygenated ACSF , a method commonly used for electrophysiological recordings ( Lee et al . , 2008 ) . Studies with primary OLs suggest that newly synthesized myelin proteins are initially transported to the PM near the cell soma where they interact with lipids and other myelin proteins ( Winterstein et al . , 2008 ) . These myelin-like structures are then thought to be endocytosed and trafficked to specific subdomains of the nascent myelin membrane sheath . Using acute brain slices combined with genetic labeling of cell in the OL lineage and confocal microscopy , we show that antibody-labeled MAG on the PM becomes rapidly endocytosed and is found in small vesicles in the OL cell soma and long processes that form internodes . Since sorting and trafficking of myelin building blocks are key components of myelinogenesis , future studies using acute brain slices may be productively combined with pharmacological and genetic manipulations to obtain detailed understanding of membrane trafficking in developing OLs . All mice were housed and cared for in accordance with NIH guidelines , and all research conducted was done with the approval of the University of Michigan Committee on Use and Care of Animals . The spontaneous Fig4-/- null mutation plt ( Chow et al . , 2007 ) is maintained as two congenic lines , C57BL/6J . plt/+ and C3HeB/FeJ . plt/+ . F1 plt/plt homozygotes obtained from crosses between these lines survive to 30–45 days , permitting analysis of myelination , and these were used for most experiments . A subset of in vitro experiments was carried out on cells from the C3HeB/FeJ . plt congenic mice . The conditional Fig4flox allele was described elsewhere ( Ferguson et al . , 2012 ) and is maintained on strain C3HeB/FeJ from which the retinal degeneration locus rd was removed by repeated backcrossing and selection . Neuron-specific conditional knockout mice ( Fig4-/flox , SynCre ) were generated and maintained as previously described ( Ferguson et al . , 2012 ) . The Olig2Cre/+ line ( Schüller et al . , 2008 ) and the PdgfraCre-ER/+ ( Kang et al . , 2010 ) ( Jackson Laboratory stock # 018280 ) were used to delete Fig4 in the OL lineage . For inducible gene ablation in Fig4-/flox , PdfrαCreER mice , 4-hydroxytamoxifen ( 4OH-tamoxifen ) ( Sigma-Aldrich , MO ) was injected directly into the stomach of P5 pups , which is easily identified by its milky-white color . 4OH-tamoxifen was dissolved in 100% ethanol at 10 mg/ml and 5 μl/day were administered for 2 days . Fig4-/flox , Hb9Cre ( Fig4-/flox , Mnx1Cre ) mice have been described previously ( Vaccari et al . , 2015 ) . The spontaneous point mutant VAC14L156R is deficient in PIKfyve binding ( Jin et al . , 2008 ) and was maintained on a C3HeB/FeJ strain background from which the retinal degeneration locus rd was removed by repeated backcrossing . Pikfyveflox/flox mice were generated on the C57BL/6J strain background ( Min et al . , 2014 ) and were crossed with Olig2Cre/+ mice . Mag-/- mice on a C57BL/6J background have been described elsewhere ( Pan et al . , 2005 ) . LacZ/ EGFP reporter mice ( Jackson laboratory stock #003920 ) were crossed with Olig2Cre/+ mice . Postnatal day ( P ) 21 and P60-P75 mice were deeply anesthetized with ketamine ( 200 mg/kg ) /xylazine ( 20 mg/kg body weight ) and perfused transcardially with ice-cold phosphate buffer saline ( PBS ) for 2 min , followed by 4% paraformaldehyde ( PFA ) and 2 . 5% glutaraldehyde in Sorensen’s buffer and embedded in epoxy resin as described ( Winters et al . , 2011 ) . Semi-thin sections were stained with toluidine blue for light microscopy . TEM micrographs were taken at 10 , 500–13 , 500x magnification with a Philips CM-100 or a JEOL 100CX microscope and analyzed using FIJI software . Fig4-/flox , Hb9Cre ( Vaccari et al . , 2015 ) , Fig4-/flox , Olig2Cre and Fig4-/flox , SynCre conditional mutants were analyzed and compared to littermate controls . Throughout the study , control mice are defined as mice that have at least one intact copy of the Fig4 allele and include the following genotypes ( i ) Fig4+/- , ( ii ) Fig4-/flox , ( iii ) Fig4+/flox , Olig2Cre and ( iv ) Fig4+/flox , SynCre . Mice between P10 and adulthood were perfused transcardially with ice-cold 4% PFA in PBS . Brains were post-fixed in perfusion solution for 2 hr at 4°C for in situ hybridization . For immunofluorescence labeling , brains were postfixed overnight and cryoprotected in 30% sucrose in PBS . For FluoroMyelin staining , brains were cryosectioned at 25–40 μm . Free-floating sections were rinsed 3x 5 min in PBS and then stained with FluoroMyelin Green ( Millipore , MA , 1:200 ) in PBS for 20 min . Sections were washed with PBS , mounted onto microscope slides , coverslipped with Prolong Gold antifade supplemented with DAPI ( Life Technologies , CA ) and imaged with an Olympus IX71 microscope attached to a DP72 camera . For immunofluorescence labeling of optic nerves , nerves were rapidly dissected , kept in perfusion solution for 30 min and cryoprotected in 30% sucrose in PBS . Cross sections ( 12–20 µm ) were mounted onto microscope slides , rinsed 3x for 5 min in PBS and incubated for 1 hr in blocking solution: 1% horse serum and 0 . 1% Triton-X100 in PBS ( anti-Olig2 ) or 4% normal goat serum and 0 . 3% Triton-X100 in PBS ( anti-NG2 ) . Primary antibody incubation was done overnight at 4°C in blocking solution with rabbit anti-Olig2 ( 1:1000 Millipore ) or rabbit anti-NG2 ( 1:800 , Abcam , UK ) . The next day , sections were rinsed 3x 5 min with PBS , incubated with appropriate secondary antibodies for 1 hr at room temperature ( 1:1000 , Alexa-conjugated , Life technologies ) , rinsed in PBS and mounted in Prolong Gold supplemented with DAPI . cDNA fragments of Mbp and Plp1 ( Ye et al . , 2009 ) were used to produce digoxigenin-labeled cRNA probe by run-off in vitro transcription . Brains were cryosectioned at 25 μm and mounted directly onto Superfrost+ microscope slides ( Fisher Scientific , MA ) . Optic nerve sections were prepared as described above and postfixed in 4% PFA/PBS overnight at 4°C . The following day , sections were rinsed with 1x PBS and dehydrated with series of ethanol dilutions ( 50% , 70% , 95% , and 100% ) . Sections were then treated with 50µg/ml proteinase K in PBS/5mM EDTA for 15 min ( optic nerves ) and 30 min for brain sections . All subsequent steps were performed as described previously ( Winters et al . , 2011 ) . P21 mouse brains were homogenized in a Wheaton Dounce tissue homogenizer cooled on ice . Brain membranes were isolated by centrifugation in a discontinuous sucrose gradient as described previously ( Winters et al . , 2011 ) . P21 control littermate and Fig4-/flox , PdfrαCreER brains were extracted and rapidly dissected on ice . Tissue was separated into two groups: 1 ) cerebellum + brainstem and 2 ) neocortex + hippocampus + thalamus ( 'forebrain' ) . Tissue was lysed in a radio-immunoprecipitation assay buffer ( RIPA ) using a tissue homogenizer and triturated with a 16G needle . Lysates were spun at 14 , 000 rpm for 15 min at 4°C and supernatants were analyzed by Western blotting as described below . Equal amounts of protein ( 7 . 5–15 µg ) from brain membranes were separated by SDS-PAGE and transferred onto PVDF membranes ( Millipore ) . Membranes were blocked in 3% dry milk powder dissolved in Tris-HCl pH 7 . 4 buffered saline containing 0 . 3% Triton X-100 for at least 1 hr and incubated with primary antibody overnight at 4°C . Primary antibodies included mouse anti-βIII tubulin ( 1:20 , 000; Promega , WI ) , rabbit anti-MAG ( 1:1000; Winters et al . , 2011 ) , rat anti-MBP ( 1:1000; Millipore ) , mouse anti-CNPase ( 1:1000 , Abcam ) , anti-PLP ( 1:1000 , Abcam ) , and mouse anti-Fig4 ( 1:200 , NeuroMab , CA ) . Primary antibodies were detected using either horseradish peroxidase ( HRP ) -conjugated secondary antibodies ( 1:2000–15000; Millipore Bioscience Research Reagents ) or Alexa-conjugated secondary antibodies ( 1:20 , 000 , Molecular Probes ) . The Licor C-DiGit and Odyssey imaging systems and software were used for visualization and quantification of protein bands ( Licor , NE ) . Recordings were carried out as described elsewhere ( Carbajal et al . , 2015 ) . Briefly , juvenile ( P21-P23 ) and adult ( 3–4 months ) mice were sacrificed by CO2 inhalation . Optic nerves were rapidly dissected , incubated at room temperature in oxygenated artificial cerebrospinal fluid ( ACSF ) for 45 min and then transferred to a temperature-controlled recording chamber ( held at 37 ± 0 . 4°C ) with oxygenated ACSF . Each end of the nerve was drawn into the tip of a suction pipette electrode . The stimulating electrode was connected to a constant-current stimulus isolation unit ( WPI , FL ) driven by Axon pClamp 10 . 3 software and a 50 μs pulse was applied to the retinal end of the nerve . The recording electrode was applied to the chiasmatic end of the nerve and connected to the input of a differential AC amplifier ( custom-made ) . A second pipette , placed near the recording pipette but not in contact with the nerve , served to subtract most of the stimulus artifact from the recordings . Signals were digitized at 100 kHz through a data acquisition system ( Axon Digidata 1440A , Axon pClamp 10 . 3 , Molecular Devices , CA ) . OPCs were isolated from P6-14 mouse pups with the following genotypes ( i ) Fig4+/+ , ( ii ) Fig4+/- , ( iii ) Fig4-/- , ( iv ) Fig4-/flox , Olig2Cre or ( v ) Pikfyveflox/flox , Olig2Cre . For immunopanning , anti-PDGFRα ( BD Biosciences , CA ) or O4 antibody ( hybridoma cells kindly provided by Dr Jonah Chan ) coated plates were used , as described ( Emery and Dugas , 2013 ) . For the first two days in vitro , OPCs were cultured on poly-D-lysine ( Sigma-Aldrich ) coated glass coverslips in DMEM-SATO medium supplemented with forskolin ( Sigma , 10 ng/ml ) , PDGF ( 20 ng/ml , Peprotech , NJ ) , CNTF ( 10 ng/ml , Peprotech ) , and NT3 ( 1 ng/ml , Peprotech ) . For differentiation studies , OPCs were switched to medium supplemented with T3 ( 40 ng/ml , Sigma-Aldrich ) without growth factors . Cells were allowed to differentiate for 4–6 days prior to fixation in 4% PFA/PBS at RT for 15 min . For immunofluorescence labeling , cells were rinsed 3x 5 min each in PBS , permeabilized with 0 . 1% Triton-X100 in PBS for 30 min and blocked for 60 min in 3% BSA in PBS . The following primary antibodies were used: rabbit anti-NG2 ( 1:500 , Millipore ) , rat anti-PDGFRα ( 1:1000 , BD Biosciences , CA ) , rabbit anti-PDGFRα ( 1:500 , Cell Signaling , MA ) , rat anti-MBP ( 1:300 , Millipore ) , rabbit anti-CNPase ( 1:1000 , Assay Biotech , CA ) , rabbit anti-Ki67 ( 1:1000 , Abcam ) , mouse anti-MAG ( 1:300 , Millipore ) , rat anti-Lamp1 ( 1:1000 , Abcam ) , mouse anti-GFAP ( 1:2000 , Sigma-Aldrich ) . Cells were incubated with primary antibodies overnight at 4°C . The following day , cells were rinsed 3x 5 min each with PBS , and incubated with secondary antibodies for 1 hr in blocking solution . Following several rinses in PBS , cells were incubated with the nuclear markers Hoechst 33 , 342 or ToPro3 dye ( Life Technologies ) and imaged with an Olympus IX71 inverted microscope ( Olympus , JP ) with a DP72 camera or a Leica SP5 confocal microscope ( Leica , DE ) . Representative confocal images were taken at 63x magnification as z-stacks with 1 µm intervals . Maximum intensity z projections were generated using Fiji . For cell viability experiments , the Live/Dead kit ( Life Technologies ) was used following the manufacturer’s instructions . For actin staining , Actin Red 555 ( Life Technologies ) was used following the manufacturer’s instructions . For live cell imaging , OPCs were switched to T3 supplemented differentiation medium and kept at 37°C in a 5% CO2 incubator equipped with an IncuCyte Zoom imaging system ( Essen Bioscience , MI ) . Images were taken with a 20x objective every 2 hr for 3 days . Data were analyzed using the IncuCyte Zoom software and Fiji . O4+ primary OLs were isolated by immunopanning as described above and cultured in 35 mm glass bottom dishes ( Mattek , MA ) . After 2–3 days under differentiation conditions , anti-MAG-Alexa488 conjugated antibody ( 1:500 , Millipore , MAB1567A4 ) was added to the culture medium for 12–14 hr . The following day , LysoTracker Deep Red ( 1:2000 , Life Technologies ) was added to the culture medium for 30–45 min . Fifteen minutes before imaging , the culture medium was replaced by 1x HBSS ( Life Technologies ) containing Prolong Live Antifade reagent ( Life Technologies , 1:100 ) and Hoechst dye 33 , 342 ( 1:50 , 000 ) or NucRed Live 647 ( Life Technologies ) . Cells were imaged at 37°C and ambient CO2 for 15–20 min/dish using a Leica SP5 confocal microscope . Confocal Z-stacks , xyt , and xyzt videos were acquired . As a specificity control for the anti-MAG-Alexa488 antibody , OLs were prepared from Mag-/- and age-matched Mag+/+ pups and imaged under identical conditions . Mouse monoclonal anti-MOG antibody ( Millipore ) was conjugated with Alex555 using the Antibody Labeling Kit ( Life Technologies ) . Some OL cultures were incubated with anti-MOG-Alexa555 ( 1:250 ) and anti-MAG-Alexa488 as described above . To some cultures 1 μM apilimod ( Axon 1369; Axon Medchem BV ) in DMSO was added 90–120 min prior to imaging . Images and videos were processed using Leica AS LF and Fiji . Tracking and movement analysis of anti-MAG-Alexa488+ particles in live cells was performed using Imaris ( Bitplane , UK ) . To monitor MAG trafficking in acute brain tissue , sagittal slices were prepared from P13-P14 pups with the following genotypes , ( i ) Pikfyve control mice , ( ii ) littermates Pikfyve flox/flox , Olig2Cre mice , ( iii ) Mag-/- , mice and ( iv ) P18 LacZ/EGFP , Olig2Cre reporter mice ( Toth et al . , 2013 ) . Briefly , mice were decapitated , brains rapidly dissected and submerged in ice-cold ACSF ( Toth et al . , 2013 ) . From forebrain tissue , hippocampi were removed and discarded . Cortex and striatum were sectioned at 300 µm using a tissue slicer ( WPI , FL ) . Brain slices were kept in oxygenated ( 95% O2 , 5% CO2 ) ACSF at RT for 40–60 min prior to incubation with anti-MAG-Alexa-555 ( 1:500 ) in oxygenated ACSF at 32°C for 2 hr . Brain slices were then fixed in 4% PFA for 25 min , rinsed 3 times for 10 min each in PBS and incubated overnight with a goat anti-mouse Alexa-488 secondary antibody ( 1:1000 ) in 3% BSA at 4°C . The following day , slices were rinsed 3 times for 10 min each in PBS , incubated with LiveRed 647 for 25 min at RT , rinsed 3 times for 10 min each in PBS , and mounted in Prolong antifade with DAPI . Individual MAG+ cells in deep cortical layers and striatum were imaged using a Leica SP5 confocal microscope . For transfection of primary OPC/OLs , Lipofectamine2000 ( Life Technologies ) was used , following a protocol previously established for transfection of primary neurons ( Duan et al . , 2014 ) . Briefly , 250 ng of LAMP1-mCherry or Rab7-YFP plasmid DNA were combined with 1 µl of Lipofectamine2000 ( Invitrogen , CA ) in optiMEM and mixed thoroughly . Transfection solution was added to OL culture medium and cells were incubated for 2 . 5 hr . Afterwards , the medium was completely replaced with fresh T3 supplemented medium . To visualize MAG trafficking , anti-MAG-Alexa-488 antibody was added to the culture medium as described above . The following day , live imaging of LAMP1-mCherry+/anti-MAG-Alexa-488+ OLs was carried out as described above . OPCs were allowed to expand in PDGF supplemented culture medium for 7–8 days , passaged and plated in 6-well culture dishes at a density of 200 , 000–300 , 000 cells/well and kept for 3 days in T3 supplemented medium . Cells were then processed for Western blotting as previously described ( Raiker et al . , 2010 ) . Capillary immunoassays were performed using the automated Wes system ( ProteinSimple , San Jose CA ) . All procedures were performed according to manufacturer’s protocol . In brief , 0 . 8 µg of lysate ( 4 µl ) were mixed with 2 µl of 5x fluorescent master mix and boiled for 5 min . These samples were dispensed into microplates along with blocking solution , primary and secondary antibodies and chemiluminescent substrate . After centrifugation , microplate was loaded into the Wes instrument for subsequent protein separation on capillaries and immuno-detection using the standard electrophores , immunolabeling , detection scheme of Wes . Data were analyzed by using Compass software ( ProteinSimple ) and peak areas were used for quantification . Erk1 peak area was used for normalization between samples . Three independent preparations were processed . To assess myelination in the optic nerve , ten non-overlapping TEM images were randomly selected and the fraction of myelinated axons quantified as described ( Winters et al . , 2011 ) . At least 600 axons were quantified per nerve . G-ratio analysis was performed as described previously ( Winters et al . , 2011 ) . At least 100 axons per optic nerve were analyzed . For Western blot analysis , Western band intensity was measured using LI-COR Studio Image Software . All band intensities were normalized either to βIII-tubulin ( brain lysates and membranes ) or actin ( OPC cultures ) . Normalized Western blot band intensity for control samples was set as 1 for each experiment . For optic nerve electrophysiology , data analysis was performed offline using Clampfit software . In order to analyze individual peaks , each trace was fitted as a sum of three or four Gaussians using Origin Pro software ( Chen et al . , 2004 ) . A peak with the largest amplitude in each trace was used for conduction velocity analysis . For quantification of Plp1 , Olig2 , and NG2 labeled cells , the number of respective positive cells was quantified per optic nerve cross section and normalized to the section area ( arbitrary units in FIJI ) . At least four sections per nerve were analyzed . For quantification of OL markers in vitro , ten non-overlapping images were taken at random positions for each coverslip/well and cells positive for a marker of interest counted and normalized to the number of Hoechst 33 , 342 dye positive cells in the same image . A minimum of 900 cells was quantified for each individual experiment with Fig4 cultures and a minimum of 120 cells was quantified for each individual experiment with Pikfyve cultures . GFAP+ astrocytes were excluded from quantification . The analysis of actin/MBP postmitotic OL morphology was performed as characterized previously ( Zuchero et al . , 2015 ) . For cell viability experiments , the Live/Dead kit was used the number of live ( green ) and dead ( red ) cells was quantified and the live/total cell ratio was calculated . For all experiments , Hoechst 33 , 342 normalized cell density in control groups was set as 1 . At least three independent experiments with duplicate coverslips were used for the analysis . For live imaging of MAG+ vesicles in primary OLs , Imaris software ( Bitplane ) was used to calculate individual particle speed and size . Four independent experiments were analyzed for Fig4+/+and Fig4-/- cultures . MAG+ particles of at least 0 . 01 µm3 in volume were included in data analysis . One-way ANOVA followed by Tukey posthoc was used for TEM optic nerve analysis . One-way ANOVA followed by Dunnett’s posthoc was used for Western blot analysis and electrophysiology with more than two groups . The unpaired Student t-test was used for analysis in all experiments with two groups .	Neurons communicate with each other through long cable-like extensions called axons . An insulating sheath called myelin ( or white matter ) surrounds each axon , and allows electrical impulses to travel more quickly . Cells in the brain called oligodendrocytes produce myelin . If the myelin sheath is not properly formed during development , or is damaged by injury or disease , the consequences can include paralysis , impaired thought , and loss of vision . Oligodendrocytes have complex shapes , and each can generate myelin for as many as 50 axons . Oligodendrocytes produce the building blocks of myelin inside their cell bodies , by following instructions encoded by genes within the nucleus . However , the signals that regulate the trafficking of these components to the myelin sheath are poorly understood . Mironova et al . set out to determine whether signaling molecules called phosphoinositides help oligodendrocytes to mature and move myelin building blocks from the cell bodies to remote contact points with axons . Genetic techniques were used to manipulate an enzyme complex in mice that controls the production and turnover of a phosphoinositide called PI ( 3 , 5 ) P2 . Mironova et al . found that reducing the levels of PI ( 3 , 5 ) P2 in oligodendrocytes caused the trafficking of certain myelin building blocks to stall . Key myelin components instead accumulated inside bubble-like structures near the oligodendrocyte’s cell body . This showed that PI ( 3 , 5 ) P2 in oligodendrocytes is essential for generating myelin . Further experiments then revealed that reducing PI ( 3 , 5 ) P2 in the neurons themselves indirectly prevented the oligodendrocytes from maturing . This suggests that PI ( 3 , 5 ) P2 also takes part in communication between axons and oligodendrocytes during development of the myelin sheath . A key next step will be to identify the regulatory mechanisms that control the production of PI ( 3 , 5 ) P2 in oligodendrocytes and neurons . Future studies could also explore what PI ( 3 , 5 ) P2 acts upon inside the axons , and which signaling molecules support the maturation of oligodendrocytes . Finally , it remains unclear whether PI ( 3 , 5 ) P2signaling is also required for stabilizing mature myelin , and for repairing myelin after injury in the adult brain . Further work could therefore address these questions as well .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	PI(3,5)P2 biosynthesis regulates oligodendrocyte differentiation by intrinsic and extrinsic mechanisms
Constraints on phenotypic variation limit the capacity of organisms to adapt to the multiple selection pressures encountered in natural environments . To better understand evolutionary dynamics in this context , we select Escherichia coli for faster migration through a porous environment , a process which depends on both motility and growth . We find that a trade-off between swimming speed and growth rate constrains the evolution of faster migration . Evolving faster migration in rich medium results in slow growth and fast swimming , while evolution in minimal medium results in fast growth and slow swimming . In each condition parallel genomic evolution drives adaptation through different mutations . We show that the trade-off is mediated by antagonistic pleiotropy through mutations that affect negative regulation . A model of the evolutionary process shows that the genetic capacity of an organism to vary traits can qualitatively depend on its environment , which in turn alters its evolutionary trajectory . In nature organisms adapt to complex environments where many biotic and abiotic factors affect survival . For microbes these factors include demands on metabolism ( Savageau , 1983 ) , motility ( Celani and Vergassola , 2010 ) and antibiotic resistance ( Vetsigian et al . , 2011 ) . In this context , evolution involves the simultaneous adaptation of many phenotypic traits . Organisms under complex selection pressures often cannot vary traits independently and instead exhibit trade-offs ( Shoval et al . , 2012 ) . Trade-offs constrain adaptive responses to selection . For example , phage exhibit a trade-off between fecundity and virulence which depends on the relative duration of periods of horizontal and vertical transmission ( Messenger et al . , 1999 ) . Bacterial populations selected for efficient conversion of nutrients to biomass exhibit a trade-off between yield and growth rate ( Bachmann et al . , 2013 ) . Predicting evolution in complex environments requires quantifying both trade-offs and selection pressures ( Lande , 1979 ) . In wild populations of birds ( Grant and Grant , 1995 ) and fish ( Ghalambor et al . , 2003 ) , phenotypic constraints and selection pressures have been inferred from measurements of phenotypic variation . However , in wild populations of higher organisms it is challenging to observe evolution , determine selection pressures and elucidate mechanisms constraining phenotypes . To better understand the interplay between trade-offs , selection and evolution , it is necessary to study genetically tractable , rapidly evolving microbial populations in the laboratory . However , laboratory-based experimental evolution of microbes typically selects for a single phenotype such as growth rate ( Lang et al . , 2013 ) . There is evidence that metabolic trade-offs arise in these experiments from the decay of traits that are not subject to selection ( Cooper and Lenski , 2000 ) rather than a compromise between multiple selection pressures . Other experiments explore how phenotypes restricted by trade-offs evolve under alternating selection for individual traits ( Yi and Dean , 2016; Messenger et al . , 1999 ) . Less is known about evolutionary dynamics in the naturally relevant regime where selection pressures are multifaceted . To address this , we selected Escherichia coli for faster migration through a porous environment . We showed that the evolution of faster migration is constrained by a trade-off between swimming speed and growth rate . Evolution of faster migration in rich medium is driven by faster swimming despite slower growth , while faster migration in minimal medium is achieved through faster growth despite slower swimming . Sequencing and genetic engineering showed that this trade-off is due to antagonistic pleiotropy through mutations that affect negative regulation . Finally , a model of multi-trait selection supports the hypothesis that the direction of evolution when phenotypes are constrained by a trade-off is determined by the genetic variance of each trait . Our results show that when selection acts simultaneously on two traits governed by a trade-off , the environment determines the evolutionary trajectory . E . coli inoculated at the center of a low viscosity agar plate consume nutrients locally , creating a spatial nutrient gradient which drives chemotaxis through the porous agar matrix ( Righetti et al . , 1981; Maaloum et al . , 1998 ) and subsequent nutrient consumption ( Adler , 1966; Wolfe and Berg , 1989; Croze et al . , 2011 ) . As a result , the outermost edge of the expanding colony is driven by both growth and motility ( Koster et al . , 2012 ) . The result is a three-dimensional bacterial colony that expands radially across the plate as individuals swim and divide in the porous environment . We refer to the outermost edge of an expanding colony as the migrating front . We tracked these migrating fronts using webcams and light-emitting diode ( LED ) illumination ( Materials and methods ) . The front migrates at a constant speed s after an initial growth phase ( Adler , 1966; Wolfe and Berg , 1989 ) . We performed experimental evolution by repeating rounds of allowing a colony to expand for a fixed time interval , selecting a small population of cells from the migrating front and using them to inoculate a fresh low viscosity agar plate ( Figure 1a ) . By isolating cells from the migrating front , our procedure selects both for motility and growth rate . We performed selection experiments in this way for two distinct nutrient conditions . First , we used rich medium ( lysogeny broth ( LB ) , 0 . 3 % w/v agar , 30°C ) where all amino acids are available . In this medium the population forms concentric rings ( Figure 1b ) that consume amino acids sequentially . The outermost ring consumes L-serine and most of the oxygen ( Adler , 1966 ) . Second , we used minimal medium ( M63 , 0 . 18 mM galactose , 0 . 3 % w/v agar , 30°C ) where populations migrate towards and metabolize galactose with a single migrating front . 10 . 7554/eLife . 24669 . 003Figure 1 . E . coli evolves faster migration through a porous environment in rich and minimal media . ( a ) A schematic of the selection procedure . E . coli are inoculated into the center of a low viscosity ( 0 . 3 % w/v ) agar plate where they form an expanding colony driven by metabolism and motility . After a fixed period of incubation , samples are taken from eight locations around the outer edge of the expanded colony , mixed , and used to inoculate a fresh plate . ( b ) Shows expanded colonies in rich medium ( LB ) plates after 12 hr of incubation over five successive rounds of selection . The color bar to the right applies to all panels in ( b ) , with darker gray indicating higher cell density . Image intensity is assumed to be monotonic but not linear with cell density in the plate . Scale bar in the left panel is 1 cm and applies to all panels in ( b ) . ( c ) Shows the rate of migration as a function of round of selection over 15 rounds for five replicate selection experiments in rich medium . No rate is reported for replicate 1 round 8 due to failure of the imaging device . Errors in measured rates of migration are smaller than the size of the markers . ( d ) Shows colonies ( gray regions ) in minimal medium ( M63 , 0 . 18 mM galactose ) after 48 hr of incubation . The color bar to the right applies to all panels in ( d ) . The scale bar in the left panel is 1 cm . ( e ) Shows the rate of migration as a function of round of selection over 10 rounds for five replicate selection experiments in minimal medium . Errors in migration rates were smaller than the size of markers . See Materials and methods for details of image processing in both experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 00310 . 7554/eLife . 24669 . 004Figure 1—figure supplement 1 . Selection with non-chemotactic ( ΔcheA-Z ) mutant . Front migration rates of non-chemotactic mutants in 0 . 3 % w/v agar at 30°C with LB ( left panel ) and M63 0 . 18 mM galactose ( right panel ) . Errors are smaller than the size of the markers , except for the red replicate in rich medium at round 2 . Red and black correspond to two independent selection experiments . Note the vertical scales . In minimal medium , zero migration rate denotes plates where density increased in the vicinity of the site of inoculation but no migration was observed . In these cases no measurable migration rate was obtained . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 00410 . 7554/eLife . 24669 . 005Figure 1—figure supplement 2 . Change in migration rate during long-term liquid culture . ( left ) The founder strain ( Figure 1c , main text , s= 0 . 3 ± 0 . 01 cm h−1 ) was inoculated into a turbidostat and continuously cultured in LB at 30°C for approximately 200 generations . Samples were periodically drawn from the turbidostat and used to inoculate 0 . 3 % w/v agar LB plates in duplicate . Migration was recorded via webcam as described in the main text . Error bars are standard errors from regression of radius with time . Note the scale on the y-axis . ( right ) Identical experiment in minimal medium conditions . Founding strain was grown in a single chemostat ( doubling time 5 . 7 hr ) in minimal medium for 120 generations . Plates were inoculated from samples drawn from the chemostat , two plates at each time point for the first four time points and then one plate at each time point . The last four time points ( where the rate appears to saturate ) exhibit a slower migration rate than the round 10 migration rates in Figure 1e ( p=0 . 02 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 00510 . 7554/eLife . 24669 . 006Figure 1—figure supplement 3 . Adaptation in rich medium depends on sampling location . Migration rate as a function of the round of selection . Colored traces are reproduced from Figure 1 in the main text . Black circles and squares are two replicate selection experiments where populations are sampled halfway between the center of the colony and the outer edge after each round of selection . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 00610 . 7554/eLife . 24669 . 007Figure 1—figure supplement 4 . Comparison of founding and evolved strains to RP437 . Single-cell swimming in rich medium: ( left ) Run duration distributions identical to those shown in Figure 3a–b of the main text . 77 RP437 individuals were tracked from a culture at the same optical density as founder and round 15 ( replicate 1 ) . A total of 9218 run events were recorded . The average ± standard deviation in run duration for RP437 is 0 . 76 ± 0 . 82 s . ( right ) Comparison of run speeds for the same three strains . RP437 has an average ± standard deviation in run speed of 18 . 6 ± 6 . 4 μm s−1 . The average run duration for RP437 exceeds that of round 15 ( p<10−4 ) , and the average run speed is smaller than that of round 15 ( p<10−4 ) . For the RP437 strain in rich medium , we measure a migration rate of 0 . 15 ± 0 . 01 cm h−1 and a liquid culture growth rate of 1 . 1 ± 0 . 02 h−1 . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 00710 . 7554/eLife . 24669 . 008Figure 1—figure supplement 5 . Persistence of rich medium fast migrating phenotype in liquid culture . A strain isolated after 15 rounds of selection in rich medium ( Figure 1c , replicate 1 , main text , s= 0 . 6 cm h−1 ) was inoculated into a turbidostat and continuously cultured in LB at 30°C for approximately 140 generations . The number of generations was estimated assuming a constant generation time of 36 min . Samples were periodically drawn from the turbidostat and used to inoculate 0 . 3 % w/v agar LB plates . Migration was recorded via webcam as as described in the main text . Error bars are standard errors from regression of radius with time . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 008 In rich medium , colonies of wild-type bacteria ( MG1655-motile , founding strain ) expand with a front migration speed s≈ 0 . 3 cm h−1 and cells were sampled from the front after 12 hr ( Figure 1b ) . A portion of this sample was used to immediately inoculate a fresh plate while the remainder was preserved cryogenically . The process was repeated every 12 hr for 15 rounds . We observed a nearly 50% increase in s over the course of the first 5 rounds of selection . The increase in s was largely reproducible across five independent selection experiments ( Figure 1c ) . We estimate that plate-to-plate variation in agar concentration due to evaporative loss could change the migration rate by up to 0 . 06 cm h−1 in later rounds ( Appendix 1 ) . However , independent replicate selection experiments exhibit fluctuations in migration rate that exceed this estimate . For example , replicate 4 declines in later rounds of selection , and this decline may reflect the unique low abundance mutation that appears in this replicate by round 15 ( Figure 5a ) . In addition , replicate 3 exhibits substantially faster migration than replicates 1 , 2 and 4 in round 7 , and this may reflect the distinct mutations observed in this replicate at round 5 ( Figure 5a ) . So , while migration rates increased in all replicates , the magnitude of the increase differed between replicates . To check whether chemotaxis was necessary for increasing s , we performed selection experiments using a motile but non-chemotactic mutant ( ΔcheA-Z , Materials and methods ) . Motility in this strain was confirmed by single-cell imaging in liquid media . As observed previously ( Wolfe and Berg , 1989 ) , the non-chemotactic strain formed dense colonies in low viscosity agar that remained localized near the site of inoculation and expanded ∼1 cm in a 24 hr period: a rate 10-fold slower than the wild-type . To allow sufficient time for colony expansion , we performed selection experiments using this strain with 24 hr incubation times and observed an increase in s from approximately 0 . 03 cm h−1 to 0 . 04 cm h−1 ( Figure 1—figure supplement 1 ) . We did not observe fast migrating spontaneous mutants which have been reported previously in multiple species ( Wolfe and Berg , 1989; Mohari et al . , 2015 ) , likely because our plates were incubated for a shorter period of time . To determine the number of generations transpiring in our selection experiments , we measured the number of cells in the inoculum and the number of cells in the colony after 12 hr of growth and expansion ( Materials and methods ) . We estimated that 10 to 12 generations occurred in each round of selection . We then tested whether prolonged growth in well mixed liquid medium for a similar number of generations could lead to faster migration by growing the founding strain for 200 generations in continuous liquid culture and periodically inoculating a low viscosity agar plate ( Figure 1—figure supplement 2 ) . We observed only a 3 . 5% increase in the rate of migration , demonstrating that selection performed on spatially structured populations results in more rapid adaptation for fast migration than growth in well mixed conditions . We then performed selection experiments in a minimal medium where growth and migration are substantially slower than in rich medium ( Figure 1d ) . In this condition we allowed 48 hr for each round of expansion and took precautions to limit evaporative loss in the plates over this longer timescale ( Materials and methods ) . In the first round , the population formed small ∼1 . 5 cm diameter colonies without a well defined front . Populations formed well defined fronts in subsequent rounds of selection ( Figure 1d ) , reflecting a transition from growth and diffusion dominated transport to chemotaxis dominated migration ( Croze et al . , 2011 ) . We observed an approximately 3-fold increase in s over the course of 10 rounds of selection . Variation across replicate experiments was substantial , and exceeded our estimate of systematic error due evaporative losses changing the agar concentration ( Appendix 1 ) . So while all replicates increased their migration rate , the magnitude of the increase in migration rate varied substantially . This variation may be due to the different mutations present across replicates ( Figure 5b ) . When we performed selection in minimal medium using the non-chemotactic mutant ( ΔcheA-Z ) , we found little or no migration and only a very small increase in the migration rate over 10 rounds of selection ( Figure 1—figure supplement 1 ) . We concluded that chemotaxis is also necessary for increasing s in this medium . Using the same technique described for rich medium , we estimated the number of generations per round of selection in minimal medium to be <10 . We tested whether approximately 120 generations of growth in liquid was sufficient to evolve faster migration in minimal medium . Here we found that prolonged growth in well mixed conditions resulted in ∼2-fold faster front migration . Despite the increase in migration rate , selection in well mixed conditions resulted in slower migration than selection in low viscosity agar plates for a similar number of generations ( Figure 1—figure supplement 2 ) . To characterize the adaptation we observed in Figure 1c , e , we studied a reaction-diffusion model of migrating bacterial fronts of the type pioneered by Keller and Segel ( 1971 ) and reviewed in Tindall et al . ( 2008 ) . We model the bacterial density ρ⁢ ( 𝐫 , t ) and a single chemo-attractant that also permits growth c⁢ ( 𝐫 , t ) . Our model includes only a single nutrient since the growth and chemotaxis of the outermost ring in rich media is driven by L-serine ( Adler , 1966 ) and our minimal media conditions contain only a single carbon source/attractant . The dynamics of ρ⁢ ( 𝐫 , t ) and c⁢ ( 𝐫 , t ) are governed by ( 1 ) ∂ρ∂t=Db∇2ρ−∇⋅ ( k0KD ( KD+c ) 2ρ∇c ) +g ( ρ , c ) and ( 2 ) ∂c∂t=Dc∇2c−f ( ρ , c ) , where the spatial and temporal dependence of ρ and c have been suppressed for clarity . The three terms on the right hand side of Equation 1 describe diffusion , chemotaxis and growth respectively . Db is the bacterial diffusion constant , which describes the rate of diffusion of bacteria due to random , undirected motility . k0 is the chemotactic coefficient , which captures the strength of chemotaxis in response to gradients in attractant . KD is the equilibrium binding constant between the attractant and its associated receptor ( Brown and Berg , 1974 ) . Growth is modeled using the Monod equation g ( ρ , c ) =kgρcKg+c , where kg is the maximum growth rate and is the concentration of nutrient allowing half-maximal growth . f ( ρ , c ) describes the nutrient consumption and has an identical form to g ( ρ , c ) since we assume the yield ( Y , cells mL-1mM-1 ) is a constant . Dc is the diffusion constant of small molecules in water . The physiological parameters describing growth and attractant-receptor binding ( kg , Kg , Y and KD ) were either measured here or have been reported in the literature and can be applied directly in our simulation of migration in both nutrient conditions . Table 1 describes each parameter used in this study . 10 . 7554/eLife . 24669 . 009Table 1 . Reaction-diffusion model parameters: Columns indicate parameter , explanation of parameter , units , value used in simulation of founder strain in rich medium , and the value used in simulation of founder strain in minimal medium . Parameters marked with an m were measured in this study . Db , k0 and Dc in rich medium were estimated as described in Appendix 1 using the methods of Croze et al ( 2011 ) . Dc is assumed to be the same in minimal medium as rich medium . Identical k0 and Db were used in rich and minimal media since Ford and Lauffenburger ( 1992 ) find nearly identical values for galactose as Ahmed and Stocker ( 2008 ) do for serine . KD for both nutrient conditions was taken from Hazelbauer Adler et al . , 1973 . For minimal medium Kg and Y were taken from Lendenmann et al . , 1999 . The values cited for s were measured from numerical simulation of the reaction-diffusion model as outlined in Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 009ParameterExplanationUnitsFounder value RMFounder value MMSingle-cell swimming ( this study ) τrrun durations0 . 67m0 . 47mτttumble durations––\|vr\|run speedμm s−118 . 7m22 . 2mReaction-diffusion modelρ⁢ ( 𝐫 , t ) cell densitym−3––c⁢ ( 𝐫 , t ) nutrient densitymM––c0nutrient concentration in mediummM1m0 . 18mDbbacterial diffusion constantcm2h−10 . 05760 . 0576Dcnutrient diffusion constantcm2h−10 . 0360 . 036k0chemotactic coefficient in liquidcm2h−16 . 126 . 12KDreceptor-nutrient binding constantmM20 . 1kgmaximum growth rateh−11 . 23m0 . 125mKgc concentration for half-maximum growth ratemM0 . 13P <10-4Yyield biomass per unit nutrientscells mL-1mM-15×107m3×108Cagar concentration% ( w/v ) 0 . 3m0 . 3msfront migration ratecm h−10 . 610 . 09 The bacterial diffusion constant and the chemotactic coefficient depend on motility and the physical structure of the agar matrix . Motility in E . coli consists of runs , segments of nearly straight swimming ∼0 . 5 to 1 s long at ∼20 μm s-1 , and tumbles that rapidly reorient the cell over a period of ∼0 . 1 s ( Berg and Brown , 1972 ) . Rivero et al . showed how the reaction-diffusion parameters Db and k0 depend on run speed and duration ( Rivero et al . , 1989 ) . Croze et al . ( 2011 ) modified these results to account for the presence of the agar matrix . The approach treats interactions between cells and agar as scattering events where the cell is forced to tumble . We estimated Db and k0 using the method developed by Croze et al . for our conditions . With these parameters we simulated the model in Equations 1 and 2 with parameters appropriate for rich media ( chemotaxis towards L-serine ) and minimal media ( chemotaxis towards galactose ) . For the founder strain , these simulations predicted a migration rate of 0 . 61 cm h−1 for rich media and 0 . 08 cm h−1 for minimal media compared to measured rates of 0 . 30 ± 0 . 01 cm h−1 and 0 . 0163 ± 0 . 0038 cm h−1 respectively . We note that this comparison involves no free parameters . In rich medium our model describes the dynamics of a single metabolite/attractant ( L-serine ) , and therefore fails to account for secondary fronts behind the outermost front , which arise from the metabolism of other amino acids ( Adler , 1966 ) ( Figure 1b , Figure 2—figure supplement 2 ) . This is a reasonable approximation since we select cells only from the outermost front of the colony . In minimal medium , where only a single nutrient is available , we observe only a single migrating front as our model predicts ( Figure 2—figure supplement 2 ) . Other limitations of this model include the fact that it does not describe the process of adaptation by chemoattractant receptors ( Berg and Tedesco , 1975 ) , nor does it describe stochastic processes at the single-cell level such as trapping in the agar matrix and cell-to-cell variability . The discrepancy between predicted migration rate and our observed migration rate most likely arises from the fact that cells are transiently trapped in the agar matrix ( Wolfe and Berg , 1989 ) rather than simply being scattered . While more sophisticated models have been developed to include these processes ( Vladimirov et al . , 2008; Frankel et al . , 2014 ) , the model in Equations 1 , 2 captures the essential features of bacterial front migration with fewer adjustable parameters . See Appendix 1 for further discussion . To understand how changes in motility and growth could contribute to the evolution of migration , we studied how the migration rate ( s ) varied with the parameters of our model through numerical simulation ( Appendix 1 ) . We found that increases in run speed ( \|vr\| ) and growth rate ( kg ) had the largest impact on s ( Figure 2 ) . Consistent with previous reports , our model indicates that only small gains in migration rate can be achieved through increases in tumble frequency ( Wolfe and Berg , 1989 ) ( ∼10% , Figure 2—figure supplement 3 ) . 10 . 7554/eLife . 24669 . 010Figure 2 . Migration rate increases with run speed and growth rate . ( a ) Front migration rate ( heatmap ) as a function of run speed ( \|vr\| ) and maximum growth rate ( kg ) simulated using the reaction-diffusion model discussed in the text with parameters appropriate for rich medium conditions ( Table 1 ) . Model parameters were estimated using the method developed by Croze et al ( Appendix 1 ) . Black square shows the run speed and growth rates measured for the founding strain in rich medium ( Figure 3 ) . Standard error in \|vr\| is smaller than the size of the marker; error bar in kg is the standard deviation across three replicate measurements . ( b ) Identical to panel ( a ) except for minimal medium . The abrupt change in migration rate around kg=0 . 2 h−1 corresponds to a transition from diffusion dominated front migration to a traveling wave ( Appendix 1 ) . The founding strain’s phenotype is shown as a black circle , error bars are constructed identically to those in ( a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01010 . 7554/eLife . 24669 . 011Figure 2—figure supplement 1 . Reaction-diffusion model recapitulates qualitative features of colony expansion . Results from numerical simulations of the reaction-diffusion model in the main text . Simulations for founding strain in rich medium ( a ) , founding strain in minimal medium ( b ) , and round 5 strain in minimal medium ( c ) are shown . Three snapshots of ρ ( r , t ) for each simulation are shown as greyscale heatmaps ( note independent color maps ) . The panel on the right in ( a–c ) shows the location of the front in time ( black trace ) and the time points corresponding to the three snapshots are labeled by the colored points . The parameters for each simulation are given in Table 1 . The founding strain in minimal medium exhibits diffusive transport due to slow growth , this is also observed experimentally ( Figure 1 , main text ) . Scale bars on the left of each panel are 1 cm . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01110 . 7554/eLife . 24669 . 012Figure 2—figure supplement 2 . Comparison of front profiles from simulation and experiment . Upper four panels show front density profiles from simulation and experiment for the rich medium condition . Left column shows founder and right column round 15 . Simulation profiles are taken from time points after a constant rate of expansion has been attained . Experimental front profiles are taken at the end of colony expansion ( 12 hr ) . In the experimental front profiles , the high-density regions arise from metabolism of amino acids other than serine . The lower four panels are identical to the upper four but are taken from minimal medium simulations and experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01210 . 7554/eLife . 24669 . 013Figure 2—figure supplement 3 . Simulation of migration rate versus tumble frequency . Using the formalism of Croze et al . , migration rate as a function of tumble frequency ( Appendix 1 ) was computed using the reaction-diffusion model presented in the main text . Panels show migration rate ( s ) as a function of tumble frequency ( α0 ) for rich medium and minimal medium conditions . Red dots indicate measured tumble frequency for founder in each condition ( Figure 3 , main text ) . Error bars in the left panel are smaller than the size of the markers . Error bars in the right panel are standard errors from a linear regression on the front location in time . The non-monotonic variation of migration rate with tumble frequency in minimal medium results from the slight curvature in the front location as a function of time in these conditions ( Figure 2—figure supplement 1 [right panel] ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 013 Figure 2 shows how the front migration rate ( heatmap ) varies with run speed and growth rate for both nutrient conditions studied in Figure 1 . Our model predicts that the fastest migrating strain should be the one that increases both its run speed and growth rate relative to the founder . Therefore , in the absence of any constraints on accessible phenotypes , we expect both run speed and growth rate to increase with selection . To test the predictions of the reaction-diffusion model , we experimentally interrogated how the motility and growth phenotypes of our populations evolved over the course of selection . We performed single-cell tracking experiments using a microfluidic method similar to one described previously ( Jordan et al . , 2013 ) . This method permitted us to acquire 5 min swimming trajectories from hundreds of individuals from strains isolated prior to selection ( founder ) and after 5 , 10 and 15 rounds of selection in rich media ( replicate 1 , Figure 1c ) and for the founder and strains isolated after 5 and 10 rounds of selection in minimal media ( replicate 1 , Figure 1e ) . For tracking , cells were grown in the medium in which they were selected . This technique permitted us to capture more than 280 , 000 run-tumble events from approximately 1500 individuals . Tracking code is available ( Mickalide et al . , 2017 ) . We identified run and tumble events for all individuals ( Berg and Brown , 1972; Taute et al . , 2015 ) ( Materials and methods ) . Figure 3a–b shows that run durations declined over the course of selection in both rich and minimal media . We show the complementary cumulative distribution function ( c⁢ ( τr ) ) of run durations ( τr ) aggregated across all run events detected for the founding or evolved strains ( c⁢ ( τr ) =1-∫-∞τrd⁢τr′⁢P⁢ ( τr′ ) , where P⁢ ( τr′ ) is the distribution of run durations ) . c⁢ ( τr ) quantifies the fraction of all runs longer than a time τr . These distributions show that the evolved strains exhibited a reduction in the probability of executing long runs . We observed opposite trends for tumble duration , with decreasing tumble duration in rich medium and increasing duration in minimal medium ( Figure 3—figure supplement 2 ) . To summarize these changes in run-tumble statistics , we computed the tumble bias ( fraction of time spent tumbling ) and the tumble frequency ( tumbles per second , Figure 3c–d ) . In both conditions , we observe an increase in the tumble frequency . This is expected since previous studies showed that mutants with increased tumble frequencies have faster migration rates through agar , likely due to tumbles freeing cells from being trapped in the agar ( Wolfe and Berg , 1989 ) . In rich medium we observed a decline in tumble bias , while selection in minimal medium increased the tumble bias . Tumble bias and frequency are reported in Table 2 for all tracked strains . 10 . 7554/eLife . 24669 . 014Table 2 . Tumble bias and frequency for additional strains . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 014Tumble bias and frequenciesstrainTumble biasTumble frequency [Hz]Rich mediumfounder0 . 197 ± 0 . 0060 . 59 ± 0 . 0215 ( r3 ) 0 . 174 ± 0 . 0060 . 78 ± 0 . 0215 ( r4 ) 0 . 2 ± 0 . 010 . 79 ± 0 . 01clpXE1850 . 19 ± 0 . 010 . 66 ± 0 . 02Minimal mediumfounder0 . 29 ± 0 . 010 . 41 ± 0 . 0310 ( r2 ) 0 . 25 ± 0 . 020 . 44 ± 0 . 03galSL22R0 . 3 ± 0 . 020 . 44 ± 0 . 0510 . 7554/eLife . 24669 . 015Figure 3 . Dynamics of phenotypic evolution in rich and minimal media . ( a–f ) Show single-cell swimming phenotypes ( run duration ( τr ) , run speed ( \|vr\| ) , tumble bias and tumble frequency , see Materials and methods ) . Tracking was performed for founding strain ( 140 cells , 19 , 597 run events ) , strains isolated after 5 ( 79 cells , 12 , 217 run events ) , 10 ( 97 cells , 18 , 505 run events ) and 15 ( 96 cells , 15 , 928 run events ) rounds in rich media and in minimal media for the founding strain ( 72 cells , 7556 run events ) , round 5 ( 45 cells , 9724 run events ) and round 10 ( 25 cells , 4892 run events ) . ( a ) Shows the fraction of runs longer than a given τr for strains evolved in rich media ( 95% confidence intervals from bootstrapping ) . The mean and standard deviation in run duration for founder is 0 . 66 ± 0 . 78 s , for round 5: 0 . 63 ± 0 . 61 s , for round 10: 0 . 58 ± 0 . 50 s and for round 15: 0 . 65 ± 0 . 57 s . Round 5 , 10 and 15 strains exhibit shorter average run durations than founder ( p<0 . 05 ) . ( b ) Shows the same distribution for strains in minimal medium with founder exhibiting average run duration 0 . 49 ± 0 . 52 s , round 5: 0 . 44 ± 0 . 48 s and round 10: 0 . 33 ± 0 . 28 s . Rounds 5 and 10 exhibit shorter average run durations than founder ( p<10−8 ) . ( c–d ) Show average fraction of time spent tumbling ( tumble bias ) and tumble frequency ( tumbles per second ) for rich medium and minimal medium respectively . Note the two vertical axes . In rich medium only the round 15 tumble bias is significantly different from founder ( p<0 . 001 ) , but the tumble frequency is higher than founder for both rounds 10 and 15 ( p<0 . 001 ) . In minimal medium all tumble biases and frequencies are significantly different from founder for all strains ( p<0 . 001 ) . ( e ) Shows run speed distributions for strains evolved in rich medium , legend in ( a ) applies . The average ± standard deviation run speeds are , for founder: 18 . 7 ± 7 . 1 μm s−1 , round 5: 24 . 9 ± 7 . 1 μm s−1 , round 10: 27 . 6 ± 7 . 0 μm s−1 , and for round 15: 28 . 7 ± 6 . 8 μm s−1 . Average run speeds for rounds 5 , 10 and 15 are greater than founder ( f ) Shows the same distributions for strains evolved in minimal medium , average run speed for founder: 20 . 7 ± 10 . 8 μm s−1 , for round 5: 11 . 2 ± 4 . 8 μm s−1 and for round 10: 13 . 3 ± 4 . 4 μm s−1 . Both rounds 5 and 10 exhibit slower average run speeds than founder , the legend in ( b ) applies . ( g–h ) Show growth rates in well mixed liquid culture for all strains studied in panels ( a–f ) in the medium in which the strains were selected . ( g ) Shows triplicate measurements from each of the four strains isolated in rich medium . Rounds 5 , 10 and 15 exhibit slower growth than founder ( p<0 . 01 ) . ( h ) Shows growth rates for strains isolated from minimal medium selection experiment . Four replicate measurements were made for founder and round 10 and three replicate measurements for round 5 . Squares and circles demarcate measurements made on separate days . Rounds 5 and 10 have higher growth rates than founder ( p<10-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01510 . 7554/eLife . 24669 . 016Figure 3—figure supplement 1 . Microfluidic device and single-cell swimming trajectory . ( left ) Bright-field image at 20× magnification of the PDMS microfluidic chamber used to trap single bacteria . The boundary of the chamber can be seen as the high contrast circle . Scale bar is 50 μm . ( right ) A segmented trajectory of a single cell in a chamber like the one shown on the left . Dots indicate locations of the centroid . Black portions indicate running events and red portions tumbles . Image processing and run-tumble detection are described in the Materials and methods section of the main text . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01610 . 7554/eLife . 24669 . 017Figure 3—figure supplement 2 . Tumble durations and run lengths for evolved strains . Tumble durations ( τt ) and run lengths ( lr ) for single-cell tracking shown in Figure 3 of the main text . ( a ) Shows the complementary cumulative distribution of tumble durations for rich media evolved strains . Shaded regions are 95% confidence intervals from bootstrapping . Averages and standard deviations are: 0 . 18 ± 0 . 20 s , 0 . 17 ± 0 . 16 s , 0 . 14 ± 0 . 13 s , 0 . 14 ± 0 . 12 s for founder , round 5 , 10 and 15 respectively . ( b ) Identical to ( a ) except constructed for run lengths . The run length is found by computing the arc-length between tumble events for each run . The averages and standard deviations are 13 . 5 ± 17 . 7 μm , 16 . 5 ± 17 . 4 μm , 16 . 5 ± 16 . 0 μm , 19 ± 17 . 8 μm respectively . ( c ) and ( d ) are identical to ( a ) and ( b ) for minimal medium evolved strains ( replicate 1 , Figure 1e ) . The tumble durations are 0 . 17 ± 0 . 17 s , 0 . 25 ± 0 . 28 s , 0 . 20 ± 0 . 21 s for founder , round 5 and 10 . The respective run lengths are 10 . 0 ± 13 . 0 μm , 5 . 0 ± 7 . 5 μm and 4 . 6 ± 4 . 6 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01710 . 7554/eLife . 24669 . 018Figure 3—figure supplement 3 . Reproducibility of the evolved phenotype . Single-cell tracking and growth rate measurements were performed on independently selected strains in rich medium ( 15 rounds , ( a–c ) ) and minimal medium ( 10 rounds , ( d–f ) ) . Panels show run durations ( a , d ) , run speeds ( b , e ) and growth rates ( c , f ) . The founder population is shown in black in all panels . Single-cell tracking experiments were performed on two additional round 15 strains from the rich medium experiment ( replicates 3 and 4 , Figure 1c main text ) . For replicates 1 , 3 and 4– 96 , 85 and 98 individuals were tracked for a total of 15 , 928 , 16 , 639 and 18 , 171 run events respectively . ( a ) shows the run duration distributions for these three strains with mean ± standard deviations: 0 . 65 ± 0 . 57 s , 0 . 60 ± 0 . 53 s , 0 . 57 ± 0 . 49 s respectively . ( b ) Run speed ( \|vr\| ) distributions for the same three strains with means 28 . 7 μm s−1 , 26 . 2 μm s−1 and 26 . 7 μm s−1 respectively . ( c ) maximum growth rates ( kg ) for the same two independently evolved strains ( with 15 ( 3 ) denoting replicate 3 and 15 ( 4 ) denoting replicate 4 ) . The decline in growth rate relative to founder is significant for both replicate 3 ( p<10−3 ) and replicate 4 ( p<10−3 ) . ( d–f ) show swimming statistics and growth rates for independently evolved strains in minimal medium , replicate 1 and 2 correspond to Figure 1e in the main text . ( d ) Run duration distributions for constructed for 25 individuals from replicate 1 and 80 individuals from replicate 2 corresponding to 4892 and 9357 run events respectively . The mean ± standard deviations are: 0 . 33 ± 0 . 26 s and 0 . 65 ± 0 . 87 s . ( e ) Run speed distributions for independently evolved minimal medium strains . Means for replicates 1 and 2 are 13 . 3 μm s−1 and 15 . 25 μm s−1 respectively . ( f ) Growth rates for founder , rounds 5 and 10 reproduced from Figure 1e , main text ( circles ) along with growth rate measurements for strain isolated from round 5 of replicate 2 ( dark red triangles ) and round 10 of replicate 2 ( light red triangles ) . Means are 0 . 3h-1and 0 . 24h-1 Round 5 growth rates do not differ significantly ( p=0 . 24 ) while round 10 growth rates do ( p<0 . 02 ) . Both replicate 2 strains from rounds 5 and 10 exhibit growth rates larger than founder ( p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 01810 . 7554/eLife . 24669 . 019Figure 3—figure supplement 4 . Swimming statistics as a function of culture density . ( a–d ) Show swimming statistics ( τr , στr , τt and \|vr\| ) as a function of culture optical density for rich medium founding ( black ) and evolved ( green , round 15 , replicate 1 ) . Each point corresponds to a single individual tracked for up to 5 min . 141 individuals were tracked from founder and 96 individuals were tracked from round 15 . Trend lines are from non-parametric kernel regressions and shaded regions represent 95% confidence intervals from bootstrapping . The shorter run duration in round 15 is apparent in the reduced στr relative to founder . ( e–h ) Show identical plots for minimal medium founding ( black ) strain ( 38 cells ) and evolved ( green , 64 cells , round 10 replicate 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 019 Figure 3e–f show the probability distributions of run speeds for founding and selected strains in both nutrient conditions . In rich medium we observed a nearly 50% increase in the run speed ( \|vr\| ) between founder and rounds 10 to 15 . Tracking strains isolated after 15 rounds from independent selection experiments ( replicates 3 and 4 , Figure 1c ) showed that this increase in run speed was reproducible across independent evolution experiments ( Figure 3—figure supplement 3 ) . Finally , to check that the phenotype we observed after 15 rounds of selection in rich medium was distinct from standard laboratory strains used in chemotaxis studies , we tracked RP437 and found that its swimming speed was slower than the round 15 strain ( Figure 1—figure supplement 4 ) . Surprisingly , when we performed single-cell tracking for strains evolved in minimal media we observed the opposite trend . In these conditions we observed a 50% reduction in run speed ( Figure 3f ) . Again , we found that this result was reproducible across independently evolved strains ( Figure 3—figure supplement 3 ) . While the overall trend in minimal medium was towards reduced run duration , one replicate showed an increase in run duration ( Figure 3—figure supplement 3 ) . The strain where we observed long runs after 10 rounds of selection ( replicate 2 , Figure 1e ) also exhibited a slower migration rate than the strain isolated from replicate 1 , and the long run durations may be responsible for this difference . We then measured the growth rates of founding and evolved strains from both selection conditions in well mixed liquid corresponding to the medium used for selection ( Appendix 1 ) . We observed a decline of about 10% in the maximum growth rate with selection in rich medium and a three-fold increase in the maximum growth rate after 10 rounds of selection in minimal medium ( Figure 3g-h ) . We found that these changes in growth rate are reproducible across independently evolved strains in both environmental conditions ( Figure 3—figure supplement 3 ) . Since motility is known to depend on the growth history of the population ( Staropoli and Alon , 2000 ) , we checked whether the phenotypic differences between founding and evolved strains shown in Figure 3 remained valid when we tracked cells over a range of optical densities during population growth . We performed these measurements for the founding strain in both rich and minimal media , and for a round 15 strain in rich medium and a round 10 strain in minimal medium ( Figure 3—figure supplement 4 ) . For both rich and minimal media , we found that the differences in run speed ( \|vr\| ) between founding and selected strains were retained across the growth curve ( Figure 3—figure supplement 4d ) . Likewise , in minimal medium , the average run duration was shorter for the selected strain than for the founder across the growth curve . For rich medium , average run durations for the round 15 strain were not consistently shorter than founder , but the round 15 strain exhibited smaller variability in run duration ( Figure 3—figure supplement 4b ) . Combining growth rate measurements with single-cell motility measurements allowed us to predict the front migration rate for strains in rich and minimal media using the reaction-diffusion model described above . We found that the model qualitatively recapitulated the increase in front migration rate that we observed experimentally ( Tables 3 , 4 , Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 24669 . 020Table 3 . Reaction-diffusion model parameters estimated from measurements of tumble frequency ( α0 ) and run speed ( \|vr\| ) for rich medium evolved strains in C= 0 . 3% agar . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 020Evolution of population level migration parametersstrainα0 [s−1]\|vr\| [ μm s−1]Db [ cm2h−1]k0 [ cm2h−1]founder1 . 4518 . 70 . 020 . 65round 51 . 5624 . 90 . 0270 . 90round 101 . 7227 . 60 . 0291 . 04round 151 . 5428 . 70 . 0311 . 0410 . 7554/eLife . 24669 . 021Table 4 . Reaction-diffusion model parameters estimated from measurements of tumble frequency ( α0 ) and run speed ( \|vr\| ) for minimal medium evolved strains in C=0 . 3% agar . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 021Evolution of population level migration parametersstrainα0 [s−1]\|vr\| [ μm s−1]Db [ cm2h−1]k0 [ cm2h−1]founder220 . 70 . 0210 . 66round 52 . 511 . 20 . 0110 . 39round 10313 . 30 . 0110 . 5 We conclude that there is a trade-off between run speed and growth rate in E . coli which constrains the evolution of faster migration through low viscosity agar . Figure 4 , which summarizes this trade-off for both conditions , shows the measured growth rates and swimming speeds for all strains presented in Figure 3 overlaid on the predicted migration rates from our reaction-diffusion model . The curves in Figure 4a–b show that the evolved phenotypes lie near a Pareto frontier in the phenotypic space of run speed and growth rate . 10 . 7554/eLife . 24669 . 022Figure 4 . Trade-off between growth rate and run speed constrains evolution of faster migration . ( a ) Shows run speeds and growth rates for strains evolving faster migration in rich medium overlaid on a heatmap of the prediction for front migration rate from the reaction-diffusion model ( Figure 2 ) . Phenotypes for strains from Figure 3 are shown along with two independently evolved strains ( replicates 3 ( 15 ( r3 ) ) and 4 ( 15 ( r4 ) ) , Figure 1c ) . In addition , the red ‘x’ marks the phenotype for the mutation clpXE185 in the founding strain background ( Figure 5 ) . ( b ) Shows run speeds and growth rates for strains evolved in minimal medium overlaid on the predicted from migration rate from the reaction-diffusion model . Growth rate and run speed for an independently evolved round 10 strain is shown ( 10 ( r2 ) , Figure 1e ) as well as the phenotype for the galSL22R mutation in the founder background ( black ‘x’ ) . Predicted front migration rates assume no change in run duration . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 02210 . 7554/eLife . 24669 . 023Figure 4—figure supplement 1 . Predicted migration rates for evolved strains . Using the reaction-diffusion model ( Main text ) , we simulated colony expansion using the parameters shown in Table 1 of the main text . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 02310 . 7554/eLife . 24669 . 024Figure 4—figure supplement 2 . Swimming statistics , growth rates and migration rates for mutants . Run durations ( τr ) and speeds ( \|vr\| ) , growth rates ( kg ) and migration rates ( s ) for four mutations reconstructed in the founder background ( see Main Text ) . Three mutants were studied in rich medium ( a , c , e , g ) - clpXE185 , a single base pair deletion at position 523 , 086 ( Δ1bp ) and the double mutant ( clpX+Δ 1 bp ) . One mutant was studied in minimal medium: galSL22R . In all panels , phenotypes of mutants are compared to founder and the population isolated after the final round of selection in the appropriate environment . ( a ) shows c⁢ ( τr ) in rich medium , means and standard deviations are: 0 . 63 ± 0 . 60 s , 0 . 66 ± 0 . 91 s and 0 . 59 ± 0 . 55 s for clpX , Δ1 bp and clpX+Δ1bp respectively . clpX and clpX+Δ1bp have shorter average run durations than founder ( p<10−4 ) . ( b ) c⁢ ( τr ) in minimal medium , where galSL22R exhibits longer runs than founder with 0 . 55 ± 0 . 75 s ( p<10−5 ) . ( c ) gives P⁢ ( \|vr\| ) in rich medium . Means ± standard deviations are 24 . 2 ± 7 . 8 μm s−1 , 18 . 2 ± 7 . 3 μm s−1 and 23 . 4 ± 7 . 6 μm s−1 for clpX , Δ1 bp and clpX+Δ1bp respectively . All mutants except Δ1 bp exhibit faster runs on average ( p<10−5 ) . ( d ) gives P⁢ ( \|vr\| ) in minimal medium . galSL22R has a mean of 17 . 6 ± 8 . 7 μm s−1 , which is lower than founder ( p<10−5 ) . ( e ) Growth rates for rich medium mutants . clpX and clpX+Δ1bp have lower growth rates than founder ( p=0 . 0087 and p=0 . 0069 ) . The Δ1 bp mutation alone does not have a statistically significant difference in growth rate from founder ( p=0 . 53 ) . ( f ) shows growth rate for the galS mutant relative to founder and round 10 . The mutant growth rate is larger than founder ( p<0 . 001 ) . ( g ) shows colony migration rates for mutants in rich medium . clpX and clpX+Δ1bp differ significantly from the migration rate of founder ( p=0 . 0021 and p=0 . 0017 ) . Δ1bp does not have a statistically significant change in migration rate . Comparisons are made between duplicate measurements for each genotype and the migration rates of all five replicate experiments in Figure 1 of the main text . ( f ) Shows migration rate measurements for the galS mutant in minimal medium compared to founder and round 10 in minimal medium . The mutant is faster than the founding strain ( p<10−3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 024 To investigate the mechanism of the phenotypic evolution and trade-off we observed , we performed whole genome sequencing of populations for the founding strain as well as strains isolated after rounds 5 , 10 and 15 in rich medium for four of five selection experiments and rounds 5 and 10 in minimal medium for four of five selection experiments ( Materials and methods ) . Figure 5 shows de novo mutations observed in each strain sequenced . Since we sequenced populations , we report the frequency of each mutation observed ( see legend , Figure 5a , middle panel ) . 10 . 7554/eLife . 24669 . 025Figure 5 . Genomic evolution . ( a ) De novo mutations observed in strains isolated after 5 , 10 and 15 rounds of selection in rich medium . Abscissa denotes position along the genome . Colors of the markers indicate independently evolved replicates and correspond to traces in Figure 1c . Circles denote single nucleotide polymorphisms ( SNP ) in coding regions , squares denote intergenic SNPs , and triangles denote larger insertions or deletions . The size of the marker is proportional to the frequency of the mutation in the population . Only mutations with a frequency above 0 . 2 in the population are shown . Genes of interest are labeled . The operons coding for motility and chemotaxis are near flhD . ( b ) Identical to ( a ) but shows de novo mutations for strains evolved in minimal medium . The marker near icd corresponds to multiple SNPs in close proximity to each other . See Tables 5–12 for a list of all mutations observed and details of the sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 025 In the rich medium experiment we observed parallel evolution across replicate selection experiments , with a mutation in clpX ( E185* ) and an intergenic single base pair deletion both rising to fixation within approximately 5 rounds of selection . In this condition we observed transient mutations in genes regulating chemotaxis or motility ( near flhD , Figure 5a ) in two of four replicates . A previous study showed that mutations in clpX alter flhDC expression and motility ( Girgis et al . , 2007 ) . We therefore focused attention on the mutation in clpX , which converted position 185 from glutamic acid to a stop codon in the 424 residue ClpX protein . ClpX is the specificity subunit of the ClpX-ClpP serine protease . ClpX forms a homohexamer that consumes ATP to unfold and translocate target proteins to the ClpP peptidase ( Baker and Sauer , 2012 ) . The ClpXP protease has many targets in the cell including FlhDC , the master regulator of flagellar biosynthesis ( Tomoyasu et al . , 2003 ) . We found that this mutation in clpX was at high abundance ( >70% ) in all populations after 5 rounds of selection and fixed by round 10 in all four replicates ( Figure 5a ) . To determine the phenotypic effects of clpXE185* , we used scarless recombineering to reconstruct this mutation in founding strain genetic background ( Kuhlman and Cox , 2010 ) ( Materials and methods ) . We then performed migration rate , single-cell tracking and growth rate measurements on this strain . We observed a statistically significant increase in migration speed for the clpXE185* mutant ( 0 . 39 ± 0 . 01 cm h−1 , mean and standard error ) relative to founder ( 0 . 30 ± 0 . 01 cm h−1 , p=0 . 002 ) . We also found that clpXE185* resulted in a statistically significant increase in run speed relative to founder ( 24 . 2 μm s−1 compared to 18 . 7 μm s−1 , p< 10-10 ) . Finally , in well mixed batch culture in rich medium , the clpXE185* mutant exhibited a maximum growth rate kg= 1 . 19 ± 0 . 009 h−1 ( standard error for triplicate measurements ) with founder exhibiting a maximum growth rate of 1 . 23 ± 0 . 01 h−1 ( p=0 . 0174 , Figure 4—figure supplement 2 ) . Knocking out clpX from founder resulted in very slow front migration ( s= 0 . 0036 ± 0 . 001 cm h-1 ) , suggesting that the stop codon mutation we observe has a more subtle effect on the enzyme’s function than a simple loss of function . Finally , we reconstructed the intergenic single base pair deletion which fixed in all four replicate selection experiments but observed no phenotypic effects of this mutation when placed in the founder or clpXE185* background ( Figure 4—figure supplement 2 ) . These results suggest that this intergenic mutation is neutral . We conclude that the clpX mutation observed in all four replicate experiments drives faster front migration through increasing run speed , despite decreasing growth rate . Since the mutant exhibits both faster swimming and slower growth rate relative to founder we conclude that the trade-off between growth rate and swimming speed is driven by antagonistic pleiotropy ( Cooper and Lenski , 2000 ) . Figure 5b shows the mutations observed in rounds 5 and 10 for four of five replicate selection experiments in minimal medium . In all experiments , we observed mutations in the transcriptional regulator galS which fixed in just five rounds . In one of four experiments , we observed a mutation in the gene encoding the motor protein FliG , otherwise the observed mutations appear to be metabolic in nature . In minimal medium we also observed a substantial number of synonymous mutations rising to fixation ( see Table 5–8 ) . The role of these synonymous mutations is not known , but may be due to tRNA pool matching ( Stoletzki and Eyre-Walker , 2007 ) . 10 . 7554/eLife . 24669 . 026Table 5 . Minimal medium replicate 1: All mutations detected in rounds 5 and 10 of minimal medium replicate 1 . The galSL22R mutation in rounds 5 and 10 was confirmed by Sanger sequencing . See Table 9 caption . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 026Minimal medium replicate 1round , ( coverage ) 5 , ( 71× ) 10 , ( 214× ) Mutation ( loc . , mut . , frac . , cov . ) 1196220 , icd H366H , 78 . 4% , 461196220 , icd H366H , 100% , 1411196232 , icd T370T , 71 . 1% , 341196232 , icd T370T , 100% , 1011196247 , icd L375L , 72 . 0% , 251196247 , icd L375L , 100% , 751196277 , icd N385N , 47 . 1% , 172015871 , fliG V331D , 100% , 1111196280 , icd A386A , 47 . 1% , 172241604 , galS L22R , 100% , 1841196283 , icd K387K , 47 . 2% , 172685013 , glyA H165H , 100% , 1971196292 , icd T390T , 46 . 2% , 133815859 , rph Δ82 bp , 100% , 2601196304 , icd E394E , 46 . 2% , 132015871 , fliG V331D , 70 . 0% , 602241604 , galS L22R , 100% , 452685013 , glyA H165H , 100% , 6210 . 7554/eLife . 24669 . 027Table 6 . Minimal medium replicate 2: All mutations detected in rounds 5 and 10 of minimal medium replicate 2 . The galSL22R mutation in rounds 5 and 10 was confirmed by Sanger sequencing . See Table 9 caption . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 027Minimal medium replicate 2round , ( coverage ) 5 , ( 67× ) 10 , ( 64× ) Mutation ( loc . , mut . , frac . , cov . ) 2241604 , galS L22R , 100% , 701757419 , IG +17 bp insertion , 94 . 9% , 372685013 , glyA H165H , 100% , 652241604 , galS L22R , 100% , 472685013 , glyA H165H , 100% , 793815828 , IG T→G , 43 . 5% , 6210 . 7554/eLife . 24669 . 028Table 7 . Minimal medium replicate 3: All mutations detected in rounds 5 and 10 of minimal medium replicate 3 . See Table 9 caption . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 028Minimal medium replicate 3round , ( coverage ) 5 , ( 208× ) 10 , ( 229× ) Mutation ( loc . , mut . , frac . , cov . ) 1291079 , rssB A280T , 29 . 7% , 542241595 , galS Δ1bp , 100% , 2182241595 , galS Δ1bp , 64 . 7% , 1023277264 , prlF +CATTCAA insertion , 93 . 6% , 1093762200 , rhsA A6A , 23 . 5% , 1813350529 , IG T→C , 100% , 1173762212 , rhsA G10G , 23 . 1% , 1643762200 , rhsA A6A , 45 . 8% , 3203762212 , rhsA G10G , 42 . 0% , 29210 . 7554/eLife . 24669 . 029Table 8 . Minimal medium replicate 4: All mutations detected in rounds 5 and 10 of minimal medium replicate 4 . See Table 9 caption . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 029Minimal medium replicate 4round , ( coverage ) 5 , ( 256× ) 10 , ( 230× ) Mutation ( loc . , mut . , frac . , cov . ) 2241232 , galS R146L , 72 . 4% , 2742020519 , fliM E145K , 100% , 2052241665 , galS I2L , 100% , 304 To understand how these mutations drive phenotypic evolution , we focused on the galSL22R mutation . galS encodes the transcriptional repressor of the gal regulon . The coding mutation we observe occurs in the highly conserved N-terminal helix-turn-helix DNA binding region of this protein , we therefore expect that this mutation alters the expression of the gal regulon ( Weickert and Adhya , 1992 ) . To assay the phenotypic effects of this mutation , we reconstructed it in the genetic background of the founder . The migration rate of the galSL22R mutant showed a statistically significant increase relative to founder ( s= 0 . 039 ± 0 . 001 cm h−1 for galSL22R and 0 . 0163 ± 0 . 0038 cm h−1 for founder , p < 10−3 ) . We found that the growth rate of the mutant was approximately 2 . 5-fold larger than founder in minimal medium ( 0 . 23 ± 0 . 005 h−1 for galSL22R and 0 . 089 ± 0 . 03 h−1 for founder , p=4×<10−4 ) . Further , this mutation reduced the mean swimming speed relative to founder by approximately 15% ( Figure 4b , Figure 4—figure supplement 2 ) . However , when we knock out the galS gene from founder we do not observe a significant increase in the migration rate ( ΔgalS s= 0 . 0165 ± 0 . 002 cm h−1 , p=0 . 92 ) . Therefore , as shown in Figure 4b , we conclude that galSL22R alone drives faster growth and slower swimming . As with the rich medium condition , this trade-off is governed by antagonistic pleiotropy . To understand why we observe divergent phenotypic trajectories in the rich and minimal medium conditions ( Figure 4a–b ) , we studied a simple model of the evolution of correlated traits ( Lande , 1979; Mezey and Houle , 2005 ) . We consider a vector of the two phenotypes of interest , run speed and maximum growth rate , normalized to the values of the founder ( Hansen and Houle , 2008 ) , ϕ→=[\|v~r\| , k~g]T ( \|v~r\|=⟨\|vr\|⟩/⟨\|vr\|f⟩ , k~g=⟨kg⟩/⟨kgf⟩ , where ⟨⟩ denotes an average across the population ) . The model describes the evolution of the mean phenotype ( ϕ→ ) under selection by ( 3 ) ϕ→=Gβ→+ϕ→0 where G , the genetic covariance matrix , describes the genetically driven phenotypic covariation in the population , which is assumed to be normally distributed ( 𝒩⁢ ( ϕ→ , G ) ) . β→ is the selection gradient which captures the change in migration rate with respect to phenotype since we are selecting for faster migration . The matrix G is given by ( 4 ) G=[σ\|v~r\|2ρ⁢σ\|v~r\|⁢σk~gρ⁢σ\|v~r\|⁢σk~gσk~g2] , where σ~2 describes the ( fractional ) variance in the phenotype due to genetic variation and captures the correlation between the two traits . Therefore , the diagonal elements of G describe the capacity for mutations to vary each trait while the off-diagonal elements describe the capacity for mutations to vary both traits . In our experiment we do not have a direct measurement of G . However , we do observe how ϕ→ changes over the course of selection , our data suggest that ρ<0 and our reaction-diffusion model permits us to estimate how migration rate depends on the two traits of interest . In particular , β→=[∂log⁡ ( s ) ∂ ( \|υr\| ) , ∂log⁡ ( s ) ∂kg]T . We approximate β→ in both rich and minimal media by fitting a plane to the heatmap shown in Figure 2a–b ( Appendix 1 ) . The resulting selection gradient is shown in Figure 6—figure supplement 1 for both conditions . Using this formalism , we asked what values of σk~g and σ\|v~r\| would result in the directions of phenotypic evolution we observed experimentally in rich and minimal media . We found that the direction of phenotypic evolution in rich medium agreed well with our experimental observations so long as σ\|υ~r\|/σkg~≥1 for ρ<−0 . 1 . This implies that our observed phenotypic evolution is consistent with a genetic variance in run speed that is no smaller than the genetic variance in growth rate ( Figure 6—figure supplement 2 ) . In contrast , in minimal medium the model predicts the direction of observed phenotypic evolution only if σ\|υ~r\|/σkg~≲0 . 3 for ρ<−0 . 1 . This result indicates that our observed phenotypic evolution is consistent with at least three-fold larger propensity for mutations to alter growth rate compared to run speed in minimal medium ( Figure 6—figure supplement 2 ) . Figure 6 shows these geometric relationships between selection , genetic covariance and phenotypic evolution . 10 . 7554/eLife . 24669 . 030Figure 6 . Evolution of correlated traits . The evolutionary model describes the change in phenotype relative to the founder ( ϕ→=[\|v~r\| , k~g]T ) under selection described by β→ . Panels show unit vectors in the direction of observed phenotypic evolution ( ϕ^ ) and the direction of selection inferred from the reaction-diffusion model ( β^ ) . Ellipses show quartiles for a normal distribution of phenotypes with covariance matrix G that is consistent with ϕ→ and β→ . In both panels , we set the correlation coefficient between k~g and \|v~r\| is ρ=−0 . 75 but our conclusions hold for ρ<−0 . 1 . In rich medium ( a ) σ\|v~r\|/σk~g=1 and in minimal medium σ\|v~r\|/σk~g=0 . 3 . In rich medium β^R⁢M=[0 . 78 , 0 . 61] and in minimal medium β^M⁢M=[0 . 87 , 0 . 49] . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 03010 . 7554/eLife . 24669 . 031Figure 6—figure supplement 1 . Determining β→ from reaction-diffusion model . Reaction-diffusion model ( main text ) was used to simulate migration rates . Panels ( a ) and ( b ) plot the normalized ( to the founder ) predicted migration rate ( s~ ) for both rich medium ( a ) and minimal medium ( b ) . ( a–b ) are surface plots of the heatmaps shown in Figures 2 , 4 of the main text . To infer the selection pressure ( β→ ) we fit a plane ( black circles ) to the surfaces shown in ( a ) and ( b ) . The residuals of this fit are shown in ( c ) and ( d ) respectively . The fit for rich medium is good , while the residual is large in minimal medium . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 03110 . 7554/eLife . 24669 . 032Figure 6—figure supplement 2 . Direction of phenotypic evolution with σ\|v~r\| and σk~g . The dot product ϕ^o⁢b⁢s⋅ϕ^p⁢r⁢e⁢d is plotted as a heatmap as a function of genetic variances in growth rate and run speed . Each row corresponds to a different value of the correlation coefficient ( ρ ) between run speed and growth rate as labeled . The left column is for rich medium and the right column for minimal medium . When ϕ^o⁢b⁢s⋅ϕ^p⁢r⁢e⁢d→1 ( dark red ) this indicates regions where the predicted direction of evolution ( ϕ^p⁢r⁢e⁢d ) coincides with the observed direction of evolution ( ϕ^o⁢b⁢s ) . Note our qualitative conclusions are robust to large variation in ρ . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 03210 . 7554/eLife . 24669 . 033Figure 6—figure supplement 3 . Stochastic simulations of selection in minimal medium . Stochastic simulations of phenotypic evolution in minimal medium . Simulations were carried out as described above . For all simulations σ\|v~r\|=0 . 1 . Each colored line represents a single simulation which initiates at [1 , 1] . Each point is the mean phenotype for a round of selection . Colors represent different values of σk~g as shown in the legends . The green-yellow heatmap is the ‘fitness landscape’ interpolated from the heatmap shown in Figure 2b of the main text . Each panel shows a simulation for different , fixed , values of the trait correlation coefficient ρ . The red line and circles show the observed phenotypic evolution in minimal medium ( Figure 4b , main text ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 033 This suggests that the capacity of mutations to alter run speed or growth rate relative to founder depends on the nutrient conditions and that changes in this capacity qualitatively alter the direction of evolution along a Pareto frontier ( Shoval et al . , 2012 ) . This result captures the intuition that mutations that can increase growth rate in rich medium are few while in minimal medium the propensity for mutations increasing growth rate is substantially larger . The model presented here relies on a linear approximation to β→ , which is a good assumption for rich medium but not for minimal medium , where the dependence of s on \|υr\| and kg is strongly nonlinear . Using simulations of the evolutionary process described by Equation 3 , we relaxed the assumption of linearity in the selection coefficient and found that our qualitative conclusions were not altered ( Figure 6—figure supplement 3 ) . We note that the structure of G inferred above reflects the capacity for mutations to change phenotypes at the outset of the experiment . As evolution proceeds in rich medium , we observe a saturation in both run speed and growth rate ( Figure 4a ) , suggesting that further variation is constrained , either genetically or through biophysical constraints on swimming speed . Similarly , in minimal medium , saturation in the growth rate occurs after 5 rounds of selection , suggesting that mutations to further improve growth rate are either not available or fundamental constraints on growth inhibit further increases ( Scott et al . , 2010 ) . The most striking observation of our study is the divergent trajectories of phenotypic evolution shown in Figure 4a–b . This observation shows that the evolution of faster migration results in environmentally dependent phenotypic outcomes . This result has important implications for interpreting phenotypic variation in natural populations . When trade-offs are observed in wild populations , it is sometimes proposed that phenotypes at the extrema of a Pareto frontier reflect the outcome of selection for a specific task ( Shoval et al . , 2012 ) . Our study shows that when selection pressures place demands on multiple traits simultaneously , evolution along the frontier can reflect differing genetic capacity for adaptation of each phenotype rather than simply the fitness benefit of improving each trait . This result suggests a cautious approach to interpreting phenotypes in nature , where selection pressures and mechanisms constraining phenotypes are often not known ( Gould and Lewontin , 1979 ) . Our results point to the potential predictive power of determining the directions in phenotype space in which genetic variation can most readily change phenotypes – so called , ‘genetic lines of least resistance’ ( Schluter , 1996 ) . These directions may be related to genetic regulatory architecture . The mutations we observe in both rich and minimal media alter negative regulators ( a protease in the case of clpX and a transcriptional repressor in the case of galS ) . This supports the hypothesis that microevolution is dominated by the disruption of negative regulation ( Lind et al . , 2015 ) and suggests that the direction of phenotypic evolution can be predicted by determining where negative regulatory elements reside in genetic and proteomic networks . The mutations we examined appear to be more subtle than simple loss of function , since knockout mutants for both clpX and galS do not exhibit fast migration , therefore a detailed understanding of how mutations disrupt negative regulation will be essential . Previous experimental evolution studies have revealed a similar trade-off to the one presented here . Comparing the results of these studies to our own demonstrates the impact of how selection is performed on the phenotypic outcomes . For example , Yi and Dean , 2016 selected E . coli alternately for growth in well mixed conditions and chemotaxis using a capillary assay and observed a trade-off between growth rate and swimming speed which was circumvented by phenotypic plasticity . We observe no evolution beyond the Pareto frontier in our study , possibly because our conditions simultaneously select for growth and motility rather than alternating between selection pressures . This suggests that evolutionarily persistent trade-offs may reflect selection pressures that occur simultaneously in nature . In addition , van Ditmarsch et al . ( 2013 ) and Deforet et al . ( 2014 ) select Pseudomonas aeruginosa for a hyperswarming phenotype on hard agar . Rather than sampling from the population at a specific location in a swarming colony , they allow the population to swarm for a fixed time interval , remove the entire colony from the plate and inoculate a second plate from a mixed sample of the entire colony . This procedure likely selects both for swarming speed and for growth in the bulk of the colony . Phenotypically , hyperswarmers selected in this way exhibit a decline in growth rate and swimming speed in liquid and a deficit in biofilm formation ( van Ditmarsch et al . , 2013; Deforet et al . , 2014 ) . In light of our study , these results suggest that evolved phenotypes can depend on whether selection occurs at well defined spatial locations in a structured population ( e . g . migrating fronts ) or through periodic removal of spatial structure . A more precise understanding of the selection pressure applied by van Ditmarsch et al . might emerge from the application of Lande’s ( Lande , 1979 ) formalism to the observed genetic and phenotypic variation . Interestingly , both Yi and Dean ( 2016 ) and van Ditmarsch et al . ( 2013 ) observe mutations that alter regulation of motility and chemotaxis genes . None of the mutations observed in our experiment were found by Yi and Dean , despite evolution along similar Pareto frontiers . This suggests that determining the allowed directions of phenotypic variation may be a more powerful approach to predicting evolution than cataloging mutations alone . The mechanism of the trade-off between growth rate and swimming speed has , to our knowledge , not been determined . However , over-expression of motility operons could drive the reductions in growth rate we observe in rich medium . Subsequent increases in speed could then arise passively from reductions in cell size which reduce hydrodynamic drag ( Taheri-Araghi et al . , 2015 ) . Similarly , increases in growth rate in minimal medium should increase cell size and hydrodynamic drag . Using the data of Taheri-Araghi et al . ( 2015 ) , we estimated changes in cell size due to measured changes in growth rate for populations evolved in rich and minimal medium . We could not account for the large change in swimming speed we observe through growth rate mediated changes in cell size alone ( Appendix 1 ) . Since we have not measured cell size directly , we cannot conclusively rule out this mechanism . To definitively characterize the mechanism of this trade-off will require measurements of cell size , gene expression , flagellar length and proton motive force . Our study shows how evolutionary dynamics are defined by the complex interplay between genetic architecture , phenotypic constraints and the environment . Our hope is that a general approach to predicting evolution can emerge from a more complete understanding of this interplay . Webcam acquired images of migrating fronts were analyzed by custom written software ( Matlab , Mathworks , Natick , MA ) . A background image was constructed by median projecting six images from the beginning of the acquisition before significant growth had occurred . This image was subtracted from all subsequent images prior to further analysis . The location of the center of the colony was determined by first finding the edges of the colony using a Canny edge detection algorithm . A circular Hough transform ( Hough , 1959 ) was applied to the resulting binary image to locate the center . In rich medium , where signal to background was >10 , radial profiles of image intensity were measured from this center location and were not averaged azimuthally due to small departures from circularity in the colony . The location of the front was determined by finding the outermost peak in radial intensity profiles . Migration rate was determined by linear regression on the front location in time . Imaging was calibrated by imaging a test target to determine the number of pixels per centimeter . The results of the calibration did not depend on the location of the test target in the field of view . In minimal medium , where the signal to background is reduced due to low cell densities , background subtraction was employed as described above but radial density profiles were not always reliable for locating the front . Instead , a circular Hough transform was applied to each image to locate the front at each point in time . Single-layer microfluidic devices were constructed from polydimethyl-siloxane ( PDMS ) using standard soft-lithography techniques , ( Quake and Scherer , 2000 ) following a design similar to the one used previously ( Jordan et al . , 2013 ) , and were bonded to coverslips by oxygen plasma treatment ( Harrick plasma bonder , Harrick Plasma , Ithaca , NY ) . Bonded devices formed a circular chamber of diameter 200 μm and depth 10 μm ( Figure 3—figure supplement 1 ) . Devices were soaked in the medium used for tracking ( LB for rich medium strains , M63 0 . 18 mM galactose for minimal medium strains ) with 1% Bovine Serum Albumin ( BSA ) for at least 1 hr before cells were loaded . Bacteria were inoculated directly from frozen stocks into medium containing 0 . 1% BSA in a custom continuous culture device . BSA was necessary to minimize cells adhering to the glass cover slip . For rich medium tracking experiments , cells grew to a target optical density and the continuous culture device was run as a turbidostat . In minimal medium experiments the device was run as a chemostat at an optical density of ∼0 . 15 . The culture was stirred by a magnetic stir bar at 775 RPM and the temperature was maintained at 29 . 75°C by feedback . To perform single-cell tracking , cells were sampled from the continuous-culture device and diluted appropriately ( to trap one cell in the chamber at a time ) before being pumped into the microfluidic chamber . Video was acquired at 30 frames per second with a Point Grey model FL3-U3-32S2M-CS camera ( Point Grey , Richmond , Canada ) and a bright-field microscope ( Omano OM900-T inverted ) at 20x magnification . Movies were recorded for 5 min before a new cell was loaded into the chamber . An example movie is available at https://doi . org/10 . 13012/B2IDB-4912922_V2 . Two microscopes were operated in parallel . The stock microscope light source was replaced by a high-brightness white LED ( 07040-PW740-L , LED Supply , Randolf , VT ) to avoid 60 Hz flickering that was observed with the stock halogen light source . All experiments were performed in an environmental chamber maintained at 30°C . Movies were segmented and tracked with custom written Matlab routines described previously ( Jordan et al . , 2013; Jaqaman et al . , 2008 ) . Code is available at https://github . com/dfraebel/CellTracking ( Mickalide et al . , 2017; copy archived at https://github . com/elifesciences-publications/CellTracking ) . At times when two individuals are present in the chamber , ambiguous crossing events can lead to loss of individual identities . All crossing events were inspected manually to prevent this . To identify runs and tumbles , we utilized a method based on reference ( Taute et al . , 2015 ) which was modified from the approach used by Berg and Brown ( 1972 ) . Briefly , for each cell the segmentation routine results in a matrix of spatial locations x→⁢ ( t ) . We compute the velocity by the method of central differences resulting in v→⁢ ( t ) from which we compute an angular velocity between adjacent velocity vectors ( ω⁢ ( t ) ) . We then define α , a threshold on ω . Tumbles are initiated if ω⁢ ( t ) & ω ( t+1 ) >α or if ω⁢ ( t ) >α and the angle defined between the vectors x→⁢ ( t-2 ) -x→⁢ ( t ) and x→ ( t ) −x→ ( t+2 ) is greater than α . The latter condition detects tumbles that occur on the timescale of the imaging ( 0 . 033 s ) . Runs are initiated only when ω⁢ ( t ) & ω ( t+1 ) & ω ( t+2 ) <α . As a result , tumbles can be instantaneous and runs are a minimum of four frames . α was determined dynamically for each individual by initializing α0 and then detecting all runs for a cell . A new αi=c×median ( ωruns ) was computed with c a constant and ωr⁢u⁢n⁢s is the angular velocity during runs . The process was iterated ten times but typically converged to a final αf in less than five iterations . c=5 was determined by visual inspection of resulting classified trajectories . Approximately , 1% of cells exhibited sustained tumbling and had average tumble durations greater than 0 . 4 s and were excluded from further analysis . We only considered run events that were in the bulk of the chamber and were not interrupted by interactions with the circular boundary of the chamber . We computed tumble bias by measuring the total time spent tumbling when the cell was not interacting with the chamber boundaries . Tumble frequency was computed by counting the number of tumble events that occurred in the bulk of the chamber and dividing by the total time the cells spent swimming in the bulk . Tumble bias and frequency were computed for each individual over the duration tracked . Averages across individuals are reported in Figure 3c–d . Due to interactions with the chamber floor and ceiling ( boundaries perpendicular to the optical axis ) , we intermittently observed cells circling . We developed a method to detect this behavior automatically and found that our results are unchanged when we consider individuals that are not interacting with the chamber boundaries ( Appendix 1 ) . Data presented in the main text excludes cells determined to be circling . Whole genome sequencing was performed using the Illumina platform with slight variations between four independent runs . For all sequencing , cultures were grown by inoculating fresh medium from frozen stocks isolated during the course of selection and growing to saturation at 30°C . For sequencing of rich medium strains from replicate 1 , DNA was extracted and purified using a Bioo Scientific NEXTprep-Bacteria DNA Isolation Kit . Libraries were prepared from these strains with the Kapa HyperLibrary Preparation kit ( Kapa Biosystems , Wilmington MA ) , pooled and quantified by qPCR and sequenced for 101 cycles from each of the fragments on a HiSeq 2500 ( Illumina , San Diego , CA ) . This HiSeq run was performed by the Biotechnology Core Facility at the University of Illinois at Urbana-Champaign and included additional strains not presented here . All other sequencing was performed on a locally operated and maintained Illumina MiSeq system . For MiSeq runs which generated data for all minimal medium evolved strains and replicates 2 to 4 of the rich medium selection experiments , DNA was extracted with either the Bioo Scientific NEXTprep kit or the MoBio Ultraclean Microbial DNA isolation kit . Different isolation kits were used due to the discontinuation of the Bioo Scientific kit . DNA was quantified by Qubit and Bioanalyzer and libraries were prepared using the NexteraXT kit from Illumina . Sequencing adapters for the HiSeq generated data were trimmed using flexbar ( http://sourceforge . net/projects/flexbar/ ) . MiSeq runs were demultiplexed and trimmed using the onboard Illumina software . Analysis was performed using the breseq platform http://barricklab . org/twiki/bin/view/Lab/ToolsBacterialGenomeResequencing in polymorphism mode . Breseq uses an empirical error model and a Bayesian variant caller to predict polymorphisms at the nucleotide level . The algorithm uses a threshold on the empirical error estimate ( E-value ) to call variants ( Barrick and Lenski , 2009 ) . The value for this threshold used here was 0 . 01 , and at this threshold , with the sequencing coverage for our samples , we report all variants present in the population at a frequency of 0 . 2 or above ( Barrick and Lenski , 2009 ) . All other parameters were set to their default values . Reads were aligned to the MG1655 genome ( INSDC U00096 . 3 ) . We note that breseq is not well suited to predicting large structural variation . Since we sequence populations at different points during selection , observation of the same mutations at different points in time significantly reduces the probability of false positives ( Lang et al . , 2013 ) . The founder strain was sequenced at an average depth of 553× when aggregating reads from four separate sequencing reactions . Any mutations observed in this strain were excluded from further analysis . Tables 5–12 document mutations , important mutations were confirmed by Sanger sequencing as noted in the captions to these tables . Since these genomes were sequenced at very high depth , we did not confirm every mutation by Sanger sequencing . All mutation calls made by breseq were inspected manually and found to be robust or they were excluded . We also manually inspected the founder strain reads aligned to regions where frequent mutations were observed in the evolved strains ( clpX E185 , the Δ1 bp mutation at position 523086 and galS L22R ) to confirm that those mutations were not present in the founder . Sequencing data are available at https://doi . org/10 . 13012/B2IDB-3958294_V1 . 10 . 7554/eLife . 24669 . 034Table 9 . Rich medium replicate 1: All mutations detected above a frequency of 0 . 2 in rounds 5 , 10 and 15 of rich medium selection replicate 1 . The first number in each cell denotes the distance in base pairs from ori ( location ) . The second entry ( mutation ) identifies the mutations with ‘IG’ denoting an intergenic mutation . The third entry ( fraction ) is the fraction of the population carrying this mutation ( as inferred by breseq in polymorphism mode ) . The fourth entry ( coverage ) is the number of reads that aligned to this location . In the round 15 strain , the clpX SNP and Δ1 bp deletion at position 523 , 086 were confirmed by Sanger sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 034Rich medium replicate 1Round , ( coverage ) 5 , ( 172× ) 10 , ( 213× ) 15 , ( 180× ) Mutation ( loc . , mut . , frac . , cov . ) 457978 , clpX E185* , 75 . 2% , 179457978 , clpX E185* , 100% , 199457978 , clpX E185* , 100% , 164523086 , IG Δ1 bp , 100% , 194523086 , IG Δ1 bp , 100% , 266523086 , IG Δ1 bp , 100% , 168950518 , pflA T188I , 22 . 2% , 144990379 , IG A→C , 100% , 201663115 , dacA Δ1 bp , 100% , 1501978458 , IG G→T , 21 . 2% , 156990379 , IG A→C , 100% , 1563618863 , nikR H92H , 20 . 7% , 18910 . 7554/eLife . 24669 . 035Table 10 . Rich medium replicate 2: All mutations detected in rounds 5 , 10 and 15 of rich medium replicate 2 . See Table 9 caption . Note low coverage on Δ1 bp mutation at 523086 noted in bold . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 035Rich medium replicate 2Round , ( coverage ) 5 , ( 218× ) 10 , ( 100× ) 15 , ( 166× ) Mutation ( loc . , mut . , frac . , cov . ) 457978 , clpX E185* , 100% , 220457978 , clpX E185* , 100% , 109457978 , clpX E185* , 100% , 184950518 , pflA T188I , 27 . 2% , 210523086 , IG Δ1 bp , 100% , 16523086 , IG Δ1 bp , 100% , 24523086 , IG Δ1 bp , 100% , 10/18667259 , mrdA R320H , 39 . 5% , 159794472 , modE L58* , 42 . 4% , 13610 . 7554/eLife . 24669 . 036Table 11 . Rich medium replicate 3: All mutations detected in rounds 5 , 10 and 15 of rich medium replicate 3 . See Table 9 caption . Note low coverage on Δ1 bp mutation at 523086 noted in bold . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 036Rich medium replicate 3Round , ( coverage ) 5 , ( 291× ) 10 , ( 45× ) 15 , ( 186× ) Mutation ( loc . , mut . , frac . , cov . ) 457978 , clpX E185* , 100% , 300457978 , clpX E185* , 100% , 43457978 , clpX E185* , 100% , 185523086 , IG Δ1 bp , 50% , 38523086 , IG Δ1 bp , 100% , 8523086 , IG Δ1 bp , 100% , 16950518 , pflA T188I , 26 . 3% , 332950518 , pflA T188I , 53 . 3% , 53950518 , pflA T188I , 30 . 6% , 190321263 , IG T→C , 25% , 161968653 , cheR Q238K , 29 . 6% , 190382794 , yaiX +9bp insertion , 64% , NA382794 , yaiX +9bp insertion , 25 . 8% , NA4161562 , fabR Δ⁢17bp , 46 . 2% , 6710 . 7554/eLife . 24669 . 037Table 12 . Rich medium replicate 4: All mutations detected in rounds 5 , 10 and 15 of rich medium replicate 4 . See Table 9 caption . Note low coverage on Δ1 bp mutation at 523086 noted in bold . DOI: http://dx . doi . org/10 . 7554/eLife . 24669 . 037Rich medium replicate 4Round , ( coverage ) 5 , ( 384× ) 10 , ( 555× ) 15 , ( 333× ) Mutation ( loc . , mut . , frac . , cov . ) 457978 , clpX E185* , 100% , 370457978 , clpX E185* , 100% , 559457978 , clpX E185* , 100% , 339523086 , IG Δ1 bp , 50% , 72523086 , IG Δ1 bp , 100% , 34/83523086 , IG Δ1 bp , 100% , 19/33950518 , pflA T188I , 31 . 7% , 4463619915 , rhsB W242G , 24 . 9% , 20 Knockout mutants ( ΔclpX , ΔgalS ) were constructed by P1 transduction from KEIO collection mutants ( Baba et al . , 2006 ) . Mutations were confirmed by PCR . Antibiotic markers were not removed prior to phenotyping . Three commonly observed single nucleotide polymorphisms ( SNPs ) observed across evolution experiments were reconstructed in the chromosome of the ancestral background ( founder ) using a recombineering method presented previously ( Kuhlman and Cox , 2010; Tas et al . , 2015 ) . These mutations were the clpXE185* mutation , the single base pair deletion between ybbP and rhsD ( which we refer to as ‘Δ⁢1bp’ ) and galSL22R . For full details of the recombineering we performed see Appendix 1 . Briefly , recombineering proficient cells were prepared by electroporation of the helper plasmid pTKRED ( Kuhlman and Cox , 2010 ) and selection on spectinomycin . A linear ‘landing pad’ fragment consisting of tetA flanked by I-SceI restriction sites and homologies to the desired target site was synthesized from the template plasmid pTKLP-tetA and site specific primers . The landing pad was inserted by electroporation into recombineering proficient cells and transformants were selected by growth on tetracycline . Successful transformants were confirmed by PCR . A second transformation was then performed using a 70 bp oligo containing the desired mutation near the center and flanked by homologies to target the landing pad . Counterselection for successful transformants was performed with NiCl2 ( 6 mM for the ClpX and GalS mutations , 6 . 5 mM for Δ1 bp ) . Successful recombination at this step resulted in removal of the landing pad and integration of the 70 bp oligo containing the desired mutation . The helper plasmid pTKRED was cured by growth at 42°C and confirmed by verifying spectinomycin susceptibility . The presence of desired mutations in the final constructs was confirmed by Sanger sequencing .	In nature organisms face many challenges , and species adapt to their environment by changing heritable traits over the course of many generations . How organisms adapt is often limited by trade-offs , in which improving one trait can only come at the expense of another . In the laboratory , scientists use well-controlled environments to study how populations adapt to specific challenges without interference from their natural habitat . Most experiments , however , only look at simple challenges and do not take into account that organisms in the wild face many pressures at the same time . Fraebel et al . wanted to know what happens when an organism’s performance depends on two traits that are restricted by a trade-off . The experiments used populations of the bacterium Escherichia coli , which can go through hundreds of generations in a week , providing ample opportunity to study mutations and their impact on heritable traits . Through a combination of mathematical modeling and experiments , Fraebel et al . found that the environment is crucial for determining how bacteria adapt when their swimming speed and population growth rate are restricted by a trade-off . When nutrients are plentiful , E . coli populations evolve to spread faster by swimming more quickly despite growing more slowly . Yet , if nutrients are scarcer , the bacteria evolve to spread faster by growing more quickly despite swimming more slowly . In each scenario , the experiments identified single mutations that changed both swimming speed and growth rate by modifying regulatory activity in the cell . A better understanding of how an organism’s genetic architecture , its environment and trade-offs are connected may help identify the traits that are most easily changed by mutations . The ultimate goal would be to be able to predict evolutionary responses to complex selection pressures .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "computational", "and", "systems", "biology" ]	2017	Environment determines evolutionary trajectory in a constrained phenotypic space
Innate lymphoid cells ( ILCs ) contribute to host defence and tissue repair but can induce immunopathology . Recent work has revealed tissue-specific roles for ILCs; however , the question of how a small population has large effects on immune homeostasis remains unclear . We identify two mechanisms that ILC3s utilise to exert their effects within intestinal tissue . ILC-driven colitis depends on production of granulocyte macrophage-colony stimulating factor ( GM-CSF ) , which recruits and maintains intestinal inflammatory monocytes . ILCs present in the intestine also enter and exit cryptopatches in a highly dynamic process . During colitis , ILC3s mobilize from cryptopatches , a process that can be inhibited by blocking GM-CSF , and mobilization precedes inflammatory foci elsewhere in the tissue . Together these data identify the IL-23R/GM-CSF axis within ILC3 as a key control point in the accumulation of innate effector cells in the intestine and in the spatio-temporal dynamics of ILCs in the intestinal inflammatory response . Innate lymphoid cells ( ILCs ) are a recently defined family of evolutionarily ancient cells involved in many facets of host defence . As with conventional Th1 , Th2 and Th17 T cells , ILCs can be functionally classified based on the expression of transcription factors and associated signature cytokines ( Spits et al . , 2013 ) . ILC1 are defined by Th1-like and ILC2 by Th2-like cytokine responses , whereas ILC3 express RORγt , and can secrete IL-17 and/or IL-22 upon activation . The ILC3 family includes the prototypic foetal lymphoid tissue inducer ( LTi ) cells that play a non-redundant role in lymphoid tissue development ( Mebius et al . , 1997 ) via lymphotoxin-dependent interactions with stromal cells . This pathway is required not only for lymph node development but also for organised lymphoid structures in the gut such as dendritic cell ( DC ) and ILC containing cryptopatches ( CP ) ( Eberl and Sawa , 2010 ) , B cell , DC and ILC containing isolated lymphoid follicles ( ILFs ) ( Tsuji et al . , 2008 ) , and small intestinal Peyer’s Patches ( PP ) that have a very similar structure to lymph nodes ( Cornes , 1965; Mebius et al . , 1997 ) . Postnatally , the LTi population can also make IL-17 and IL-22 ( Cupedo et al . , 2009 ) and contribute to host defence against pathogens , particularly in the gut ( Sonnenberg et al . , 2011 ) . The multi-functional roles of ILCs in disease development and pathogenesis ( Buonocore et al . , 2010 ) , as well as host defence ( Moro et al . , 2010; Neill et al . , 2010; Sonnenberg et al . , 2011 ) and repair ( Monticelli et al . , 2011 ) have been the focus of much interest . Polyfunctional ILCs have also been described that do not fall into distinct ILC1 or ILC3 phenotypes but express both ILC1 and ILC3 lineage defining transcription factors Tbet and RORγt and secrete multiple cytokines such IL-17A , IL-22 as well as IFNγ ( Buonocore et al . , 2010 ) . We previously identified a critical role for a phenotypically distinct population of IL-23R+ RORγt+ CD4- NKp46- ILCs in the development of innate intestinal inflammation in Rag-/- mice following Helicobacter hepaticus infection or αCD40 stimulation ( Buonocore et al . , 2010 ) . Similar ILC populations were enriched in the colonic mucosa of patients with inflammatory bowel disease ( IBD ) ( Geremia et al . , 2011 ) , implicating IL-23-responsive , RORγt expressing ILCs in the pathogenesis of inflammatory gut disease in mice and humans . However , it remains unclear how ILCs , that are numerically sparse in vivo can initiate inflammatory processes that lead to colitis . Despite advances in understanding of the functions of ILCs , little is known about their location in tissue at different stages of the inflammatory response , and how putative structural and cytokine-mediated functions are co-ordinated . Since its description in 2006 ( Uhlig et al . , 2006 ) , the induction of colitis by injecting agonistic anti-CD40 antibody has become an important tool to assess ILC-driven acute colitis ( Buonocore et al . , 2010; Vonarbourg et al . , 2010; Fuchs et al . , 2013; Kim et al . , 2013; Song et al . , 2015 ) . By contrast with other models , anti-CD40 induced colitis follows discrete phases at well-defined time points following initiation , offering the opportunity to probe the role of leukocytes in the development and amplification of the inflammatory response . Experiments have demonstrated that intestinal inflammation was mediated via Thy1+ ILCs in a rorc dependent manner , making it an ideal system to study how ILCs contribute to pathogenesis ( Buonocore et al . , 2010 ) . A recent study investigating potential biomarkers for anti-IL-23 therapy described similar changes in the colons of both anti-CD40-treated mice and patients with active Crohn’s disease ( Cayatte et al . , 2012 ) . Many recent publications have focused on the specific functions of ILC subsets within effector sites , and the location of ILCs has been proposed to contribute to their ability to affect systemic cytokine levels ( Nussbaum et al . , 2013 ) . Despite histological and flow cytometry data demonstrating the presence of ILCs within lymphoid structures in the gut ( Eberl and Sawa , 2010 ) , it isn’t clear whether they function as sedentary , cytokine producing cells or play a more active role in cell interactions and organization . In vivo microscopy is a tool that provides an opportunity to look at the behaviour of ILCs within the tissue . By combining anti-CD40 stimulation with intra-vital microscopy we are able to reliably track cellular changes at discrete phases of disease and capture cell movement at key timepoints . Our results show two novel mechanisms through which the small number of ILCs found in vivo orchestrate the intestinal inflammatory response . IL-23-driven GM-CSF production by ILC3s is critical for the development of colitis , and ILCs mobilise from cryptopatches after activation in a GM-CSF-dependent manner . Both of these behaviours likely contribute to the ability of ILCs to coordinate the immune response in the gut . Initiation and perpetuation of disease occur in distinct anatomical compartments , indicating both a temporal and spatial switch of ILC function during inflammatory conditions . Anti-CD40 induced colitis is dependent on a RORγt/IL-23 axis but key downstream cytokines are less well understood ( Uhlig et al . , 2006; Buonocore et al . , 2010 ) . As IL-17 and IL-22 are major downstream effectors of the IL-23 signalling axis ( Zheng et al . , 2007; McGeachy et al . , 2009 ) we first investigated their role in anti-CD40 colitis . However , blockade of IL-17A failed to modify anti-CD40-induced systemic or intestinal disease ( Figure 1A , B ) , indicating that IL-17A is dispensable for development of acute colitis in this model . Blocking the closely related molecule IL-17F also failed to modify disease ( Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 10066 . 003Figure 1 . GM-CSF is a critical cytokine mediator of ILC-driven colitis . ( A ) Weight loss and ( B ) proximal colon histopathology scores in untreated B6Rag1-/- mice ( control , n=3 ) or mice injected with anti-CD40 and treated with blocking antibody to IL-17A ( n=5 ) , IL-22 ( n=7 ) or isotype ( n=8 ) for 7 days . Representative photomicrographs of H&E stained proximal colon sections are shown . ( C ) Weight loss and ( D ) proximal colon histopathology scores in untreated B6Rag1-/- mice ( control , n=8 ) or mice injected with anti-CD40 and treated with blocking antibody to GM-CSF ( n=7 ) , blocking antibody to IL-23R ( n=5 ) or isotype ( n=6 ) for 7 days . Representative photomicrographs of H&E stained proximal colon sections are shown . ( E ) Number of innate immune cell populations in cLP at three days following anti-CD40 injection . Data are shown as means and SEM . Results are representative of n=2–5 independent experiments . * , p<0 . 05 , , p<0 . 01 , One-way ANOVA with Bonferroni’s post test . ( Figure 1—figure supplement 1 ) shows no effect of IL-17A and IL-22 double blockade and IL-17F blockade on systemic disease or colitis . ( Figure 1—figure supplement 2 ) shows improved systemic and intestinal disease with anti-GM-CSF treatment in Helicobacter hepaticus driven innate colitis . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 00310 . 7554/eLife . 10066 . 004Figure 1—figure supplement 1 . IL-17A and IL-22 combination blockade or anti-IL-17F does not protect from anti-CD40 mediated colitis . ( A ) Weight loss and ( B ) proximal colon histopathology scores in untreated B6Rag1-/- mice ( control , n=6 ) or mice injected with anti-CD40 and treated with blocking antibody to IL-17F ( n=5 ) , a combination of IL-17A and IL-22 blockade ( n=6 ) or isotype ( n= 6 ) . Representative photomicrographs of H&E stained proximal colon sections are shown . Data are shown as mean and SEM . Results are a combination of 2 independent experiments . One-way ANOVA with Bonferroni’s post test . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 00410 . 7554/eLife . 10066 . 005Figure 1—figure supplement 2 . Helicobacter hepaticus driven innate colitis depends on GM-CSF . ( A ) Spleen weight from uninfected ( n=4 ) , isotype treated H . h , infected ( n=7 ) , or anti-GM-CSF treated H . h , infected mice ( n=7 ) 6 weeks after infection . ( B ) Colitis score of uninfected ( n=4 ) , isotype treated H . h , infected ( n=6 ) , or anti-GM-CSF treated H . h , infected mice ( n=7 ) 6 weeks after infection . Data are shown as mean and SEM . Results are representative of 2 independent experiments . * , p<0 . 001 . One-way ANOVA with Bonferroni’s post testDOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 005 Like IL-17 blockade , treatment with a neutralising anti-IL-22 mAb had no effect on systemic disease or colitis in the proximal colon ( Figure 1A&B ) , indicating that in our facility , IL-22 is also redundant for disease induction in this model . However , blockade of IL-17A and anti-IL-22 reduced intestinal inflammation but not systemic disease indicating redundant roles for IL-17A and IL-22 in the intestinal inflammatory response ( Figure 1—figure supplement 1 ) . As we have described an important role for GM-CSF in IL-23 driven T cell dependent colitis we next investigated the role of GM-CSF in innate colitis models . Strikingly , administration of an anti-GM-CSF blocking mAb reduced both weight loss ( Figure 1C ) and severity of colitis ( Figure 1D ) . Indeed amelioration of colitis with GM-CSF blockade was similar to that observed with anti-IL-23R mAb treatment . Similar effects of GM-CSF blockade were also observed in bacteria-induced innate colitis following Helicobacter hepaticus infection of 129SvEv Rag2-/- mice ( Figure1—figure supplement 2 ) supporting a pivotal role for this cytokine in both T cell dependent and innate colitis . GM-CSF has been shown to promote CNS inflammation through effects on inflammatory monocytes ( Croxford et al . , 2015 ) . Consistent with this and in line with a recent report ( Song et al . , 2015 ) GM-CSF blockade led to a marked reduction in the number of inflammatory monocytes in the colon of anti-CD40 treated mice ( Figure 1E ) . The number of neutrophils and eosinophils also decreased , but this was not significant . These changes were accompanied by an increase in both the percentage and total number of CD11b- CD103+ dendritic cells ( Figure 1E ) . We next investigated GM-CSF expression by flow cytometry to identify the cellular source in intestinal inflammation . Only a small proportion of epithelial , stromal or lineage positive ( CD11b , CD11c , Gr-1 , B220 , CD49b ) cells expressed GM-CSF ( Figure 2A ) . The majority of these lineage positive GM-CSF producers were NK cells , and they expressed a lower amount of GM-CSF than ILCs ( Figure 2—figure supplement 1A & B ) . However , using Rag and IL-15 receptor double deficient mice that lack NK cells ( Lodolce et al . , 1998 ) , we found , as previously reported ( Vonarbourg et al . , 2010 ) , that NK cells were not required for the development of anti-CD40 mediated colitis ( Figure 2—figure supplement 1C–F ) . By contrast , a significant proportion of ILCs were capable of producing GM-CSF , even in the healthy intestine , and this proportion increased during colitis ( Figure 2B ) . Consistent with an IL-23-dependent ILC3 phenotype , GM-CSF+ ILCs were CCR6- , cKit- and NKR- , and expressed only low amounts of T-bet ( Figure 2A ) . 10 . 7554/eLife . 10066 . 006Figure 2 . ILCs are a major source of GM-CSF . ( A ) Representative flow cytometric analysis of ILC populations at day 3 following anti-CD40 treatment . Surface marker and GM-CSF expression is shown following 3 hr stimulation with PMA , ionomycin , monensin , and brefeldin A ( n=6 ) . Cells are gated on a live cell gate with doublets excluded . GM-CSF expression gating is based on an FMO control . Red marks ILCs and black marks NK cells . Lower left panel shows IL-7Rα staining in live CD45+ cells compared with FMO control . ( B ) Representative flow cytometric analysis showing GM-CSF production for ILC populations in untreated or anti-CD40 treated mice . The solid grey histogram represents the isotype control . Following doublet exclusion , cells are gated on a live cell gate , then gated on lineage- , CD45+ Thy1 . 2+ IL-7Rα+ cells . ( C ) Percent and median fluorescence intensity of GM-CSF expression in CD4- NKp46- ILCs isolated from mice 7d following anti-CD40 injection . Total cLP cells were stimulated over night with IL-12 , IL-23 or in medium alone , followed by 3 hr stimulation with PMA , ionomycin , monensin and brefeldin A ( n=7 ) . Results are representative of 2–3 independent experiments . ( D ) mRNA expression of Il23a and Csf2 in proximal colonic lamina propria at various time points following anti-CD40 injection . Results are shown as fold change in target gene relative to hprt compared with day 0 uninjected control mice . Data are shown as means and SEM . Results are pooled from 2 independent experiments with n=4–6 mice per group per experiment . * , p<0 . 05 , ** , p<0 . 01 , one-way ANOVA with Bonferroni’s post test . ( Figure 2—figure supplement 1 ) shows analysis of lack of NK cell contribution to colitis . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 00610 . 7554/eLife . 10066 . 007Figure 2—figure supplement 1 . NK cells are not required for anti-CD40 induced colitis . ( A ) Representative flow cytometry plot of GM-CSF+ cells showing lineage+ cells are NKp46+ . ( B ) Representative flow cytometric analysis of GM-CSF producing cells gated on single , live , CD45+ cells showing that non-ILCs are NK cells based on perforin expression and that total NK cells produce less GM-CSF than total ILCs . ( C ) Representative NK cell gating from B6Rag1-/- and B6Rag1-/-Il15ra-/- mice . ( D ) Percent NK1 . 1 positive cells of live leukocytes in B6Rag1-/- and B6Rag1-/-Il15ra-/- mice . ( E ) Weight loss and ( F ) proximal colon histopathology scores in untreated B6Rag1-/- mice ( n=13 ) or B6Rag1-/-Il15ra-/- ( n=12 ) mice injected with anti-CD40 . Representative photomicrographs of H&E stained proximal colon sections are shown . Results are a combination of 3 independent experiments . * , p<0 . 05 Mann Whitney . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 007 We next investigated the cytokine milieu that regulates GM-CSF production by ILCs in vivo . To determine whether IL-23 signalling induces GM-CSF , we cultured total colonic lamina propria cells overnight with IL-12 or IL-23 and measured GM-CSF production in the cultures . IL-12 had no effect on either the proportion of GM-CSF-expressing ILCs or the amount of GM-CSF per cell ( Figure 2C ) . By contrast and contrary to a previous study using sorted ILC3s ( Mortha et al . , 2014 ) , culture with IL-23 increased both the proportion of GM-CSF expressing ILCs as well as GM-CSF production per cell ( Figure 2C ) . Analysis of Il23a mRNA expression within colonic tissue revealed a peak in expression at day 1 following anti-CD40 treatment that rapidly declined but was followed by increased expression of Csf2 , the gene for GM-CSF , at day 3 ( Figure 2D ) . The increase in both Il23a and Csf2 was abrogated in mice treated with a blocking antibody to GM-CSF , indicating that GM-CSF may be required to maintain and amplify expression of Il23a . To assess whether GM-CSF production may also be relevant in human IBD , we first analyzed GM-CSF production from blood ILCs . PMA and ionomycin stimulated blood ILC3s produced GM-CSF with very little production from other ILC subsets ( Figure 3A&B , gating strategy in Figure 3—figure supplement 1 ) . Indeed , a greater proportion of ILCs produced GM-CSF than did T cells from the same healthy donors ( Figure 3C ) . ILCs secreting GM-CSF are further enriched in the colon ( Figure 3D ) . Critically , the proportion of blood ILCs capable of GM-CSF production along with other disease-associated cytokines such as IFNγ and TNFα was greater in IBD patients than in healthy controls ( Figure 3E ) . 10 . 7554/eLife . 10066 . 008Figure 3 . Human ILCs are a source of GM-CSF , which increases in IBD . ( A ) Representative flow cytometric analysis of GM-CSF in human blood ILC subsets ( following doublet exclusion and gated on live cells , ILCs are lineage- . ILC1 are IL-7Rα+ cKit- CRTH2- , ILC2 are IL-7Rα+ CRTH2+ , ILC3 are IL-7Rα+ cKit+ CRTH2- . cNK cells are CD56+ IL-7Rα- CD45RO- ) stimulated with PMA , ionomycin and brefeldin A for 4 hr . ( B ) Quantification of percent of GM-CSF+ blood ILC populations ( n=3–6 ) . ( C ) Percent of human blood and colon ILCs and T cells expressing GM-CSF following stimulation with PMA , ionomycin and brefeldin A for 4 hr ( n=5 ) . ( D ) Comparison of ILCs in blood and colon expressing GM-CSF ( n=5 ) . ( E ) Percent of ILCs expressing pro-inflammatory cytokines in the blood of control and IBD patients following stimulation with PMA , ionomycin and brefeldin A for 4 hr ( n=10–15 ) . Results are representative of n=2–5 independent experiments . * , p<0 . 05 , , p<0 . 01 , * , p<0 . 001 , *** , p<0 . 0001 one-way ANOVA with Bonferroni’s post test . ( F ) Relative expression of CSF2 from the publicly available Leuven cohort ( GSE16879 ) in control , Crohn’s disease , and ulcerative colitis patients before infliximab treatment . ( G ) Relative expression of CSF2 in the Oxford IBD cohort . Data for individual genes were normalized to the median value of healthy control patients , converted to log2 ratios , and analyzed by one-way ANOVA with Tukey’s multiple comparisons test ( data were found to be normally distributed using the D’Agostino and Pearson omnibus normality test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 00810 . 7554/eLife . 10066 . 009Figure 3—figure supplement 1 . Gating strategy for human ILC subsets . ( A ) Representative flow cytometric analysis of human blood ILC subsets . Doublet exclusion and live cell gate , ILCs are lineage- . ILC1 are IL-7Rα+ cKit- CRTH2- , ILC2 are IL-7Rα+ CRTH2+ , ILC3 are IL-7Rα+ cKit+ CRTH2- . ( B ) Representative staining for GM-CSF and TNFα within ILC subsets stimulated with PMA , ionomycin and brefeldin A for 4 hr ( n=3–6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 009 Analysis of a publically available dataset of colon tissue from healthy controls and IBD patients showed a significant increase in CSF2 gene expression in the colons of patients with either Crohn’s disease or ulcerative colitis compared with controls ( Figure 3F ) . This was confirmed by gene expression analysis of colon biopsies from the Oxford cohort of IBD patients ( Figure 3G ) . CSF2 expression was also higher in uninflamed biopsies from IBD patients compared with controls . There was significantly greater expression again in lesional areas in active disease , suggesting that GM-CSF expression correlates with , and may be a driver of , inflammation in IBD . To date attention has focussed on the capacity of ILCs in tissue to produce immune modulatory cytokines with little emphasis on how ILC positioning within a tissue may impact on functional outcomes . Under non-inflammatory conditions , ILCs are found within lymphoid structures in the colon ( Eberl and Sawa , 2010 ) , and in Rag-deficient hosts these structures are limited to cryptopatches as there are no B cells to form ILFs . Under homeostatic conditions ILCs primarily reside within lymphoid aggregate cryptopatch ( CP ) structures ( Figure 4A ) and are found only rarely within the non-inflamed lamina propria . Within CP , RORγt+ and IL-7Rα+ ILC3s are present at high density surrounded by CD11c and MHCII expressing cells . Some of these ILCs within CP also express CD4 , suggesting that some but not all are classical LTi cells that are known to be involved in lymphoid organogenesis ( Figure 4A ) . In addition to RORγt expression , ILCs are known to express IL-23R ( Buonocore et al . , 2010 ) . FACS analysis of Rag-deficient mice that express GFP under control of the Il23r demonstrates that these mice can be used to study ILC behaviour in vivo ( Figure 4B ) . Two-photon imaging of Il23rgfp/+ Rag-/- mice shows ILCs ( green ) present within a cryptopatch ( Figure 4C , and Video 1 ) . Quantification of ILC localization in 3D indicates that greater than 90% of ILCs are present within cryptopatches under steady-state conditions ( Figure 4D ) . This suggests that IL-23R+ ILCs are perfectly positioned in close proximity to a high density of CD11c+ DCs that can be activated by colitogenic stimuli . Indeed Il23r mRNA increased in the colon 6 hr following anti-CD40 injection indicating that the first ILC changes occur in the hours following CD40 stimulation ( Figure 4—figure supplement 1A ) . 10 . 7554/eLife . 10066 . 010Figure 4 . IL-23R marks ILCs that are present in cryptopatches within the gut . ( A ) Representative H&E and immunofluorescence staining of cryptopatches in transverse proximal colon sections in B6Rag1-/- mice from the steady state at 2 . 5x and 20x magnification . ( B ) Flow cytometry staining of Thy1 . 2 in the colon LPLs of steady-state Il23rgfp/+ Rag1-/- . ( C ) Representative image of intact tissue explant of proximal colon from Il23rgfp/+ Rag1-/- mice from the steady state . Left and middle panels show IL-23R ( green ) and collagen ( blue ) . Right panel shows IL-23R alone ( top ) and collagen alone ( bottom ) . ( D ) Quantification of steady-state Il23rgfp/+ Rag1-/- ILCs in clusters within the proximal colon from explant imaging . ( Figure 4—figure supplement 1 ) shows modulation of RORγt and IL-23R after anti-CD40 treatment . Results are representative of 3-4 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01010 . 7554/eLife . 10066 . 011Figure 4—figure supplement 1 . RORγt and IL23R expression are affected by anti-CD40 treatment . ( A ) Tissue Il23r expression relative to Hprt ( n=6 ) . This is a combination of 2 independent experiments . ( B ) Representative flow cytometric analysis of IL-23R ( using Il23rgfp/+ Rag1-/- mice ) on cells gated on live cells , CD45+ , IL-7Rα+ , and Thy1 . 2+ 0 and 3 days after anti-CD40 treatment . ( C ) Quantification of IL-23R+ cells as a percent of single , live , CD45+ , IL-7Rα+ , Thy1 . 2+ cells ( n=4 ) . This is representative of 2 independent experiments . ( D ) Representative flow cytometric analysis of RORγt staining in cells gated on live , single , CD45+ , IL-7Rα+ , and Thy1 . 2+ 0 and 3 days after anti-CD40 treatment . ( E ) Quantification of the percent of RORγt+ cells relative to Day 0 ( n=8–12 ) . This is a combination of 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01110 . 7554/eLife . 10066 . 012Video 1 . 3D rotation of Il23rgfp/+ Rag1-/- ILCs ( green ) within the collagen matrix of the gut ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 012 We further examined the changes in IL-23R expression following anti-CD40 stimulation , and found a decrease in IL-23R+ ILCs at day 3 ( Figure 4—figure supplement 1B&C ) , which may indicate a change in ILC phenotype . In support of this hypothesis , analysis of RORγt protein revealed a transient decrease in the proportion of ILCs expressing it 3 days following anti-CD40 stimulation ( Figure 4—figure supplement 1D&E ) , similar to a previous report ( Vonarbourg et al . , 2010 ) . Whilst these data suggested that ILCs are reacting to changes in the environment , it remained unclear how this leads to the downstream inflammatory cascade . It is possible that ILCs that are resident within lymphoid structures produce cytokines locally to carry out their effects . Alternatively , ILCs may be able to mobilize into the tissue to coordinate the immune response . To understand which mechanism ILCs utilize , we adopted the strategy of McDole et al ( McDole et al . , 2012 ) used to study leukocytes in the gut . In the steady-state , ILCs were much more motile than expected . While little motility could be observed within the cryptopatch , ILCs can be observed to enter and exit the cluster near the basal collagen layer ( Figure 5A and Videos 2 and 3 ) . Interestingly , the cryptopatch did not have an obvious point of entry or exit , implying that trafficking into and out of the structure may be governed by mechanisms different to that of the lymph node . By gating on motile cells outside the cluster , we could see that the average velocity of motile ILCs ( approximately 9 μm/min ) is grossly similar to that of other lymphocytes and while cell velocity decreases after treatment with anti-CD40 ( Figure 5B ) , this difference does not appear to affect overall cell trafficking . To determine whether ILC trafficking changed in colitis , we plotted the displacements of ILCs in steady-state and after anti-CD40 treatment ( Figure 5C , Video 4 , and Video 5 ) . While no gross changes appeared in track displacements , the overall movement of cells shifted . In steady-state , roughly equal numbers of cells move toward and away from the cluster; however , just 4–6 hr after anti-CD40 treatment , a greater proportion of cells exit the tissue and a smaller proportion enter ( Figure 5D ) . The ratio of entering to exiting cells indicates a significant skewing toward egress from cryptopatches ( Figure 5E ) . To assess whether GM-CSF may be playing a role in ILC mobilization , we treated mice with anti-GM-CSF 24 hr before imaging . The shift toward exit that is normally observed after anti-CD40 treatment was strikingly reversed when mice were treated with a GM-CSF blocking antibody 24 hr before imaging ( Figure 5D and E , Video 6 ) . It is possible that many of these cells move through blood vessels , but by labelling the blood vessels with texas-red dextran , ILCs appeared to be adjacent to blood vessels , not within them ( Figure 5F ) . This indicates that ILCs may be mobilizing into the adjacent tissue from cryptopatches . 10 . 7554/eLife . 10066 . 013Figure 5 . ILC3s are dynamic and mobilize after anti-CD40 treatment . ( A ) Tracks of motile ILCs in the centre ( left ) and superficial ( right ) 15 μm of a representative cryptopatch . ( B ) Track speed average of ILCs combined from 5 independent experiments before and 4-–6 hr after anti-CD40 treatment . Mean steady-state 8 . 9 μm/min anti-CD40 7 . 5 mm/min , p=0 . 01 . ( C ) Displacement vectors of ILCs moving into and out of representative cryptopatches from steady state ( left ) , 4–6 hr anti-CD40 treated mice ( middle ) , and 4–6 hr anti-CD40 treated mice given anti-GM-CSF 24 hr before imaging ( right ) . ( D ) Quantification of displacement of motile ILCs from control , anti-CD40 treated , and anti-GM-CSF/anti-CD40 treated mice . ( E ) Ratio of entering to exiting ILCs from control , anti-CD40 , and anti-GM-CSF/anti-CD40 treated mice . Motile ILCs are defined as cells with tracks lasting more than 5 min , within 75 µm of a cryptopatch , and displacing more than 14 µm ( approximately two cell lengths ) . * , p<0 . 05 One-way ANOVA with Tukey’s post test . Data are combined ( n=3–5 ) from at least 3 independent experiments . ( F ) Representative image of blood vessels ( red ) and ILCs ( green ) showing ILCs present adjacent to but outside blood vessels . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01310 . 7554/eLife . 10066 . 014Video 2 . Timelapse of cells within the cryptopatch show little motility . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01410 . 7554/eLife . 10066 . 015Video 3 . ILCs within the top 15 μm of the cryptopatch are motile under steady-state conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01510 . 7554/eLife . 10066 . 016Video 4 . Timelapse of steady-state ILCs showing tracks of motile cells . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01610 . 7554/eLife . 10066 . 017Video 5 . Timelapse of ILCs 6 hr after anti-CD40 treatment showing tracks of motile cells . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 01710 . 7554/eLife . 10066 . 018Video 6 . Timelapse of ILCs pretreated with anti-GM-CSF 24 hr and anti-CD40 6 hr before imaging . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 018 The initial characterisation of the anti-CD40 model identified an influx of myeloid cells , particularly CD40+ MHC-II+ CD11chi DCs , in specific inflammatory foci within the tip of the villi of the proximal colon during colitis ( Uhlig et al . , 2006 ) . We hypothesised that development of colitis is driven by initiation of the immune response in the cryptopatch , followed by accumulation of immune cells within anatomically distinct villus inflammatory foci . To understand what impact the early mobilization of ILCs has on the tissue , we looked at several chemokines that could be influencing the development of inflammatory foci . 24 hr after anti-CD40 treatment the chemokines Ccl19 and Ccl2 were increased in the tissue . The receptor for Ccl19 , Ccr7 , was also increased ( Figure 6A ) . Blocking GM-CSF prevented the increase of Ccr7 and Ccl2 , supporting the idea that there is cross-talk between the GM-CSF and migration axes . Staining for CCR7 and CCR2 on lineage+ and ILC subsets suggested that it is the lineage positive cells that are capable of responding to the increased tissue chemokines ( Figure 6B ) . Histological analysis of the proximal colon 7 days after anti-CD40 treatment showed immune foci at the tips of villi ( Figure 6C ) . Immunofluorescence staining for CD11c and RORγt showed colocalization of ILCs with the CD11c positive immune infiltrate . The latter may reflect the inflammatory monocytes identified by flow cytometry from colon digests ( Figure 6D and 1E ) . Quantification of cryptopatch cellularity from H&E after anti-CD40 treatment showed a decrease in cellularity after 24 hr ( Figure 6E ) . The net egress of cells from cryptopatches over a short imaging window ( Figure 5D ) corresponds to the loss of cells from the tissue over 24 hr . This decrease is followed by an increase in inflammatory infiltrates starting one day later ( Figure 6F ) . Blocking GM-CSF resulted in maintenance of cryptopatch structure as would be predicted from the imaging data ( Figure 6G ) . Taken together , these data indicate that ILC mobilization from cryptopatches is associated with a decrease in cryptopatch cellularity , increase in tissue chemokines , recruitment of inflammatory cells , and development of inflammatory foci associated with active disease . 10 . 7554/eLife . 10066 . 019Figure 6 . ILC movement precedes tissue reorganization . ( A ) mRNA expression of Ccl19 , Ccl20 , and Ccl2 with the receptor Ccr7 in proximal colonic lamina propria at various time points following anti-CD40 and isotype or anti-GM-CSF injection at days -1 , +1 , and +3 . Results are shown as fold change in target gene relative to hprt compared with day 0 uninjected mice . ( B ) Representative flow cytometry staining of CCR7 and CCR2 in single live lineage+ cells or ILCs from anti-CD40 or anti-GM-CSF and anti-CD40 treated mice . ( C ) Representative H&E staining of inflammatory foci ( IF ) in transverse proximal colon sections in B6Rag1-/- mice at day 7 following anti-CD40 injection at 2 . 5x and 20x magnification . ( D ) Representative immunofluorescence of IF in the proximal colon of anti-CD40 treated mice showing CD11c ( red ) , RORγt ( green ) and DAPI ( blue ) . ( E ) Number of cells in cryptopatches following anti-CD40 injection as assessed by H&E stained transverse colon sections . ( F ) Mean number of normal cryptopatches ( CP ) and areas containing inflammatory infiltrate in H&E stained transverse colon sections . ( G ) Representative immunofluorescence of a cryptopatch from an anti-GM-CSF and anti-CD40 treated mouse at day 3 showing MHCII ( red ) , IL-7Rα ( green ) , and DAPI ( blue ) . Data are shown as means and SEM . Results are representative of 3–6 independent experiments with n=2–5 mice per experiment . * , p<0 . 05 One-way ANOVA with Bonferroni’s post test . Data are shown as means and SEM . Results are pooled from 2 independent experiments with n=4–6 mice per group per experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 10066 . 019 The results of this study reveal a new IL-23-driven GM-CSF axis that promotes ILC3-mediated acute colitis . Our data indicate a spatiotemporal model for innate colitis that involves cellular activation , GM-CSF dependent mobilization of IL-23R+ ILCs from cryptopatches , and subsequent perpetuation of the response through separate ILC-containing inflammatory foci and GM-CSF dependent accumulation of inflammatory monocytes . Although IL-23 dependent , we found that anti-CD40 induced colitis was IL-17A and IL-17F independent in keeping with earlier reports ( Figure 1 and Eken et al . , 2014 ) . Two recent reports indicated a functional role for IL-22 , possibly through neutrophil recruitment ( Eken et al . , 2014; Song et al . , 2015 ) . However our data show no impact of IL-22 blockade or neutrophil depletion ( data not shown ) on the incidence or severity of disease in this model . By contrast , double blockade of IL-17A and IL-22 reduced colitis severity , indicating that these cytokines may play a redundant role in driving intestinal inflammation possibly through shared effects on intestinal epithelial cells ( Liang et al . , 2006 ) . The pathogenic role of IL-22 in CD40-triggered colitis may depend on the presence of microbial factors that also influence IL-17A production in the intestine . Consistent with this , IL-22 alone was required for the maintenance of bacteria driven innate colitis-associated cancer ( Kirchberger et al . , 2013 ) . Together these results suggest that context-dependent effector functions mediate the colitogenic potential of IL-23 . Recent studies have focussed on the role of GM-CSF as an important mediator of IL-23 driven tissue inflammation ( Codarri et al . , 2011; El-Behi et al . , 2011; Griseri et al . , 2012 ) . We found that GM-CSF played a non-redundant role in both systemic and intestinal inflammation in the anti-CD40 model and was required for bacteria-induced innate colitis . The role of GM-CSF in anti-CD40 mediated systemic disease has been recently reported ( Song et al . , 2015 ) but effects on colitis were not assessed . Our results indicate that IL-23-responsive ILCs can act as the main source of GM-CSF in an innate model . Importantly we also find that ILCs are potent sources of GM-CSF in humans , consistent with a previous report describing cytokine-driven GM-CSF production from an NKp44- ILC population ( Glatzer et al . , 2013 ) . Our study goes further to show that GM-CSF is produced by ILC3s in the blood and colon of patients . The percent of GM-CSF+ ILCs is larger than the percent of GM-CSF+ T cells . This GM-CSF+ ILC3 population is increased in the blood and colonic tissue of patients with IBD , indicating commonalities between mouse models and human disease pathogenesis . IL-23 and GM-CSF may act in an autocrine loop early in the early inflammatory response as we found that anti-CD40-induced early colonic IL-23 expression was GM-CSF dependent and that in turn , IL-23 was required for sustained GM-CSF production by ILCs . A similar positive feedback loop has been described for T cell derived GM-CSF in CNS inflammation ( El-Behi et al . , 2011 ) . Such a feedback loop would promote more GM-CSF production from ILCs as IL-23 production increases in disease . The role of GM-CSF may be to recruit myeloid cells to sites of inflammation , and impact their ability to coordinate inflammatory responses , as has recently been shown in EAE ( Croxford et al . , 2015 ) . Indeed , we found blocking GM-CSF prevented the accumulation of Ly6C+ inflammatory monocytes following anti-CD40 stimulation , in line with recent evidence for a key role for these cells in driving pathogenesis during anti-CD40 induced colitis ( Song et al . , 2015 ) . However , GM-CSF expression also has wider consequences since it affects the behaviour of ILCs within the cryptopatch . Functional analyses of ILC populations have mainly been restricted to cytokine production , and recent evidence that they do not recirculate suggested a primarily sedentary role for these cells ( Gasteiger et al . , 2015 ) . However , that study did not assess ILC movement within tissues as a mechanism through which ILCs may modulate the immune and inflammatory response . To investigate this , we utilized intra vital microscopy to image ILC movement within cryptopatches under homeostatic conditions and in anti-CD40 induced intestinal inflammation . Cryptopatches may provide a platform for the rapid amplification of the immune response and may provide a potential mechanism by which a very small number of ILCs could initiate an immune cascade that culminates in colonic inflammation . Because of the large overlap of cell surface molecule and transcription factor expression between ILCs and T cells , imaging Rag-/- mice provided the cleanest system to study ILC motility . The anti-CD40 model provided temporal synchronization , which is a feature unique to this model of colitis . To our surprise , a subset of ILCs associated with the cryptopatch was migratory under steady-state conditions . After anti-CD40 treatment , ILC motility shifted toward egress from the cryptopatch . In agreement with data that suggests they do not recirculate through the blood ( Gasteiger et al . , 2015 ) , ILCs were not observed migrating into the peripheral blood but appeared to crawl directly through the tissue , possibly following a local chemokine gradient . Future studies will be required to determine what factors govern entrance and exit from lymphoid structures within the tissue , which appears very different from the lymph node . Further understanding of this response may identify pathways that can be targeted to break the proinflammatory cycle within the tissue . As tools become available , studying ILC behaviour in lymphorepleate hosts will further expand these findings and allow increased interrogation of cellular interactions between ILCs and other immune cells as well as spatio-temporal control points in the inflammatory response . Consistent with chemokine-induced migration of ILCs , we found an early peak in mRNA expression of the chemokines Ccl2 , Ccl19 and its receptor Ccr7 in the tissue , suggesting an early role for CCR7 and CCR2 . Despite data showing that migration of LTi ILC3s from the gut to the mesenteric lymph nodes is dependent on CCR7 expression ( Mackley et al . , 2015 ) , expression of CCR7 on ILCs was not assessed in that study and we were unable to identify CCR7 or CCR2 on intestinal ILC3s in anti-CD40 induced colitis . While this argues against CCR7 driven chemotaxis of the ILCs , it does not preclude an organizational function for ILCs in the tissue driven by other mediators . Indeed , our data indicate a role for GM-CSF in the net egress of ILCs from cryptopatches in anti-CD40 induced colitis . Similarly , increased amounts of ccl2 and ccr7 mRNA in the colon were also GM-CSF-dependent , as was accumulation of inflammatory monocytes which also express CCR7 and CCR2 ( Tsou et al . , 2007; Förster et al . , 2008 ) . Together , the data are compatible with a model in which GM-CSF-dependent activation of ILCs leads to exit from the cryptopatch to the villus tip . This coincides with the recruitment of CCR7 and CCR2 expressing inflammatory monocytes through a CCL19 and CCL2 gradient , culminating in tissue damage . Sustained production of GM-CSF by ILCs in inflammatory foci could recruit inflammatory progenitor cells and contribute to differentiation of myeloid effector cells . GM-CSF has been shown to control the differentiation of inflammatory monocytes ( Lenzo et al . , 2012 ) , and this population was markedly reduced following anti-GM-CSF treatment . GM-CSF may also impact on this pathway through altered haematopoiesis and accumulation of granulocyte macrophage progenitors as described in T cell dependent colitis ( Griseri et al . , 2012 ) . Given the multiple activities of GM-CSF on myeloid cell development and function , further studies will be required to delineate the key downstream pathways . IL-23 has been suggested as a therapeutic target in IBD ( Bowman et al . , 2006; Uhlig et al . , 2006 ) . However , our studies indicate that neutralisation of GM-CSF is as efficacious as blockade of IL-23 in mouse models . In addition , increases in GM-CSF secreting ILCs are a feature of IBD . However , targeting GM-CSF may not be straightforward in IBD as there are clear host protective functions of GM-CSF in the intestine ( Sainathan et al . , 2008; Bernasconi et al . , 2010; Hirata et al . , 2010 ) . Indeed neutralising anti-GM-CSF autoantibodies are increased in IBD , particularly in patients with ileal and stricturing disease ( Han et al . , 2009 ) . Furthermore , administration of recombinant GM-CSF ( sargramostim ) has been used to treat Crohn’s disease ( Korzenik et al . , 2005 ) although further trials did not show any substantial benefit of GM-CSF treatment ( Roth et al . , 2012 ) . Despite these issues , in light of recent developments testing anti-GM-CSF in clinical trials for the treatment of other inflammatory conditions such as multiple sclerosis ( Constantinescu et al . , 2015 ) it will be interesting to see if subgroups of patients with hyperinflammatory innate immune activation-driven IBD could benefit from GM-CSF blockade . IBD and colorectal cancer ( CRC ) patients were recruited through the Oxford IBD cohort study , and samples were obtained from Oxford GI Biobank in collaboration with Oxford Radcliffe Biobank . Colon specimens were collected from patients undergoing surgery for severe disease . Macroscopically healthy sections from CRC patients were used as controls . For gene expression data , intestinal mucosal pinch biopsies were taken from patients undergoing routine endoscopy . From 20 patients with evidence of active intestinal inflammation , matched biopsies were collected from both inflamed tissue and regions of the intestine with no apparent inflammation . IBD blood samples were collected from patients attending the outpatient clinic and in some cases from patients undergoing surgery . Control blood samples were obtained from healthy donors . Ethical approval was obtained from the Oxfordshire Research Ethics Committee ( Reference numbers: 11/YH/0020 and 09/H0606/5 ) and informed written consent was given by all study patients . A previously published study with a publically available dataset ( Arijs I . , Van Lommel L . , Van Steen K . , De Hertogh G . , Geboes K . , Schuit F . and Rutgeerts P . 2009 . Mucosal expression profiling in patients with inflammatory bowel disease before and after first infliximab treatment . Accession number GSE16879 , Gene Expression Omnibus repository ) was used to confirm mRNA observations of the Oxford cohort , using colonic mucosal biopses obtained at endoscopy from healthy controls or patients refractory to corticosteroids and/or immunosuppression . Gene expression data were generated using the Affymetrix Human Genome U133 Plus 2 . 0 array as described in ( Arijs et al . , 2009 ) . All mice were bred and maintained under specific pathogen-free conditions in accredited animal facilities . C57BL/6 Rag1-/- , 129SvEvS6 . Rag2-/- and C57BL/6 Rag1-/-Il23rgfp/+ ( from Daniel Cua , Merck Research Laboratories , Palo Alto , USA ) strains used in this study were bred and maintained at the University of Oxford . C57BL/6 Rag1-/-Il15ra-/- and C57BL/6 Rag1-/- were bred and maintained at the National Institute for Medical Research . Experiments were conducted in accordance with local animal care committees ( UK Scientific Procedures Act of 1986 ) . Mice were routinely screened for the absence of pathogens , and were kept in individually ventilated cages with environmental enrichment . For cLP isolation , colon tissue was cut into 1cm pieces , and epithelial cells were removed by incubation ( 2x ) in RPMI containing 5% heat-inactivated FCS and 5mM EDTA . Remaining tissue was incubated in RPMI containing 5% FCS , 15mM HEPES and 300U/ml of Collagenase VIII ( Sigma-Aldrich , St Louis , MO ) to digest the remaining tissue . Cell populations were purified by 75%:40%:30% Percoll ( GE Healthcare , Little Chalfont , U . K . ) gradient centrifugation ( 600 x g , 20 min ) . Lymphocytes were isolated from the 40%/30% interface . Peripheral blood was diluted in an equal volume of PBS and layered onto Ficoll-Hypaque ( GE-Healthcare ) . Following centrifugation , PBMCs were collected from the Ficoll-plasma interface and washed 3x in PBS prior to culture or flow cytometry . Specimens were incubated in 1 mM DTT solution , followed by three washes in 0 . 75 mM EDTA solution . Tissues were then digested overnight in 0 . 1 mg/ml collagenase A solution ( Roche Diagnostics , Basel , Switzerland ) . Cell populations were purified by 100%:60%:40%:30% Percoll ( GE Healthcare ) gradient centrifugation ( 900 x g , 30 min ) . Lymphocytes were isolated from the 60%/40% interface . For human biopsy samples , tissue was immediately placed in RNAlater solution ( ThermoFisher Scientific , Waltham , MA ) and stored at -20oC until processed . Human and murine cLP cells were lysed in RLT buffer ( Qiagen , Hilden , Germany ) containing ß-mercaptoethanol . RNA was isolated using the RNEasy kit ( Qiagen ) including a DNAse I digestion . Content and purity of RNA was measured using a Nanodrop spectrophotometer ( Thermo Fisher Scientific ) . cDNA was synthesised using the Superscript III reverse transcription kit ( Life Technologies , Paisley , U . K . ) or the High Capacity cDNA kit ( ThermoFisher Scientific , Waltham , MA ) . Quantitative real-time PCR for the candidate genes was performed using the Taqman system . cDNA samples were analysed in triplicate ( technical replicates ) using the CFX96 detection system ( Bio-Rad Laboratories , Hercules , CA ) or the Viia7 system ( Applied Biosystems , Foster City , CA ) . Values were normalised to Hprt expression , and analysed using the d-Ct method . Taqman Gene Expression Assays ( Life Technologies ) for mouse Ccl19 , Ccl20 Ccr7 , Ccl2 , Il23r and Csf2 , and human RPLPO and CSF2 were used . To induce acute innate colitis , 25-–50 mg of αCD40 IgG2a mAb ( clone FGK45 , BioXCell , West Lebanon , NH ) was administered via i . p . injection to mice on a C57BL/6 Rag-/- background . Unless otherwise indicated , mice were sacrificed 7 days post αCD40 injection , and weight loss was monitored throughout the course of the experiment . To induce colitis using Helicobacter hepaticus , Hh NCI-Frederick isolate 1A ( strain 51449 ) was grown as previously described ( Maloy et al . , 2003 ) . Mice on a 129SvEv Rag2-/- background were fed 3x on consecutive days with Hh 1A ( 1 . 0x108 CFU ) by oral gavage . Animals were analysed at week 6 following infection . To block specific cytokine or cellular activity in vivo , aCD40-treated mice received 0 . 375 mg IL-17A blocking mAb ( UCB Celltech , Slough , U . K . ) , 0 . 15 mg IL-22 blocking mAb ( Genentech , San Francisco , CA , clone 8E11 ) , 0 . 8 mg IL-23R blocking mAb ( D . Cua , Merck , Kenilworth , NJ ) , 0 . 125 mg GM-CSF blocking mAb ( CSL Ltd , Parkville , Australia ) , 100 mg IL-17F blocking antibody ( eBioscience , San Diego , CA ) or isotype controls ( GP120 10E7 . 1D2 , Genentech or 22E9 . 11 , CSL Ltd ) . Treatment started on day -1 and antibodies were injected i . p . at day -1 and day 3 , except blocking mAb to IL-22 which was administered at days -1 , 1 , 3 and 5 . Blocking mAb to GM-CSF during Helicobacter hepaticus infection was administered i . p . 2x per week starting at day 0 . The following antibodies were purchased from eBioscience: fixable viability dye , GM-CSF-PE ( MP1-22e9 ) , CD127-FITC/PE/PECy5 ( A7R34 ) , Thy1 . 2-PECy7 ( 53 . 2 . 1 ) , T-bet-eFluor660 ( eBio4B10 ) , RORγt-PE ( AFKJS-9 ) , B220-PerCP-Cy5 . 5 ( RA3-6B2 ) , CD11b-PerCP-Cy5 . 5 ( M1/70 ) , CD11c-PerCP-Cy5 . 5/APC ( N418 ) , Gr-1-PerCP-Cy5 . 5 ( RB6-8C5 ) , CD64-PE ( X54-5/7 . 1 ) , F4/80-PerCP-Cy5 . 5 ( BM8 ) , MHCII-AlexaFluor700 ( M5/114 . 15 . 2 ) , CCR7-PE ( 4B12 ) , perforin-APC ( eBioOMAK-D ) and cKit-Alexa-Fluor700 ( ACK2 ) . The following antibodies were purchased from Biolegend ( San Diego , CA ) : NKp46-FITC ( 29A1 . 4 ) , CD49b-PerCP-Cy5 . 5 ( DX5 ) , Ly6C-PECy7 ( HK1 . 4 ) , Ly6G-BV570 ( 1A8 ) , CCR6-BV421 ( 29-2L17 ) , CD4-BV605 ( RM4-5 ) , CCR2-AF647 ( SA203G11 ) and NK1 . 1-BV650 ( PK136 ) . The following antibodies were purchased from BD Horizon ( Franklin Lakes , NJ ) : NKp46-V450 ( 29A1 . 4 ) , CD11b-V500 ( M1/70 ) , CD45-V500/BV650 ( 30-F11 ) , SiglecF-BV421 ( E5-2440 ) and NKG2D-PE CF594 ( CX5 ) . Lineage staining was performed using PerCP-Cy5 . 5-conjugated antibodies against B220 , CD11b , CD11c , Gr-1 , F4/80 and CD49b . The following antibodies were purchased from eBioscience: CD1a-FITC ( HIH9 ) , CD3-BV650 ( OKT3 ) , CD8-FITC ( RPA-T8 ) , CD19-FITC ( HIB19 ) , IL-17A-PE ( eBio64DEC17 ) , IL-22-PECy7 ( 22URT1 ) . The following antibodies were purchased from Biolegend: CD3-FITC ( UCHT1 ) , CD4-FITC ( OKT4 ) , CD11c-FITC ( B-LY63 . 9 ) , CD14-FITC ( MSE2 ) , CD16-FITC ( 3G8 ) , CD34-FITC ( 581 ) , CD123-FITC ( 6H6 ) , TCRαβ-FITC ( IP26 ) , TCRγδ-FITC ( B1 ) , CD45-AlexaFluor700 ( HI30 ) , CD56-PerCP-Cy5 . 5/BV510 ( HCD56 ) , CD127-BV421 ( A019D5 ) , GM-CSF-PerCP-Cy5 . 5 ( BVD2-21C11 ) , IL-5-APC ( TRFK5 ) and IFNγ-BV605 ( 4S . B3 ) . Lineage staining was performed using FITC-conjugated antibodies against CD1a , CD8 , CD19 , CD3 , CD4 , CD14 , CD16 , CD34 , CD123 , TCRαβ and TCRγδ . Single cell suspensions were prepared from mouse spleen or colonic LP or human LPMC or PBMC cells as described above . For intracellular cytokine staining mouse or human were directly cultivated for 3 or 4 hr in the presence of phorbol 12-myristate 13-acetate ( PMA , Sigma-Aldrich ) ( 100ng/ml ) , ionomycin ( Sigma-Aldrich ) ( 1μg/ml ) , monensin ( BD Biosciences ) and brefeldin A ( eBioscience ) . Where indicated , mouse and human cLP cells were stimulated overnight with 10 ng/ml of recombinant IL-12 ( Peprotech , Rocky Hill , NJ ) or IL-23 ( mouse - R&D Systems ( Minneapolis , MN ) , human - Peprotech ) , and then for the last 4 hr PMA , ionomycin and brefeldin A were added . Cells were incubated with CD16/32 ( eBioscience ) to block Fcγ receptors , prior to incubation with fixable viability dye ( eFluor780 , eBioscience ) , and mAb conjugates in 100 μl PBS containing 2% BSA and 1 mM EDTA ( PBS-E ) for 30 min . Cells were subsequently washed twice in PBS-E . For cytokine staining , cells were stained with antibodies for 1 hr using the Cytofix/Cytoperm kit ( BD Biosciences ) . For transcription factor staining , cells were stained with transcription factor antibodies for 1 hr using the Foxp3 staining kit ( eBioscience ) . Cells were washed twice in PBS-E . All labelled cells were then acquired on a LSRII SORP ( BD Biosciences ) and analysed using Flowjo software ( Tree star , Ashland , OR ) . To assess the severity of colitis , samples of the proximal colon were taken and immediately fixed in buffered 10% formalin ( 3 . 6% w/v formaldehyde ) . 4–5 μm paraffin-embedded sections were cut and stained with haematoxylin and eosin ( H&E ) . Inflammation was assessed using previously published criteria ( Izcue et al . , 2008 ) . Each sample was graded semiquantitatively from 0 to 4 in each of the 4 following features: epithelial hyperplasia and goblet cell depletion , cellular infiltration of the lamina propria , percentage of the section affected , and markers of severe inflammation ( submucosal inflammation , crypt abscesses ) . Three separate colon sections for each sample were examined . Scores for each criterion were added to give an overall score for each sample of 0–12 . Samples were scored in a blinded fashion by two individuals . Gut imaging was performed as described in McDole et al . ( McDole et al . , 2012 ) . Briefly , mice were anesthetized with ketamine/xylazine , the gut was exposed and immobilized with an imaging window , and imaging proceeded using isoflurane to maintain anesthesia . Images were collected using a 20x water dipping lens and the spectral detector of a Zeiss 780 upright microscope ( Carl Zeiss , Oberkochen , Germany ) . Images were linearly unmixed using the Zen software ( Carl Zeiss ) to separate autofluorescence , collagen , EGFP , and Texas red-dextran based on single color controls . Mice were anesthetized for imaging and injected intravenously with 50 mg of 70 kDa texas-red labelled dextran ( Life Technologies ) . Imaging proceeded immediately as described above . Intravital microscopy was analyzed using Imaris 8 ( Bitplane , Belfast , U . K . ) after linear unmixing using Zen ( Carl Zeiss ) . Images were drift corrected based on mucus or collagen signal . ILCs were tracked using autoregressive motion algorithm of the spot-tracking module . Motile ILCs were gated based on distance less than 75 μm from a cryptopatch , track duration greater than 5 min , and displacement greater than 14 μm ( approximately two cell lengths ) . Images were smoothed using a Gaussian filter for display . For immunofluorescence , samples of the proximal colon were taken and immediately embedded in OCT compound ( Tissue-Tek , Sakura , Alphen aan den Rijn , The Netherlands ) and frozen in a bath of isopentane on dry ice prior to storage at -80oC . 6 μm frozen sections were cut and collected on frosted glass slides . For staining , slides were fixed in 2% formalin . For RORγt staining , slides were subsequently also treated with Fix-Perm buffer ( eBioscience ) . Endogenous peroxidase activity was blocked with 1% H2O2 ( Sigma-Aldrich ) and 2% sodium azide and non-specific binding was blocked with 10% donkey serum . Sections were incubated with the following antibodies CD4 ( RM4-5 , BD Biosciences ) IL-7Rα ( A7R34 , eBioscience ) in 10% donkey serum or RORγt ( AFKJS-9 , eBioscience ) in Perm buffer ( eBioscience ) . Sections were then incubated with donkey anti-rat HRP secondary antibody ( Jackson ImmunoResearch Laboratories , West Grove , PA ) . Tyramide signal amplification was the performed ( Perkin Elmer , Waltham , MA ) . Then sections were stained with CD11c ( N418 , eBioscience ) in 10% goat serum followed by goat anti-hamster Cy3 , or with MHCII-FITC or AF647 ( both M5/114 . 15 . 2 , Biolegend ) . Sections were mounted with Vectashield containing DAPI . Images were collected using a 710 microscope ( Carl Zeiss ) , and analysed using ImageJ open access software . The colons from C57BL/6 Rag1-/-Il23rgfp/+ mice were freshly isolated and maintained in room temperature PBS before imaging . Opened sections of colon were immobilized using Vetbond tissue adhesive ( 3M , Saint Paul , MN ) in dishes containing PBS . CPs were imaged from the muscularis side of the colon using a 20x water dipping objective and spectral detector on a Zeiss 780 upright multiphoton microscope . Images were analysed using the linear unmixing module of Zen ( Carl Zeiss ) and Imaris ( Bitplane ) . Statistical analysis and graphical representations were performed using Prism 5 . 0a software ( GraphPad , La Jolla , CA ) . For comparison of qPCR results and histopathological analysis , the nonparametric Mann-Whitney U test was used , and for multiple samples , one-way ANOVA with Bonferroni’s post-test was used unless otherwise stated . Technical replicates are from the same sample , analysed on the same day . Biological replicates are from independent experiments . Sample size was computed using an estimate of colitis score between 6–8 , to detect a score reduction by 50% in a treated group with a standard deviation of 30% , false positive of 0 . 05 , and power of 0 . 80 . Sample size was therefore calculated to be between 6–7 . Where smaller changes were expected , sample sizes were increased accordingly .	Crohn’s disease and ulcerative colitis are diseases in which the body’s own immune system causes inflammation of the large intestine . These autoimmune diseases can be severely debilitating and difficult to treat . However an improved understanding of the factors that contribute to the intestinal inflammation may lead to new and effective treatments . Immune cells called innate lymphoid cells were discovered recently , and shown quickly to play a role in host defense , tissue repair and inflammation regulation . Several groups of innate lymphoid cells are now known; each group is characterized by the genes that control the cell’s development and the small proteins ( called cytokines ) that the cells release . One group of innate lymphoid cells , the ILC3s , are generally found in the intestinal tract , albeit in small numbers . Given that innate lymphoid cells are known to manage inflammatory responses , it is possible that ILC3s contribute to intestinal inflammation . However , it remains unclear how such a small population of cells could so dramatically inflame the gut . Pearson et al . now reveal two mechanisms that these innate lymphoid cells use to amplify the inflammatory response and exacerbate intestinal inflammation . First , in both mice and humans , ILC3s were found to be a key source of a cytokine called GM-CSF , which recruits additional immune cells that further promote intestinal inflammation . Secondly , while ILC3s were traditionally regarded as immobile immune cells , Pearson et al . discovered that these cells can move within the intestinal tissue and mobilize from their starting points within this tissue if they are activated . These two mechanisms could explain how ILC3s can trigger inflammation that occurs throughout the gut . The experiments suggest that blocking production of the GM-CSF cytokine or altering ILC3 movement or activity may help reduce intestinal inflammation . However , the use of GM-CSF blocking drugs to protect against colitis and similar conditions could be problematic , because GM-CSF also plays an important protective role in the intestines . Nevertheless , clinical trials are underway to investigate the use of anti-GM-CSF drugs to treat other inflammatory conditions ( such as rheumatoid arthritis ) . These studies could offer insight into whether these drugs provide relief to trial participants who suffer from intestinal inflammation as well .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2016	ILC3 GM-CSF production and mobilisation orchestrate acute intestinal inflammation
Bone cells sense and actively adapt to physical perturbations to prevent critical damage . ATP release is among the earliest cellular responses to mechanical stimulation . Mechanical stimulation of a single murine osteoblast led to the release of 70 ± 24 amole ATP , which stimulated calcium responses in neighboring cells . Osteoblasts contained ATP-rich vesicles that were released upon mechanical stimulation . Surprisingly , interventions that promoted vesicular release reduced ATP release , while inhibitors of vesicular release potentiated ATP release . Searching for an alternative ATP release route , we found that mechanical stresses induced reversible cell membrane injury in vitro and in vivo . Ca2+/PLC/PKC-dependent vesicular exocytosis facilitated membrane repair , thereby minimizing cell injury and reducing ATP release . Priming cellular repair machinery prior to mechanical stimulation reduced subsequent membrane injury and ATP release , linking cellular mechanosensitivity to prior mechanical exposure . Thus , our findings position ATP release as an integrated readout of membrane injury and repair . The mechanical environment is an important determinant of bone health , as emphasized by a consistent bone loss in astronauts exposed to microgravity or in paralyzed or bedridden patients ( Nagaraja and Jo , 2014 ) . Gravitational and muscle forces act on the skeleton during physical activity resulting in a complex combination of shear forces , strains and pressures . Bone-embedded osteocytes and bone-forming osteoblasts are widely regarded as the mechanosensitive cells in the skeletal system ( Weinbaum et al . , 1994 ) . Following mechanical stimulation of rodent and human osteoblasts , transient intracellular free calcium ( [Ca2+]i ) elevations and ATP release are among the earliest detectable events , which result in autocrine and paracrine purinergic ( P2 ) receptor signaling ( Robling and Turner , 2009; Romanello et al . , 2001; Genetos et al . , 2005 ) . Vesicular release of lysosomes or secretory vesicles , or conductive release via channels such as Maxi anion channels , Volume-regulated anion channels ( VRAC ) , Connexins or Pannexins are the main mechanisms of regulated ATP release in mammalian cells ( Mikolajewicz et al . , 2018 ) . Pathological ATP spillage also occurs from traumatically damaged cells ( Burnstock and Verkhratskii , 2012 ) . Of interest , non-lethal membrane injury has been demonstrated in vivo in several tissues ( McNeil and Steinhardt , 2003 ) , including bone ( Yu et al . , 2017 ) , under physiological conditions . The mechanism of facilitated cell membrane repair has been described and involves Ca2+/PKC-dependent vesicular exocytosis ( Togo et al . , 1999 ) . However , the contribution of non-lethal cell injury to ATP release and related mechanotransductive purinergic signaling remains unclear . The goal of this study was to examine the mechanism of ATP release from mechanically stimulated cells of the osteoblastic lineage . Since we have previously demonstrated that transient membrane disruption is required to induce global [Ca2+]i elevations in osteoblasts ( Lopez-Ayon et al . , 2014 ) , we were particularly interested in understanding the contribution of membrane injury to mechanically induced ATP release . Mechanical forces were applied by local membrane deformation or turbulent fluid shear stress in vitro to BMP-2 transfected C2C12 osteoblastic cells ( C2-OB ) , primary bone marrow ( BM-OB ) and compact bone ( CB-OB ) -derived osteoblasts and changes in [Ca2+]i , vesicular exocytosis , membrane permeability and ATP release were assessed . The prevalence of membrane injury in osteocytes at physiological and supraphysiological mechanical strain levels was investigated following in vivo cyclic compressive tibial loading of 10-week-old female C57Bl/6J mice . Osteoblasts from three different sources , C2-OB , CB-OB , and BM-OB , were loaded with [Ca2+]i dye Fura2 and mechanically stimulated with a glass micropipette , which induced qualitatively similar transient global [Ca2+]i elevations , consistent with prior work ( Robling and Turner , 2009; Romanello et al . , 2001; Genetos et al . , 2005 ) ( Figure 1A–C , Figure 1—video 1 ) . L-type voltage-sensitive calcium channel ( VSCC ) inhibitor Nifedipine and P2 antagonist PPADS significantly reduced the amplitude of mechanically-stimulated [Ca2+]i transients ( Figure 1D ) . L-type VSCC activation occurred gradually ( Figure 1E ) while the P2 receptor-driven component of the response peaked within seconds of stimulation ( Figure 1F ) . Together , L-type VSCC and P2 receptor-driven component accounted for ~50% of the mechanical stimulated [Ca2+]i transient . Consistent with previous reports ( Robling and Turner , 2009; Romanello et al . , 2001; Genetos et al . , 2005 ) , shortly after a single osteoblast was mechanically stimulated , neighboring cells exhibited delayed secondary [Ca2+]i responses ( Figure 1G ) . Pharmacological interventions revealed that P2 receptors mediated the secondary response in all three osteoblast models , while a tendency for GAP junction involvement was observed in BM-OB responses ( Figure 1H ) . Puff application of 10 µM ATP mimicked the appearance of secondary responders in C2-OB ( Figure 1I ) . We used a luciferin fluorescence-based imaging technique ( Sørensen and Novak , 2001 ) to directly measure ATP release ( [ATP]e ) ( Figure 2—figure supplement 1 ) . Simultaneous recording of [ATP]e and [Ca2+]i demonstrated that ATP release occurred within seconds after mechanical stimulation ( time to peak release: 1 . 64 ± 0 . 36 s ) and preceded the onset of [Ca2+]i responses in neighboring cells ( Figure 2 ) . We measured pericellular [ATP]e concentrations in the range of 0 . 05–80 . 5 µM corresponding to a release of 70 ± 24 amole ATP from a single mechanically stimulated osteoblast ( Figure 2B ) . Pericellular [ATP]e significantly correlated with the percentage of secondary responders ( Figure 2C ) . Vesicular ATP release was proposed to be a major source of extracellular ATP ( 5 ) . Using confocal microscopy , we found that an acidophilic marker quinacrine and a fluorescent ATP analog MANT-ATP co-localized in intracellular granular compartments ( Figure 3A; Figure 3—video 1–2 ) . Quinacrine-positive vesicular pool was entirely released after treatment with Ca2+ ionophore ionomycin , and only partially by mechanical stimulation ( Figure 3B; Figure 3—video 3–5 ) , while negligible vesicular release was observed in unstimulated cells . Interestingly , vesicular release peaked within 20 s of mechanical stimulation ( Figure 3C , D ) , while ATP release within ~2 s ( Figure 2A ) . Basal vesicular density was 70 ± 4×10−3 vesicles/µm2 ( Figure 3E ) and mechanical stimulation resulted in the cumulative release of 5 ± 1×10−3 vesicles/µm2 over 100 s post-stimulation ( 7 . 2 ± 1 . 2% of the basal vesicular pool ) ( Figure 3F ) . In [Ca2+]e-free conditions vesicular density was unaffected , whereas mechanically evoked vesicular release was markedly suppressed ( Figure 3E , F ) . Treatment with N-ethylmaleimide ( NEM ) caused a 33 ± 2% increase in the vesicular density ( Figure 3E ) , consistent with a NEM-induced increase in the readily releasable vesicular pool ( Lonart and Südhof , 2000 ) , and a 55 ± 8% increase in mechanically-evoked vesicular exocytosis ( Figure 3F ) . Surprisingly , the percentage of secondary responders and activation rates for Ca2+ responses ( response amplitude divided by time for signal to increase from 10 to 90% of peak ) were unaffected in [Ca2+]e-free conditions , whereas NEM induced a significant decline in the percentage of secondary responders and activation rates ( Figure 3G , H ) . We measured bulk ATP release following mechanical stimulation by media displacement , a model for turbulent fluid shear stress ( tFSS , Rumney et al . , 2012 , Figure 3I ) . Proportional to the extent of tFSS , 21 ± 11 to 422 ± 97 amol ATP/cell were released , approaching 31 . 5% of total intracellular ATP at high tFSS ( Figure 3J ) . tFSS ( 10x media displacements ) induced release of 324 ± 50 amol ATP/cell , while ionomycin only led to release of 94 ± 5 amol ATP/cell . Moreover , tFSS-induced ATP release was significantly potentiated under [Ca2+]e-free conditions ( 961 ± 88 amol ATP/cells ) and inhibited in NEM-treated cells ( 41 ± 24 amol ATP/cells ) ( Figure 3K ) . We also investigated conductive channels , however , pharmacological inhibition of Maxi-anion channels , P2X7 , Connexins , Pannexins , T-type VSCC , TRPV4 or Piezo1 channels did not affect tFSS-stimulated ATP release ( Figure 3—figure supplement 1 ) . Although inhibition of L-Type VSCC had a small inhibitory effect ( Figure 3—figure supplement 1 ) , our data suggest that conductive channels cannot account for ATP release observed from mechanically stimulated osteoblasts . Importantly , our data show that increases in vesicular release were consistently associated with a reduction in mechanically stimulated ATP release , while inhibition of vesicular exocytosis increased ATP release from mechanically-stimulated osteoblasts . We hypothesized that mechanically stimulated ATP release is related to membrane injury . Adopting a Fura2-based dye-leakage assay ( Togo et al . , 1999 ) , we examined changes in 340 ex/510 em Fura2 fluorescence following mechanical stimulation of osteoblasts ( Figure 4A ) . In ~40% of micropipette-stimulated osteoblasts , intracellular Fura2 fluorescence returned to pre-stimulation baseline ( minor or no injury; mIn ) . However , in 30–40% of osteoblasts post-stimulation fluorescence was partially reduced ( intermediate injury; iIn ) and in 15–20% of osteoblasts completely lost ( severe injury; sIn ) ( Figure 4B ) . The extent of injury sustained by a micropipette-stimulated cell correlated with the magnitude and duration of [Ca2+]i elevations ( Figure 4C–E ) , as well as the percentage of secondary responders ( Figure 4F ) , suggesting that ATP release is related to the extent of membrane disruption . In a dye uptake assay , we added membrane impermeable dyes , trypan blue ( TB ) or rhodamine-conjugate dextran ( R-dextran ) , to osteoblasts before or after mechanical stimulation ( Figure 4G ) . When the dyes were applied prior to tFSS , the proportion of dye-permeable osteoblasts positively correlated with the degree of tFSS . The extent of dye uptake was higher for TB ( ~900 Da ) compared to R-dextran ( ~10 kDa ) ( Figure 4H , I ) . However , when membrane permeability was examined 300 s after stimulation , a significantly lower percentage of cells were permeable to the dyes ( Figure 4H , I ) . Hemichannel involvement was excluded , since hemichannel blockers did not affect mechanically-stimulated [Ca2+]i elevations , membrane permeability or ATP release ( Figure 4—figure supplement 1 ) . Consistent with transient membrane disruption , lactate dehydrogenase ( LDH , ~140 kDa ) leaked into the extracellular media proportionally to tFSS magnitude ( Figure 4J ) , even though cell viability was minimally affected 1 hr after tFSS stimulation ( Figure 4K ) . Examining TB uptake following micropipette stimulation , we have found that all C2-OB were TB-permeable immediately upon stimulation , decreasing to 75 ± 22% within 10 s , to 40–50% by 180 s and to 18 ± 12% by 300 s ( Figure 4L ) . Comparing these data to dye leakage experiments suggested that minor injuries resealed within 10 s , intermediate injuries within 20–180 s and severe injuries did not reseal by 300 s . ( Figure 4L ) . Based on predicted molecular radii of the dyes ( Erickson , 2009 ) , we estimated the membrane lesion radius to be on the nanometer scale , sufficiently large to permit the efflux of smaller ATP molecules ( ~507 Da ) ( Figure 4M ) . To determine whether mechanically induced repairable membrane disruptions occur in vivo , 10-week-old female C57Bl/6J mice were injected with lysine-fixable Texas Red-conjugated dextran ( LFTR-Dex , ~10 kDa ) either 30 min prior to or 20 min after cyclic compressive tibial loading to strain magnitudes at the upper level of physiological activities ( 600 µε ) or at supraphysiological osteogenic levels ( 1200 µε ) ( Willie et al . , 2013 ) ( Figure 5A–D , Figure 5-video 1 ) . In the contralateral non-loaded control tibiae ( which experienced habitual loads only ) , sclerostin-positive osteocytes were observed at a density of 3212 ± 287 cells/mm2 ( Figure 5—figure supplement 1 ) , ~4–7% of which demonstrated LFTR-Dex uptake , while only 2% of calvarial osteocytes were LFTR-Dex positive ( Figure 5E ) . Cyclic compressive tibial loading in the presence of LFTR-Dex resulted in significantly increased osteocyte dye uptake compared to unloaded bone , demonstrating that in vivo mechanical loading results in cellular membrane disruption ( Figure 5F , G , red ) . Importantly , when LFTR-Dex was administered 20–50 min after 600 µε loading , the levels of cellular dye uptakes were similar to those in unloaded tibia , suggesting that damage evident immediately after loading was repaired within this time period ( Figure 5F , light blue ) . Of interest , when LFTR-Dex was administered 20–50 min after 1200 µε loading , the proportion of osteocyte exhibiting dye uptake was significantly lower compared to non-loaded controls , suggesting that adaptive improvements in membrane integrity occurred in response to the supraphysiological strain ( Figure 5F , G , dark blue ) . Vesicular exocytosis has been demonstrated to facilitate membrane repair and improve membrane integrity in a PKC-dependent manner ( Togo et al . , 1999 ) . We examined the role of Ca2+/PLC/PKC pathway during mechanically induced membrane injury and resealing . Pharmacological activation of PKC ( Figure 6A , Figure 6—figure supplement 1A ) and NEM-induced potentiation of vesicular release ( Figure 6—figure supplement 2A ) both attenuated tFSS-induced membrane disruption . Depletion of [Ca2+]e , or pharmacological inhibition of PLC ( Figure 6—figure supplement 2A , B ) or PKC ( Figure 6A , B; Figure 6—figure supplement 1A , B ) increased membrane disruption in osteoblasts following tFSS or micropipette stimulation . Together these findings suggest that membrane repair was regulated by extracellular calcium influx , PLC/PKC signaling and vesicular exocytosis . We next investigated the link between Ca2+/PLC/PKC pathway and mechanically stimulated [Ca2+]i elevations , vesicular exocytosis and ATP release . Depletion of [Ca2+]e or PLC inhibition consistently reduced the amplitude of mechanically-stimulated [Ca2+]i elevations ( Figure 6—figure supplement 2C ) and vesicular release ( Figure 6—figure supplement 2D ) , and potentiated ATP release ( Figure 6—figure supplement 2E ) . PKC activation had minor effect on calcium responses ( Figure 6—figure supplement 2C ) , augmented vesicular release ( Figure 6C ) and reduced ATP release ( Figure 6D , Figure 6—figure supplement 1C ) . In contrast , PKC inhibition potentiated calcium responses ( Figure 6—figure supplement 2C ) , suppressed vesicular release ( Figure 6C ) and increased ATP release ( Figure 6D , Figure 6—figure supplement 1C ) . PMA-induced PKC activation rescued vesicle pool depletion following PLC inhibition , thereby positioning PKC signaling downstream of PLC ( Figure 6—figure supplement 3 ) . Together these data suggest that mechanically stimulated membrane injury and repair , vesicular release and ATP release are regulated by Ca2+/PLC/PKC-dependent signaling . We next examined which PKC isoform is involved in the mechanoresponse . Immunoblot analysis demonstrated that mechanical stimulation ( using tFSS ) of CB-OB activated conventional PKCα/β and PKD/PKCμ , but not atypical PKCζ/λ or novel PKCδ/θ isoforms ( Figure 6E , Figure 6—figure supplement 4 ) . PMA stimulated phosphorylation of conventional and PKD/PKCμ isoforms , but not atypical or novel PKC isoforms , while only basal PKD/PKCμ phosphorylation had a tendency ( p=0 . 09 ) to be inhibited by Bis . Total PKC levels were unaffected by pharmacological interventions or mechanical stimulation ( Figure 6—figure supplement 4G ) . These findings suggest that the PKD/PKCμ isoform regulates membrane resealing in osteoblasts . To establish the relationship between vesicular release , membrane disruption and ATP release , we pooled data for the mechanically induced responses from all treatments studied in CB-OB ( Figure 6F–H ) . Mechanically induced vesicular release negatively correlated with the extent of membrane disruption ( Figure 6F ) and the amount of ATP released ( Figure 6G ) , while membrane disruption positively correlated with amount of ATP released in response to mechanical stimulation ( Figure 6H ) . These data support a model in which mechanical disruption of the cell membrane leads to ATP spillage , while the extent of membrane resealing , regulated by Ca2+/PLC/PKC-dependent vesicular exocytosis , limits the amount of total ATP released ( Figure 7 ) . We have demonstrated that non-lethal cell membrane injury is routine in mechanically stimulated cells . In vitro , single cell membrane deformation or fluid shear stress transiently compromised cell membrane integrity in 2–30% of cells , which was repaired within 10–100 s . Consistent with findings observed using in vitro models of mechanical stimulation , in vivo cyclic compressive tibial loading engendering physiological and supraphysiological strains resulted in transient osteocyte membrane disruption in 5–10% of cells . Interestingly , supraphysiological mechanical strains improved membrane integrity post-stimulation compared to baseline levels . Previously , non-lethal reversible cell wounding was shown to occur in mechanically active tissues , with membrane-impermeable marker uptake reported in 3–20% of cells in skeletal muscle , 3–6% in skin , 25% in cardiac muscle , 2–30% in lung and aorta ( McNeil and Steinhardt , 2003 ) . Consistent with our findings , membrane disruption was reported in 20–60% of long bone osteocytes after in vivo treadmill loading ( Yu et al . , 2017 ) . The higher prevalence of disruption reported by Yu et al . in long bones was likely a consequence of cumulative tracer uptake over 18 hr ( Yu et al . , 2017 ) , while our study examined membrane disruption over a 20–50 min period . Our data demonstrated that reparable membrane injuries are common and significantly contribute to ATP release from mechanically stimulated osteoblasts and osteocytes . We observed that osteoblasts release 70 ± 24 amol ATP/cell in response to direct membrane deformation and 21 ± 11 to 422 ± 97 amol ATP/cell in response to tFSS , which is consistent with previously reported estimates ranging from 24 ± 1 to 324 ± 59 amol ATP/cell ( Genetos et al . , 2005; Wang et al . , 2013; Kringelbach et al . , 2015; Kringelbach et al . , 2014; Genetos et al . , 2007; Pines et al . , 2003; Romanello et al . , 2005; Manaka et al . , 2015 ) . Osteoblasts have been reported to release ATP through mechanisms related to vesicular exocytosis ( Genetos et al . , 2005; Romanello et al . , 2005 ) and L-type VSCC ( Genetos et al . , 2005 ) . In osteocytes , hemichannels ( Kringelbach et al . , 2015; Seref-Ferlengez et al . , 2016 ) , P2X7 ( Seref-Ferlengez et al . , 2016 ) and T-type VSCC ( Thompson et al . , 2011 ) have also been implicated . We visualized ATP-containing quinacrine-positive vesicles and directly demonstrated their release upon mechanical stimulation . Surprisingly , the total vesicular ATP content was ~94 ± 5 amol ATP/cell , which , considering that only 7 . 2 ± 1 . 2% of vesicles were released upon mechanical stimulation , is much less ( <10 amol ATP/cell ) than mechanically induced ATP release . Moreover , treatments that potentiated vesicular release resulted in a surprising decrease in ATP release , while treatments interfering with vesicular release enhanced ATP release . In search for an alternative route of ATP release , we found that mechanically stimulated ATP release was proportional to the extent of membrane injury . Vesicular exocytosis is believed to promote membrane repair via membrane insertion , thereby reducing membrane tension and allowing exposed hydrophobic residues to reseal ( McNeil and Steinhardt , 2003 ) . We have shown that vesicular release was critical for the repair of mechanically induced membrane injury and that the Ca2+/PLC/PKC pathway regulates vesicular exocytosis in osteoblasts . We interrogated the role of vesicular release in the mechanoresponse using NEM , a modulator of vesicular docking machinery ( Lonart and Südhof , 2000 ) . While NEM is commonly used as an inhibitor of vesicular release ( Genetos et al . , 2005; Kowal et al . , 2015 ) , it has also been shown to increase vesicular release probability ( Kirmse and Kirischuk , 2006 ) through accumulation of readily-releasable vesicles ( Lonart and Südhof , 2000 ) . Using real-time imaging , we demonstrated that NEM increased the basal vesicular density and potentiated mechanically stimulated vesicular exocytosis in osteoblasts . In the presence of NEM , mechanically induced membrane injury and ATP release were significantly reduced . Vesicular exocytosis has been implicated as a mechanism of ATP release from osteoblasts and osteocytes ( Genetos et al . , 2005; Kringelbach et al . , 2015; Romanello et al . , 2005 ) based on the assumed inhibition of vesicular exocytosis by NEM or vesicular trafficking by brefeldin A , bafilomycin or monensin . However , our data demonstrated that the role of vesicles cannot be inferred without a direct validation of the effect of NEM on vesicular release . In our study , [Ca2+]e depletion abolished mechanically stimulated vesicular release , interfered with membrane repair and increased ATP release . Calcium is one of the first mechanically induced signals that initiates downstream responses ( Robling and Turner , 2009 ) , including exocytosis-mediated membrane repair ( Cooper and McNeil , 2015 ) . It was previously shown that during in vivo loading of mouse bone , the number of osteocytes exhibiting [Ca2+]i elevations increased proportionally to the magnitude of applied strain ( Lewis et al . , 2017 ) . We found that PLC and PKCµ , known downstream targets of Ca2+ signaling ( Mochly-Rosen et al . , 2012 ) , regulated basal vesicle abundance and membrane resealing . Our study suggests that the osteoporotic phenotype reported in PKCµ-deficient bones ( Li et al . , 2017 ) may be associated with defective mechanoresponsiveness . Consistent with mechanisms of facilitated membrane resealing ( Togo et al . , 1999; Cooper and McNeil , 2015 ) , we propose that mechanical stimulation results in membrane disruption , which leads to ATP efflux and Ca2+ influx , triggering vesicular exocytosis and membrane repair that then limits ATP release ( Figure 7 ) . Our findings suggest that osteoblast mechano-adaptive status may be influenced by PKCµ/vesicular signaling . Activation of PKC prior to mechanical stimulation or priming the vesicular pool for release with NEM resulted in drastic decreases in membrane injury and ATP release in response to subsequent mechanical stimulation . Prior work has demonstrated that mechanical loading induces the production and release of LAMP-1-positive vesicles from osteocytes ( LAMP-1 colocalizes with the quinacrine tracer-dye used in this study ( Cao et al . , 2014 ) , which was critical for mechano-adaptive bone formation ( Morrell et al . , 2018 ) . This mechano-adaptive effect may underlie cellular accommodation in the bone response to mechanical loading previously reported ( Schriefer et al . , 2005 ) , and explain the reduced responsiveness of osteocytes exposed to higher frequency in vivo mechanical loading , compared to lower frequency loading of the same magnitude ( Lewis et al . , 2017 ) . Our findings suggest that during repeated mechanical stimulation , the priming of PKCµ/vesicular pathway by earlier stimuli will determine osteoblast resilience and responsiveness to subsequent cycles of mechanical stimulation , representing a mechanism for cellular accommodation . In this study , we established the regulatory mechanisms involved in controlling the amount of ATP released following reversible cellular injury . We have demonstrated that membrane injury and repair are common under physiological stresses in vitro and in vivo , and determined the critical role for PKC-regulated vesicular exocytosis in bone cell membrane repair . We suggest that rather than delivering ATP to extracellular space , exocytosis of ATP-containing vesicles limits the much larger efflux of intracellular ATP through damaged membranes . We propose a model of biological adaptation to mechanical forces , which directly links mechanosensation through reversible membrane injury and ATP release to the development of adaptive resilience against the destructive potential of mechanical forces . PKC-mediated vesicular release provides a target for potential therapeutic interventions to modulate mechano-responsiveness and mechano-resilience in humans habitually experiencing altered mechanical environment , such as astronauts or paralysis patients . Phosphate-buffered saline ( PBS; 140 mM NaCl , 3 mM KCl , 10 mM Na2HPO4 , 2 mM KH2PO4 , pH 7 . 4 ) , sterilized by autoclave . Physiological solution ( PS; 130 mM NaCl; 5 mM KCl; 1 mM MgCl2; 1 mM CaCl2; 10 mM glucose; 20 mM HEPES , pH 7 . 6 ) , sterilized by 0 . 22 µM filtration . In [Ca2+]-free physiological solution , CaCl2 was excluded and 10 mM EGTA was added to chelate any residual calcium , sterilized by 0 . 22 µm filtration . Bioluminescence reaction buffer ( 0 . 1 M DTT , 25 mM tricine , 5 mM MgSO4 , 0 . 1 mM EDTA , 0 . 1 mM NaN3 , pH 7 . 8 ) , sterilized by 0 . 22 µm filtration . RIPA lysis buffer ( 50 mM Tris , pH 7 . 4 , 150 mM NaCl , 1% Nonidet P-40 , 1 mM EDTA , 1 mg/mL aprotinin , 2 mg/mL leupeptin , 0 . 1 mM phenylmethylsulfonyl fluoride , 20 mM NaF , 0 . 5 mM Na3VO4 ) . TBST buffer ( 10 mM Tris-HCL , pH 7 . 5 , 150 mM NaCl , 1% Tween 20 ) . The pharmacological interventions ( final concentration , name ( abbreviation ) : molecular target ) used in this study were: 10 µM Gadolinium ( Gd3+ ) : maxi-anion channel , 10 µM Carbenoxolone ( Cbx ) : connexins/pannexins , 10 µM Flufenamic Acid ( FFA ) : connexins , 1 mM 1-Octanol ( Oct ) : connexins , 100 nM GSK 2193874 ( GSK ) : TRPV4 , 100 nM HC 067047 ( HC ) : TRPV4 , 10 µM Nifedipine ( Nif ) : L-type VSCC , 10 µM ML218 ( ML ) : T-type VSCC , 100 µM Suramin ( Sur ) : P2 receptors ( P2R ) , 100 µM PPADS: P2 receptors , 1 µM A740003 ( A7 ) : P2X7 receptor , 10 µM U73122 ( U73 ) : PLC , 100 nM Phorbol 12-myristate 13-acetate ( PMA ) : PKC , 1 µM Bisindolylmaleimide II ( Bis ) : PKC , 1 mM N-Ethylmaleimide ( NEM ) : NSF , 1 µM GsMTx-4 ( GsM ) : Piezo1 . Unless stated otherwise , cells were treated with drugs for 10 mins prior to experiments . We validated the effects of hemichannel blockers ( Figure 4—figure supplement 1 ) and evaluated the effects of the inhibitors on ATP- or citrate-induced [Ca2+]i responses in Fura2-loaded C2-OB cells and on luciferin/luciferase bioluminescence ( Figure 6—figure supplement 5 ) . Three osteoblast models were used in this study . All key experiments required to support our proposed model were performed using primary CB-OB cells; however , additional data from another primary source ( bone-marrow-derived; BM-OB ) and cell-line ( C2-OB ) were reported to emphasize the validity of our findings and demonstrate how different osteoblast models perform in this study . Cells were fixed with formalin ( pH 7 . 4 , 8 min ) and rinsed with PBS ( 3x ) . Alkaline phosphatase ( ALP ) staining solution was prepared by combining 4 . 5 mg fast red violet salt dissolved in 3 . 75 mL H2O and 3 . 75 mL 0 . 2 M Tris-HCl , pH 8 . 3 with 0 . 75 mg Naphthol AS-MX phosphate disodium salt dissolved in 30 µL N , N-dimethylformamide . Sufficient staining solution was added to each well to ensure entire surface was covered ( approx . 500 µL in 35 mm dish ) and incubated at room temperature for 15 min , or until red precipitate formed . Cells were washed with H2O and observed under bright-field microscopy . Alkaline phosphate-positive cells were stained pinkish-red , while yellow hue was observed in alkaline phosphatase-negative cells . Cells plated on glass-bottom plates ( MatTek Corporation ) were loaded with a ratiometric fluorescent calcium dye Fura2-AM ( r . t . , 30 min ) , washed with PS and allowed to acclimatize for 10 min to reduce the effects of mechanical agitation . [Ca2+]i was recorded at a sampling rate of 2 images per second ( 2 Hz ) using a Nikon T2000 fluorescent inverted microscope with 40x objective ( Tiedemann et al . , 2009 ) . Recordings consisted of ~10 s baseline [Ca2+]i and additional 110–170 s following application of mechanical stimulation or treatment . The excitation wavelength was alternated between 340 and 380 nm using an ultra-high-speed wavelength switching illumination system ( Lambda DG-4 , Quorum Technologies ) . Regions of interest ( ROI ) were manually defined and the ratio of the fluorescent emission at 510 nm , following 340 and 380 nm excitation ( f340/f380 ) , was calculated and exported using the imaging software ( Volocity , Improvision ) . All data were imported into an excel spreadsheet for characterization using a MATLAB algorithm previously described ( Mackay et al . , 2016 ) . The following parameters were obtained for statistical analysis: amplitude ( amp; magnitude of response ) , activation time ( t10%-90%; time between 10% and 90% of maximum response ) , and decay constant ( τdecay ) of exponential decay region of deactivation phase of transient response . For micropipette stimulation experiments , secondary responsiveness was calculated as the percentage of neighboring cells exhibiting [Ca2+]i elevations . 10-week-old female C57Bl/6J mice were injected intraperitoneally with LFTR-Dex either 30 min before or 20 min after loading . In vivo cyclic compressive loading at 600 µε or 1200 µε was applied for 5 min ( Willie et al . , 2013 ) . Mice were randomized into two groups , anesthetized , and in vivo cyclic compressive loading was applied to the left tibia ( 216 cycles at 4 Hz the mean mouse locomotor stride frequency [Clarke and Still , 1999] ) , delivering a maximum force of −5 . 5 N and −11 N , which engenders 600 µε or 1200 µε , respectively at the periosteal surface of the tibia mid-diaphysis in these mice determined by prior in vivo strain gauging studies ( Birkhold et al . , 2014 ) . Previous reports have shown strain levels of 200–600 µε are engendered on the medial tibia during walking in the mouse ( De Souza et al . , 2005; Sugiyama et al . , 2012 ) . Strain levels of 1200 µε are considered supraphysiological resulting in a robust bone formation response after 5 days of mouse tibial loading ( Willie et al . , 2013; Brodt and Silva , 2010 ) and altered gene expression after only a single bout of loading ( Zaman et al . , 2010; Kelly et al . , 2016 ) . Mice ambulated freely after controlled loading , and were euthanized 50 min post-dye-injection , weighed ( Figure 5—figure supplement 1A ) and tibiae and calvariae were dissected and fixed in formalin . Tibia lengths were measured ( Figure 5—figure supplement 1B ) and mid-shaft tibiae were cut into fragments , which were immunofluorescently labeled for sclerostin , counterstained with DAPI , and visualized with Nikon T2000 fluorescence inverted microscope . Right non-loaded tibiae were used as an internal control . Female mice were used instead of males to minimize in-cage aggression and related variations in background mechanical-loading . Experimenter was blinded during image analysis . Cell lysates were extracted in RIPA lysis buffer and centrifuged ( 12 000 rpm , 10 min , 4°C ) . Supernatants were collected and protein concentration were determined using Quant-iT protein assay kit ( Invitrogen ) . 70 µg cell lysates were separated on a 10% SDS-PAGE and transferred to a nitrocellulose membrane ( 0 . 45 µm , 162–0115 , Bio-Rat ) using a 10 mM sodium borate buffer . The membranes were blocked in 5% BSA in TBST buffer ( 1 hr , room temperature ) followed by overnight incubation at 4°C with primary antibodies ( 1:1000 dilution , 5% BSA in TBST ) for phosphorylated PKC isoforms: p-PKC ( pan; βII Ser660 ) , p-PKCδ/θ ( Ser643/676 ) , p-PKD/PKCµ ( Ser744/748 ) and p-PKCζ/λ ( Thr410/403 ) , or for total PKC isoforms: PKCα , PKCδ , PKD/PKCµ and PKCζ . Blots were washed and incubated with horseradish peroxidase ( HRP ) -conjugated secondary antibodies for 1 hr at room temperature ( 1:1000 dilution , 5% BSA in TBST ) and visualized with a chemiluminescence system . Cells were scraped in the presence of GAP-junction permeable dye Lucifer yellow ( LY , 10 µM ) , incubated for 2 min ( 37°C ) , rinsed and fixed with formalin ( pH 7 . 4 , 8 min ) . Dye transfer was visualized using wide-field epifluorescence microscopy and cell-cell coupling was visually determined by bright-field microscopy . Cells that were not initially scraped but were physically coupled and LY-positive after 2 min of staining were reported as a percentage of total cells that were physically coupled to scraped neighbors . Data are representative results or means ± standard errors ( S . E . M ) . Curve fittings and [Ca2+]i transient characterization was done in MATLAB ( MathWorks ) . Sample sizes for in vitro experiments indicate the total number of biological replicates ( specified in figure legends as number of stimulated cells or number of independent cultures , which were isolated from different mice for primary OB , or represent different plating dates for C2-OB ) pooled across a minimum of three independent experiments . For in vivo experiments , sample sizes indicate number of animals . Statistical significance was assessed by ANOVA followed by Bonferroni post-hoc test , significance levels are reported as single symbol ( p<0 . 05 ) , double symbol ( p<0 . 01 ) or triple symbol ( **p<0 . 001 ) .	Athletes' skeletons get stronger with training , while bones weaken in people who cannot move or in astronauts experiencing weightlessness . This is because bone cells thrive when exposed to forces . When a bone cell is exposed to a physical force , the first thing that happens is the release of the energy-rich molecule called ATP into the space outside the cell . This molecule then binds to the neighboring cell to unleash a cascade of responses . ATP can exit the cell either through special canals in the cell membrane or released in tiny pod-like structures called vesicles . It is known that strong forces can injure the cell membrane and cause ATP to spill out . However , it is less clear how ATP is released when cells are subjected to regular forces . Mikolajewicz et al . investigated whether ATP exits through injured membranes of cells experiencing regular forces . Bone cells grown in the laboratory were gently poked with a glass needle or placed in a turbulent fluid to simulate forces experienced in the body . Dyes and fluorescent imaging techniques were used to observe the movement of vesicles and calculate the concentration of ATP in these cells . The experiments showed that regular forces in the body do indeed injure the cell membranes and cause ATP to spill out . But importantly , the cells repaired the injuries quickly by releasing vesicles that patch the wound . As soon as the membrane is sealed , ATP stops coming out . From the first injury , cells adapted and quickly strengthened their membrane and repair system to be more resilient against future forces . This process was also seen in the shin bones of mice . These results are important because knowing how bone cells sense , respond and convert physical forces can help us develop treatments for astronauts , the injured and aged .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "physics", "of", "living", "systems" ]	2018	Mechanically stimulated ATP release from murine bone cells is regulated by a balance of injury and repair
CRISPR-Cas systems protect bacteria and archaea from phages and other mobile genetic elements , which use small anti-CRISPR ( Acr ) proteins to overcome CRISPR-Cas immunity . Because Acrs are challenging to identify , their natural diversity and impact on microbial ecosystems are underappreciated . To overcome this discovery bottleneck , we developed a high-throughput functional selection to isolate ten DNA fragments from human oral and fecal metagenomes that inhibit Streptococcus pyogenes Cas9 ( SpyCas9 ) in Escherichia coli . The most potent Acr from this set , AcrIIA11 , was recovered from a Lachnospiraceae phage . We found that AcrIIA11 inhibits SpyCas9 in bacteria and in human cells . AcrIIA11 homologs are distributed across diverse bacteria; many distantly-related homologs inhibit both SpyCas9 and a divergent Cas9 from Treponema denticola . We find that AcrIIA11 antagonizes SpyCas9 using a different mechanism than other previously characterized Type II-A Acrs . Our study highlights the power of functional selection to uncover widespread Cas9 inhibitors within diverse microbiomes . CRISPR-Cas adaptive immune systems are present in many bacterial and archaeal genomes ( Makarova et al . , 2015; Burstein et al . , 2016 ) , where they protect their hosts from infection by phages ( Barrangou et al . , 2007 ) , plasmids ( Marraffini and Sontheimer , 2008 ) , and other mobile genetic elements ( MGEs ) ( Zhang et al . , 2013 ) . CRISPR-Cas systems mediate this defense by incorporating short ( ~30 bp ) spacer sequences from invading genomes into an immunity locus in the host genome . These spacer sequences are then expressed and processed into CRISPR RNAs ( crRNAs ) that , together with various Cas nucleases , mediate homology-dependent restriction of invading genomes . CRISPR-Cas systems are classified into six types ( I , II , III , etc . ) and 30 subtypes ( I-A , I-B , II-C , etc . ) on the basis of functional differences , phylogenetic relatedness , and locus organization ( Makarova et al . , 2018 ) . In response to restriction , phages and other MGEs have evolved dedicated CRISPR-Cas antagonists , called anti-CRISPRs ( Acrs ) ( Bondy-Denomy et al . , 2013 ) , which can promote phage infection , enable horizontal gene transfer ( HGT ) , and thus shape microbial ecosystems ( Borges et al . , 2017; Pawluk et al . , 2017a ) . Acrs that inhibit the type II Cas9 ( Pawluk et al . , 2016a; Rauch et al . , 2017 ) and type V Cas12 systems ( Marino et al . , 2018; Watters et al . , 2018 ) have also garnered significant interest in biotechnology , with demonstrated utility for reducing off-target Cas9 lesions ( Shin et al . , 2017 ) , suppressing gene drives ( Basgall et al . , 2018 ) , and precisely controlling synthetic gene circuits ( Nakamura et al . , 2019 ) . Work over the past decade has revealed much about the activity ( Hille et al . , 2018 ) , distribution ( Makarova et al . , 2015 ) , and evolution ( Koonin and Makarova , 2017 ) of CRISPR-Cas systems . In contrast , comparatively little is known about Acr diversity and function . Known Acrs inhibit only seven of the 30 CRISPR-Cas subtypes ( Bondy-Denomy et al . , 2018; Makarova et al . , 2018 ) , though it is quite likely that unidentified Acrs antagonize the remaining 23 groups ( Pawluk et al . , 2017a ) . In cases where antagonists of CRISPR-Cas systems have been identified , it is also likely that many additional , undiscovered Acrs exist to antagonize those systems ( Watters et al . , 2018 ) . These undiscovered Acrs likely act via a diversity of mechanisms to influence gene flow and phage dynamics in microbial communities ( van Belkum et al . , 2015; Westra et al . , 2016; Borges et al . , 2017 ) , and may unlock new modes of manipulating phage- and CRISPR-Cas-enabled technologies ( Sheth et al . , 2016; Knott and Doudna , 2018 ) . Previous efforts to discover acrs have relied on phage genetics ( Bondy-Denomy et al . , 2013; Pawluk et al . , 2014; Hynes et al . , 2017; He et al . , 2018 ) , linkage to conserved genes ( Pawluk et al . , 2016a; Pawluk et al . , 2016b; Hynes et al . , 2018; Lee et al . , 2018; Marino et al . , 2018 ) or the presence of a self-targeting CRISPR spacer , which would create an unstable autoimmune state if not for the presence of an Acr protein ( Rauch et al . , 2017; Marino et al . , 2018; Watters et al . , 2018 ) . These clever strategies have revealed many candidate acrs . However , they have also highlighted the difficulty of finding new acr genes based on homology , since acrs share little sequence conservation ( Sontheimer and Davidson , 2017 ) . As a result , most acrs almost certainly lie unrecognized among the many genes of unknown function in phages , plasmids , and other MGEs ( Hatfull , 2015 ) . To overcome the challenges associated with anti-CRISPR discovery , we devised a functional metagenomic selection that identifies acr genes from any cloned DNA , based on their ability to protect a plasmid from CRISPR-Cas-mediated destruction . Recently , Uribe et al . ( 2019 ) independently developed a similar Acr search strategy . Because functional metagenomics selects for a function of interest from large clone libraries ( Handelsman , 2004 ) , it is well-suited to identify individual genes like acrs that have strong fitness impacts ( Iqbal et al . , 2014; Forsberg et al . , 2015; Forsberg et al . , 2016; Genee et al . , 2016 ) . This approach may be particularly useful for Acr discovery because Acrs are expressed from single genes and function readily in many genetic backgrounds ( Pawluk et al . , 2016a; Rauch et al . , 2017 ) . Using this functional selection , we find that many unrelated metagenomic clones from human oral and gut microbiomes protect against Streptococcus pyogenes Cas9 ( SpyCas9 ) , the variant used most commonly for gene editing applications ( Knott and Doudna , 2018 ) . We identify a broadly distributed but previously undescribed Acr from the most potent SpyCas9-antagonizing clone in our libraries . This Acr , named AcrIIA11 , binds both SpyCas9 and double-stranded DNA ( dsDNA ) and exhibits a novel mode of SpyCas9 antagonism , protecting both plasmids and phages from immune restriction . We designed a functional selection that can isolate rare acrs from complex metagenomic libraries . We based this selection on the ability of an acr gene product to protect a plasmid , which bears an antibiotic resistance gene , from being cleaved by SpyCas9 ( Figure 1A ) . By screening metagenomes , our selection interrogates core bacterial genomes as well as DNA from the phages , plasmids , and other mobile genetic elements that infect these bacteria , which must contend with CRISPR-Cas immunity . Because most DNA inserts in large metagenomic libraries lack an acr , most clones in a library should be susceptible to SpyCas9-mediated destruction . However , those few DNA inserts that encode and express a functional Acr will resist SpyCas9 and can be recovered using the antibiotic resistance conferred by the plasmid they protect . In our acr selection scheme ( Figure 1—figure supplement 1 ) , we targeted an inducible SpyCas9 nuclease to a kanamycin resistance ( KanR ) gene on a plasmid used to construct the metagenomic libraries . Anticipating that resistance to SpyCas9 cleavage could arise readily via point mutations in target sites , we targeted SpyCas9 to two distinct loci within the KanR gene ( Figure 1B ) to reduce the frequency of these ‘escape’ plasmids . We transformed metagenomic libraries into an Escherichia coli strain that contained SpyCas9 under an arabinose-inducible promoter . After the cells were allowed to recover , we grew them overnight with arabinose to induce SpyCas9 expression . During this phase , cells were not exposed to kanamycin and thus were not under selection to maintain the metagenomic library bearing the KanR gene . We then plated cells on solid media with kanamycin , killing cells in which SpyCas9 cleaved the KanR gene and allowing the recovery of metagenomic DNA from the few remaining KanR colonies . Our analysis revealed that SpyCas9 loss-of-function mutants ( Figure 1C ) dominated the KanR population in early , single-iteration experiments , occurring in approximately 10−4 to 10−5 transformants ( Figure 1—figure supplement 2 , Supplementary file 1 table S1 ) . We therefore added a second iteration of SpyCas9 selection , reasoning that the loss-of-function rate would remain constant across iterations , whereas acr-encoding clones would be enriched by a factor of ~104 with each iteration . In theory , this enables us to select for an acr gene even if it is present just once in a library of ~107 clones . To accomplish these two rounds of SpyCas9 selection , we purified total plasmid DNA from KanR colonies following one iteration and removed the original SpyCas9 expression plasmid via digestion with I-SceI , a homing endonuclease that cleaves an 18 bp target sequence . We then transformed the surviving metagenomic plasmids into a fresh SpyCas9-expressing strain and exposed them to SpyCas9 selection a second time ( Figure 1—figure supplement 1 ) . As expected , adding a second iteration of SpyCas9 selection resulted in a significant enrichment for bona fide SpyCas9 antagonists above background ( Figure 1D ) . We used our functional selection to search for acrs in five metagenomic libraries: two oral metagenomes from Yanomami Amerindians ( Clemente et al . , 2015 ) and three fecal DNA metagenomes from peri-urban residents of Lima , Peru ( Pehrsson et al . , 2016 ) . We subjected each of these libraries , with an estimated 1 . 3 × 106 - 3 . 4 × 106 unique clones per library , to SpyCas9 selection ( Supplementary file 1 table S2 ) . For each library , we observed a 104 to 105-fold reduction in the proportion of KanRcolony forming units ( CFU ) following one iteration of SpyCas9 selection . This value matches the reduction in KanR CFU seen for a GFP control ( Figure 1D ) and the empirically determined frequency of SpyCas9 loss-of-function mutations ( Figure 1—figure supplement 2 , Supplementary file 1 table S1 ) . Since KanR is a measure of plasmid retention , this result indicates that most clones in each library cannot inhibit SpyCas9 . Following a second round of SpyCas9 selection for each library , metagenomic inserts were amplified from pooled KanR colonies by PCR , deep-sequenced , and assembled de novo using PARFuMS ( Parallel Assembly and Re-annotation of Functional Metagenomic Selections ) ( Forsberg et al . , 2012 ) . We used read-coverage over each assembled contig to estimate its abundance following selection . After quality-filtering ( e . g . removal of low-abundance contigs ) , we recovered a total of 51 contigs across all five libraries that putatively antagonize SpyCas9 ( Figure 1—figure supplement 3 , Supplementary file 1 tables S3 , S4 ) . After two rounds of SpyCas9 selection , two libraries ( Oral_5 , Fecal_01A ) seemed more likely to contain new acrs than the other three ( Oral_3 , Fecal_01E , Fecal_03G ) . The Oral_3 library poorly withstood SpyCas9 targeting , so was not studied further ( Figure 1D ) . Although the Fecal_03G library completely resisted SpyCas9 ( Figure 1D ) , subsequent analysis revealed that this was almost entirely due to a single clone that acquired mutations in both Cas9 target sites ( Figure 1B , Figure 1—figure supplement 3 , Supplementary file 1 table S3 ) . Accordingly , just one contig from this library passed quality filters . In contrast , the Fecal_01E library showed intermediate SpyCas9 resistance and was largely devoid of ‘escape’ mutations ( Figure 1D , Supplementary file 1 table S3 ) . However , only one of the 18 contigs from this library was found to be phage-associated ( Supplementary file 1 table S4 ) . Because acrs are expected to originate in phages and MGEs ( Borges et al . , 2017; Pawluk et al . , 2017a; Sontheimer and Davidson , 2017 ) , we predicted that contigs from Fecal_01E were unlikely to contain bona fide acrs and did not prioritize them in this study . These contigs may nonetheless encode anti-SpyCas9 activity , perhaps via Cas9 regulatory factors employed by host bacteria rather than MGEs ( Høyland-Kroghsbo et al . , 2017; Faure et al . , 2019 ) , and therefore could represent a useful resource for probing host regulation of SpyCas9 activity . Contigs from the Oral_5 and Fecal_01A libraries looked most promising for acr discovery . These libraries conferred SpyCas9 resistance at levels 10-fold to 1 , 000-fold above background ( Figure 1D ) . Moreover , plasmids containing these contigs had few escape mutations in Cas9 target sites , so they likely withstood SpyCas9 due to functions encoded by their DNA inserts ( Supplementary file 1 table S3 ) . Finally , many contigs from these libraries were found to be phage-associated ( 30%; Figure 2A , Supplementary file 1 table S4 ) and many encoded genes of unknown function ( Figure 2—figure supplement 1 ) ; these are both hallmarks of known acrs ( Borges et al . , 2017; Pawluk et al . , 2017a; Sontheimer and Davidson , 2017 ) . The contigs identified belonged to 19 bacterial genera . Intriguingly , Cas9 homologs are found in 15 of these 19 bacterial genera ( 79% , Supplementary file 1 table S4 ) , a significant enrichment over the 12% of bacterial genera that harbor Cas9 ( 567/4822 ) in a representative set of ~24 , 000 bacterial genomes ( Mendler et al . , 2018 ) ( p=8×10−20 , χ² test ) . This enrichment strongly supports our hypothesis that many of the recovered contigs encode true acrs rather than artifacts of functional selection , since phages and MGEs must encounter Cas9 to benefit from the protective effect of acrs . To confirm that the Oral_5 and Fecal_01A libraries encoded Acrs , we re-cloned the most abundant contigs from these libraries into new plasmid and strain backgrounds and tested them individually for SpyCas9 resistance . This step eliminated potential mutations to the plasmid backbone , the spycas9 gene , or the host genome , which may have accounted for SpyCas9 resistance in the original screen . For 10 of 14 re-cloned contigs , the DNA insert still protected its parent plasmid from SpyCas9 , confirming the power of our selection to identify novel acrs from metagenomic DNA ( Figure 2A and B ) . Intriguingly , none of the contigs recovered by functional selection encoded homologs of previously identified acrs , nor did they contain homologs of any known acr-associated ( aca ) genes ( Bondy-Denomy et al . , 2018 ) . Among the 10 contigs we confirmed as inhibiting SpyCas9 , several features made us focus on a single contig , F01A_2 ( Figure 2A ) . First , the F01A_2 contig was recovered from our iterative selection of the Fecal_01A library , which conferred near-complete protection against SpyCas9 ( Figure 1D ) . Second , the F01A_2 contig was by far the most abundant contig from this library ( with coverage 216-fold above the median coverage in the library ) , suggesting that it outperformed other contigs during selection . In our re-testing , we confirmed that F01A_2 completely inhibited SpyCas9 activity ( Figure 2B ) . Finally , F01A_2 shares near-perfect nucleotide identity ( >99 . 9% ) with a Siphoviridae phage . This phage infects bacteria from the genus Clostridium_Q , where it is found as a prophage in the genomes of some strains but not in those of close relatives ( Figure 2C ) , suggesting that it is an actively circulating phage . Clostridium_Q is a member of the family Lachnospiraceae described recently in a reassessment of bacterial taxonomy ( Parks et al . , 2018 ) and includes the species Clostridium symbiosum , an important biomarker associated with colorectal cancer progression ( Xie et al . , 2017; Thomas et al . , 2019 ) . As is typical for Acr-encoding loci , the five open reading frames ( ORFs ) on F01A_2 are small , map to accessory regions of phage genomes , and appear to routinely undergo HGT ( Figure 2—figure supplement 2 ) . To identify the ORF ( s ) in F01A_2 responsible for SpyCas9 antagonism , we introduced an early stop codon into each of the five predicted ORFs in the contig and re-tested this set of null mutants for anti-Cas9 activity . Orf_3 completely accounted for SpyCas9 inhibition: a null mutation in orf_3 reduced the frequency of KanR105-fold , to the level of an empty-vector control ( Figure 3A ) . This ORF , which we named acrIIA11 , was sufficient for SpyCas9 antagonism , protecting a target plasmid ( Figure 3B ) from SpyCas9 approximately as well as acrIIA4 , a potent inhibitor of SpyCas9 used in multiple gene-editing applications ( Rauch et al . , 2017; Shin et al . , 2017; Basgall et al . , 2018; Nakamura et al . , 2019 ) . We also investigated the ability of acrIIA11 to restore phage infection in the face of SpyCas9-mediated immunity . We tested SpyCas9’s ability to prevent Mu phage infection in the absence of any Acr or in the presence of either acrIIA4 or acrIIA11 . When SpyCas9 is equipped with a non-targeting crRNA that does not recognize Mu’s genome , the phage successfully infects E . coli . However , when SpyCas9 contains a Mu-targeting crRNA , its expression robustly inhibits phage infection , provided that no Acr is present . When either acrIIA11 or acrIIA4 is expressed , SpyCas9 immunity is nearly completely abolished , confirming the ability of these Acrs to inhibit SpyCas9 and restore phage infection ( Figure 3C , Figure 3—figure supplement 1 ) . AcrIIA11 bears no discernible homology to any previously identified acr ( Borges et al . , 2017; Pawluk et al . , 2017a; Lee et al . , 2018; Uribe et al . , 2019 ) . Since acrIIA11 is a newly discovered acr , we wished to determine its distribution in nature . We therefore searched for homologs in NCBI and in IMG/VR , a curated database of cultured and uncultured DNA viruses ( Paez-Espino et al . , 2017 ) . We identified many proteins homologous to AcrIIA11 in both phage and bacterial genomes and focused on a high-confidence set of homologs , those with ≥35% amino acid identity that cover ≥75% of AcrIIA11’s 182 amino acid sequence ( Figure 4—figure supplement 1 ) . AcrIIA11 homologs have a wider phylogenetic distribution than most previously identified type II-A anti-CRISPRs ( Figure 4A ) and span multiple bacterial phyla ( Figure 4—figure supplement 2 , Supplementary file 1 table S5 ) . We made a phylogenetic tree of AcrIIA11 homologs and found that they clustered into three monophyletic groups ( Figure 4B ) . AcrIIA11 homologs from each group were identified in a variety of MGEs ( Figure 4C ) , consistent with our observation that AcrIIA11 routinely undergoes HGT ( Figure 2—figure supplement 2 , Figure 4—figure supplement 3A ) . Despite these signatures of HGT , the three groups on the AcrIIA11 phylogenetic tree ( Figure 4B ) correspond directly to the three bacterial taxonomic clades in which AcrIIA11 is found ( Figure 4A , Figure 4—figure supplement 2 ) . This concordance between gene and species clusters indicates that , while HGT of acrIIA11 may readily occur across short phylogenetic distances , intra-class and intra-phylum gene flow is rare , in contrast to what has been observed for some other acrs ( Uribe et al . , 2019 ) . AcrIIA11 homologs are only found in bacterial taxa that are highly diverged from Streptococcus . Nevertheless , AcrIIA11 potently inhibits SpyCas9 , even though SpyCas9 is quite divergent from the type II-A Cas9 proteins found in AcrIIA11-encoding taxa . For instance , the only Cas9 protein in Clostridium_Q ( CqCas9 , NCBI accession CDD37961 ) , AcrIIA11’s genus-of-origin , shares just 32% amino acid identity with SpyCas9 . Intrigued by this observation , we tested divergent AcrIIA11 homologs from all three phylogenetic groups against SpyCas9 and found that homologs from each group could inhibit its activity ( Figure 4D ) . Because diverse AcrIIA11 homologs inhibit SpyCas9 , we hypothesized that AcrIIA11 may intrinsically possess the capacity to antagonize a broad set of Cas9 orthologs . To test this hypothesis , we asked whether a panel of AcrIIA11 homologs could also inhibit the type II-A Cas9 protein from Trepenoma denticola ( TdeCas9 , 30/42% amino acid identity to SpyCas9/CqCas9 ) in a plasmid protection assay . Consistent with our predictions , we found that many homologs of AcrIIA11 can inhibit TdeCas9 ( Figure 4—figure supplement 4 ) . This indicates that AcrIIA11 can inhibit divergent Cas9 proteins , as has been observed for other Acrs ( Harrington et al . , 2017; Lee et al . , 2018; Marshall et al . , 2018 ) . Because homologs of AcrIIA11 are found in many genera prevalent within human gut microbiomes ( Figure 4—figure supplement 2 ) , its broad inhibitory range suggests that AcrIIA11 homologs will pose a meaningful barrier to Cas9 activity in this habitat – both in the context of natural phage infections as well as Cas9-based interventions to manipulate microbiome composition ( Sheth et al . , 2016; Pursey et al . , 2018 ) . Type II-A CRISPR-Cas systems are enriched in the Lachnospiraceae relative to other bacterial families ( Figure 4—figure supplement 2 , p=5×10−27 , χ² test ) , which suggests that MGEs in these bacteria regularly encounter type II-A systems . This could explain why all tested AcrIIA11 homologs in the ‘a’ group ( Figure 4B ) inhibited SpyCas9 ( Figure 4D ) . Consistent with their anti-CRISPR activity , two genes from this group , acrIIA11a . 1 and acrIIA11a . 2 , are found within 2 Kb of a putative aca4 homolog ( Figure 4—figure supplement 3A ) , which has been previously associated with multiple acr loci ( Marino et al . , 2018 ) . Because aca-linked loci often encode hot spots of multiple acrs ( Pawluk et al . , 2016b; Borges et al . , 2017; Pawluk et al . , 2017a ) , we also tested 17 genes from these loci for their capacity to inhibit SpyCas9 in a plasmid protection assay . We found that none of these genes encoded strong inhibitors of SpyCas9 ( Figure 4—figure supplement 3B and C ) , although it remains possible that they exhibit anti-CRISPR activity – perhaps against type I-C or III-A systems , as many of the tested genes are found in prophages which encounter these CRISPR-Cas subtypes ( e . g . in Ruminococcus_B lactaris , NCBI accession QSQN01 ) . Considering their relatively recent discovery , it is not surprising that very few Acrs have been mechanistically characterized . Nonetheless , a common theme has emerged from biochemical studies of AcrIIA2 and AcrIIA4 , the only type II-A Acrs for which a mechanism of Cas9 inhibition has been elucidated . These studies have shown that both AcrIIA2 and AcrIIA4 are dsDNA mimics; they inhibit SpyCas9 by binding to its gRNA-loaded form and preventing association with a dsDNA target ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017; Jiang et al . , 2019; Liu et al . , 2019 ) . Given that AcrIIA11 is an unrelated inhibitor , we sought to address whether it has the same mode of SpyCas9 antagonism as these previously-studied type II-A Acr proteins . We therefore purified recombinant AcrIIA11 from E . coli to analyze its biochemistry , interactions with SpyCas9 , and possible modes of inhibition . We first observed that untagged AcrIIA11 migrates at approximately twice its predicted molecular weight by size exclusion chromatography ( Figure 5A ) , suggesting that it may function naturally as a dimer . Consistent with its ability to protect plasmids and phage from SpyCas9 in vivo , purified AcrIIA11 inhibited the dsDNA cleavage activity of SpyCas9 in-vitro , in a concentration-dependent manner ( Figure 5B and C ) . These data demonstrate that AcrIIA11 is sufficient for SpyCas9 inhibition and does not require additional host or phage factors , at least in vitro . This autonomous activity is consistent with acrIIA11’s phylogenomic signature , as it shows no linkage to any other gene across phage genomes ( Figure 2—figure supplement 2 , Figure 4—figure supplement 3A ) . We considered several possibilities for AcrIIA11’s mode of antagonism . AcrIIA11 did not prevent SpyCas9 from binding a single-guide RNA ( sgRNA ) and we did not observe a strong AcrIIA11/sgRNA interaction ( Figure 5—figure supplement 1 ) . Notably , AcrIIA11 did limit migration of the SpyCas9/sgRNA ribonucleoprotein through a native gel , suggesting that AcrIIA11 may bind this complex ( Figure 5—figure supplement 1 ) . So , we next tested whether AcrIIA11 directly binds SpyCas9 ( Figure 5D ) . We found that AcrIIA11 was unable to bind the I-SmaMI meganuclease ( used as a negative control ) but bound both the apo ( without sgRNA ) and sgRNA-loaded forms of SpyCas9 . The addition of sgRNA enhanced AcrIA11 binding to SpyCas9 ( Figure 5D , Figure 5—figure supplement 2 ) but to a lesser extent than previously documented for both AcrIIA2 and AcrIIA4 ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017; Jiang et al . , 2019; Liu et al . , 2019 ) . Given the precedents set by AcrIIA2 and AcrIIA4 , we next tested whether AcrIIA11 acts as a DNA mimic to antagonize SpyCas9 . Both AcrIIA2 and AcrIIA4 have the key property of preventing sgRNA-complexed SpyCas9 from binding target dsDNA ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017; Jiang et al . , 2019; Liu et al . , 2019 ) . To test whether AcrIIA11 functions similarly , we performed an electrophoretic mobility shift assay ( EMSA ) in which we tracked the migration of 6-FAM labeled dsDNA upon SpyCas9 binding . Consistent with previous work ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017 ) , we found that AcrIIA4 prevents the gel-shift caused by SpyCas9/sgRNA binding to its target dsDNA ( Figure 6A , lane 3 ) . In contrast , we observe that AcrIIA11 does not prevent this gel-shift , even at molar ratios that abolish SpyCas9 cleavage activity ( compare lanes 2 through 5; Figure 6A ) . Instead of preventing a SpyCas9-induced gel-shift , we see that AcrIIA11 creates the opposite effect , giving rise to a super-shifted SpyCas9/sgRNA/dsDNA ternary complex ( lane 4 , Figure 6A ) . A similar super-shift occurs using both short and long dsDNA substrates , which differ in the amount of dsDNA unprotected by SpyCas9’s footprint and thus exposed to AcrIIA11 ( Figure 6A , Figure 6—figure supplement 1A ) . Because the shorter substrate has minimal dsDNA overhangs ( five bp ) , AcrIIA11 binding to adjacent dsDNA is not likely to result in the super-shift observed; instead , we conclude that AcrIIA11 binds SpyCas9 to trigger this super-shift ( Figure 6B ) . Consistent with this model , we observe that AcrIIA11 retards the migration of apo- , sgRNA-loaded , and dsDNA-bound SpyCas9 through a native gel in a concentration-dependent manner ( Figure 6C , Figure 6—figure supplement 1B and C ) . These data suggest that AcrIIA11 binds a motif on SpyCas9 preserved across all three conformations of the enzyme . This also distinguishes AcrIIA11 from AcrIIA2 and AcrIIA4 , which bind to the dsDNA binding site of SpyCas9 only available in its sgRNA-loaded form ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017; Jiang et al . , 2019; Liu et al . , 2019 ) . In addition to binding SpyCas9 , AcrIIA11 binds dsDNA weakly in the absence of SpyCas9 ( lanes 8 and 9 , Figure 6A ) , but strongly in the presence of either apo- or sgRNA-loaded SpyCas9 ( compare ‘band #2’ in lanes 4 and 12 with 8 and 11 , Figure 6A ) . We verified this behavior by Western blot , demonstrating that SpyCas9 does not co-migrate with ‘band #2’ , thereby confirming that AcrIIA11 binds dsDNA to generate this gel-shift ( Figure 6C , Figure 6—figure supplement 1A and B ) . Based on these findings , we hypothesize that AcrIIA11 undergoes a conformational change that enhances its dsDNA-binding upon interaction with SpyCas9 . Consistent with this hypothesis , a Coomassie-stained native gel shows that AcrIIA11 forms a discrete , bright band in the presence of SpyCas9 but is more diffuse in its absence ( Figure 6—figure supplement 1D ) , indicating that SpyCas9 may stabilize a conformation of AcrIIA11 . In contrast to our result with dsDNA , we do not detect binding of AcrIIA11 to single-stranded DNA ( Figure 6—figure supplement 1A ) . In summary , we demonstrate that AcrIIA11 can directly bind multiple conformations of SpyCas9 and that SpyCas9 induces AcrIIA11 to bind dsDNA . Though our results leave uncertain which of these behaviors , in isolation or combination , contribute to SpyCas9 antagonism , they make clear that AcrIIA11 uses a different mechanism of inhibiting SpyCas9 than the dsDNA mimicry employed by AcrIIA2 or AcrIIA4 . Because AcrIIA11 functioned in-vitro and in bacteria , we next sought to determine if it inhibited SpyCas9 activity in mammalian cells , which has been shown for a short list of other Acrs ( Rauch et al . , 2017; Hynes et al . , 2018 ) . We therefore transfected HEK293T cells with plasmids expressing SpyCas9 and either a genome-targeting or non-targeting sgRNA . In addition , we co-transfected a second plasmid expressing AcrIIA4 ( as a positive control ) or either of two AcrIIA11 homologs ( AcrIIA11a . 1 or AcrIIA11b . 1 ) . We first tested if these Acrs inhibited SpyCas9 cleavage at the CACNA1D target locus . As expected , we found that SpyCas9 was able to robustly cleave the CACNA1D target locus in the presence of the cognate sgRNA . Additionally , AcrIIA4 was able to potently block this SpyCas9 activity , consistent with previous studies ( Rauch et al . , 2017; Shin et al . , 2017 ) . We also found that AcrIIA11a . 1 , but not AcrIIA11b . 1 , inhibited SpyCas9’s ability to generate genomic lesions as potently as AcrIIA4 , ( Figure 7A and B , Figure 7—figure supplement 1 ) , consistent with the results from our plasmid protection assays ( Figure 4D ) . These findings demonstrate that AcrIIA11 homologs can inhibit SpyCas9 in human cells . Next , we asked whether AcrIIA11’s inhibition of SpyCas9 was universal or locus-specific . We examined SpyCas9 cleavage at a second target locus: the EMX1 gene ( Figure 7C and D ) . As expected , we found that SpyCas9 robustly cleaves the EMX1 target in the presence of the cognate sgRNA . Like with the CACNA1D locus , EMX1 cleavage was also strongly inhibited by AcrIIA4 . In contrast , we found that neither AcrIIA11 homolog could inhibit SpyCas9 activity at the EMX1 target locus . These results further support our in vitro findings that the mechanism of AcrIIA4 and AcrIIA11 are substantially different . For example , AcrIIA4 binds SpyCas9 to prevent DNA target recognition , and can therefore inhibit SpyCas9 at all loci tested ( Figure 7 ) . In contrast , AcrIIA11 homologs can robustly inhibit SpyCas9 only at some loci . This indicates that some feature of the target site ( for instance , its chromatin state ) might impact AcrIIA11’s activity . The identification and characterization of CRISPR-Cas systems has outpaced the discovery of anti-CRISPRs , even though Acr diversity almost certainly exceeds CRISPR-Cas diversity in nature ( Pawluk et al . , 2017a; Watters et al . , 2018 ) . These undiscovered Acrs are likely to be important influences on both phage infection outcomes and horizontal gene transfer in natural microbial communities ( Borges et al . , 2017 ) . Determining the impact of Acrs on these processes is a challenge , however , because acrs lack shared sequence features and so are difficult to detect . To overcome this obstacle and enable function-based acr discovery , we developed a high-throughput selection that identifies acrs based solely on their activity . Applying our functional selection to human fecal and oral metagenomes , we recovered 51 DNA fragments that putatively antagonize SpyCas9 , with 10 confirmed for activity; many additional host- or phage-encoded inhibitors are expected to be among the remaining 41 sequences . In parallel work to ours , Uribe and colleagues recently described a functional metagenomic scheme to identify acrs ( Uribe et al . , 2019 ) . They used a single round of Cas9 selection and one targeting crRNA to identify putative SpyCas9-antagonizing clones , though the authors suspected many of these to be false positives . To find acrs , they selected 39 ideal candidates from their initial dataset for re-testing , confirming 11 DNA fragments for anti-Cas9 activity , from which they identified four new acrs ( acrIIA7-10 ) . Although the premise of both of our approaches is similar , our strategy uses two iterations of SpyCas9 selection and two targeting crRNAs to identify acrs , which suppresses both Cas9 loss-of-function mutations and escape mutations in the target plasmid ( Figure 1D , Figure 1—figure supplement 1 , Figure 1—figure supplement 2 ) . As a result , we confirmed that 10 of the 14 most abundant clones following selection could antagonize SpyCas9 ( Figure 2B ) . More generally , we show that iterative selection can enable high-confidence discoveries from enormously diverse input libraries , which will be critical in the search for even rarer classes of acrs and other phage counter-defense strategies . One of the major lessons from our work and that of Uribe et al is that the vast majority of acrs in nature remain undiscovered . Neither study encountered any previously described acrs and no acrs were shared across datasets , though more than 107 unique clones were screened for antagonists of the same Cas9 allele ( SpyCas9 ) in the same selection host ( E . coli ) . This lack of overlap emphasizes that the few acrs which have been catalogued are a very minor subset of those that are found in nature . This tremendous diversity exists because selection favors a highly-varied repertoire of acrs among even related MGEs ( Bondy-Denomy et al . , 2013; Borges et al . , 2017 ) and because Acrs have evolved independently many times from a variety of progenitor proteins ( Rollins et al . , 2018; Stone et al . , 2018; Uribe et al . , 2019 ) . With extreme diversity favored , and the relative ease by which new Acr function is born , it is almost certain that much more mechanistic variety among Acrs remains to be discovered . In this study , we have focused on the most potent SpyCas9 antagonist from our set — AcrIIA11 — extensively characterizing its phylogenetic distribution , functional range , and mechanism . SpyCas9 is unlikely to be the natural target of AcrIIA11 , as it is quite divergent from the Cas9 homologs found in AcrIIA11-encoding bacteria ( Chylinski et al . , 2014 ) . Yet , diverse AcrIIA11 homologs , which are distributed across multiple phyla , retain inhibitory activity against SpyCas9 . This suggests that AcrIIA11 is intrinsically predisposed to act against a wide variety of Cas9 sequences , a prediction we verified by confirming that many AcrIIA11 homologs can inhibit a highly-diverged Cas9 ortholog from T . denticola ( Figure 4—figure supplement 4 ) . Such inhibitory breadth may be particularly useful to MGEs that infect bacteria with diverse CRISPR-Cas systems ( Makarova et al . , 2015; Crawley et al . , 2018 ) and could explain AcrIIA11’s broad phylogenetic distribution . This combination of prevalence and inhibitory breadth may also impact many Cas9-based interventions ( Sheth et al . , 2016; Pursey et al . , 2018 ) , especially in the gut microbiome , where AcrIIA11 homologs are particularly common . Broad inhibition implies that AcrIIA11 might interact with conserved residues on Cas9 . There is precedent for such biochemical activity in an Acr . For example , AcrIIC1 , a broad-spectrum inhibitor of type II-C Cas9 enzymes , binds to conserved Cas9 residues to prevent nuclease activity but not target DNA binding ( Harrington et al . , 2017 ) . The same outcome is achieved by AcrIE1 and AcrIF3 , two inhibitors of type I CRISPR-Cas systems , as each Acr impairs Cas3 nuclease activity but does not affect target DNA recognition ( Bondy-Denomy et al . , 2015; Pawluk et al . , 2017b ) . No type II-A Cas9 inhibitor has been shown to act similarly , though AcrIIA11’s behavior in-vitro is consistent with such a mechanism; it binds SpyCas9 , inhibits DNA cleavage , but does not prevent target recognition . Alternatively , AcrIIA11’s inhibition of SpyCas9 could be related to dsDNA-binding ability . Though other Acrs can bind nucleic acids , such as AcrIIA1 and AcrIIA6 ( Hynes et al . , 2018; Ka et al . , 2018 ) , AcrIIA11 is the first Acr reported to bind both dsDNA and its target Cas protein ( Stanley and Maxwell , 2018; Trasanidou et al . , 2019 ) . Furthermore , interaction with SpyCas9 dramatically improves AcrIIA11’s dsDNA binding , linking these key behaviors . A role for dsDNA-binding in AcrIIA11’s mode of inhibition could explain why it inhibited SpyCas9 at only some genomic loci ( Figure 7 ) . Chromatin state ( or other locus-specific factors ) would alter AcrIIA11’s access different SpyCas9 target sites which , in turn , would influence its ability to inhibit SpyCas9 at these loci . In comparison , AcrIIA4 interacts with SpyCas9 but not target DNA and , consistent with this model , inhibited SpyCas9 at all tested target sites . While AcrIIA11’s exact mechanism still awaits elucidation , its mode of antagonism is clearly distinct from the other two mechanistically-characterized type II-A Acrs , AcrIIA2 and AcrIIA4 , which act as dsDNA mimics to prevent target recognition ( Dong et al . , 2017; Shin et al . , 2017; Yang and Patel , 2017; Jiang et al . , 2019; Liu et al . , 2019 ) . A combination of Acrs that act via different mechanisms may be more effective at inhibiting CRISPR-Cas activity than a combination of Acrs that act redundantly ( Borges et al . , 2017 ) . Thus , AcrIIA11’s novel mode of antagonism could enable more potent SpyCas9 inhibition than can be achieved using only Acrs that act via dsDNA mimicry . AcrIIA11’s effectiveness against SpyCas9 at some loci in human cells reinforces this possibility and highlights its potential use for modifying new Cas9-based tools in medicine and research . Excitingly , AcrIIA11 represents just the tip of the iceberg . We have identified 51 metagenomic clones that putatively antagonize SpyCas9 ( with 10 confirmed for activity ) ; many of these sequences are likely to contain new SpyCas9 inhibitors , that is , ones not homologous to any known Acr . Functional metagenomics not only offers a powerful means to identify new acr genes , but also enables discovery of fundamentally new ways to inhibit Cas9 . More broadly , our functional metagenomic approach allows us to detect acrs beyond curated sequence databases and can enable the discovery of Acrs against any CRISPR-Cas system of interest , in any microbial habitat , strain collection , or phage bank . This information will improve our understanding of phage and MGE dynamics in microbial ecosystems and holds significant promise for not only describing the longstanding evolutionary arms race between phages and bacteria , but also for improving phage- and CRISPR-Cas-based technologies used in therapeutic and engineering applications . Metagenomic libraries were generously shared by Gautam Dantas ( Washington Univ . in St . Louis ) and have been described previously . Briefly , these libraries were constructed using oral swabs from Yanomami Amerindians ( Clemente et al . , 2015 ) or fecal samples from periurban residents of Lima , Peru ( Pehrsson et al . , 2016 ) . These libraries contain 1 . 3–3 . 9 × 106 unique clones with an average DNA insert size of 2 Kb ( see Supplementary file 1 table S2 for further details ) . To process and store each metagenomic library , the original freezer stock was inoculated into 52 ml of lysogeny broth ( LB; 10 g/L casein peptone , 10 g/L NaCl , 5 g/L ultra-filtered yeast powder ) and grown to an OD600 value of ~0 . 7 ( 300 µl/100 µl inoculum for oral/fecal libraries ) . All libraries were previously constructed in a pZE21 MCS1 plasmid backbone ( henceforth , pZE21 ) marked by kanamycin resistance ( KanR ) ( Lutz and Bujard , 1997; Clemente et al . , 2015; Pehrsson et al . , 2016 ) . Each library was then titered on agar plates containing lysogeny broth with 50 ug/ml Kanamycin; 10–12 ml was used to create replicate freezer stocks . The remaining cells were pelleted by centrifugation at 4100 rcf for 6 . 5 min , purified using Qiagen miniprep kits ( four columns per library ) , and quantified using the Qubit BR dsDNA quantification kit . The libraries Oral_3 and Oral_5 are combinations of two smaller libraries ( Clemente et al . , 2015 ) ; each smaller library was processed independently , the libraries combined in proportion to their number of unique clones , and then were transformed into E . coli . Different plasmids ( generically , pSpyCas9 ) were used to express S . pyogenes Cas9 during the development of the Acr selection scheme and follow-up work . Plasmid composition , applications , and resultant data are detailed in Supplementary file 1 tables S1 , S6 , and Figure 1—figure supplement 2 . Early versions of pSpyCas9 were built by modifying a previously described SpyCas9 expression vector ( addgene #48645 ) ( Esvelt et al . , 2013 ) to target pZE21 with new crRNAs . This vector expresses SpyCas9 , its trans-acting crRNA ( tracrRNA ) , and all crRNAs from the same backbone . The targeting crRNAs were engineered into the original locus ( Esvelt et al . , 2013 ) by Gibson cloning with long-tailed primers . All Gibson assembly was performed using the NEBuilder HiFi DNA assembly mastermix [New England Biolabs ( NEB ) cat# E2621] per manufacturer’s recommendations . The arabinose-inducible araC and pBad regulon was amplified from pU2 ( Lee et al . , 2015 ) and inserted into pSpyCas9 in place of the constitutive proC promoter using Gibson assembly to regulate SpyCas9 expression . For experiments using T . denticola Cas9 ( TdeCas9 ) , the cas9 gene and tracrRNA locus from addgene plasmid #48648 were cloned into the same inducible vector . The crRNA locus was generated by two rounds of Gibson assembly using long-tailed primers: the first round generated the repeat sequence and the second round added the spacer indicated in Supplementary file 1 table S6 . In between these rounds , we noticed a frameshift in the tdeCas9 gene , which we corrected using the Q5 site-directed mutagenesis kit ( NEB cat #E0554 ) . To create the first 2-crRNA pSpyCas9 construct , we ordered a gBlock from Integrated DNA Technologies ( IDT ) containing an I-SceI cut site and second crRNA ( sequence crRNA ‘Z’ in Supplementary file 1 table S6 ) under the control of pJ107106 and a T7TE-LuxIA terminator . This gBlock was cloned into pSpyCas9 by Gibson assembly . Swapping crRNA spacers at this locus was also performed by Gibson cloning with long-tailed primers . The pJ107106 promoter was converted to pJ107111 ( Zucca et al . , 2015 ) using the Q5 site-directed mutagenesis kit ( NEB ) , with suggested protocols . All pSpyCas9 constructs were transformed into Escherichia coli ( strain: NEB Turbo ) by heat-shock and made electrocompetent as described ( New England Biolabs , 2015 ) . In the first iteration of SpyCas9 selection , 360 ng of each metagenomic library was electroporated into a strain of ultra-competent E . coli ( NEB Turbo ) containing the spectinomycin-marked SpyCas9 expression plasmid ( pSpyCas9 ) . As a control , 360 ng of pZE21 expressing GFP was also transformed . We performed all electroporations in a 1 mM cuvette using a Bio-Rad Gene Pulser Xcell with the following settings: 2 . 1kV , 100 Ω , 25 µF . Following electroporation , cells recovered for three hours at 37°C in SOC media ( NEB , cat . # B9020S ) . An aliquot was then taken to titer transformants and the remaining cells ( 860 µl ) were inoculated into 25 ml of LB broth with 50 ug/ml spectinomycin ( Spec ) and 2 mg/ml arabinose . The GFP control was split across two 25 ml flasks , one with and one without arabinose ( 430 µl inoculum per flask ) . After 20 hr in selective conditions , all samples were titered on LB-Spec or LB-Kan/Spec plates . Figure 1D depicts the population proportion of KanRcolony forming units ( cfu ) at 20 hr after SpyCas9 induction relative to their proportion before SpyCas9 induction . To isolate plasmids after one round of selection , colonies from LB-Kan/Spec titer plates ( Supplementary file 1 table S2 ) were collected in 3 ml of LB-Spec by scraping colonies with an L-shaped cell scraper ( Fisher Scientific cat . # 03-392-151 ) to gently remove them from the agar . Colonies were collected from all metagenomic libraries and from the GFP control exposed to arabinose . One-third of the cells were used to make −80°C freezer stocks and plasmids from the remaining 2 ml were purified across two Qiagen miniprep columns into a combined 100 µl of nuclease-free H2O . An I-SceI restriction site was engineered into pSpyCas9 to enable its removal via treatment with the homing endonuclease . For all samples , 51 µl of miniprep eluate was combined with 6 µl NEB cutsmart buffer and 3 µl ( 15 units ) I-SceI , incubated for 20 hr at 37°C , and the reaction heat-killed at 65°C for 20 min . After withholding 5 µl of each sample for gel electrophoresis , 2 . 98 µl of the E . coli RecBCD enzyme ( 149 units , given generously by Andrew Taylor ) , 4 . 9 µl 10 mM ATP , and 1 . 62 µl NEB cutsmart buffer were added to the remaining 55 µl of each sample and incubated for one hour at 37°C . To stop the RecBCD reaction , EDTA was added to a final concentration of 20 mM and the sample incubated at 70°C for 30 min . Linearization of pSpyCas9 by I-SceI and the subsequent removal of linear DNA by RecBCD was confirmed by visualization with gel electrophoresis . Plasmid preparations were then purified through a Zymo Research DNA Clean and Concentrator column and the metagenomic DNA libraries electroporated into the original SpyCas9 selection strain a second time ( Supplementary file 1 table S2 ) . For the second iteration of SpyCas9 selection , electroporation , recovery , induction , titering , and outgrowth was performed exactly as described for the first iteration . From twice-selected metagenomic libraries , LB-Kan/Spec titer plates were used to collect KanR clones from the population . Two titer plates were processed independently for each library and were chosen to have approximately 100 and 10 , 000 colonies . Only one titer plate was collected for the GFP control . For each plate , colonies were collected , stored , and plasmids then purified as described above . No enzymatic removal of pSpyCas9 was performed . Instead , purified plasmids were diluted ten-fold and metagenomic DNA fragments amplified via PCR using 45 µl Platinum HiFi polymerase mix ( Thermo Fischer cat . # 12532016 ) , 1 . 67 µl plasmid template , and 5 µl of a custom primer mix . The custom primer mix contained three forward and three reverse primers , each targeting the sequence immediately adjacent the metagenomic clone site in pZE21 , staggered by one base pair . The stagger enabled diverse nucleotide composition during early Illumina sequencing cycles . A sample PCR contained the following primer volumes , each from a 10 µM stock: ( primer F1 , 5′-CCGAATTCATTAAAGAGGAGAAAG , 0 . 83 µl ) ; ( primer F2 , 5′- CGAATTCATTAAAGAGGAGAAAGG , 0 . 83 µl ) ; ( primer F3 , 5′- GAATTCATTAAAGAGGAGAAAGGTAC , 0 . 83 µl ) ; ( primer R1 , 5′- GATATCAAGCTTATCGATACCGTC , 0 . 36 µl ) ; ( primer R2 , 5′- CGATATCAAGCTTATCGATACCG , 0 . 71 µl ) ; ( primer R3 , 5′-TCGATATCAAGCTTATCGATACC , 1 . 43 µl ) . PCRs were then executed using the following conditions: 94°C for 2 min , 30 cycles of: 94°C for 10 s + 55°C for 30 s + 68°C for 5 . 5 min , and 68°C for 10 min . The amplified metagenomic fragments were cleaned using a Zymo Research DNA Clean and Concentrator column and eluted in 80 µl of Qiagen elution buffer . For each sample , 50 µl of PCR amplicons were diluted to a final volume of 130 µl in Qiagen elution buffer and sheared to a fragment size of approximately 500 bp using a Covaris LE220 sonicator . Sheared DNA was purified and concentrated using a Zymo Research DNA Clean and Concentrator column and eluted in 25 µl nuclease-free H2O . DNA was then end-repaired for 30 min at 25°C using 250 ng sample in a 25 µl reaction with 1 . 5 units T4 Polymerase ( NEB cat . # M0203 ) , five units T4 polynucleotide kinase ( NEB cat . # M0201 ) , 2 . 5 units Taq DNA polymerase ( NEB cat . # M0267 ) , 40 µM dNTPs , and 1x T4 DNA ligase buffer with 10 mM ATP ( NEB cat . # B0202 ) . Reactions were then heat-killed at 75°C for 20 min and 2 . 5 µl of 1 µM pre-annealed , barcoded sequencing adapters added with 0 . 8 µl of NEB T4 DNA ligase ( cat . # M0202T ) . This reaction was incubated at 16°C for 40 min and heat-killed at 65°C for 10 min . The barcoded adapters contained a 7 bp sequence specific to each sample , which allowed mixed-sample sequencing runs to be de-multiplexed . Forward and reverse sequencing adapters were stored at 1 µM in TES buffer ( 10 mM Tris , 1 mM EDTA , 50 mM NaCl , pH 8 . 25 ) and annealed by heating to 95°C and cooling at a rate of 0 . 1 °C/sec to a final temperature of 4°C . Adapters were stored at −20°C and thawed slowly on ice before use . After adapter-ligation , samples were purified through a Zymo Research DNA Clean and Concentrator column and eluted in 12 µl of the supplied elution buffer . Next , 6 µl from each sample was pooled together and the mixture size-selected to ~500–900 bp using a 1 . 9% agarose gel visualized with SYBR-Safe dye ( Thermo Fisher cat . # S33102 ) . DNA was purified from gel slices using a Zymo Research gel DNA recovery kit , eluting in 28 µl nuclease-free H2O . Purified DNA was then enriched by PCR; a sample 25 µl reaction ( in 1x Phusion HF buffer ) contained 6 µl purified DNA template , 0 . 5 µl dNTPs ( 10 mM ) , 0 . 25 µl Phusion polymerase ( two units/µl , Thermo Fisher cat . # F530 ) , and 1 . 25 µl each of the following primers ( 10 µM stock ) : ( primer F , 5’ – AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT ) ; ( primer R , 5’ – CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTGAACCGCTCTTCCGATCT ) . PCRs were amplified for 30 s at 98°C , subjected to 18 cycles of 98°C for 10 s , 65°C for 30 s , and 72°C for 30 s , and then incubated at 72°C for 5 min . Following enrichment PCR , samples were size-selected a second time to ~500–900 bp as previously described and purified libraries quantified using the Qubit fluorimeter ( HS assay ) . Finally , 300 cycles of paired-end sequence data were generated at the Fred Hutchinson Cancer Research Center Genomics Core using a 17 pM loading concentration on an Illumina MiSeq machine with the v3 reagent kit . Illumina paired-end reads were binned by barcode ( perfect match required ) such that data from each titer plate was assembled and annotated independently . Metagenomic DNA fragments from each sample were assembled using the first 93 bp of each read with PARFuMS , a tool developed specifically for assembling small-insert functional metagenomic selections and which was optimized for shorter read lengths ( Forsberg et al . , 2012 ) . The PARFuMS assembler uses three iterations of Velvet ( Zerbino and Birney , 2008 ) with variable job size , two iterations of PHRAP ( de la Bastide and McCombie , 2007 ) , and custom scripts to clean reads , remove assembly chimeras , and link contigs by coverage and shared annotation , as previously described ( Forsberg et al . , 2012 ) . In sum , 206 contigs were assembled from all ten titer plates ( encompassing five selections , Figure 1—figure supplement 3 ) . We predicted open reading frames ( ORFs ) from each contig with MetaGeneMark ( Zhu et al . , 2010 ) using default parameters . Proteins were annotated by searching their amino acid sequences against the TIGRFAMs ( Haft et al . , 2001 ) and Pfam ( Bateman et al . , 2000 ) profile HMM databases using HMMER3 ( Finn et al . , 2011 ) . The highest-scoring profile was used to annotate each ORF ( minimum e-value 1e−5 ) ; these automated annotations were curated by hand into the functional groups shown in Figure 2—figure supplement 1 . The NCBI taxonomies of these contigs were determined using BLASTn against the ‘nt’ database; all contigs could confidently be assigned NCBI taxa at the resolution of order ( Figure 2—figure supplement 1 , Supplementary file 1 table S4 ) . Seven assembled contigs matched pSpyCas9 ( >99% nucleotide identity ) and were removed from the dataset . We also discarded contigs covered by less than two reads/base-pair/million reads ( RPBM ) to ensure contigs assembled from background DNA ( i . e . plasmid backbone , host genome ) were not considered ( the coverage maximum among pSpyCas9-derived contigs was 1 . 67 RPBM , which informed our 2 RPBM cutoff ) . Then , contigs which could be linked to a mutation in either protospacer or protospacer adjacent motif ( PAM ) were eliminated ( see Materials and methods section below ) . Note that we use the term ‘target site’ to refer to a particular protospacer/PAM sequence . Subsequently , redundant contigs assembled across both titer plates from a given library were removed with CD-HIT ( Li and Godzik , 2006 ) , using the following parameters: -c 0 . 95 , -aS 0 . 95 , -g 1 ( the shorter contigs among clusters sharing 95% nucleotide identity over 95% the length of the shorter contig were removed ) . We checked for cross-contamination among the non-redundant contigs with a BLASTn search against the unfiltered contig set , detecting ( and discarding ) just one cross-contaminating contig ( in library Fecal_01E ) . Manual curation of this near-final dataset prompted removal of one chimeric contig and the fusion of two incomplete assemblies . Note that the final dataset may exclude some metagenomic clones that inhibit SpyCas9 , specifically those with mild antagonism or high fitness costs ( rare among surviving colonies , with coverage <RPBM = 2 ) and those which have acquired target site mutations yet also encode SpyCas9 antagonists ( partial antagonism may potentiate target site mutation ) . Indeed , one contig ( O5_7 ) contains a mutation in the PAM of target site B yet still confers SpyCas9 protection when re-cloned into a wild-type vector backbone ( Figure 2B ) . On the basis of empirical antagonism , we include this contig in our final dataset ( to give 51 final contigs across five selections , Supplementary file 1 tables S3 , S4 ) , but exclude all untested contigs linked to target site mutations . While these exclusions may omit bona-fide SpyCas9 antagonism , their exclusion yields a dataset with higher-confidence enrichment for SpyCas9 antagonists . All final assembled contigs have been submitted to GenBank under accession numbers MK637556 - MK637606 ( Supplementary file 1 table S4 ) . Reads used as input for each assembly [i . e . those after adapter-trimming , vector-masking , and spike-in removal by PARFuMs ( Forsberg et al . , 2012 ) were also used to calculate coverage across each contig . For each sample , these reads were mapped to assembled contigs with bowtie2 ( local alignment , unpaired reads , very-fast mode ) and RPBM calculated for each contig . To estimate the proportion of wild-type target sites in each library after two rounds of SpyCas9 selection , sample-specific reads ( before vector-cleaning ) were concatenated by library and mapped to the pZE21 vector with bowtie2 ( end-to-end alignment , unpaired reads , sensitive mode ) . Base frequencies at each protospacer and PAM position were calculated using bam-readcount ( https://github . com/genome/bam-readcount ) , requiring minimum mapping and base quality scores of 25 and 30 , respectively . The proportion of wild-type target sites was conservatively calculated as the product of wild-type base calls at each protospacer/PAM position . In two libraries , ( Oral_3 and Fecal_01E ) , sanger sequencing of individually picked colonies revealed a recombination event between a metagenomic DNA fragment and target site B , which would remove this SpyCas9 target from these clones ( but avoid a frameshift within the kanamycin resistance gene ) . For these two libraries , reads were mapped to a variant of the pZE21 backbone corresponding to each recombination mutant . The proportion of reads mapping across the recombination site was compared to the proportion mapping to the wild-type vector backbone to estimate the frequency of recombination mutants in the population ( 0 . 4% and 24 . 2% for Oral_3 and Fecal_01E , respectively ) . The overall proportion of wild-type target sites in these libraries was then adjusted to reflect the proportion of these recombinants ( Supplementary file 1 table S3 ) . For some of the most abundant contigs in the final dataset , the genotype of the target site sequences in the corresponding plasmid backbone could be determined by Sanger sequencing individual KanR clones following two rounds of SpyCas9 selection . To link additional contigs with target site genotype in the clones not sampled by our Sanger data , we examined paired-end reads that mapped to target site sequences and an assembled contig . Because we used PCR to amplify metagenomic fragments before Illumina library creation , only a small minority of reads mapped to the vector backbone . Additionally , target sites A and B are 514 bp and 365 bp from the clone site of pZE21 , respectively , so only read pairs from long sequencing inserts could link target site and contig genotype . Accordingly , many assembled contigs remain without links to target site genotype ( Figure 2—figure supplement 1 ) , though the overall proportion of wild-type target sites serves as a proxy for the number of SpyCas9 antagonists versus escape mutants in a given library ( Supplementary file 1 table S3 ) . For each titer plate , paired-end reads were mapped to a reference database containing all assembled contigs in both forward and reverse orientations within the pZE21 plasmid . We used bowtie2 in ‘no-mixed’ mode to consider only paired reads with an insert size ≥515 bp when mapping against this reference database ( additional mapping options: end-to-end alignment , paired reads , sensitive mode ) . Only reads with a mapping quality score over ten were considered . We determined the true orientation of a metagenomic DNA fragment by counting whether more reads mapped across the clone-site junctions of the contig in the forward or reverse orientation . If more than one mapped read supported a particular protospacer/PAM variant , the contig linked to that target site was removed from the final dataset . Similarly , if fewer than ten reads mapped to a target site , a single read with any deviation from a wild-type protospacer/PAM was sufficient to eliminate the corresponding contig from further consideration . The assembled metagenomic contigs ( N50 length 2 . 2 Kb ) were too short to provide sufficient information content for phage prediction based on their nucleotide sequence ( Roux et al . , 2015 ) . Instead , protein sequences were used: those with close relatives in predicted phages were used to classify their parent contig as phage-associated . Predicted proteins from each contig were queried against NCBI’s NR database using BLASTp ( on September 12th , 2017 ) and the top five unique hits retained from each query . For each BLAST hit , we downloaded sequence 25 Kb upstream and downstream of the corresponding gene sequence and used these ~50 KB sequences as input for phage prediction with VirSorter ( Roux et al . , 2015 ) . Sequences were compared against VirSorter’s more encompassing ‘virome’ database ( using default parameters ) and hits to bacteriophage of any confidence category ( from ‘possible’ to ‘most confident’ ) were considered phage-associated ( Roux et al . , 2015 ) . The proteins that seeded these bacteriophage predictions were used then to classify metagenomic DNA contigs . If any predicted protein from our assembled contigs had a top-five BLAST hit found in a predicted bacteriophage , the whole metagenomic contig was classified as being phage-associated . These classifications are referenced in Figure 2A and Supplementary file 1 table S4 . All contigs assembled from libraries Fecal_01A and Oral_5 with coverage above the sample median ( RPBM >3 . 6 , Fecal_01A and RPBM >8 . 2 , Oral_5 ) were chosen for validation , totaling 15 contigs . Five contigs were represented in a sparsely-sampled set of individually-picked and Sanger-sequenced colonies . In these cases , the corresponding plasmids were re-transformed into NEB Turbo , target site sequences confirmed as wild-type by additional Sanger Sequencing , and the plasmids co-transformed with pSpyCas9 into NEB Turbo for functional testing . The remaining contigs were amplified by PCR from the plasmid preparations used for Illumina library construction . PCR primers ( Supplementary file 1 table S7 ) targeting the junction between pZE21 and the assembled contigs were used to facilitate re-cloning of the amplified fragments into pZE21 by Gibson assembly . PCRs used the high-fidelity Q5 polymerase ( NEB ) per suggested protocols . Cycling conditions are as follows: 98°C for 3 min , 35 cycles of 98°C for 15 s + 61°C or 65°C for 30 s + 72°C for 3 min , and 72°C for 10 min . Reaction-specific annealing temperatures are listed in Supplementary file 1 table S7 and PCR products of the expected size were purified from gel slices using a Zymo Research gel DNA recovery kit . Nine of ten PCRs were successful [a contig from Fecal_01A ( RPBM = 3 . 66 ) did not amplify well] . All Gibson assembly was performed using the NEBuilder HiFi DNA assembly mastermix per manufacturer’s recommendations . Following Gibson-assembly , sequence-verified clones were co-transformed with pSpyCas9 into NEB Turbo , resulting in 14 total strains re-tested for SpyCas9 antagonism . These 14 strains and a control carrying an empty pZE21 vector were grown overnight in LB with 50 µg/ml kanamycin and spectinomycin ( LB-Kan/Spec ) to maintain the pZE21 variants and pSpyCas9 , respectively . The next morning , cultures were diluted 30-fold into LB-Kan/Spec and grown to log phase ( absorbance readings at 600 nm ranged from 0 . 2 to 0 . 4 ) . Mid-log cultures were then inoculated into LB with 50 ug/ml spectinomycin and either 0 . 2 mg/ml arabinose ( to induce SpyCas9 ) or no arabinose . Kanamycin was omitted to allow for elimination of the pZE21 plasmid . Inocula were normalized by optical density ( a 1:40 inoculum was used for an absorbance reading of 0 . 4 ) . Cultures were then arrayed in triplicate across a 96-well plate ( Greiner cat#655083 , 200 µl per well ) and grown in a BioTek Cytation three plate reader at 37°C with linear shaking at 1096 cycles per minute ( cpm ) . After six hours of growth ( at which time cultures had reached stationary phase ) , each well was serially-diluted ten-fold in peptone-NaCl ( 1 g/L each , pH 7 . 0 ) . Spots ( 5 µl ) of each dilution were plated on agar plates with LB-Kan/Spec ( to count KanR colonies ) and LB-Spec ( to count total colonies ) . Figure 2B depicts the log-transformed proportion of KanR/total cfu with and without SpyCas9 induction . Null alleles of each predicted ORF in F01A_2 were created by inserting an early stop codon into each predicted protein . Stop codons were inserted by PCR using the high-fidelity Q5 polymerase ( NEB ) per suggested protocols with the following cycling conditions: 98°C for 3 min , 35 cycles of 98°C for 15 s + [annealing temperature] for 30 s + 72°C for [extension time] , and 72°C for 10 min . Reaction-specific primers and conditions are listed in Supplementary file 1 table S7 . Constructs were then prepared using the Q5 site-directed mutagenesis kit ( NEB ) or by Gibson assembly ( using the NEBuilder HiFi assembly mastermix , cat# E2621 ) per manufacturer’s recommendations . Final constructs were sequence-verified , co-transformed with pSpyCas9 into NEB Turbo , and re-tested for SpyCas9 antagonism alongside an empty-vector control and the parent pZE21-F01A_2 construct . Plasmid protection assays were performed exactly as described in the section titled ‘Validating contigs with SpyCas9 protection’ , except that SpyCas9 expression was induced using 2 mg/ml arabinose ( rather than 0 . 2 mg/ml ) . The third ORF ( acrIIA11 ) from F01A_2 and acrIIA4 were codon-optimized for E . coli and sub-cloned into the KpnI ( 5’ ) and HindIII ( 3’ ) sites of pZE21 containing the tetR gene to allow for doxycycline-induced expression of the candidate Acrs from the pLtetO-1 promoter . To generate this plasmid , tetR ( and its pLac promoter + rrnB terminator ) was amplified from pCRT7 ( addgene #52053 ) with NEB’s Q5 polymerase per suggested protocols and using the conditions listed in Supplementary file 1 table S7 . This amplicon was then cloned by Gibson assembly into the pZE21 backbone . To clone acrIIA4 , acrIIA11 , and most of the candidate acrs near acrIIA11a . 1 ( depicted in Figure 4—figure supplement 3 ) , KpnI and HindIII sites were added to each gene by PCR with the Q5 polymerase ( see Supplementary file 1 table S7 ) , vectors and genes digested using these enzymes , and sticky-end ligations performed using the Fast-Link ligation kit ( epicentre cat# LK0750H ) per suggested protocols . Genes flanking acrIIA11a . 1 were amplified from the F01A_2 contig , as this contig had proteins identical to those encoded by the genome depicted in Figure 4—figure supplement 3 ( one notable exception: the final 20 aa of orf_d in this figure were truncated in F01A_2 and a different 12 aa C-terminus was cloned , see Supplementary file 1 table S8 ) . This gene was cloned by Gibson assembly rather than KpnI/HindIII digests . The acrIIA11 and acrIIA4 gene sequences were codon-optimized and synthesized as gBlocks from IDT ( the AcrIIA4 amino acid sequence is identical to NCBI accession AEO04689 . 1 ) . So were the candidate acr genes flanking acrIIa11a . 2 ( depicted in Figure 4—figure supplement 3 ) , which we cloned by Gibson assembly into the inducible pZE21_tetR backbone , using vector digested with KpnI and HindIII ( via recommended protocols from NEB ) . AcrIIA11 homologs were codon-optimized for E . coli and synthesized by GenScript . They were also cloned into the pZE21_tetR vector . All constructs were co-transformed with Cas9-expressing plasmids ( Supplementary file 1 table S6 ) into NEB Turbo prior to performing plasmid protection assays . Because NEB Turbo contains the laciq mutation , 0 . 5 mM IPTG was used throughout cloning and handling of the pZE21_tetR plasmid in this strain . IPTG was used because TetR is driven by the pLac promoter in this construct . Including IPTG ensured that TetR was at sufficient intracellular concentrations to maintain repressed transcription from the pLtetO-1 promoter prior to Acr induction ( Acrs were expressed from the pLtetO-1 promoter ) . Overnight cultures of candidate Acr constructs were grown in LB-Kan/Spec + 0 . 5 mM IPTG ( LB-Kan/Spec/IPTG ) to maintain pCas9 and the Acr variants in an uninduced state . The next morning , cultures were diluted 1:50 into LB-Kan/Spec/IPTG and grown at 37°C for approximately two hours to mid-log with 100 ng/ml doxycycline to control Acr expression ( final absorbance readings ranged from 0 . 2 to 0 . 6 ) . In Figure 3B , doxycline was omitted from the indicated sample . Mid-log cultures were then inoculated into LB with 50 ug/ml spectinomycin , 0 . 5 mM IPTG , and either 2 mg/ml arabinose ( to induce Cas9 ) or no arabinose . Doxycycline was included in this medium at 100 ng/ml to maintain Acr expression for previously induced cultures . Kanamycin was omitted to allow for elimination of the pZE21 target plasmid and inocula were normalized by optical density ( a 1:40 inoculum was used for an absorbance reading of 0 . 4 ) . Cultures were grown in a 96-well plate reader and the proportion of KanR cfu determined exactly as described in the section ‘Validating contigs with SpyCas9 protection’ . All figures depicting these data show the log-transformed proportion of KanR/total cfu , both with and without Cas9 induction , for each candidate Acr . Overnight cultures with pSpyCas9 and pZE21_tetR expressing either GFP or an Acr were diluted 1:50 in LB-Kan/Spec/IPTG supplemented with 5 mM MgSO4 and grown shaking at 37C for three hours until late-log phase ( OD600 range 0 . 5–0 . 8 , see Supplementary file 1 table S6 for exact crRNA sequences ) . GFP and the candidate Acrs were induced by adding doxycycline to a final concentration of 100 ng/ul and cultures grown for another two hours before SpyCas9 was induced by adding 0 . 2 mg/ml arabinose . After three additional hours of growth in the presence of both inducers , 200 µl of culture was used in a top-agar overlay , allowed to harden , and ten-fold serial dilutions of phage Mu spotted on top . The top and bottom agar media were made with LB-Kan/Spec/IPTG supplemented with 5 mM MgSO4 , 0 . 02 mg/ml arabinose , and 100 ng/ul doxycycline and contained 0 . 5% and 1% Difco agar , respectively . Plates were incubated at 37°C overnight and plaques imaged the following day . To determine the distribution of AcrIIA1 - AcrIIA11 across bacteria ( Figure 4A and Figure 4—figure supplement 2 ) , homologs of each Acr were identified via a BLASTp search against NCBI nr; all sequences with ≥35% amino acid identity that cover ≥75% of the query length were retrieved . For AcrIIA11 , these thresholds were roughly equivalent to an e-value threshold of 1e−50 following three iterative psi-BLAST searches ( Figure 4—figure supplement 1 ) . We chose to use percent-identity and query-length thresholds to retrieve Acr homologs , rather than psi-BLAST e-values , because the position-specific scoring matrix used in psi-BLAST searches are not comparable across queries . We found that many AcrIIA11 homologs were derived from bacteria with phylogenetically-ambiguous taxonomic labels in NCBI . For instance , the NCBI genus ‘Clostridium’ is extremely polyphyletic , appearing in 121 genera and 29 families in a rigorous re-assessment of bacterial phylogeny performed by the genome taxonomy database , or GTDB ( Parks et al . , 2018 ) . Importantly , the GTDB taxonomy ensures that taxonomic labels are linked to monophyletic groups , applies taxonomic ranks ( phylum , class , order , etc . ) at even phylogenetic depths , and substantially improves problematic taxonomies ( Parks et al . , 2018 ) . To make phylogenetically meaningful inference with respect to AcrIIA11 , we adopted the GTDB taxonomic scheme which , crucially , is largely congruent with NCBI taxonomy outside a few key problem areas ( e . g . Clostridiales , see Supplementary file 1 table S4 ) . The taxonomic assignments for AcrIIA1-6 homologs are unchanged between NCBI and GTDB schema . To map protein homologs onto a bacterial genus tree , the NCBI genome assemblies corresponding to each homolog were downloaded and classified according to the GTDB scheme using the ‘classify_wf’ workflow within GTDB toolkit ( v0 . 1 . 3 ) ( Parks et al . , 2018 ) ; default parameters were used . To visualize these classifications on a tree of life , the minimal phylogenetic tree encompassing all lineages encoding AcrIIA1-6 or AcrIIA11 homologs was downloaded from AnnoTree ( v1 . 0; node ID 31285 ) . The AnnoTree phylogeny additionally includes KEGG and PFAM annotations for nearly 24 , 000 bacterial genome assemblies and is classified according to GTDB taxonomy ( Mendler et al . , 2018 ) . Node 31285 includes four bacterial phyla which are identified with letters on Figure 4—figure supplement 2 ( a: Firmicutes , b: Firmicutes_D , c: Firmicutes_A , d: Firmicutes_F ) . The KEGG identifier K09952 was used to determine which GTDB genera encoded SpyCas9 while the PFAM IDs PF09711 and PF16813 were used to identify taxa which encode Csn2 , the signature protein of type II-A CRISPR-Cas systems ( Makarova et al . , 2015 ) . Enrichment for type II-A CRISPR-Cas systems was determined via a chi-squared test using the number of Csn2-encoding genomes within each GTDB family , with p-values corrected for multiple hypothesis testing using the Bonferroni method . Reciprocal best blast hits were used to assign homology to genes near AcrIIA11 loci ( e . g . in Figure 2—figure supplement 2 and Figure 4—figure supplement 3 ) , with an e-value threshold of 10−4 . In some cases , additional homologous gene pairs were identified via shared annotation , consistent operon structure , and conserved genome organization . The gene tree in Figure 4 and Figure 4—figure supplement 4 was built using the homologs in NCBI ( see above section ) and additionally a set of homologs identified via a BLASTP search of IMG/VR ( the January 1 , 2018 release ) , a curated database of cultured and uncultured DNA viruses ( Paez-Espino et al . , 2017 ) . An e-value cutoff of 1 × 10−10 was used in the IMG/VR homolog search . A preliminary phylogeny of all AcrIIA11 homologs was used to select genes that optimally sample AcrIIA11 diversity for gene synthesis and anti-Cas9 activity screening . The final phylogeny contains unique NCBI homologs and the viral homologs which were selected for gene synthesis ( see Supplementary file 1 table S8 for sequences and accession numbers ) . Alignments were performed using the Geneious ( v8 ) alignment tool and a maximum-likelihood tree was generated with PhyML using the LG substitution model and 100 bootstraps . Codon-optimized AcrIIA4 and AcrIIA11 were cloned into pET15b to contain thrombin-cleavable 6XHis N-terminal tags , with and without C-terminal 2xStrep2 tags . All plasmids were transformed into E . coli BL21 ( DE3 ) RIL cells except the C-terminally-tagged AcrIIA11 variant , which was transformed into a BL21 ( DE3 ) pLysS strain . Overnight cultures were grown in LB/Ampicillin ( 100 µg/mL ) , diluted 100-fold into 1 L pre-warmed LB/Amp media , grown until an OD600 of 0 . 6–0 . 8 , then incubated on ice for 30 min . IPTG was added to 0 . 2 mM , and the cultures where shaken for 18–20 hr at 18°C . The cells were pelleted and stored at −20°C . Cell pellets were resuspended in lysis/wash buffer ( 500 mM NaCl , 25 mM Tris , pH 7 . 5 , 20 mM Imidazole ) , lysed by sonication , and centrifuged for 25 min in an SS34 rotor at 18 , 000 rpm for 30 min . The soluble fraction was filtered through a 5 µm filter and incubated in batch with Ni-NTA resin ( Invitrogen , Cat# R90115 ) at 4°C for 1 hr . The resin was transferred to a gravity filtration column and washed with at least 50 volumes of wash buffer , followed by elution in 200 mM NaCl , 25 mM Tris , pH 7 . 5 , 200 mM Imidazole . The buffer was exchanged into 200 mM NaCl/25 mM Tris , pH 7 . 5 by concentrating and diluting using an Amicon filter ( EMD Millipore , 10 , 000 MWCO ) . Biotinylated thrombin ( EMD Millipore ) was added ( 1 U per mg of protein ) and incubated for 16 hr at 18°C . Streptavidin-agarose was then used to remove thrombin according to the manufacturer’s instructions ( EMD Millipore ) . The proteins were diluted to 150 mM NaCl , 25 mM Tris pH 7 . 5 , loaded onto a 1 mL HiTrapQ column ( GE Life Sciences ) , and eluted by a sodium chloride gradient ( 150 mM to 1 M over 20 mL ) . Peak fractions were pooled and concentrated , bound to Ni-NTA to remove any uncleaved proteins , and the flow-through was purified via size exclusion chromatography ( using Superdex75 16/60 ( GE HealthCare ) or SEC650 ( BioRad ) columns ) in 200 mM NaCl , 25 mM Tris , pH 7 . 5 , 5% glycerol . Figure 5A depicts size exclusion chromatography data for thrombin-cleaved AcrIIA11 lacking a 2xStrep2 tag . Peak fractions were pooled , concentrated , flash frozen as single-use aliquots in liquid nitrogen , and stored at −80°C . Plasmid pMJ806 ( addgene #39312 ) was used to express SpyCas9 containing an N-terminal 6XHis-MBP tag followed by a TEV protease cleavage site . The protein was expressed and purified over Ni-NTA as described above . The eluate from the Ni-NTA resin was dialyzed into 25 mM Tris , pH 7 . 5 , with 300 mM NaCl , 1 mM DTT , and 5% glycerol . It was simultaneously cleaved overnight with homemade TEV protease at 4°C . The cleaved protein was loaded onto a 1 mL heparin HiTrap column ( GE ) in 300 mM NaCl , 25 mM Tris , 7 . 5 and eluted in a gradient extending to 1 M NaCl/25 mM Tris , 7 . 5 . Pooled fractions were concentrated and buffer-exchanged with 200 mM NaCl , 25 mM Tris ( pH 7 . 5 ) , and 20 mM imidazole , and bound to Ni-NTA to remove uncleaved fusion proteins . The flow-through was purified over a Superdex 200 16/60 column ( GE Healthcare ) equilibrated in buffer containing 200 mM NaCl , 25 mM Tris ( pH 7 . 5 ) , 5% glycerol , and 2 mM DTT . Peak fractions were pooled , concentrated to 3–6 mg/mL , flash frozen as single-use aliquots in liquid nitrogen , and stored at −80°C . A second SpyCas9 variant was also expressed and purified from a pET28a backbone ( addgene #53261 ) as described above except that tags were not removed . This second SpyCas9 variant includes an N-terminal 6xHis tag , a C-terminal HA-tag , and a C-terminal NLS sequence . DNA cleavage assays and Acr pulldowns were performed using tagged SpyCas9 , whereas EMSAs used the untagged version . EMSAs were also performed using the tagged SpyCas9 variant and the results did not differ between SpyCas9 purifications . sgRNA was generated by T7 RNA polymerase using Megashortscript Kit ( Thermo Fisher #AM1354 ) . Double-stranded DNA template was generated by a single round of thermal cycling ( 98°C for 90 s , 55°C for 15 s , 72°C for 60 s ) in 50 µl reactions using Phusion PCR polymerase mix ( NEB ) containing 25 pmol each of the following ultramers ( the protospacer-matching sequence is underlined ) : GAAATTAATACGACTCACTATAGGTAATGAAATAAGATCACTACGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCG and AAAAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGC . The dsDNA templates were purified using an Oligo Clean and Concentrator Kit ( ZymoResearch ) and quantified by Nanodrop . Transcription reactions were digested with DNAse , extracted with phenol-chloroform followed by chloroform , ethanol precipitated , resuspended in RNase free water and stored at −20°C . RNA was quantified by Nanodrop and analyzed on 15% acrylamide/TBE/UREA gels . The buffer used in DNA cleavage reactions was NEB buffer 3 . 1 ( 100 mM NaCl , 50 mM Tris-HCl , pH 7 . 9 , 10 mM MgCl2 , 100 µg/mL BSA ) ; proteins were diluted in 130 mM NaCl , 25 mM Tris , pH 7 . 4 , 2 . 7 mM KCl . SpyCas9 ( 0 . 4 µM ) and AcrIIA11 ( 0 . 4–12 . 8 µM ) were incubated for 10 min at room temperature before the reaction was started by simultaneously adding 0 . 4 µM sgRNA and 4 nM linearized plasmid ( 2 . 6 Kb ) and transferring reactions to a 37°C water bath . After 10 min at 37°C , the reaction was stopped by adding 0 . 1% SDS and 50 mM EDTA . Reactions were then run on a 1 . 25% agarose gel containing ethidium bromide at 115V for 2 hr at room temperature . Gels were imaged using the ethidium bromide detection protocol on a BioRad Chemidoc gel imager . The binding buffer for pull-down assays was 200 mM NaCl , 25 mM Tris ( pH 7 . 5 ) ; protein dilutions were made in the same buffer . In 20 µl binding reactions , 160 pmol of SpyCas9 and sgRNA were incubated for 20 min at room temperature , followed by incubation with 210 pmol of strep-tagged Acr for an additional 20 min at room temperature . 50 µl of a 10% slurry of Streptactin Resin ( IBA biosciences #2-1201-002 ) equilibrated in binding buffer was added to the binding reactions and incubated at 4°C on a nutator . Thereafter all incubations and washes were carried out at 4°C or on ice . The beads were washed a total of four times , including one tube transfer , by centrifuging 1 min at 2000 rpm , carefully aspirating the supernatant with a 25 gauge needle and resuspending the beads in 100 µl binding buffer . After the final bead aspiration , Strep-tagged proteins were eluted by resuspending in 40 µl of 1X BXT buffer ( 100 mM Tris-Cl , 150 mM NaCl , 1 mM EDTA , 50 mM Biotin , pH 8 . 0 ) and incubated for 15 min at room temperature . The beads were spun and 30 µl of the supernatant was carefully removed and mixed with 2X SDS Sample Buffer ( Novex ) . Proteins were then separated by SDS PAGE on BOLT 4–12% gels in MES buffer ( Invitrogen ) , followed by Coomassie staining . Reactions were carried out in EMSA binding buffer ( 56 mM NaCl , 10 mM Tris , pH 7 . 4 , 1 . 2 mM KCl , 5% glycerol , 1 mM DTT , 2 mM EDTA , 50 µg/ml heparin , 100 µg/ml BSA , 0 . 01% Tween-20 ) ; proteins were diluted in 130 mM NaCl , 25 mM Tris , pH 7 . 4 , 2 . 7 mM KCl . Omitting MgCl2 ensured that SpyCas9 did not cleave target DNA , as previously described ( Lee et al . , 2018 ) . DNA gel shifts used a 6FAM-labeled 60-mer or 36-mer target dsDNA ( see Supplementary file 1 table S7 for sequence ) and was visualized on a BioRad Chemidoc gel imager . All incubations were carried out at room temperature . SpyCas9 and sgRNA ( each at 2 µM or , in the case of Figure 6—figure supplement 1A , 2 . 25 µM ) were incubated for 25 min , followed by addition of Acrs ( 2-16x molar excess over SpyCas9 ) for 20 min , followed by addition of 20 nM dsDNA template and incubation for 20 min . Samples were loaded on an 8% acrylamide/0 . 5X TBE gel that was pre-run ( 30 min , 90 V , 4°C ) . Reactions were resolved for 160 or 120 min ( Figure 6 and Figure 6—figure supplement 1A , respectively ) at 4°C , 90 V in 0 . 5X TBE buffer . For sgRNA EMSA experiments ( Figure 5—figure supplement 1 and Figure 6—figure supplement 1C and D ) , AcrIIA11 ( at 1–32 µM ) was incubated with 2 µM SpyCas9 for 15 min , followed by incubation with 0 . 2 µM sgRNA for 20 min . The sgRNA was melted at 95°C for five minutes and then slowly cooled at 0 . 1 °C/s to promote proper folding prior to use . sgRNAs used in EMSAs were verified for function in SpyCas9 cleavage assays . The same buffers were used as in DNA EMSAs , except that 3 mM MgCl2 was included . Samples were run for 225 min under the gel conditions described above . The gels were post-stained with a 1:10 , 000 dilution of SYBR-Gold ( Invitrogen ) in 0 . 5X TBE to visualize RNA . Native gels with only apo-SpyCas9 and/or AcrIIA11 ( Figure 6—figure supplement 1C and D ) were prepared and run exactly as described for the sgRNA EMSAs , with nuclease-free water replacing sgRNA . They were then Coomassie-stained or analyzed by SpyCas9 Western blot . To determine the extent to which SpyCas9 migrated through native gels , we transferred total protein to a 0 . 2 µM nitrocellulose membrane using the Bio-Rad Trans-Blot Turbo system ( 25 V , 1 . 3 A for 10 min ) . Membranes were washed with wash buffer ( PBS/0 . 1% Triton-X ) before incubation with a 1:5000 dilution of primary antibody ( monocolonal , N-terminal anti-SpyCas9 , Diagenode cat #C15200229-50 ) in Licor Odyssey Blocking Solution ( part no . 927–40000 ) . Membranes were left shaking for either two hours at room temperature or overnight at 4°C . Then , the membrane was washed four times ( ten-minute washes ) with wash buffer before a 30 min , room-temperature incubation with a secondary antibody conjugated to an infrared dye ( IR800 donkey , anti-mouse IgG , Licor cat# 926–32212 ) . Following three additional washes , blots were imaged on a Licor Odyssey CLx . To verify Acr expression in 293 T cells , we performed Western blots as follows on cells collected 72 hr post transfection . Samples were transfected as described in ‘Mammalian Genome Editing by SpyCas9’ – the CACNA1D or non-targeting sgRNAs were used to load SpyCas9 . Equal cell numbers were harvested across samples and proteins extracted using RIPA buffer with protease inhibitors . Sample input was normalized using a Bradford protein assay that was calibrated with BSA standards . For each sample , 20 ug of total protein was run on a 4–20% Mini-PROTEAN TGX Precast Protein Gel ( Bio-Rad ) . Proteins were then transferred onto a PVDF membrane using a wet Mini Trans-Blot Cell per the manufacturer’s instructions ( Bio-Rad ) . Membranes were blocked in LICOR Odyssey Blocking Buffer ( Neta Scientific ) incubated with the following primary antibodies overnight: rabbit anti-HA antibody ( 1:5000; ICL Lab ) for anti-CRISPRs , mouse anti-FLAG M2 antibody ( 1:2000; Sigma-Aldrich ) for SpyCas9 , and mouse anti-beta tubulin antibody ( 1:2000; Thermo Fisher ) as a loading control . The membrane was washed three times with PBST ( 1X PBS , 0 . 1% Tween 20 ) for 10 min and then incubated with the following secondary antibodies for 2 hr: goat anti-mouse IRDye680 conjugated antibody ( 1:10 , 000 ) and goat anti-rabbit IRDye800 conjugated antibody ( 1:10 , 000 ) . The membrane was washed again with PBST three times for 10 min and imaged on the Odyssey Li-Cor . All oligonucleotides and human codon-optimized gene fragments were purchased from IDT and are listed in Supplementary file 1 table S7 . The EF1a promoter-driven SpyCas9 expression vector was purchased from Addgene [Plasmid #98293; ( Stringer et al . , 2019 ) . The CMV promoter-driven AcrIIA4 expression vector was also purchased from Addgene [Plasmid #113038; ( Bubeck et al . , 2018 ) . Previously-published sgRNAs for CACNA1D ( GCAGGAGUAUUUCAGUAGUG ) and EMX1 ( GAGUCCGAGCAGAAGAAGAA ) were incorporated into the SpyCas9 expression vector using Golden Gate cloning via BsmBI cut sites ( Wang et al . , 2018 ) . The non-target sgRNA ( GUAUUACUGAUAUUGGUGGG ) was cloned identically . The AcrIIA4 expression vector was modified to express AcrIIA11 variants with N-terminal NLS and HA tags using the HiFi Assembly Kit ( NEB ) , via suggested protocols . HEK293T cells were maintained in DMEM ( Thermo Fisher/Gibco ) containing phenol red , 4 mM L-glutamine , 110 mg/L sodium pyruvate , 4 . 5 g/L D-glucose , and supplemented with 10% ( v/v ) FBS ( Thermo Fisher/Gibco ) and 100 U/mL penicillin + 100 μg/mL streptomycin ( Thermo Fisher/Gibco ) . Cell lines were authenticated and tested for mycoplasma contamination before use via the Mycoplasma Detection Kit ( Southern Biotech ) . Transient transfections were performed with Lipofectamine 2000 ( Life Technologies ) , according to the manufacturer’s instructions . Approximately 350 , 000 cells were seeded in each well of a 12-well plate 24 hr prior to transfection to allow cells to become 60–70% confluent at the time of transfection . Wells were transfected with either the anti-CRISPR expression vector , the SpyCas9/sgRNA expression vector , or both using 500 ng for each vector ( 3:1 Acr:SpyCas9/sgRNA plasmid ratio ) and either 1 . 5 µL or 3 µL of Lipofectamine ( 3 μL per μg of DNA ) . Cells were collected and pelleted 72 hr post transfection for genomic DNA extraction using the Wizard Genomic DNA Purification Kit ( Promega ) . The target locus was PCR-amplified using AccuPrime Pfx high-fidelity DNA polymerase ( Thermo Fisher ) and the following PCR conditions: 95°C for 2 min , 35 cycles of 98°C for 15 s + 64°C for 30 s + 68°C for 2 min , and 68°C for 2 min . Reaction-specific primers and conditions are listed in Supplementary file 1 table S7 . Indel frequencies at the SpyCas9 target site were assessed via a T7E1 assay with the EnGen Mutation Detection Kit ( NEB ) , using manufacturer’s recommendations . Reaction products were analyzed on a 1 . 5% SeaKem GTG agarose gel ( Lonza ) and imaged with the InGenuis3 ( Syngene ) . For calculating indel percentages from gel images , bands from each lane were quantified with GelAnalyzer ( version 2010a freeware ) . Peak areas were measured and percentages of insertions and deletions [Indel ( % ) ] were calculated using the formula: Indel ( % ) =100 × ( 1 – ( 1 – Fraction cleaved ) *0 . 5 ) , where Fraction cleaved = ( Σ ( Cleavage product bands ) ) / ( Σ ( Cleavage product bands + PCR input band ) ) .	Viruses that attack bacteria are known as bacteriophages , or phages for short . Bacteria have developed an antiviral immune system called CRISPR-Cas that works by targeting particular genetic sequences , such as those of an invading phage , for destruction . To counteract this immune system , phages have evolved proteins that can block CRISPR-Cas known as anti-CRISPRs . Researchers have studied the CRISPR-Cas bacterial defense systems intensively over the past decade but much less is known about anti-CRISPRs . For example , the natural diversity and prevalence of anti-CRISPRs is still unknown , and identifying these proteins has proven difficult . To address this gap , Forsberg et al . developed a technique to identify new anti-CRISPRs based on their ability to inhibit CRISPR-Cas activity . The method relies on three elements . First , a piece of DNA that lets bacteria resist a specific antibiotic . Second , a test piece of DNA that might code for an anti-CRISPR . Third , a CRISPR-Cas system designed to target and destroy the antibiotic resistance DNA . The three elements are put into bacteria , and two things can happen . If the ‘test DNA’ does not code for an anti-CRISPR , then the CRISPR-Cas system destroys the antibiotic resistance DNA and the bacteria die when exposed to the antibiotic . On the other hand , if the test DNA does code for an anti-CRISPR , it will inhibit the CRISPR-Cas system and the antibiotic resistance DNA will remain intact . This means that the bacteria will survive when grown in the antibiotic , and new anti-CRISPRs can be found by examining the test DNA in those bacteria . Forsberg et al . employed this strategy to screen a huge library of DNA pieces , uncovering several new anti-CRISPRs . They then focused on an anti-CRISPR that was very common in the human gut called AcrIIA11 . Biochemical characterization showed that AcrIIA11 inhibited CRISPR-Cas via a different mechanism from other known anti-CRISPRs . Moreover , it could inhibit CRISPR-Cas systems from many different bacteria . The potential to systematically identify anti-CRISPRs able to resist any bacterium’s CRISPR-Cas defense system could lead to the design of phages that can infect bacteria which are otherwise difficult to destroy . In the future , these phages could be used to clear antibiotic-resistant infections . Beyond its role as an antiviral system in bacteria , CRISPR-Cas is a widely used tool for genetic modification in biomedical research . Using anti-CRISPRs to regulate where , how , and when CRISPR-Cas systems act could make their many emerging applications safer and more effective .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2019	Functional metagenomics-guided discovery of potent Cas9 inhibitors in the human microbiome
CRISPR/Cas9-based genome editing has yet to be reported in species of the Platyhelminthes . We tested this approach by targeting omega-1 ( ω1 ) of Schistosoma mansoni as proof of principle . This secreted ribonuclease is crucial for Th2 polarization and granuloma formation . Schistosome eggs were exposed to Cas9 complexed with guide RNA complementary to ω1 by electroporation or by transduction with lentiviral particles . Some eggs were also transfected with a single stranded donor template . Sequences of amplicons from gene-edited parasites exhibited Cas9-catalyzed mutations including homology directed repaired alleles , and other analyses revealed depletion of ω1 transcripts and the ribonuclease . Gene-edited eggs failed to polarize Th2 cytokine responses in macrophage/T-cell co-cultures , while the volume of pulmonary granulomas surrounding ω1-mutated eggs following tail-vein injection into mice was vastly reduced . Knock-out of ω1 and the diminished levels of these cytokines following exposure showcase the novel application of programmed gene editing for functional genomics in schistosomes . Schistosomiasis is considered the most virulent of the human helminth diseases in terms of morbidity and mortality ( Gryseels et al . , 2006; Hotez , 2014a; Hotez , 2014b; Hotez et al . , 2008; Colley , 2014; Colley et al . , 2014 ) . The past decade has seen major advances in knowledge and understanding of the pathophysiology , developmental biology , evolutionary relationships and genome annotation of the human schistosomes ( Berriman et al . , 2009; Young et al . , 2012; Lepesant et al . , 2011; Geyer et al . , 2011; Vanderstraete et al . , 2014; Rinaldi et al . , 2012a; Rinaldi et al . , 2012b; Protasio et al . , 2012; Valentim et al . , 2013; Wang et al . , 2013; Collins et al . , 2013; Hagen et al . , 2014 ) . Establishing CRISPR/Cas9 genome editing in schistosomiasis would greatly enable effective functional genomics approaches . Stable CRISPR/Cas9-based site-specific gene mutation and phenotyping will drive innovation and a deeper understanding of schistosome pathogenesis , biology and evolution ( Hoffmann et al . , 2014 ) . The schistosome egg plays a central role in disease pathogenesis , causation and transmission ( Gryseels et al . , 2006 ) . The appearance of S . mansoni eggs in host tissues by 6 to 7 weeks after infection coincides with profound polarization to a granulomatous , T helper type 2 ( Th2 ) cell phenotype ( Pearce et al . , 2012; Everts et al . , 2012; Fairfax et al . , 2012; Steinfelder et al . , 2009; Everts et al . , 2009; Pearce et al . , 2004 ) . Numerous egg proteins have been characterized , with >1000 identified in a well-studied fraction termed soluble egg antigen ( SEA ) ( Dunne et al . , 1991; Curwen et al . , 2004; Ashton et al . , 2001; Cass et al . , 2007; Mathieson and Wilson , 2010 ) . In viable eggs , about 30 of the SEA proteins are located outside the developing miracidium and encompass the complement of secreted antigens ( egg-secreted proteins , ESP ) that interact with host tissues to facilitate the passage of the egg from the mesenteric veins to the intestinal lumen ( Mathieson and Wilson , 2010 ) . The T2 ribonuclease omega-1 ( ω1 ) is the principal Th2-inducing component of ESP with its Th2-polarizing activity dependent upon both its RNase activity and glycosylation ( Steinfelder et al . , 2009; Everts et al . , 2009; Wilbers et al . , 2017 ) . This RNase is hepatotoxic ( Fitzsimmons et al . , 2005 ) , and its secretion by eggs into the granuloma regulates the pattern recognition receptor signaling pathways in dendritic cells that , in turn , prime Th2 responses from CD4+ T cells ( Ferguson et al . , 2015 ) . Secreted ω1 provokes granulomatous inflammation around eggs traversing the wall of the intestines , and trapped in hepatic sinusoids and other host organs , driving fibrosis that eventually results in hepatointestinal schistosomiasis ( Gryseels et al . , 2006; Wynn et al . , 2004 ) . As ω1 drives distinctive immunological phenotypes including Th2 polarization and granuloma formation , we investigated the use of programmed CRISPR/Cas9-mediated genome editing ( Jinek et al . , 2012; Hsu et al . , 2014 ) to alter the ω1 locus by both gene knockout and knock-in approaches . The investigation revealed that programmable genome editing catalyzed by the bacterial endonuclease Cas9 was active in schistosomes , with chromosomal double stranded breaks ( DSB ) repaired by homology directed repair ( HDR ) using a donor , single-stranded oligonucleotide template bearing short homology arms and/or by non-homologous end joining ( NHEJ ) . The programmed mutagenesis decreased levels of ω1 mRNA and induced distinct in vitro and in vivo phenotypes , including a substantial loss of capacity of SEA from ω1-mutated eggs to polarize Th2 cytokine responses ( IL-4 and IL-5 ) in co-cultured macrophages and T cells and loss of capacity to provoke formation of pulmonary granulomas in vivo . Functional knock-out of ω1 and the resulting immunologically impaired phenotype showcase the novel application of CRISPR/Cas9 and its utility for functional genomics in schistosomes . Five genomic copies of ω1 were identified in the S . mansoni reference genome , version 5 ( Figure 1—figure supplement 1 ) , although the repetitive nature of the ω1 locus on chromosome one presents a challenge for genome assembly . A single copy of ω1 selected for genome editing , Smp_193860 , included nine exons separated by eight introns and spanned 6 , 196 nt ( Figure 1A ) . Several other copies shared similar exon/intron structure and conserved coding sequences ( Figure 1—figure supplement 1A and B ) . The predicted coding sequence ( CDS ) of Smp_193860 encoded a ~27 kDa protein , of similar mass to the 31 kDa reported for ω1 ( Murare et al . , 1992 ) . The gene encodes a secreted ribonuclease of the T2 family of transferase-type endoribonucleases with conserved catalytic regions ( Luhtala and Parker , 2010 ) . We designed a sgRNA targeting residues 3 , 808 to 3 , 827 of Smp_193860 within exon 6 , adjacent to an AGG protospacer adjacent motif ( PAM ) and with the predicted Cas9 cleavage site located at three residues upstream of the PAM . Figure 2—figure supplement 1 provides the nucleotide sequence of the Smp_193860 copy , and indicates the UTR , coding exons and introns; 6 , 196 nt . The AGG and the nucleotide sequence complementary to this sgRNA were also present in the Smp_179960 and Smp_184360 copies of ω1 . These three copies shared >99% identity in the 202 bp PCR amplicon region subjected to next generation sequencing ( NGS ) although they differed by several substitutions ( Figure 2—figure supplement 2A , B and C ) . All three copies of ω1 display a tight profile of developmental stage expression: expression is restricted to the mature egg with expression not apparent elsewhere during the developmental cycle of this schistosome ( Figure 1—figure supplement 1C ) ( Lu et al . , 2018 ) . The draft genome of S . mansoni was surveyed for key proteins of the non-homologous end joining ( NHEJ ) and homology-directed repair ( HDR ) pathways . Artemis and DNA-PKcs are essential NHEJ factors in vertebrates ( Deriano and Roth , 2013; Lee et al . , 2014; Pardo et al . , 2009 ) . Candidate homologues for six of seven human NHEJ pathway genes and for two key HDR pathway genes were identified by searching for matches to Pfam ( Supplementary file 1 ) . A putative homologue of Cernunnos/XLF was not apparent in S . mansoni ( Deriano and Roth , 2013 ) based on searching for the Pfam XLF domain ( PF09302 ) found in human Cernunnos/XLF . The domain appears to be absent from all flatworm species studied by the International Helminth Genomes Consortium ( International Helminth Genomes Consortium , 2019 ) . The activity and efficiency of CRISPR/Cas9 to edit the schistosome genome , by targeting the ω1 locus , was explored using two approaches . First , a ribonucleoprotein complex ( RNP ) comprising of sgRNA mixed with recombinant Cas9 endonuclease was delivered into schistosome eggs isolated from livers of experimentally infected mice ( eggs termed ‘LE’ , for ‘liver eggs’ ) by electroporation . In addition , homology directed repair ( HDR ) of CRISPR/Cas9-induced double stranded breaks ( DSBs ) at ω1 in the presence of a donor DNA template was investigated ( Lok et al . , 2017; Gang et al . , 2017; Chen et al . , 2015 ) . A single-stranded oligodeoxynucleotide ( ssODN ) of 124 nt in length was delivered to some LE as a template for HDR of chromosomal DSBs ( Figure 1B ) . The ssODN included a short transgene encoding six stop codons flanked by 5’- and 3’-homology arms , each arm 50 nt in length , complementary to the genome sequence of exon 6 on the 5´ and 3´ sides of the Cas9 cleavage site ( Figure 1A and B ) . In a second approach , a lentivirus vector ( pLV-ω1X6; Figure 1A and B ) that included Cas9 , driven by the mammalian translational elongation factor one promoter , and the exon 6-targeting sgRNA ( 20 nt ) , driven by the human U6 promoter ( Duvoisin et al . , 2012 ) was engineered . LE were transduced with pseudotyped lentiviral virions ( pLV ) by exposure in culture to LE for 24 hr ( Rinaldi et al . , 2012a; Suttiprapa et al . , 2016 ) and , thereafter , transfected with the ssODN repair template . In both approaches , expression of ω1 in LE after 72 hr in culture was ascertained . Given that the donor ssODN included a short transgene that facilitates genotyping , PCR was performed using template genomic DNAs from the CRISPR/Cas9-treated LE ( Lok et al . , 2017 ) to reveal the site-specific knock-in ( KI ) . A forward primer termed Smω1X6–6 stp-cds-F specific for the ssODN transgene was paired with three discrete reverse primers , termed Smω1-R1 , Smω1-R2 and Smω1-R3 , at increasing distance from the predicted HDR insertion site in ω1 ( Supplementary file 2 ) . Amplicons of the expected sizes of 184 , 285 and 321 nt were observed in genome-edited eggs but not in control eggs ( Figure 1A and B , Figure 2—figure supplement 3A ) , a diagnostic pattern indicative of the ssODN transgene insertion into ω1 and , in turn , indicating that the resolution of the DSB at the ω1 locus from CRISPR/Cas9 had been mediated by HDR . Amplification using a control primer pair that spanned the predicted DSB , termed Smω1-control-F/R , yielded control amplicons of the expected 991 nt . Similar findings were observed with genome editing delivered by RNPs and by LV ( Figure 2—figure supplement 3 ) . Sanger sequence analysis of the knocked-in amplicons ( KI-R1 , KI-R2 and KI-R3 ) confirmed the presence of the transgene inserted into ω1 at the site targeted for programmed cleavage ( Figure 2C ) . The activity of CRISPR/Cas9 was first evaluated by a quantitative PCR ( qPCR ) approach that relies on the inefficient binding of a primer overlapping the gRNA target , which is where mutations were expected to have occurred , compared to the binding efficiency of flanking primers , which were outside the mutated region ( Shah et al . , 2015; Yu et al . , 2014 ) . The overlapping ( OVR ) primer pair shared the reverse primer with OUT primer ( Figure 3—figure supplement 1A ) . Genomic DNA template was used for qPCR to quantify the efficiency of CRISPR-mediated mutagenesis at the target locus; the ratio between the OVR products and OUT products estimate the relative fold amplification reduction in CRISPR/Cas9-manipulated samples compared to controls in the target sequence of the gRNA . Relative fold amplification was reduced by 12 . 5% in gDNA isolated from eggs treated with pLV-ω1X6 and ssODN , whereas a reduction in relative fold amplification of 2 . 5 , 6 . 9 , and 4 . 5 were observed in eggs treated with gRNA/Cas9 RNP complex alone , gRNA/Cas9 RNP complex and ssODN , or pLV-ω1X6 alone , respectively . A reduction in relative fold amplification was not apparent among control groups , that is untreated eggs , eggs electroporated in the presence of Opti-MEM only , Cas9 only , eggs transduced with heat-inactivated pLV-ω1X6 with ssODN donor , and eggs transfected with ssODN only ( Figure 3—figure supplement 1B ) . To further characterize and quantify the mutations that arose in the genome of ω1 gene-edited eggs , we used an amplicon next generation sequencing approach . Barcoded amplicon libraries were constructed from pooled genomic DNA of six independent exposures of LE to pLV-ω1X6 and the donor ssODN . Each amplicon was sequenced on the MiSeq Illumina platform and the CRISPResso pipeline ( http://crispresso . rocks/ ) ( Pinello et al . , 2016; Canver et al . , 2018 ) was used to analyze deep-coverage sequence reads . More than 56 million sequenced reads were compared to the reference amplicon sequence of the Smp_193860 locus ( Supplementary file 3 ) , which revealed that 71% exhibited the wild type ( WT; i . e . , unmodified DNA ) whereas 29% reads exhibited apparent mutations ( Figure 2D ) across the 202 bp amplicon , with 0 . 13% insertions , 0 . 58% deletions and 28 . 2% substitutions ( Supplementary file 3 , sample 9 ) . In addition , the deletions in treated samples compared to controls are longer around the DSB predicted site ( Figure 2D ) . In contrast , in the control eggs-only group , 76% were WT , and 24% of reads exhibited apparent mutations , with 0 . 14% insertions , 0 . 33% deletions , and 24 . 0% substitutions ( Figure 2E , sample two in Supplementary file 3 ) . Thus , subtracting the rate of apparent mutations in the control , we estimated that 0 . 25% and 4 . 2% of reads in the experimental sample carried programmed CRISPR-induced deletions and substitutions , respectively . Indels of 1–2 bp , or multiples thereof , in coding DNA cause frame-shifts , and consistent with its higher rate of indels , the CRISPR/Cas9-treated sample displayed a higher rate of frame-shifts compared to a sample from control eggs , 2 . 0% versus 1 . 4% ) . Many apparent sequence variants common to the control and edited eggs likely reflect polymorphism among copies of ω1 rather than programmed mutations . The sequence reads revealed several common variants , such as adjacent ‘TA’ substitutions instead of ‘CC’ at positions 152–153 of the amplicon , which encodes a change from K to Q at the amino acid level . The gene Smp_193860 has ‘TA’ at this position in the V5 assembly ( Protasio et al . , 2012 ) , as does the mRNA XP_018647487 . 1 from the NCBI database , whereas Smp_193860 , Smp_184360 and Smp_179960 all have ‘CC’ at this position in the V7 assembly ( Berriman and co-workers , in preparation ) ( Figure 2—figure supplement 2B and C ) . In addition , ‘CC’ was also observed in KI fragments by Sanger direct sequencing ( Figure 2C ) . A second common dinucleotide substitution from ‘AC’ to ‘TT’ at positions 60–61 encodes an amino acid change from T to F . Both dinucleotide substitutions occurred together in 8% of reads in the control group ( Supplementary file 3 , sample 2 ) and 4% of reads in the gene-edited group ( Supplementary file 3 , sample 9 ) . These non-synonymous substitutions may have functional significance given their proximity to the catalytic site of the ribonuclease ( Figure 2—figure supplement 2C ) . Along with the predicted NHEJ-catalyzed mutations , CRISPResso ( http://crispresso . rocks/ ) determined the rate of HDR-mediated ssODN knock-in ( Figure 2F and Supplementary file 3 ) . Here , insertion of the 24 bp transgene was confirmed in 0 . 19% of the reads at the sgRNA programmed CRISPR/Cas9 target site ( Figure 2F , sample nine in Supplementary file 3 ) . Some reads containing the knock-in sequence included the ‘CC’ to ‘TA’ substitutions at positions 152–153 and ‘AC’ to ‘TT’ at positions 60–61 ( Figure 2—figure supplement 2 ) . This indicates that the indels catalyzed by NHEJ and/or KI by HDR occurred in target DNA sequences which exhibited 99% identity in multiple copies of ω1 including Smp_193860 , Smp_184360 and Smp_179960 , and possibly also further copies not yet annotated in the reference genome . The qPCR approach estimated a reduction by 12 . 5% in relative fold amplification in the pLV-ω1X6 with ssODN treatment group ( Figure 3—figure supplement 1 ) , whereas CRISPResso analysis of the pooled NGS reads indicated a frequency of indel/substitution mutation of ~4 . 5% ( Supplementary file 3 ) . Liver eggs ( LE ) transfected with RNP complexes , with or without ssODN , displayed a downregulation of the ω1-specific transcript of ~45% and 81% , respectively , compared to controls ( p≤0 . 05; n = 11 by one way ANOVA ) . However , LE transduced with pLV-ω1X6 virions , with or without ssODN , showed a reduction of the ω1-specific transcripts of 67% and 83% respectively , when compared to controls ( Figure 3A ) . Similar outcomes were seen in all biological replicates undertaken ( n = 11 ) . This outcome indicated that resolution of chromosomal DSB by NHEJ plus HDR provided enhanced programmed gene knockout compared to NHEJ-mediated chromosomal repair alone . Nevertheless , both RNPs and pLV virions efficiently delivered programmed gene editing to schistosomes but lentiviral transduction delivered enhanced performance with stronger gene silencing , in particular when the donor repair template was provided ( Figure 3A ) . When examined at later time points ( days 5 and 7 following manipulation of the LE ) , further reduction in ω1 abundance was not apparent ( Figure 3B ) . Large DNA deletions have been associated with CRISPR/Cas9 mutations in another helminth species ( Gang et al . , 2017 ) . However , using qPCR to estimate relative copy number , as previously described ( Suttiprapa et al . , 2012a ) , there were no evidence that silencing of ω1 was associated with the ω1 multiple gene copy numbers ( Figure 2—figure supplement 4 ) . The ribonuclease activity of the ω1 glycoprotein in SEA is associated with the Th2-polarized immune response that underpins the appearance of schistosome egg granulomata ( Everts et al . , 2012; Steinfelder et al . , 2009; Everts et al . , 2009 ) . Ribonuclease activity of SEA from control and experimental groups on substrate yeast RNA was investigated following CRISPR/Cas9 programmed mutation of ω1 mediated by the RNP and the pseudotyped lentiviral approaches with or without ssODN . Intact yeast RNA was evident in the DNase-RNase free condition ( negative control ) , indicating absence of RNase activity in the reagents ( 200 ng yeast RNA at the outset ) . There was near complete hydrolysis of yeast RNA following exposure to RNase A ( positive control ) ;~1 . 4 ng of RNA remained intact . Wild type SEA exhibited marked RNase activity against the yeast RNA;~70 ng RNA remained intact after one hour , corresponding to >60% digestion . Incubation of the RNA with Δω1-SEA ( i . e . SEA from the gene edited eggs ) from the experimental groups , RNP , RNP +ssODN , pLV-ω1X6 , and pLV-ω1X6 + ssODN , resulted in ~30% substrate digestion , with 124 , 140 , 135 and 153 ng of RNA remaining , respectively . All conditions for programmed genome editing resulted in less digestion of the yeast RNA than for wild type SEA ( p≤0 . 0001 ) ( Figure 3C and D ) . Moreover , the Δω1-SEA with programmed knock-in exhibited less RNase activity than Δω1-SEA prepared without the donor ssODN repair template ( p≤0 . 01 , n = 6 by one-way ANOVA ) . The ω1 ribonuclease alone is capable of conditioning human monocyte-derived dendritic cells to drive Th2 polarization ( Everts et al . , 2012 ) and enhanced CD11b+ macrophage modulation of intracellular toll like receptor ( TLR ) signaling ( Ferguson et al . , 2015; Ferguson et al . , 2016 ) . Ribonuclease ω1 inhibited TLR-induced production of IL-1β and redirected the TLR signaling outcome toward an anti-inflammatory profile via the mannose receptor ( MR ) and dectin ( Everts et al . , 2012; Zaccone et al . , 2011; Ritter et al . , 2010 ) . The human monocytic cell line , THP-1 , and the Jurkat human CD4+ T cell line were employed to investigate the interaction of antigen-presenting cells and T cells ( Qin , 2012; Fuentes et al . , 2002 ) . At the outset , the THP-1 cells were differentiated to macrophages for 48 hr , and subsequently pulsed with SEA or Δω1–SEA for 48 hr . Thereafter , the Jurkat CD4+ T cells were added to the wells , and the co-culture continued for 72 hr . Representative cytokines , including IL-4 , IL-5 , IL-13 , IL-2 , IL-6 , IL-10 , TNF-α and IFN-γ , were quantified in supernatants of the co-cultures ( Figure 4 ) . SEA from ω1-mutated eggs reduced levels of Th2 cytokines , including IL-4 and IL-5 , in comparison to wild-type SEA ( p≤0 . 01 ) , and a trend toward less IL-13 production was also observed ( Figure 4 ) . Reduced levels of IL-6 and TNF-α were also observed ( p≤0 . 01 , n = 4 by one-way ANOVA ) . By contrast , significant differences in IL-10 and IL-2 were not evident between the WT- and mutant-SEA groups . IFN-γ was not detected following pulsing with the WT-SEA or mutant-SEA ( Figure 4—figure supplement 1 ) . Following the entrapment of eggs in the intestines , liver and eventually lungs , the glycosylated ω1 ribonuclease represents the principal stimulus that provokes the development of the circumoval granuloma , necessary for extravasation of the eggs ( Doenhoff et al . , 1986 ) . A long-established model of the schistosome egg granuloma employs tail vein injection of eggs into mice , which leads to formation of circumoval granuloma in the mouse lung ( Boros and Warren , 1970; Eltoum et al . , 1995; Wynn et al . , 1993 ) . The latter approach has been extensively employed for immunopathogenesis-related studies of ω1 ( Hagen et al . , 2014 ) . Accordingly , to elicit circumoval granulomas , ~3000 WT or Δω1 LE were injected into the lateral vein of the tail of BALB/c mice . The mice were euthanized 10 days after injection , and the entire left lung was removed , fixed , sectioned , and stained for histological analysis ( Figure 5 ) . Representative digital microscopic images of the whole mouse lungs acquired through high-resolution 2D digital scans are presented in Figure 5 , panels A-G . At low magnification ( 2× ) , much more severe and widespread inflammation was visible in lungs exposed to WT eggs compared to Δω1-eggs . In addition , markedly more intense and dense focal inflammation was induced by WT compared to Δω1-eggs ( Figure 5B ) . Granulomas were not seen in control naive mice not exposed to schistosome eggs ( Figure 5C ) . At 20× magnification , vast disparity in volume of the circumoval granulomas was observed for WT versus Δω1 LE ( Figure 5A1–A2 , D , E and Figure 5B1–B2 , F and G ) . The volume of granulomas surrounding single schistosome eggs was quantified; those surrounding WT eggs in lungs of the mice were 18-fold greater than for Δω1-eggs , 21×10−2 ± 1 . 61×10−3 mm3 and 0 . 34 × 10−2 ± 0 . 12×10−4 mm3 ( mean ±S . E . , 17–26 granulomas per mouse ) , respectively ( p<0 . 0001 , n = 103–130 by Welch’s t-test ) ( Figure 5H ) . The experiment was repeated with 3–4 mice per group with a similar outcome . The findings documented marked deficiency in the induction of pulmonary granulomas by the Δω1 compared to WT eggs of S . mansoni . This report , and the accompanying article on the liver fluke Opisthorchis viverrini ( Arunsan et al . , 2019 ) pioneer programmed genome editing using CRISPR/Cas9 of trematodes and indeed genome editing for species of the phylum Platyhelminthes . Using S . mansoni as an exemplar , here we have demonstrated the activity and feasibility of gene knockout and knock-in in schistosomes . Programmed on-target editing was evidenced by site-specific mutations at the ω1 locus on chromosome 1 . The chromosomal lesion was repaired by the non-homologous end joining ( NHEJ ) pathway ( Lieber , 2010 ) in the absence of a donor oligonucleotide and by homology directed repair ( HDR ) when a single-stranded oligonucleotide donor template was provided ( Paquet et al . , 2016; Zhang and Matlashewski , 2015; Yoshimi et al . , 2016 ) . To investigate the feasibility of genome editing , schistosome eggs were transfected with ribonucleoprotein particle ( RNP ) complexes and with lentiviral virions carrying the CRISPR/Cas9 components , in similar fashion to earlier reports in cell lines , tissues and entire organisms ( Kosicki et al . , 2017; Holmgaard et al . , 2017; Yu et al . , 2018; Luo et al . , 2018 ) . Delivery by RNPs facilitates immediate editing , although the short half-life of the RNP components may be disadvantageous . Delivery by plasmids and by viral-mediated infection may provide sustained Cas9 activity , transgene integration in non-dividing cells , and other advantages ( Hsu et al . , 2014; Shalem et al . , 2015 ) . Transfection by LV provides a hands-free approach to enable scaling of gene editing and the potential that less accessible and/or differentiated cells can be reached . To examine the efficiency of programmed genome editing , several parallel approaches were undertaken including NGS- and quantitative PCR-based analysis of pooled genomic DNAs , analysis of levels of ω1-transcripts and ribonuclease , and immuno-phenotypic status of cultured cells and of mice exposed to gene edited eggs . Analysis of the deep-coverage nucleotide sequence reads of amplicons spanning the predicted DSB site in the ω1 locus revealed that ~4 . 5% of the reads were mutated by insertions , deletions and substitutions . The target locus was mutated by knock-in ( KI ) of a ssODN repair template bearing short homology arms to mediate homology-directed repair following DSB at ω1 , with an efficiency for HDR of 0 . 19% insertion of the donor transgene . Numerous substitutions in addition to deletions and insertions were seen , some of which may represent single nucleotide polymorphisms ( SNPs ) among the gene copies rather than genome editing-induced changes . As well as the anticipated ease of access by the RNPs , virions and the donor template ssODN , to cells proximal to the eggshell compared to the cells deeper within the egg , other factors also may have contributed to unevenness of CRISPR efficiency among the eggs and the cells within individual eggs . Given the presence of multiple copies of ω1 in the genome , the diverse organs , tissues and cells comprising the mature egg ( Jurberg et al . , 2009 ) , the spectrum of development of the eggs in LE , and other factors , a mosaic of mutations would be expected . Some alleles might display HDR but not NHEJ , some NHEJ but not HDR , others both NHEJ and HDR , and others retain the wild-type genotype . HDR proceeds at cell division , with the cell at late S or G2 phase following DNA replication where the sister chromatid serves as the repair template; otherwise , NHEJ proceeds to repair the DSB ( Heyer et al . , 2010 ) . Sequence analysis of individual schistosomes rather than pools of the parasites would provide more information , including transfection/transduction efficiency and parasite-to-parasite mutation rates . However , the analysis would be constrained by the number of parasites that could be investigated at the genome level in terms of effort necessary for the numerous NGS groups in the computational analysis . Here , many thousands of individual eggs were simultaneously subjected to gene editing in order to provide sufficient quantities of RNA and gDNA for the downstream transcript level and NGS-based gene editing analyses . Although it is technically feasible to consider analysis of individual schistosome eggs , it would be challenging to reliably recover enough nucleic acids for the downstream analysis . Nonetheless , genotyping individual eggs or other developmental stages of the schistosome would be informative and straightforward to confirm the presence of transgenes following HDR of donor templates . Moreover , droplet digital PCR ( ddPCR ) -based analysis , a more sensitive approach than used here , is expected to provide a more sensitive and reliable quantification of gene-editing mutations . The ddPCR approach provides simultaneous assessment of both HDR and NHEJ , the repair pathways that resolve Cas9-catalyzed , double-stranded breaks , and also enable investigation of multiple , simultaneous editing conditions at the target locus ( Dibitetto et al . , 2018; Miyaoka et al . , 2018 ) . Newer technologies including single-cell genome sequencing will be able to precisely define not only individual target cells that were mutated , but also identify which kind of mutations arose in these individual cells ( Gawad et al . , 2016 ) . Expression levels of ω1 were diminished by as much as 83% relative to controls suggesting that Cas9 catalyzed the mutation of ω1 , and that programmed Cas9-induced DSBs had been resolved by NHEJ and/or HDR . Knock-in of the ssODN repair template induced 81–83% reduction in ω1-specific mRNA levels , whereas downregulation of 45% to 67% followed the exposure of eggs to RNP or lentivirus without ssODN . Ribonuclease activity in ω1-mutated eggs was likewise significantly diminished . Curiously , less than 5% efficiency ( NGS findings ) in gene editing appeared to account for this markedly reduced ( >80% ) gene expression . The possibility of large-scale deletions , as reported in Strongyloides stercoralis ( Gang et al . , 2017 ) , may provide an explanation for this apparent paradox . However , analysis of copy number by qPCR failed to reveal apparent differences for copy numbers of ω1 among treatment and control groups of eggs . An alternative explanation may be the tight , stage- and tissue-specific expression of ω1 . Within the mature egg , the fully developed miracidium is surrounded by a squamous , syncytial epithelium termed variously as the envelope , the inner envelope , or von Lichtenberg’s envelope ( Ashton et al . , 2001; Mathieson and Wilson , 2010; Jurberg et al . , 2009; Neill et al . , 1988 ) . The inner envelope is metabolically active ( Mathieson and Wilson , 2010 ) and is considered to be the site of synthesis of the ω1 T2 ribonuclease that is released from the egg into the granuloma ( Ashton et al . , 2001; Mathieson and Wilson , 2010; Fitzsimmons et al . , 2005 ) , along with other secreted/excreted proteins that facilitate egress of the egg from the venule and through the wall of the intestine ( Mathieson and Wilson , 2010 ) . Expression of ω1 is tightly , developmentally regulated: based on comparison of expression in the egg to expression in the miracidium , immunostaining and transmission electron microscopy of the mature egg , and on meta-analysis of transcriptomics sequence reads , all or most expression of ω1 occurs solely in the mature egg . Expression does not occur earlier in the embryo or immature egg or other developmental stages ( Everts et al . , 2009; Ashton et al . , 2001; Mathieson and Wilson , 2010; Fitzsimmons et al . , 2005; Lu et al . , 2018; Schramm et al . , 2009 ) ( Figure 1—figure supplement 1C ) . The virions may have transduced only a small number of the cells within the mature egg because , following entry into the egg , the virus would be expected to first contact the inner envelope , rather than traveling further to the cells of the miracidium . Accordingly , the efficiency of gene editing may have ranged from high to low – from the exterior of the egg to its center through the cells of the egg due to ease of virus access from the culture supernatant , and displayed highest efficiency in the envelope where ω1 is expressed . In any case , this apparent paradox deserves deeper inquiry , for example by confirming the site of expression of ω1 by immunolocalization in the inner envelope of the fully mature egg . Following mating of the adult female and male S . mansoni , schistosome oviposition commences at ~35 days after infection of the mouse . Thereafter the female schistosome continuously releases several hundred eggs each day . When released into the mesenteric veins from the adult female , the schistosome ovum contains a single-celled zygote surrounded by ~40 vitelline cells of maternal origin ( Mathieson and Wilson , 2010 ) . By six days later , the miracidium has developed within the eggshell into a multi-cellular , mobile , ciliated larva composed of organs , tissues , muscles and nerves ( Ashton et al . , 2001; Jurberg et al . , 2009; Neill et al . , 1988; Mann et al . , 2011 ) . Productive infection and transmission of schistosomiasis require the translocation of the egg across the wall of the bowel . However , many eggs fail to escape from the venules and are transported by the portal circulation to the sinusoids of the liver where they become entrapped . Maturation of the egg from the single-cell zygote within the eggshell at the time of oviposition to the fully mature miracidium within the eggshell takes about 7 days in vitro ( Rinaldi et al . , 2012a; Jurberg et al . , 2009 ) and presumably a similar duration in vivo . Here , at necropsy of the mice , we anticipated that the LE preparation of eggs would include a spectrum of embryonic development of the schistosome egg from the blastula stage to mature miracidium containing organs , tissues and nerves . The fully formed egg is laid still undeveloped , without any cleavage . Once released from the female schistosome , eight embryonic stages can be defined . Stage one refers to early cleavages and the beginning of yolk fusion , whereas stage eight refers to the fully formed larva , presenting muscular contraction , cilia , and flame-cell beating ( Jurberg et al . , 2009 ) . The envelope of the egg , the site of expression of ω1 ( 28 , 30 ) first appears at stage three and is developed and secretory at the later stages ( Jurberg et al . , 2009 ) . To address the effects on the hallmark Th2 polarization of the immune response to schistosomiasis ( MacDonald et al . , 2002; MacDonald and Pearce , 2002 ) , co-cultures of human macrophages and T cells were exposed to Δω1-SEA and mice were exposed to Δω1-eggs . Whereas wild type SEA polarized Th2 cytokine responses including IL-4 and IL-5 in the co-cultures , significantly reduced levels of these cytokines were observed after exposure to ω1-mutated SEA . Moreover , following introduction of eggs into the tail vein of mice , the volume of pulmonary circumoval granulomas around Δω1 eggs was enormously reduced compared to those provoked by wild-type eggs . Although this outcome extends earlier findings using lentiviral transduction of eggs of S . mansoni to deliver microRNA-adapted short hairpin RNAs aiming to silence expression of ω1 ( Hagen et al . , 2014 ) , the decrement in granuloma volume was vastly more marked in the present study . In addition to incomplete disruption of all copies of ω1 , residual granulomas containing mutant eggs may be due to the presence of other Th2-polarizing components within SEA ( Kaisar et al . , 2018 ) . Given that the T2 ribonuclease ω1 is the major type 2-polarizing protein among egg-secreted proteins ( Steinfelder et al . , 2009; Everts et al . , 2009 ) , these findings of a phenotype characterized by the absence of or diminutive pulmonary granulomas provide functional genomics support for this earlier advance ( Everts et al . , 2012 ) . Nonetheless , given that hepatointestinal disease is characteristic of the infection with S . mansoni , changes in granuloma formation in the liver and intestines triggered by mutant egg will be prioritized in future studies . This study provides a blueprint for editing other schistosome genes and those of parasitic platyhelminths at large . However , hurdles remain . We suggest the following ( non-inclusive ) list of near term priorities with respect to development of programmed gene editing for functional genomics in flatworms: the delivery of programmed genome editing to the germ line , including transgenes that confer constitutive or conditional expression of Cas9 ( Idoko-Akoh et al . , 2018; Chaverra-Rodriguez et al . , 2018 ) ; HDR-catalyzed insertion of surrogate reporter , antibiotic resistance markers ( Yan et al . , 2018 ) or antimetabolites to facilitate drug selection of the mutants ( Yoshimi et al . , 2016; Wu et al . , 2014; Yan and Finnigan , 2018 ) ; and dual guide systems ( Teixeira et al . , 2018b; Teixeira et al . , 2018a; Gong et al . , 2017 ) . Focusing on schistosomes , establishment of lines of transgenic S . mansoni by retroviral-based transgene integration into the zygote within the newly laid egg has been described ( Rinaldi et al . , 2012a ) . Beyond the zygote , the germ line is also comparatively more accessible at several developmental stages , including the daughter sporocysts ( Wang et al . , 2013 ) . Concerning somatic tissue or whole worm gene editing approaches , synchronizing the miracidial development in the cultured egg is straightforward by maintaining LE in vitro for 7 days . In this way , the reproducibility of somatic editing might be improved , as mosaicism outcomes would be reduced since each mature egg enclosed the fully developed miracidium with the full complement of cells ( Jurberg et al . , 2009 ) . For HDR , jointly supplying both anti-sense and sense long single-stranded donor DNAs in trans should increase the likelihood of obtaining mutant homozygotes rather than heterozygotes ( Paquet et al . , 2016; Gantz and Bier , 2015 ) . In turn , HDR focused programmed gene editing to generate bi-allelic mutation would hasten the heritable spread of transgenes to F2 and beyond . Gene editing the germ line followed by establishment of mutant schistosomes represents a powerful strategy to determine the role and potential fitness cost that result from deletion of the ω1 gene . However , given the established role of the ω1 T2 ribonuclease in provoking the circumoval granuloma and given the predicted role of the granuloma in translocation of the schistosome egg from the intestinal wall to the lumen of the bowel , a line of Δω1 schistosomes may not be transmissible via eggs shed in the host feces . Nonetheless , a line of Δω1 S . mansoni schistosomes might be rescued by experimentally transferring Δω1 eggs from livers of mice into water , for continuation of the developmental cycle . If so , not only would this definitively demonstrate both the pathophysiological role of ω1 but also would demonstrate essentiality of ω1 for successful parasitism . Mutant parasite lines can be predicted to enable more comprehensive understanding of the pathobiology of these neglected tropical disease pathogens and facilitate access to novel insights and strategies for disease management . To conclude , these findings reported here confirmed that somatic genome editing of schistosome eggs led to functional knockout of the ω1 T2 ribonuclease , and revealed the likely presence of genetic mosaicism in mutant cells resulting from gene editing ( Mehravar et al . , 2018 ) . The genome-edited eggs exhibited loss of function of ω1 but remained viable . Programmed mutation of ω1 using CRISPR/Cas9 not only achieved the aim of establishing the applicability of genome editing for functional genomics of schistosomes but also demonstrated manipulation of a gene expressed in the schistosome egg , the developmental stage central to the pathophysiology of schistosomiasis . Mice experimentally infected with S . mansoni , obtained from the Biomedical Research Institute ( BRI ) , Rockville , MD were housed at the Animal Research Facility of the George Washington University Medical School , which is accredited by the American Association for Accreditation of Laboratory Animal Care ( AAALAC no . 000347 ) and has an Animal Welfare Assurance on file with the National Institutes of Health , Office of Laboratory Animal Welfare , OLAW assurance number A3205-01 . All procedures employed were consistent with the Guide for the Care and Use of Laboratory Animals . The Institutional Animal Care and Use Committee ( IACUC ) of the George Washington University approved the protocol used for maintenance of mice and recovery of schistosomes . Studies with BALB/c mice involving tail vein injection of schistosome eggs and subsequent euthanasia using overdose of sodium pentobarbital was approved by the IACUC of BRI , protocol 18–04 , AAALAC no . 000779 and OLAW no . A3080-01 . Mice were euthanized 7 weeks after infection with S . mansoni , livers were removed at necropsy , and schistosome eggs recovered from the livers , as described ( Dalton et al . , 1997 ) . The liver eggs termed ‘LE’ were maintained in DMEM medium supplemented with 10% heat-inactivated fetal bovine serum ( FBS ) , 2% streptomycin/penicillin at 37°C under 5% CO2 in air for 18–24 hr , after which LE was exposed to electroporation with RNP or transduction by LV for programmed gene editing ( Mann et al . , 2010; Mann et al . , 2014 ) . Polymyxin B ( 10 μg/ml ) was added to the cultures twice daily to neutralize lipopolysaccharide ( LPS ) ( Cardoso et al . , 2007 ) . Soluble egg antigen ( SEA ) was prepared from these eggs , as described ( Dunne et al . , 1991; Boros and Warren , 1970 ) . In brief , the homogenate of eggs in 1× PBS containing protease inhibitor cocktail ( Sigma ) was frozen and thawed twice , clarified by centrifugation at 13 , 000 rpm , 15 min , 4°C , the supernatant passed through a 0 . 22 μm pore size membrane . Protein concentration of the supernatant ( SEA ) was determined by the Bradford Protein Assay ( Bradford , 1976 ) and aliquots of the SEA stored at −80°C . Single guide RNA ( sgRNA ) was designed using the web-based tools at http://bioinfogp . cnb . csic . es/tools/breakingcas/ ( Oliveros et al . , 2016 ) to predict cleavage sites for the Streptococcus pyogenes Cas9 nuclease within the genome of S . mansoni . The sgRNA targeted exon 6 of the ω1 gene , Smp_193860 , www . genedb . org , residues 3 , 808–3 , 827 , adjacent to the protospacer adjacent motif , AGG ( Figure 1A ) . This is a multi-copy gene with at least five copies of ω1 located in tandem on chromosome 1 ( Protasio et al . , 2012 ) . To infer the gene structure of Smp_193860 in the S . mansoni V5 genome assembly more accurately , the omega-1 mRNA DQ013207 . 1 sequenced by Fitzsimmons et al . ( 2005 ) was used to predict the gene structure with the exonerate software , by aligning it to the assembly using the exonerate options '--model coding2genome' and '--maxintron 1500' . The Smp_193860 copy of ω1 includes nine exons interspersed with eight introns ( 6196 nt ) ( Figure 1A ) . Synthetic gRNA ( sgRNA ) , ω1-sgRNA was purchased from Thermo Fisher Scientific ( Waltham , MA ) . A double-stranded DNA sequence complementary to the sgRNA was inserted into lentiviral gene editing vector , pLV-U6g-EPCG ( Sigma ) , which encodes Cas9 from S . pyogenes driven by the eukaryotic ( human ) translation elongation factor 1 alpha 1 ( tEF1 ) promoter and the sgRNA driven by the human U6 promoter ( Figure 1C ) . The pLV-U6g-EPCG vector is tri-cistronic and encodes the reporter genes encoding puroR and GFP , in addition to Cas9 ( Fitzsimmons et al . , 2005 ) . This gene-editing construct , targeting exon 6 of ω1 Smp_193860 , was termed pLV-ω1X6 . A single-stranded oligodeoxynucleotide ( ssODN ) ( Lok et al . , 2017 ) , which included homology arms of 50 nt each in length at the 3’ ( position 3775–3824 nt ) and 5’ ( 3825–3874 nt ) flanks and a small transgene ( 5’-TAAGTGACTAGGTAACTGAGTAGC-3’ , encoding stop codons ( six ) in all open-reading frames ) ( Figure 1B ) , was synthesized by Eurofin Genomics ( Louisville , KY ) . An oligonucelotide primer that included this sequence was employed in PCRs to investigate the presence of CRISPR/Cas9-programmed insertion of the transgene ( Supplementary file 2 ) . For the ribonucleoprotein ( RNP ) complex of the ω1-sgRNA and recombinant Cas9 from Streptococcus pyogenes , 6 μg of ω1-sgRNA and 6 μg of Cas9-NLS nuclease ( Dharmacon , Lafayette , CO ) were mixed in 100 μl Opti-MEM ( Sigma ) to provide 1:1 ratio w/w RNP . The mixture was incubated at room temperature for 10 min , pipetted into a 4 mm pre-chilled electroporation cuvette containing ~10 , 000 LE in ~150 μl Opti-MEM , subjected to square wave electroporation ( one pulse of 125 volts , 20 ms ) ( BTX ElectroSquarePorator , ECM830 , San Diego , CA ) . The electroporated eggs were incubated for 5 min at room temperature , and maintained at 37°C , 5% CO2 in air for 3 , 5 and 7 days . To investigate whether homology-directed repair ( HDR ) could catalyze the insertion of a donor repair template , 6 μg ssODN was mixed with RNP and the LE before electroporation . In a second approach ( above ) , the ssODN was delivered to LE by electroporation at ~24 hr after the lentiviral transduction of the LE . The eggs were collected 3 , 5 and 7 days later and genomic DNA recovered from LE . The negative controls included LE subjected to electroporation in the presence of only Opti-MEM , only Cas 9 , only sgRNA , and only ssODN . Escherichia coli Zymo 5α ( Zymo Research ) cells were transformed with lentiviral plasmid pLV-ω1X6 and cultured in LB broth in 100 μg/ml ampicillin at 37°C , agitated at 225 rpm for ~18 hr , after which plasmid DNA was recovered ( GenElute Plasmid purification kit , Invitrogen ) . A lentiviral ( LV ) packaging kit ( MISSION , Sigma-Aldrich ) was used to prepare LV particles in producer cells ( human 293T cell line ) . In brief , 3 . 5×105 of 293T cells/well were seeded in a six-well tissue culture plate in DMEM supplemented with 10% heat-inactivated fetal bovine serum ( FBS ) , 2 mM L-glutamine , 1% penicillin/streptomycin and cultured at 37°C , 5% CO2 for 18 hr . The producer cells were transfected using FUGENE HD ( Promega ) with pLV-ω1X6 and LV packaging mix containing two additional plasmids; one plasmid that expressed HIV structural and packaging genes and another that expressed the pseudotyping envelope protein Vesicular Stomatitis Virus Glycoprotein ( VSVG ) . Subsequently , the transfection mixture ( 50 μl; 500 ng plasmid DNA , 4 . 6 μl packaging mix , 2 . 7 μl of FUGENE HD in Opti-MEM ) was dispensed drop wise into each well on the plate . Sixteen hours later , the media were removed from the transfected cells , replaced with pre-warmed complete DMEM , and cells cultured for 24 hr . The supernatant , containing VSVG-pseudotyped LV particles was filtered through 22 μm pore size membranes ( Suttiprapa et al . , 2016 ) , and stored at 4°C . Additional pre-warmed complete DMEM was added to the well , for culture for a further 24 hr . The supernatant was collected as above , combined with the first supernatant and concentrated ( Lenti-X concentrator , Takara Bio , Mountain View , CA ) . Virion titer was estimated by two methods; first , by use of Lenti-X-GoStix ( Takara Bio ) to establish the presence of functional virions at >105 infectious units ( IFU ) /ml , and second , by reverse transcriptase assay ( Suttiprapa et al . , 2016; Marozsan et al . , 2004 ) to quantify levels of active virions . Virions with counts of ~4×106 count per minute ( cpm ) /ml were aliquoted and stored at −80°C . To transduce LE with LV , ~10 , 000 eggs were incubated for 24 hr in complete DMEM containing 500 μl of ~4×106 cpm/ml VSVG-LV virions . Thereafter , the LE were washed three times in 1× PBS and transfected with ssODN ( 6 μg ) by square wave electroporation . The further steps ( Suttiprapa et al . , 2012b ) with subsequent transfection with heat-inactivated pLV virions at 70°C for 4 hr with ssODN , transfection with ssODN in the absence of virions or Opti-MEM only served as negative controls . For each DNA sample , four separate PCR assays using four distinct primer pairs ( Supplementary file 2 ) were carried out . The first ω1 primer pair , to amplify locations 3 , 751–4 , 740 nt of Smp_193860 , was employed as positive control for the presence of genomic DNA with the Smp_193860 copy of ω1 . The other primer pairs shared one forward primer complementary to the knock-in 24 nt transgene with three reverse primers , Sm ω1-R1 , -R2 and -R3 at positions 3 , 966–3 , 984 , 4 , 066–4 , 085 and 4 , 102–4 , 121 nt , respectively , binding to three sites downstream of the ω1 predicted DSB site ( Figure 2A , Supplementary file 2 ) ( Lok et al . , 2017 ) . The PCR mix included 10 μl Green GoTaq DNA polymerase mix ( Promega ) with 200 nM of each primer and 10 ng genomic DNA . Thermal cycling conditions involved denaturation at 95°C , 3 min followed by 30 cycles of 94°C , 30 s , 60°C , 30 s and 72°C , 30 s and a final extension at 72°C for 5 min . Following agarose gel electrophoresis ( 1 . 2% agarose/TAE ) , amplicons of the expected sizes were recovered from gels and ligated into pCR4-TOPO ( Thermo Fisher ) . E . coli Zymo 5α competent cells were transformed with the ligation products , several colonies of each transformant were grown under ampicillin selection , plasmid DNA purified , and the inserts sequenced to confirm the presence and knock-in of the transgene ( Figure 1C ) . We used the egg genomic DNA templates directly for qPCR as described ( Yu et al . , 2014 ) , with slight modification which enabled estimation of the efficiency of CRISPR-mediated mutagenesis at the target locus without the need to normalize the experimental and control template DNAs . The general approach makes use of the fact that the binding of a primer overlapping the sgRNA site was compromised in programmed mutagenized egg ( LE ) genome ( s ) , resulting in delayed amplification , whereas binding of a flanking primer pair was unaffected . The ‘OUT’ ( flanking ) primer pair encompassed at least 50 bp surrounding the sgRNA binding region . The ‘OVR’ ( overlapping ) primer pair used one of the Smω1-OUT primers and another primer that bound the 20 bp of the sgRNA target sequence . The 3′ side of the Smω1-OVR primers bound immediately upstream of the NGG motif , as the majority of indels would be expected to affect positions −1 to −10 of the binding site ( Canver et al . , 2018 ) . At days 3 , 5 and 7 following transfections with or without ssODN , genomic DNA was isolated from LE . Using 5 ng of DNA template , separate 20 μl OUT and OVR qPCR reactions were undertaken . Quantitative PCR was performed with SsoAdvanced SYBR Green Supermix ( Bio-Rad , 172–5271 ) using a Bio-Rad iQ5 Real-Time PCR system , with qPCR conditions at initial 95°C for 30s , 40 cycles , 95°C for 10s , 60° C for20 s . Oligonucleotide primer sequences are provided in Supplementary file 2 . The ratio of the qPCR quantification cycle values for the control Smω1-OVR and control Smω1-OUT primers reflected the differences in amplification of the two primer pairs on control DNA template . This might have been due to inherent differences in amplification that exist even between perfectly complementary primer pairs . In contrast , the OVR/OUT ratio in mutant DNA reflected both this difference in amplification between the primer pairs and the loss of the OVR binding site due to CRISPR-introduced indels . A comparison of the OVR/OUT quantification cycle ratios of control versus mutated genomes thus reflected the efficiency of mutagenesis . The CRISPR efficiency was calculated by the Ct ratio of OVR:OUT , after which indel/substitution mutation percentage was estimated as follows: % Relative fold amplification = 100 × ( B/A ) A = Ct ratio of OVR:OUT from control group B = Ct ratio of OVR:OUT from experiment group Pooled LE DNA samples from six independent KI experiments of pLV-ω1X6 with ssODN were used as the template to amplify the on-target DNA fragment using MiSeq primers ( Figure 2A ) with High Fidelity Taq DNA polymerase ( Thermo Fisher ) . PCR reactions were set up with 10 ng LE DNA samples from the KI experiment in 25 µl reaction mix using the HiFidelity Taq DNA polymerase ( Thermo Fisher ) following the PCR program 94°C for 3 min of denaturation followed by 30 cycles of 94°C for 30s , 60°C or 54°C for 30 s , 72°C for 45s and final extension at 72°C for 2 min . The expected size of the amplicon flanking predicted DSB was 202 bp . Amplicons of this size were purified using the Agencourt AMPure XP system ( Beckman Coulter ) . Amplicons generated from four different PCR reactions from each sample were pooled , and 100 ng of amplicons from each sample was used to construct the uniquely indexed paired-end read libraries using the QIAseq 1-step Amplicon Library Kit ( Qiagen ) and GeneRead Adapter I set A 12-plex ( Qiagen ) . These libraries were pooled , and the library pool was quantified using the QIAseq Library Quant System ( Qiagen ) . Samples ( Supplementary file 3 ) were multiplexed ( 10 samples ) and each run on four MiSeq lanes . After sequencing , the fastq files for each particular sample were merged . Samples 1–6 , 8 and 10 were prepared using an annealing temperature of 54°C . Samples 7 and 9 were prepared using an annealing temperature of 60°C , and included an extra 10 bp at the start of the MiSeq sequences , ‘GTTTTAGGTC’ , present upstream of the 5’ primer in the genomic DNA . We trimmed this sequence from the reads using cutadapt v1 . 13 ( Martin , 2011 ) . To detect HDR events , computational software program CRISPResso ( http://crispresso . rocks/ ) ( Pinello et al . , 2016; Canver et al . , 2018 ) was employed using a window size of 500 bp ( -w 500 ) with the reference amplicon according to gene Smp_193860 in the S . mansoni V7 assembly , and with the --exclude_bp_from_left 25 and --exclude_bp_from_right 25 options in order to disregard the ( 24 bp ) primer regions on each end of the amplicon when indels are being quantified . A window size of 500 nt was employed to include the entire amplicon . In order to search for HDR events , CRISPResso checked for HDR events ( using –e and –d options ) in treatment groups including controls . To infer frameshifts using CRISPResso the –c option was used , giving CRISPResso the coding sequence from positions 42–179 of the amplicon . To confirm the insertions of the knock-in sequences reported by CRISPResso ( right side column in Supplementary file 3 ) , we took all insertions of 20–28 bp reported by CRISPResso , and calculated their percent identity to the expected knock-in sequence using ggsearch v36 . 3 . 5e in the fasta package ( Pearson and Lipman , 1988 ) . An insertion was considered confirmed if it shared ≥75% identity to the expected donor knock-in sequence . A quantitative PCR to estimate the relative copy number of ω1 was performed using Kapa SYBR FAST Universal qPCR mastermix ( KK4602 ) on 1 ng of gDNA templates ( pooled genomic DNA samples from six biological replicates ) isolated from control and test samples , in 20 μl volumes . A primer pair of OMGgRNA1F and OMGgRNA1R was used to amplify the ω1 gRNA target region and SmGAPDH ( Supplementary file 2 ) as a reference single-copy gene ( primers shown in Supplementary file 2 ) . The PCR efficiencies for primer pairs were estimated by titration analysis to be 100% ± 5 ( Ginzinger , 2002 ) and qPCRs were performed in triplicate in 96-well plates , with a denaturation step at 95°C of 3 min followed by 40 cycles of 30 s at 95°C and 30 s at 55°C , in thermal cycler fitted with a real time detector ( StepOnePlus , Applied Biosystems ) . The relative quantification assay 2-ΔΔCt method ( Livak and Schmittgen , 2001 ) was used to ascertain the relative copy number of ω1 . Relative copy number of ω1 in the CRISPR/Cas9 treated groups reflects the fold change of ω1 copy number normalized to the reference gene ( SmGAPDH-F and -R primers ) and relative to the untreated control group ( calibrator sample with relative ω1 copy number = 1 ) . Total RNAs from schistosome eggs were extracted using the RNAzol RT reagent ( Molecular Research Center , Inc ) , which eliminates contaminating DNA ( Chomczynski , 1993 ) , and concentration and purity determined using a spectrophotometer ( OD260/280 ~2 . 0 ) . Reverse transcription ( RT ) of the RNA ( 500 ng ) was performed using iScript Reverse Transcript ( Bio-Rad ) , after which first strand cDNA was employed as template for qPCRs using SsoAdvanced Universal SYBR Green Supermix ( Bio-Rad ) performed in triplicates in an iQ5 real-time thermal cycler ( Bio-Rad ) . RT-qPCR reaction mixtures included 2 μl first strand cDNA , 5 μl SsoAdvanced Universal SYBR Green Supermix , and 300 nM schistosome gene-specific primers . Supplementary file 2 provides details of the oligonucleotide primers . Thermal cycling included denaturation at 95°C for 30 s , 40 amplification cycles each consisting of denaturation at 95°C for 15 s and annealing/extension at 60°C for 30 s , and a final melting curve . The output was analyzed using the iQ5 software ( BioRad ) . Relative expression was calculated using the 2-ΔΔCt method and normalized to schistosome GAPDH expression ( Livak and Schmittgen , 2001 ) ; data are presented as transcript levels ( three replicates ) compared to the wild-type LE ( 100% ) , and fold change reported as mean relative expression ±SD ( n = 11 ) . A stock solution of yeast RNA ( Omega Bio-tek , Norcross , GA ) was prepared at 1 . 0 μg/μl , 50 mM Tris-HCl , 50 mM NaCl , pH 7 . 0 . Yeast RNA ( 200 ng ) was incubated with 2 μg SEA from control and experimental groups individually at 37°C for 60 min . SEA investigated here , named Δω1-SEA , was extracted from LE transduced with pLV- ω1X6 virions and ssODN , pooled from six biological replicates . RNase A , an endoribonuclease from bovine pancreas ( Thermo Fisher ) served as a positive control enzyme , whereas yeast RNA in reaction buffer only served as the negative control . The RNase activity of ω1 in wild-type SEA or Δω1-SEA was analyzed by visualizing and quantifying the substrate that remained following enzymolysis by agarose gel electrophoresis and staining with ethidium bromide . The yeast RNA digestion by control SEAs or Δω1-SEA were set up in triplicates , with quantity of residual RNA determined by densitometry ( Ke et al . , 2017 ) . Human monocytic THP-1 cells were maintained in Roswell Park Memorial Institute medium ( RPMI ) 1640 ( Thermo Fisher Scientific ) containing 10% ( v/v ) FBS with 4 mM L-glutamine , 25 mM HEPES , 2 . 5 g/L D-glucose at 37°C in 5% CO2 in air . THP-1 cells were non-adherent cells . In a 6-well plate , THP-1 monocytes ( 3×105 cells in each well ) were differentiated into macrophages ( Mϕ ) by incubation in 150 nM phorbol 12-myristate 13-acetate ( PMA ) ( Sigma ) for 48 hr ( Genin et al . , 2015 ) . Mϕ were exposed to SEA ( 50 ng/ml ) or Δω1-SEA ( 50 ng/ml ) ( from LE transduced with pLV-ω1X6 virions and ssODN ) for 48 hr . To investigate macrophage and T cell interactions , Mϕ cells were pulsed with 50 ng/ml SEA or Δω1-SEA and thereafter co-cultured in direct contact with Jurkat ( human CD4+ T ) cells . Nine ×105 Jurkat were added to Mϕ with direct contact and were co-cultured for an additional 72 hr . Cell-free supernatants from the co-cultures were collected to quantify secretion of T helper cell cytokines including IL-4 , IL-5 , IL-13 , IL-10 , TNF-α , IL-6 , IL-2 and IFN-γ by enzyme linked immunosorbent assay ( Multiplex Human Cytokine ELISA kit , Qiagen ) ( Schmid and Varner , 2010 ) . The assay included positive controls for each analyte , which were provided in the kit . Three biological replicates were undertaken . Human HEK293 cells , Jurkat and THP-1 cell lines were obtained from ATCC . All three cell lines were authenticated by ATCC using STR profiling by PCR and were confirmed in our laboratory to be Mycoplasma-free using the Lookout Mycoplasma PCR detection kit ( Sigma-Aldrich ) . For induction of circumoval , egg-induced granulomas in the lungs of mice , 8 weeks old female ( Ashton et al . , 2001 ) BALB/c mice were injected with 3 , 000 WT eggs or Δω1-eggs ( from experiment pLV-ω1X6 with ssODN ) or 1× PBS as negative control by tail vein , as described ( Wynn et al . , 1993 ) . The mice were euthanized 10 days later . Each group included 3 or five mice , two biological replicates were undertaken , totaling 6–10 mice for each treatment group . Before starting the experiment , mice were allocated randomly to the control or experimental treatment groups . For histopathological assessment of granuloma formation , the left lung was removed at necropsy and fixed in 10% formalin in pH 7 . 4 buffered saline for 24 hr , after which it was dehydrated in 70% ethanol , and clarified in xylene . The fixed lung tissue was embedded in paraffin and sectioned at 4-μm-thickness by microtome . Thin sections of the left lung lobe were mounted on glass slides and fixed at 58–60°C . Subsequently , rehydrated sections were stained with hematoxylin-eosin ( H&E ) for evaluation of inflammatory infiltrates and cellularity of granulomas . Digital images were captured using a 2D glass slide digital scanner ( Aperio Slide Scanner , Leica Biosystems , Vista , CA ) and examined at high magnification using the Aperio ImageScope software ( Leica ) ( Eltoum et al . , 1995; Cheever et al . , 1992 ) . The longest ( R ) and shortest ( r ) diameters of each granuloma containing a single egg were measured with an ocular micrometer , and the volume of the granuloma calculated assuming a prolate spheroidal shape , using 4/3πRr2 ( Ashton et al . , 2001 ) . All granulomas in all slides from the left lung of the mice , 15 slides per treatment group , were measured; in total , >100 granulomas from each treatment group . In the assays above , we performed two or more biological replicates . These biological replicates represented parallel measurements of biologically discrete samples in order to capture any random biological variation . Technical replicates were undertaken as well; these represented two or three repeated measurements of the same sample undertaken as independent measurements of the random noise associated with the investigator , equipment or protocol . Means for experimental groups were compared to controls by one-way ANOVA and , where appropriate , by two-tailed Student’s t-test and Welch’s unequal variances t-test ( GraphPad Prism , La Jolla , CA ) . Values for p of ≤0 . 05 were considered to be statistically significant .	Schistosomiasis is a tropical disease that can cause serious health problems , including damage to the liver and kidneys , infertility and bladder cancer . Nearly a quarter billion people are currently infected , mostly in poor regions of sub-Saharan Africa , the Philippines and Brazil . A freshwater worm known as Schistosoma mansoni causes the disease . These parasites enter the human body by burrowing into the skin; once in the bloodstream , they move to various organs where they rapidly start to reproduce . Their eggs release several molecules , including a protein known as omega-1 ribonuclease , which can damage the surrounding tissues . A gene editing technique called CRISPR/Cas9 allows scientists to precisely target and then deactivate the genetic information a cell needs to produce a given protein . While the tool has been used in other species before , it was unknown if it could be applied to S . mansoni . Here , Ittiprasert et al . harnessed CRISPR/Cas9 to deactivate the gene that codes for omega-1 ribonuclease and create parasites that do not produce the protein , or only very little of it . The experiments showed that mice infected with the gene-edited worm eggs displayed far fewer symptoms of schistosomiasis compared to those that carry the non-edited parasites . Alongside this work , Arunsan et al . used CRISPR/Cas9 to inactivate a gene in another species of worm that can cause liver cancer in humans . Together , these findings demonstrate for the first time that the gene editing method can be adapted for use in parasitic flatworms , which are a major public health problem in tropical climates . This tool should help scientists understand how the parasites invade and damage our bodies , and provide new ideas for treatment and disease control .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2019	Programmed genome editing of the omega-1 ribonuclease of the blood fluke, Schistosoma mansoni
Feeding behaviors require intricately coordinated activation among the muscles of the jaw , tongue , and face , but the neural anatomical substrates underlying such coordination remain unclear . In this study , we investigate whether the premotor circuitry of jaw and tongue motoneurons contain elements for coordination . Using a modified monosynaptic rabies virus-based transsynaptic tracing strategy , we systematically mapped premotor neurons for the jaw-closing masseter muscle and the tongue-protruding genioglossus muscle . The maps revealed that the two groups of premotor neurons are distributed in regions implicated in rhythmogenesis , descending motor control , and sensory feedback . Importantly , we discovered several premotor connection configurations that are ideally suited for coordinating bilaterally symmetric jaw movements , and for enabling co-activation of specific jaw , tongue , and facial muscles . Our findings suggest that shared premotor neurons that form specific multi-target connections with selected motoneurons are a simple and general solution to the problem of orofacial coordination . Behaviors are executed through coordinated activity of multiple groups of motor neurons and their muscle targets . Coordination of jaw and tongue muscles during feeding behaviors represents one of the most intricate mechanisms of the motor system and has been observed in a wide range of animals including humans ( Gerstner and Goldberg , 1991; Thexton and McGarrick , 1994; Takada et al . , 1996; Palmer et al . , 1997; Ishiwata et al . , 2000; Miller , 2002 ) . Here , coordination concerns primarily the adjustment of both timing and sequence of muscle activation to enable smooth , effective jaw and tongue movements . Three basic forms of coordination are consistently observed in feeding behaviors . First , the left and right jaw muscle activities are temporally symmetric , which is necessary because the mandible is joined by ligaments at the midline . Second , during chewing the activity of the tongue and jaw muscles is held to a similar low frequency rhythm , with the tongue positioning food between the surfaces of the teeth while the jaw moves the teeth to break down food ( Thexton and McGarrick , 1994; Takada et al . , 1996; Hiyama et al . , 2000; Naganuma et al . , 2001; Yamamura et al . , 2002 ) . Third , during chewing ( Gerstner and Goldberg , 1991; Liu et al . , 1993; Naganuma et al . , 2001 ) , licking ( Travers et al . , 1997 ) , and suckling ( Thexton et al . , 1998 ) , the tongue-protruding and jaw-opening muscles are co-active during jaw opening , while tongue-retracting and jaw-closing muscles are co-active during jaw closing . This co-activation also occurs in cortically-induced fictive mastication and occurs regardless of the frequency of cortical stimulation or the intensity of sensory stimuli applied ( Gerstner and Goldberg , 1991; Liu et al . , 1993 ) . The neural architecture enabling these different yet specific forms of jaw-tongue-facial muscle coordination remain unclear . In the more extensively studied vertebrate spinal cord , the network that generates the coordinated and rhythmic muscle activity during locomotion is referred to as the central pattern generator ( CPG ) . Separate CPGs are known to control separate limb muscles , with forelimb CPGs located in the cervical enlargement and hindlimb CPGs residing in the lumbar enlargements ( Kiehn , 2011 ) . Limb coordination occurs through interactions between these separate CPGs ( Kiehn , 2011 ) . By analogy , it is conceivable that different orofacial muscles are also controlled by different CPGs with their interaction resulting in orofacial coordination , although the evidence for well-defined jaw , face , and tongue CPGs is lacking . On the other hand , previous studies injecting two different retrograde tracers into two different cranial motor nuclei have suggested the existence of neurons projecting to both nuclei ( Amri et al . , 1990; Li et al . , 1993; Kamogawa et al . , 1994; Popratiloff et al . , 2001; Kondo et al . , 2006 ) . However , due to limitations of the neural tracer technique , such as non-specific labeling of passing fibers and labeling of the entire nucleus rather than the motoneurons innervating specific muscles , whether there are common premotor neurons simultaneously innervating specific motoneuron groups enabling the above mentioned orofacial coordination remained unclear . In this study , we employed a recently established monosynaptic circuit tracing methodology to identify premotor neurons of the jaw-closing masseter and tongue-protruding genioglossus motoneurons ( See ‘Materials and methods’ ) . Analysis of the resultant labeling reveals several premotor circuit elements that are well suited for the orofacial coordination observed in feeding behaviors . As described in the ‘Materials and methods’ and schematically illustrated in Figure 1 , we have developed a mouse line , Chat::Cre; RΦGT , that enables us to inject deficient rabies virus ( ΔG-RV ) into desired muscles resulting in transsynaptic tracing of the corresponding premotor circuitry in neonatal mice ( Takatoh et al . , 2013 ) . For further description and discussion of this method , see the ‘Materials and methods’ and associated Figure 1—figure supplement 1 . The masseter is the primary jaw-closing muscle and is innervated by motoneurons located in the trigeminal motor nucleus ( MoV ) . Masseter activity is coordinated with other muscles in multiple orofacial behaviors including suckling , chewing , biting , and vocalization ( Travers et al . , 1997 ) . To investigate the premotor circuitry controlling the masseter motoneurons , we injected ΔG-RV-EGFP into the left masseter of P1 Chat::Cre; RΦGT mouse pups ( Figure 1 ) . 7 days later at P8 ( P1→P8 tracing ) , we serial sectioned and imaged the brains to visualize the transsynaptically labeled masseter premotor neurons . 10 . 7554/eLife . 02511 . 003Figure 1 . Schematics detailing the premotor circuit tracing strategy . ( A ) Illustration of viral injection sites used in this study . Left , the jaw-closing masseter muscle; right , the genioglossus: a muscle of the tongue controlling protrusion . ( B ) Genetic cross used in this study . Arrow indicates action of Cre recombinase on the RΦGT locus enabling rabies G expression in motoneurons . ( C ) ΔG-RV injection into a selected muscle results in infection of motor axons innervating that muscle . Complementation of the virus with endogenous rabies G in motoneurons results in transsynaptic retrograde labeling of premotor neurons . Retrograde passage is halted at one synapse due to lack of complementation in premotor neurons . Inset , pups were injected at post-natal day 1 ( P1 ) , and their brainstems were analyzed at post-natal day 8 ( P8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 00310 . 7554/eLife . 02511 . 004Figure 1—figure supplement 1 . Extremely rare labeling of ChAT+ premotor neurons in masseter and genioglossus premotor tracing studies . The brains from monosynaptic rabies tracing experiments were immunostained for choline acetyltransferase ( ChAT ) . Examples of rarely labeled ChAT+ premotor neurons in the masseter ( A ) and genioglossus ( B ) premotor circuits . Such rare labeling was inconsistent between samples , and could account for some of the sporadic , inconsistent labeling observed between mice ( See Tables 1 and 2 , and ‘Results’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 004 Video 1 is a representative example of all serial sections from one such transsynaptic tracing experiment showing all viral-labeled regions . Figure 2 shows representative images from selected labeled regions . As a summary of the key findings , the masseter premotor circuitry contains: ( 1 ) extensive populations of neurons located bilaterally in the brainstem intermediate reticular ( IRt ) and lateral reticular nuclei ( MdRt , PCRt ) , extending from caudal ( Figure 2A ) to rostral ( Figure 2B , D ) regions; ( 2 ) a large number of proprioceptive trigeminal mesencephalic neurons ( MesV ) ( Figure 2F ) , and scattered second-order sensory-related neurons in the rostral and dorsal trigeminal brainstem nuclei ( SpO , dPrV , Figure 2D , E ) ; ( 3 ) numerous neurons in the region surrounding MoV ( Figure 2E ) ; ( 4 ) neurons in deep cerebellar nuclei—in particular the fastigial nucleus ( DCN , Figure 2C ) , midbrain reticular formation ( dMRf , Figure 2G ) and the red nucleus ( RN , Figure 2H ) ; and ( 5 ) sparse and sporadically labeled neurons in midline and other regions , including interneurons located in ipsi- and contralateral MoV ( Table 1 ) . A much more detailed description and quantification of the labeling results for each anatomical location are shown in Table 1 ( n = 5 mice ) . Note that due to the dense labeling of motoneurons , as well as all axon projections from labeled premotor neurons , we could not accurately quantify numbers of interneurons inside the ipsilateral MoV . Labeling in all other regions is quantified ( Table 1 ) . 10 . 7554/eLife . 02511 . 005Video 1 . Representative complete premotor circuit labeling after injection of ΔG-RV-EGFP into the left masseter muscle . Sections were obtained from the brainstem of an 8-day-old pup 7 days after peripheral rabies injection . 80-µm serial sections are shown in sequence from caudal to the hypoglossal motor nucleus ( MoXII ) to the rostral end of labeling in the dorsal midbrain reticular formation ( dMRf ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 00510 . 7554/eLife . 02511 . 006Figure 2 . Representative images of labeled jaw premotor neurons after ΔG-RV injection into the left masseter muscle . ( A ) Caudal brainstem showed labeling primarily in the caudal intermediate reticular formation ( IRt-c ) and medullary reticular formation ( MdRt ) . ( B ) Rostral brainstem at the level of the facial motor nucleus ( MoVII ) showed extensive labeling in the rostral IRt ( IRt-r ) and some labeling in the gigantocellular ( Gi ) and parvocellular ( PCRt ) reticular formation . Insets , labeled neurons in the spinal trigeminal nucleus oralis ( SpVo ) and the lateral paragigantocellular nucleus ( LPGi ) . This region of the brainstem contained extensive axon collaterals crossing the midline . ( C–H ) Labeling of other premotor neuron groups including: the deep cerebellar nuclei ( DCN , C ) ; the dorsal principal trigeminal sensory nucleus ( dPrV , D ) ; the motor trigeminal nucleus ( MoV , primary infection ) and surrounding trigeminal regions ( collectively , PeriV ) ( E ) ; the mesencephalic sensory nucleus ( MesV ) , which extended from MoV to dorsal to the periaqueductal grey ( PAG ) ( F ) ; the dorsal midbrain reticular formation ( dMRf , G ) ; and the red nucleus ( RN , H ) . Displayed side of the brainstem is indicated in each panel . Stereotaxic maps for this and all subsequent figures were obtained from the Allen Brain Institute website: www . brain-map . org . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 00610 . 7554/eLife . 02511 . 007Figure 2—figure supplement 1 . The masseter premotor circuit contains more LPGi neurons in old pups . ( A ) Schematic illustrating monosynaptic rabies-mediated tracing of the masseter premotor circuit in older pups ( P8→P15 tracing ) . Rabies containing mCherry ( red ) was injected into the left masseter of P8 Chat-Cre::RGΦT pups; brainstem samples were obtained at P15 . ( B–E ) Stereotypic labeling of the P15 masseter premotor circuit . ( B ) Premotor neurons were similarly observed in the rostral IRt and PCRt , but an increased number of neurons in the LPGi was observed in P8→P15 tracing ( arrow; compare to Figure 2B , inset ) . Other labeling included: the MdRt ( C ) ; MesV ( D ) ; and the region surrounding MoV ( E ) . ( F ) The number of LPGi neurons connected to masseter motoneurons increases during development . The ratio of LPGi neurons to primary motoneurons labeled in MoV was calculated for each sample . n = 3 mice per group , all values are averages ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 00710 . 7554/eLife . 02511 . 008Table 1 . Description and quantification of the distribution of masseter premotor neuronsDOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 008Masseter premotor neuronsPremotor region% Ipsilateral% ContralateralReticular regions Medullary reticular formation , caudal intermediate reticular formation5 . 65 ± 0 . 765 . 45 ± 0 . 69 Rostral intermediate reticular formation19 . 41 ± 1 . 6114 . 68 ± 1 . 22 Parvocellular reticular formation13 . 43 ± 0 . 985 . 32 ± 0 . 32 Lateral paragigantocellular nucleus0 . 44 ± 0 . 030 . 17 ± 0 . 02Trigeminal sensory regions Mesencephalic sensory nucleus16 . 81 ± 3 . 981 . 11 ± 0 . 35 Peri-trigeminal zone8 . 16 ± 0 . 591 . 91 ± 0 . 22 Dorsal principal trigeminal sensory nucleus2 . 47 ± 0 . 701 . 31 ± 0 . 37 Spinal trigeminal nucleus , Oralis1 . 78 ± 0 . 210 . 31 ± 0 . 08Descending control regions Dorsal midbrain reticular formation0 . 45 ± 0 . 210 . 08 ± 0 . 03 Deep cerebellar nuclei0 . 18 ± 0 . 100 . 40 ± 0 . 06 Red nucleus0 . 01 ± 0 . 010 . 48 ± 0 . 12Extensive bilateral labeling in both caudal ( level of MoXII; MdRt , IRt-c ) and rostral ( rostral to MoXII to caudal MoV; IRt-r , PCRt ) reticular regions was observed . Trigeminal sensory-related nuclei labeling primarily included MesV , comprised of jaw muscle proprioceptive and periodontal sensory neurons , and rostral trigeminal sensory nuclei ( SpVo , dPrV , and PeriV ) . Labeling in MesV and SpVo showed a strong ipsilateral bias . Nuclei implicated in descending control were labeled , consisting of contralateral DCN and RN , and ipsilateral dMRf , as well as LPGi . We also found scattered and sparse labeling of premotor neurons in the Gi , interneuron labeling in the contralateral MoV , lateral reticular formation , pre-Bötzinger complex ( pre-BötC ) , medial vestibular nucleus , raphe magnus nucleus , raphe pallidus nucleus , dorsal medial tegmental nucleus , and pontine reticular nucleus . However the labeling pattern and number of neurons in these nuclei were few and not consistent across animals . Percentage of total premotor neurons in a region was calculated within sample ( thereby normalizing values to tracing efficacy ) , and subsequent values were averaged across five mice . All values are averages ± SEM . The masseter premotor neurons revealed here through P1→P8 tracing likely reflect the circuits controlling suckling at this early stage , because chewing movements in mice emerge around post-natal day 12 ( Westneat and Hall , 1992 ) . On the other hand , previous studies have found that during the development of chewing , glycine switches from providing excitatory to inhibitory input onto motoneurons ( Inoue et al . , 2007; Nakamura et al . , 2008 ) . Thus , it is possible that the same circuitry is used to produce rhythmic suckling early in life and rhythmic chewing later in development ( Langenbach et al . , 1992; Westneat and Hall , 1992; Morquette et al . , 2012 ) . To investigate whether there might be developmental changes in the masseter premotor circuitry after chewing has begun , we conducted monosynaptic rabies-mediated tracing at P8 and sampled at P15 ( P8→P15 tracing ) ( Figure 2—figure supplement 1 ) . A much lower efficiency of motoneuron infection from peripheral injection was observed at this later stage ( 19 ± 4 motoneurons labeled in P8 injected animals; 35 ± 6 motoneurons labeled in P1 injected animals; mean ± SEM , n = 3 samples per group ) , and the overall number of labeled premotor neurons was drastically reduced . Despite this , the labeled neurons were distributed in similar locations as those observed in the P1→P8 tracing ( Figure 2—figure supplement 1B–D ) . Interestingly , in contrast to other regions , the number of labeled premotor neurons in the lateral paragigantocellular ( LPGi ) nucleus increased ( Figure 2—figure supplement 1E ) . When normalized against the number of infected motoneurons , twice as many LGPi neurons were labeled in the P8→P15 tracing , suggesting more LPGi neurons form synapses with motoneurons , or individual LPGi neurons form synapses with more motoneurons . Due to the overall inefficiency of infection and transsynaptic spreading in P8 or older animals , all subsequent results were obtained from P1→P8 tracing experiments . Because the mandible is joined by ligaments at the midline , jaw movement is obligated to be temporally symmetric on the left and right side . Trigeminal motoneurons do not themselves project bilaterally to enable such coordination ( Shigenaga et al . , 1988 ) . Previously , there have been observations of commissural interneurons located inside MoV projecting to the contralateral MoV ( Appenteng and Girdlestone , 1987; Ter Horst et al . , 1990; Juch et al . , 1993; McDavid et al . , 2006 ) , raising the possibility that these MoV interneurons might play critical role in left–right symmetry . We also observed labeled interneurons in the contralateral MoV in our tracing , however this labeling was very sparse ( see Figure 3E , F , arrow heads ) . Additionally , previous studies using retrograde dyes have labeled some reticular neurons projecting to both the left and right MoV , suggesting that bilateral coordination may also arise from inputs other than MoV interneurons ( Kamogawa et al . , 1994; Yoshida et al . , 2005 ) . Our finding that masseter motoneurons on one side receive extensive premotor inputs from both ipsi- and contralateral reticular ( Rt ) neurons ( Figure 2A–B ) suggests that premotor neurons in the reticular formation may be involved in producing synchronized and symmetric jaw motor activity on both sides . 10 . 7554/eLife . 02511 . 009Figure 3 . Evidence for the presence of bilateral-projecting masseter premotor neurons . ( A–F ) Simultaneous tracing of left ( ΔG-RV-EGFP , green ) and right ( ΔG-RV-mCherry , red ) masseter premotor neurons . Yellow cells , which indicate bilaterally projecting premotor neurons , were observed in many brainstem regions ( arrows indicate some examples ) including: IRt-c ( A ) ; IRt-r and Gi ( B ) ; PCRt and the dorsal reticular region ( dRt ) ( C ) ; MesV , with a magnified ( 1 . 5X ) view of the upper double-labeled neuron highlighting its morphology characteristic of primary afferent neurons in MesV ( D ) ; and dPrV and the peri-trigeminal region ( PeriV ) ( E and F ) . Additionally , premotor interneurons were found in the contralateral MoV ( arrow heads E and F ) . Displayed side of the brainstem is indicated in each panel . ( G–H ) ChAT-immunostained ( red ) contralateral MoV showing extensive innervation from labeled ipsilateral masseter premotor axons ( green ) . The boxed region in G , and a line scan ( right ) of an orthogonal slice at the yellow dotted line are shown in H . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 009 Since the monosynaptic rabies virus expresses fluorescent protein at high levels , axon terminals of viral-infected neurons can be clearly visualized . When premotor neurons were labeled from viral injection into the left masseter muscle , we found that the right MoV nucleus was covered by fluorescently labeled axons ( Video 1 ) , indicating that many premotor neurons project to motor nuclei on both sides . This is consistent with previous dye tracing studies ( Kamogawa et al . , 1994; Yoshida et al . , 2005; Kondo et al . , 2006 ) . However , because MoV contains both motoneurons and interneurons ( Appenteng and Girdlestone , 1987; Ter Horst et al . , 1990; Juch et al . , 1993; McDavid et al . , 2006 ) , it is possible that these bilaterally projecting premotor neurons synapse with motoneurons on the ipsilateral side , and with interneurons on the contralateral side . To determine whether at least some of the contralateral projections directly innervate motoneurons , we used anti-ChAT immunostaining to visualize motoneurons in ΔG-RV-EGFP-labeled brains . Indeed , numerous GFP+ boutons from labeled left masseter premotor neurons directly contacting ChAT+ motoneurons in the right MoV ( Figure 3G , H ) . To further confirm the existence of neurons presynaptic to motoneurons on both sides , as well as to identify the locations of these neurons in the jaw premotor circuitry , we injected the left masseter of P1 Chat::Cre; RΦGT pups with ΔG-RV-EGFP ( green ) , and the right masseter of the same pups with ΔG-RV-mCherry ( red ) . Thus , if any premotor neuron provided monosynaptic input to both the equivalent left and right masseter motoneurons , it would be labeled by both red and green ΔG-RV , thereby appearing yellow . We observed many double-labeled yellow neurons in most of the premotor nuclei , including many in the caudal and rostral reticular formation ( Figure 3A–C ) , a few primary proprioceptive neurons in the trigeminal mesencephalic nucleus ( MesV ) ( as distinguished from interneurons by their large spherical unipolar cell bodies ( Lazarov and Chouchkov , 1995; Verdier et al . , 2004 ) ) ( Figure 3D and inset ) , neurons in dorsal principle trigeminal nucleus ( dPrV ) ( Figure 3C , E , F ) , and neurons in the region above dPrV ( dRt ) ( Figure 3C , F ) . The number of double-labeled neurons was relatively few ( ∼8% ) . However , this is likely a significant under-representation . The masseter is a large muscle , and the chances of virus infecting functionally equivalent muscle fibers on both sides are very low to begin with . Additionally , when taking into account the stochastic nature of viral spreading at the synapses , the actual number of bilateral-projecting premotor neurons could be much higher than what we observed . The presence of premotor neurons innervating motoneurons on both sides throughout the jaw premotor circuitry provides a simple mechanism for directly synchronizing bilateral motoneuron activities . Suckling in neonates and chewing later in life involve not only coordination of bilateral jaw muscles , but also coordination between muscles of the jaw , the lips , and the tongue ( Naganuma et al . , 2001; Thexton et al . , 2004 ) . We next wanted to map the premotor circuitry controlling the tongue , which is innervated by motoneurons located in the hypoglossal motor nucleus ( MoXII ) . The genioglossus is the main tongue-protruding muscle . Partly due to its ease of access , the genioglossus has been a primary target for many previous studies investigating orofacial behaviors , including chewing , suckling , and licking ( Gerstner and Goldberg , 1991; Sawczuk and Mosier , 2001; Kakizaki et al . , 2002 ) . During rhythmic chewing , the genioglossus and masseter are activated with the same rhythm but in opposite phases , because activation of both at once could result in biting of the tongue . We investigated the premotor circuitry of the genioglossus motoneurons using a similar method as described for the masseter motoneurons ( Figure 1 ) . ΔG-RV-EGFP was injected into the left genioglossus muscle of the tongue . However , due to the small size of the tongue in P1 mice and the amount of virus necessary to infect motor axons , in most animals some of the virus injected to the left side unavoidably spread to the right genioglossus . Thus , it was not feasible to map the premotor inputs solely to the genioglossus on one side . Video 2 is a representative example of one serially sectioned brain after ΔG-RV-EGFP-mediated transsynaptic tracing from primarily the left genioglossus muscle . Figure 4 shows representative labeling patterns from selected brainstem regions . As a summary of the key findings , the genioglossus premotor circuitry contains: ( 1 ) a large population of premotor neurons located bilaterally in the intermediate reticular formation ( IRt ) , with fewer neurons in the lateral side of the reticular formation ( in PCRt ) ( Figure 4A–C ) ; ( 2 ) premotor neurons located in trigeminal sensory nuclei mainly in caudal brainstem ( SpC , Figure 4A ) , and a few in dPrV ( Figure 4E ) and MesV ( Figure 4F ) , as well as taste-related neurons in the nucleus of the solitary tract ( NTS , Figure 4A–B ) ; ( 3 ) neurons in the deep cerebellar nucleus ( DCN , Figure 4D ) and midbrain reticular formation ( dMRf , Figure 4G ) , which likely provide descending inputs; and ( 4 ) scattered cells in the midline and other brainstem structures ( Table 2 ) . A much more detailed description and quantification of the labeling results for each anatomical location are shown in Table 2 ( n = 5 mice ) . 10 . 7554/eLife . 02511 . 011Video 2 . Representative complete premotor circuit labeling after injection of ΔG-RV-EGFP into the genioglossus muscle . Sections were obtained from the brainstem of an 8-day-old pup 7 days after peripheral rabies injection . 80-µm serial sections are shown in sequence from caudal to the hypoglossal motor nucleus ( MoXII ) to the rostral end of labeling in the dorsal midbrain reticular formation ( dMRf ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 01110 . 7554/eLife . 02511 . 012Figure 4 . Representative images of labeled tongue premotor neurons after ΔG-RV injection into the left genioglossus muscle . ( A ) Caudal brainstem at the level of the hypoglossal motor nucleus ( MoXII ) , illustrating primary infection of the left MoXII , and extensive labeling in IRt-c , in the spinal trigeminal nucleus caudalis ( SpVc ) , and sparse labeling in the nucleus of the solitary tract ( NTS ) . ( B ) Brainstem between MoXII and MoVII showing extensive bilateral labeling in the IRt-r , with sparse labeling in the NTS , PCRt , Gi , and pre-Bötzinger complex ( pre-BötC ) below the nucleus ambiguus ( NA ) . ( C ) Rostral brainstem at the level of MoVII , showing labeling in the IRt-r and extending into the PCRt . Additionally , bilateral labeling of the LPGi is visible ( inset ) . Labeled premotor axon collaterals are visible invading the central MoVII ( arrow ) . ( D–G ) Other groups of labeled premotor neurons in: the DCN ( D ) ; dPrV , dRt and PeriV ( E ) ; MesV ( F ) ; and dMRf ( G ) . Genioglossus premotor labeling was weaker in the DCN and stronger in bilateral dMRf and LPGi as compared to masseter premotor labeling . Note the dense innervation of the anterior digastric motor nucleus ( Dig ) with premotor axons ( E , arrow ) . Displayed side of the brainstem is indicated in each panel . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 01210 . 7554/eLife . 02511 . 013Table 2 . Description and quantification of the distribution of genioglossus premotor neuronsDOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 013Genioglossus premotor neuronsPremotor region% Ipsilateral% ContralateralReticular regions Caudal intermediate reticular formation18 . 84 ± 1 . 798 . 99 ± 0 . 78 Rostral intermediate reticular formation , parvocellular reticular formation27 . 85 ± 0 . 4524 . 23 ± 1 . 78 Lateral paragigantocellular nucleus0 . 85 ± 0 . 230 . 73 ± 0 . 18Trigeminal sensory regions Spinal trigeminal sensory nucleus , caudalis4 . 21 ± 1 . 442 . 19 ± 0 . 70 Peri-trigeminal zone2 . 66 ± 0 . 471 . 88 ± 0 . 30 Mesencephalic sensory nucleus1 . 53 ± 0 . 310 . 59 ± 0 . 10 Spinal trigeminal sensory nucleus , oralis0 . 95 ± 0 . 200 . 56 ± 0 . 06 Dorsal principal trigeminal sensory nucleus0 . 73 ± 0 . 250 . 52 ± 0 . 13Descending control regions Dorsal midbrain reticular formation1 . 18 ± 0 . 381 . 21 ± 0 . 33 Deep cerebellar nuclei0 . 11 ± 0 . 030 . 08 ± 0 . 02 Red nucleus0 . 06 ± 0 . 040 . 05 ± 0 . 04Extensive bilateral labeling was observed in a concentrated band within the IRt from the medulla to the caudal border of MoVII ( IRt-c , IRt-r ) , after which it spread slightly into the PCRt ( IRt-r , PCRt ) . Labeling in trigeminal sensory related nuclei was primarily in the caudal sensory nuclei , particularly in bilateral SpVc . Additional sparse labeling of neurons in trigeminal sensory-related regions was found in SpVi , dPrV , PeriV , and MesV , with the MesV labeling occurring as far rostral as dorsal to the PAG . Nuclei implicated in descending control were labeled , consisting of contralateral DCN , bilateral dMRf , and bilateral LPGi . We also found scattered and sparse labeling of premotor neurons in the Gi , nucleus of the solitary tract ( NTS ) , rostral ventral respiratory group , lateral reticular nucleus , pre-BötC , midline raphe nuclei , superior vestibular nucleus , pontine reticular nucleus , and dorsal medial tegmental area . However , the labeling pattern and number of neurons in these nuclei were few and not consistent across animals . Percentage of total premotor neurons in a region was calculated within sample ( thereby normalizing values to tracing efficacy ) , and subsequent values were averaged across five samples . All values are averages ±SEM . Very interestingly , from the genioglossus transsynaptic tracing experiments , we observed that a large cohort of axon collaterals from these tongue-protruder premotor neurons projected to the motoneurons innervating the jaw-opening digastric muscle ( anterior portion ) , which are located in the accessory trigeminal motor nucleus ( Dig ) medial and ventral to the main MoV ( Figure 4E , arrow ) . Furthermore , many genioglossus premotor axons were observed projecting into the central MoVII , but not other divisions of MoVII , on both sides of the brainstem ( Figure 4C , arrow; Figure 5C ) . It is known that motoneurons in this central MoVII supply the posterior digastric muscle ( jaw-opening and swallowing ) , the platysma muscle ( jaw-depressing and lip-lowering ) , and the lower lip muscle ( Ashwell , 1982; Hinrichsen and Watson , 1984 ) . To further examine whether these collateral projections indeed form synapses onto Dig or central MoVII motoneurons , we conducted anti-ChAT immunostaining on brainstems after ΔG-RV-EGFP tracing from the genioglossus . Using high-resolution confocal microscopy , we observed GFP+ boutons directly contacting ChAT+ motoneurons in both the central MoVII ( Figure 5C , D ) and the Dig ( Figure 5A , B ) . These findings confirm that subsets of genioglossus premotor neurons simultaneously provide inputs to digastric and/or platysma/lower lip motoneurons . Considering many previous experimental observations of the co-activation of these muscles ( Gerstner and Goldberg , 1991; Liu et al . , 1993; Travers et al . , 1997; Thexton et al . , 1998; Naganuma et al . , 2001 ) , for example tongue protrusion always involves concomitant opening of the jaw and movement of the lower lip , such specifically shared premotor neurons are the simplest mechanism to generate co-activation of their target motoneurons . 10 . 7554/eLife . 02511 . 014Figure 5 . Premotor axon boutons onto ChAT+ motoneurons revealing direct premotor control of multiple motor groups . ( A ) ChAT-immunostained ( red ) jaw-opening digastric ( Dig ) motoneurons showing innervation from labeled genioglossus premotor axons ( green ) . ( B ) The boxed region in A , and a line scan ( right ) of an orthogonal slice at the yellow dotted line in A . ( C ) ChAT-immunostained central MoVII showing innervation from labeled genioglossus premotor axons . ( D ) The boxed region in C , and a line scan ( right ) of an orthogonal slice at the yellow dotted line in C . ( E ) ChAT-immunostained left MoXII showing innervation from labeled left masseter premotor axons . ( F ) The boxed region in E , and a line scan ( right ) of an orthogonal slice at the yellow dotted line in E . ( G ) ChAT-immunostained right MoXII showing innervation from labeled left masseter premotor axons . ( H ) The boxed region in G , and a line scan ( right ) of an orthogonal slice at the yellow dotted line in G . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 014 After observing genioglossus premotor neurons also innervating lip-lowering and jaw-opening motoneurons , we wondered if this type of multi-target connectivity was also employed by masseter premotor neurons . Previous studies using retrograde tracers discovered the existence of neurons which project to both MoV and MoXII nuclei ( Amri et al . , 1990; Li et al . , 1993; Travers et al . , 2005 ) . The masseter exhibits some degree of synchronous activity with tongue-retractor muscles during feeding behavior ( Kakizaki et al . , 2002 ) , although they are not as tightly coordinated as the digastric and genioglossus muscles . Tongue-retractor motoneurons are located in the dorsal MoXII , as compared to tongue-protruder genioglossus motoneurons which are located in the ventral MoXII ( McClung and Goldberg , 1999 ) . We thus re-examined the MoXII in ChAT-immunostained sections after ΔG-RV infection of the left masseter muscle ( Figure 5E–H ) . We found that indeed a few masseter premotor axons specifically innervated dorsal MoXII ( Figure 5E , G ) , and formed boutons opposing both contralateral and ipsilateral motoneurons located in this region ( Figure 5F , H ) . Thus , shared premotor neurons innervating both jaw-closing and some tongue-retracting motoneurons could indeed facilitate the co-activation of these two muscles . From the circuit-tracing results described above , it is immediately apparent that masseter and genioglossus motoneurons receive distinct sensory-related inputs . The masseter neurons receive extensive MesV-derived proprioceptive and SpVo-derived somatosensory inputs , and the genioglossus neurons primarily receive SpVc-derived somatosensory inputs ( compare Figure 2D , F with Figure 4A , F ) , while both motoneuron populations receive sensory-related input from dPrV . Additionally , genioglossus motoneurons receive taste-related inputs from the NTS ( Figure 4B , inset ) , while no labeling in the NTS was observed in the masseter premotor circuit . However , in the rostral reticular formation where many of the premotor neurons for both muscles are located , these populations are distributed in a similar pattern , suggestive of spatial intermingling of these two groups of premotor neurons . To more directly compare the spatial distributions of jaw and tongue premotor inputs , we injected ΔG-RV-EGFP ( green ) into the left genioglossus and ΔG-RV-mCherry ( red ) into the left masseter muscle of Chat::Cre; RΦGT pups ( Figure 6; Video 3; Video 4 ) . We focus our comparison on spatial distribution rather than the absolute numbers of labeled-neurons due to varying levels of infection in the two muscles across different samples . 10 . 7554/eLife . 02511 . 010Figure 6 . Representative images from experiments simultaneously tracing both genioglossus ( green ) and left masseter ( red ) premotor neurons . ( A ) Caudal brainstem at the level of MoXII , showing a rough spatial segregation between the two premotor populations in the IRt-c , with masseter premotor neurons more ventrally situated as compared to genioglossus premotor neurons . ( B ) Brainstem at the level between MoXII and MoVII , showing mostly genioglossus premotor neurons present in this region . ( C ) Brainstem at the level of MoVII , showing spatial intermingling of the two premotor populations . Axon collaterals crossing the midline are visible from both genioglossus and masseter premotor neurons . Genioglossus axon collaterals are visible extending into the central MoVII . ( D ) Labeling patterns at the level of MoV in dPrV , dRt ( region just above dPrV ) , supra trigeminal nucleus ( SupV ) , and inter-trigeminal region ( IntV ) . Masseter premotor axon collaterals ( red ) extend into the contralateral MoV , while genioglossus premotor axon collaterals ( green ) extend into the Dig ( arrows ) . Displayed side of the brainstem is indicated in each panel . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 01010 . 7554/eLife . 02511 . 015Video 3 . Comparison of masseter ( ΔG-RV-mCherry ) and genioglossus ( ΔG-RV-EGFP ) premotor circuitry . Sections were obtained from the brainstem of an 8-day-old pup 7 days after peripheral rabies injection . 80-µm serial sections are shown in sequence from caudal to the hypoglossal motor nucleus ( MoXII ) to the rostral end of labeling in the dorsal midbrain reticular formation ( dMRf ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 01510 . 7554/eLife . 02511 . 016Video 4 . Pseudocolored movie of masseter ( ΔG-RV-mCherry ) and genioglossus ( ΔG-RV-EGFP ) premotor circuitry . Sections were obtained from the brainstem of an 8-day-old pup 7 days after peripheral rabies injection . 80-µm serial sections are shown in sequence from caudal to the hypoglossal motor nucleus ( MoXII ) to the rostral end of labeling in the dorsal midbrain reticular formation ( dMRf ) . Images are pseudocolored such that masseter infection is visible in magenta , and genioglossus infection is visible in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 016 In the caudal-most IRt at the level of MoXII , masseter premotor neurons are situated ventrally to the labeled genioglossus premotor neurons ( Figure 6A ) . At the level just rostral to the MoXII , sparsely labeled masseter premotor neurons are spatially intermixed with densely labeled genioglossus premotor neurons in the IRt ( Figure 6B ) . At the level of the MoVII and MoV , there is extensive spatial overlap between the two populations of premotor neurons in the reticular region ( IRt and PCRt , Figure 6C ) , and in regions surrounding MoV ( Figure 6D ) . Notably , these regions in the reticular formation have previously been found to contain central pattern generating circuitry for jaw and tongue movements ( Chandler and Tal , 1986; Lund , 1991; Nakamura and Katakura , 1995; Morquette et al . , 2012 ) . However , even when the premotor pools were intermixed , we never observed double-labeled neurons; they remain distinct populations , avoiding simultaneous activation of these two muscles which could result in biting of one's own tongue . The majority of the masseter and the genioglossus premotor neurons are located in the intermediate reticular region ( IRt ) between the level of MoXII and the rostral end of MoV . Because the rostral portion of this region between the inferior olive and MoV is thought to contain CPGs for jaw and tongue movements ( Chandler and Tal , 1986; Lund , 1991; Nakamura and Katakura , 1995; Morquette et al . , 2012 ) , we wanted to determine the potential output signs of labeled premotor neurons . To do this , we conducted a series of in situ hybridization experiments to examine the neurotransmitter phenotypes while simultaneously immunostaining for EGFP expressed in ΔG-RV-EGFP-labeled premotor neurons . We focused our analysis on known markers for glutamatergic neurons ( vesicular glutamate transporter 2: vGluT2 ) , GABAergic neurons ( glutamic acid decarboxylase 1 and 2: GAD1 and GAD2 mixed probe [GAD1/2] ) , and glycinergic neurons ( glycine transporter 2: GlyT2 ) . We additionally tested neurons for tryptophan hydroxylase 2 ( Tph2 ) expression as a marker for serotonergic neurons , since we found some premotor neurons in the midline Raphé which is a source of serotonin in the brain . We performed these in situ-immuno analyses on brains 7 days post ΔG-RV injection . We found that both the masseter and genioglossus premotor neurons in IRt are of mixed transmitter phenotypes , that is they can be either excitatory ( vGluT2+ ) , or inhibitory ( GAD1/2+ or GlyT2+ ) ( Figure 7A–H ) . These mixed phenotypes are also true for premotor neurons in most other regions in brainstem ( data not shown ) . Due to viral toxicity , a portion of GFP+ neurons in these regions did not hybridize with any markers . Therefore , we could not reliably quantify the relative percentage of each type of neuron . However , in all samples examined , we only detected vGluT2+ neurons in the cerebellar DCN and the midbrain dMRf ( Figure 7I–L ) , with a notable absence of GAD1/2+ or GlyT2+ neurons in these regions ( data not shown ) suggesting that the descending inputs from cerebellum and midbrain to jaw and tongue motoneurons are primarily excitatory . 10 . 7554/eLife . 02511 . 017Figure 7 . Neurotransmitter phenotypes of labeled premotor neurons . In situ hybridization in combination with rabies tracing showing glycinergic ( A and E ) GABAergic ( B and F ) , and glutamatergic ( C and G ) premotor neurons to masseter and genioglossus motoneurons . Serotinergic neurons were found in the midline Raphé in both premotor tracing studies ( D and H ) . Premotor neurons observed in descending regions , including the dMRf ( I and J ) and DCN ( K and L ) were glutamatergic . DOI: http://dx . doi . org/10 . 7554/eLife . 02511 . 017 Because the mammalian mandible is connected by ligaments at the midline , the output of left and right jaw motoneurons must be temporally symmetric . Even when humans chew more on one side , the contralateral masseter is simultaneously activated to a comparable extent ( Moore , 1993; Peyron et al . , 2002 ) . By analogy to the studies investigating spinal locomotion circuits , one model of bilateral coordination involves the interaction between the jaw CPGs in the left and right brainstem ( Figure 8A1 ) . Studies investigating the rhythm-generating properties of the brainstem after midline transection have supported this theory through the finding that sufficient circuitry exists on both sides of the brainstem to independently generate a rhythmic output ( Chandler and Tal , 1986; Ihara et al . , 2013 ) . Previous studies have implicated commissural interneurons located inside MoV and bilaterally projecting neurons extending axons to both left and right MoV in some brainstem regions as a potential source of bilaterally coordinated inputs ( Appenteng and Girdlestone , 1987; Ter Horst et al . , 1990; Juch et al . , 1993; Kamogawa et al . , 1994; Yoshida et al . , 2005; McDavid et al . , 2006 ) . We also labeled a sparse number of interneurons in the contralateral MoV ( Figure 3E , F ) . Furthermore , we observed axon terminals of premotor neurons transsynaptically labeled from left masseter motoneurons extensively innervating the contralateral ( right ) MoV ( Figure 6C ) and forming boutons opposing motoneurons ( Figure 3G , H ) , supporting the existence of large numbers of bilaterally projecting premotor neurons . Using two-color rabies-mediated tracing , we confirmed that premotor neurons that synapse onto left and right masseter motoneurons were present in many brainstem regions . It thus appears that the jaw premotor circuit includes the simplest configuration for bilateral coordination: a single premotor neuron that synapses on equivalent ipsilateral and contralateral motoneurons . Interestingly , these neurons are especially prevalent in the reticular formation region previously identified as containing the jaw CPG ( Chandler and Tal , 1986; Chandler et al . , 1990; Lund , 1991; Nakamura and Katakura , 1995; Morquette et al . , 2012 ) . This organization could facilitate bilateral synchrony and symmetry , either independently of or in concert with interactions between CPGs , by enabling the output of the jaw CPG on one side to activate jaw muscles on both sides ( Figure 8A2 ) . In this model , the premotor neurons that transmit the CPG signals can be either part of the CPG ( dashed circle in model ) or immediately downstream of the CPG , and can include neurons projecting either unilaterally or bilaterally to the MoV ( Figure 8A2 ) . Tongue activity is tightly coordinated with jaw and facial muscle activity during feeding behaviors in a variety of mammals , including humans ( Westneat and Hall , 1992; Takada et al . , 1996; Ishiwata et al . , 2000; Yamamura et al . , 2002 ) . More specifically , the tongue-protruding genioglossus is active in phase with the jaw-opening digastric muscle and with the orbicularis oris of the lips under a wide range of conditions ( Liu et al . , 1993; Takada et al . , 1996 ) , and the tongue-retracting styloglossus is often active in phase with the jaw-closing masseter muscle ( Kakizaki et al . , 2002 ) . Previous work studying body muscle coordination in vertebrates and invertebrates showed that distinct CPGs which control motoneurons of different body segments may interact with each other to effect cross-muscle coordination ( Cang and Friesen , 2002; Briggman and Kristan , 2008; Smarandache-Wellmann et al . , 2014 ) . Because orofacial muscles are innervated by different groups of motoneurons located in MoXII , MoV , or MoVII , respectively , an analogous model would be that orofacial muscle co-activation is achieved by interaction between different CPGs that drive distinct motoneuron pools ( Figure 8B1 ) . Our study discovered that axon collaterals from labeled genioglossus premotor neurons also innervate MoV motoneurons supplying the jaw-opening anterior digastric muscles , as well as the motoneurons located in the central part of the MoVII . The central MoVII supplies the posterior digastric muscle ( jaw-opening and swallowing ) , the platysma muscle ( which depresses the jaw and draws down the lower lip ) , and the lower lip muscle ( Munro , 1974; Ashwell , 1982; Hinrichsen and Watson , 1984 ) , which are all activated when the tongue is protruding ( Figure 5 ) . Additionally , we found that axon collaterals from labeled masseter premotor neurons also innervate motoneurons located in the dorsal MoXII , previously shown to innervate tongue-retractor muscles ( McClung and Goldberg , 1999 ) . Our results indicate that specific premotor neurons simultaneously innervating multiple different groups of motoneurons provides the simplest circuit mechanism for enabling orofacial muscle co-activation ( Figure 8B2 ) . It is likely that such common premotor neurons are located in IRt because the majority of labeled jaw or tongue premotor neurons are located in that region . During chewing , the tongue-protruding genioglossus and jaw-closing masseter are generally active with the same rhythm but in opposite phases to prevent biting of the tongue . Again , this could in theory be achieved by mutual inhibition of their separate CPGs , which might have similar rhythm-generation properties ( Figure 8C1 ) . Here , we found that many premotor neurons for these two antagonist motor groups are located in a spatially intermixed pattern ( especially those situated between the caudal end of MoVII and the rostral end of MoV ) ( Figure 6C ) . As this region is within the region proposed to contain the CPG for chewing ( Chandler and Tal , 1986; Nozaki et al . , 1986; Ihara et al . , 2013 ) , our findings suggest that premotor neurons for different orofacial muscles are embedded within the same rhythm-generating network . Furthermore , previous studies examining intracellular potentials have found that masseter motoneurons receive rhythmic inhibitory and excitatory signals ( Goldberg et al . , 1982 ) , while genioglossus motoneurons receive only rhythmic excitatory signals ( Sahara et al . , 1988 ) . In the reticular region previously implicated in rhythm generation , we found that premotor neurons consisted of both excitatory and inhibitory neurons . Thus , we propose a hypothetical model consistent with these results whereby premotor neurons in an extended rhythmogenic network could have the same rhythm but could have opposite phases of activity through local circuit interactions , resulting in rhythmic excitation of masseter motoneurons and anti-phase excitation of genioglossus motoneurons and inhibition of masseter motoneurons ( Figure 8C2 ) . There were three particularly interesting findings regarding premotor nuclei labeled through our masseter and genioglossus rabies tracing experiments that deserve more detailed discussion . The NTS ( nucleus of the solitary tract ) is thought to play a major role in taste processing ( Bradley and Grabauskas , 1998 ) . Previous tracing studies have suggested that the NTS contains neurons which project directly to the hypoglossal nucleus ( Aldes , 1980; Borke et al . , 1983; Travers and Norgren , 1983; Dobbins and Feldman , 1995; Ugolini , 1995; Fay and Norgren , 1997 ) . Our results support these findings , revealing that some NTS neurons provide monosynaptic input onto tongue-protruding genioglossus motoneurons ( Figure 4B ) . While sparse , this input could play a central role in the well-established taste reactivity test ( Grill and Norgren , 1978b ) , which has been used extensively in assessing the interaction of taste processing and ingestive behaviors ( Flynn , 1995 ) . These stereotypic responses to controlled tastes have been found to be independent of cortical areas , and highly consistent across animals , suggesting that there may be a brainstem substrate for these behaviors ( Grill and Norgren , 1978a ) . Our finding of a direct innervation of tongue motoneurons by NTS neurons may be one such neural substrate . The labeling of neurons in the DCN ( deep cerebellar nuclei ) was somewhat surprising . The DCN had not previously been implicated in directly contacting masseter or genioglossus motoneurons of the brainstem in the neonate . Furthermore , where DCN axons project into the brainstem in non-murine species , the projection path is known to be primarily ipsilateral , with only sparse contralateral projections ( Cohen et al . , 1958; Ruigrok and Voogd , 1990 ) . Previous retrograde or transsynaptic tracing studies in rodents did not comment on any deep cerebellar nuclei labeling , even in polysynaptic tracing experiments ( Travers and Norgren , 1983; Fay and Norgren , 1997; Kolta et al . , 2000 ) . By contrast , in our transsynaptic studies we found that the fastigial nucleus of the DCN provides inputs onto contralateral masseter motoneurons in the early postnatal mouse ( Figure 2C ) . Additionally , we observed that a small number of neurons in both the fastigial and dentate nuclei of the DCN provide inputs onto genioglossus motoneurons ( Figure 4D , Video 2 ) . Recently , the Allen Brain Institute published their adult mouse brain connectivity study ( Oh et al . , 2014 ) . We consulted their connectivity atlas for tracing from the fastigial and dentate nuclei of the DCN . Interestingly , fastigial neurons give rise to a major contralateral projection pathway ( as well as ipsilateral projections ) and sparse fastigial axon terminals can be seen inside the contralateral MoV ( Allen Mouse Brain Connectivity Atlas , 2014a; Oh , et al . , 2014 ) . Furthermore , tracing results from either the fastigial or the dentate nucleus revealed sparse axon terminals inside MoXII ( Allen Mouse Brain Connectivity Atlas , 2014b; Oh , et al . , 2014 ) . Although these DCN innervations of contralateral MoV and MoXII are sparse in adult mice , it is possible that in newborn mice there are transiently more connections . Notably , in our P8→P15 masseter tracing experiment , we did not observe any labeling in the DCN , supporting such transient projections from DCN to masseter and genioglossus motoneurons during early development . Finally , labeling of bilaterally projecting MesV primary afferent neurons , although very few , was another unexpected finding . In our P8→P15 tracing study , we only observed ipsilateral MesV neurons labeled . Thus , such bilaterally projecting MesV neurons could be another case that only transiently exists in neonatal animals . In summary , our study revealed a set of premotor connection configurations well-suited to enable multi-muscle coordination and bilateral symmetry observed in feeding behaviors . Shared premotor neurons simultaneously providing inputs onto multiple groups of motoneurons innervating specific muscles may be a common circuit mechanism for motor coordination . It would be interesting to examine how these specific connections are established during development , and whether sensory inputs and CPG inputs converge on these coordinating premotor neurons to control both reflexes and centrally-induced movements . Additionally , we provide high-resolution maps of premotor nuclei for these jaw and tongue motoneurons including sensory-related inputs and descending inputs . An important goal of future studies is to develop molecular tools that allow specific functional manipulation of individual groups of premotor neurons to establish their roles in controlling and coordinating a variety of orofacial behaviors . We employed a previously described monosynaptic rabies-virus-based technique ( Wickersham et al . , 2007a; Callaway , 2008; Arenkiel and Ehlers , 2009 ) adapted to selectively trace neurons that directly synapse onto primary motoneurons ( Stepien et al . , 2010; Takatoh et al . , 2013 ) . Briefly , we used a knock-in mouse line containing CAG-loxP-STOP-loxP-rabies-G-IRES-TVA at the Rosa26 locus ( RΦGT ) ( Takatoh et al . , 2013 ) crossed with a mouse line that expresses Cre recombinase under the control of the choline acetyltransferase ( ChAT ) gene ( Chat::Cre mouse , JAX Stock #006410 ) . The resultant male and female pups were heterozygous for both alleles ( Chat::Cre;RΦGT ) , and thus expressed the rabies glycoprotein ( rabies-G ) in all cholinergic neurons , including motoneurons ( Figure 1B ) . Injection of a rabies virus expressing a fluorophore in place of the glycoprotein ( green: ΔG-RV-EGFP; or red: ΔG-RV-mCherry ) ( Wickersham et al . , 2007b ) into the primary muscle ( jaw-closing masseter or tongue-protruding genioglossus , Figure 1A ) infected motoneurons targeting that muscle ( Figure 1C , left ) . In the case of the masseter injections , animals which showed infection of the overlying skin as evident by fluorescence were excluded from analysis . Subsequent complementation of this virus by rabies-G in motoneurons enabled transsynaptic , retrograde travel of the virus into presynaptic partners ( Figure 1C , right ) . The monosynaptic rabies virus is deficient for the rabies-G in its genome , resulting in an inability for the virus to spread into presynaptic neurons of the source-infected cells unless it is complemented ( Etessami et al . , 2000 ) . Additionally , rabies virus infects neurons potentially through binding to the neuronal cell adhesion molecule ( NCAM ) or other neuronal receptors irrespective of neurotransmitter phenotype , size , or morphology , enabling an unbiased assessment of presynaptic populations ( Thoulouze et al . , 1998; Ugolini , 2010 ) . As a caveat to our premotor tracing strategy , if any premotor neurons expressed ChAT , rabies virus could be complemented again , resulting in spurious two-step labeling . A previous study has reported small numbers of cholinergic neurons in the intermediate reticular region of the brainstem that project to MoV and MoXII ( Travers et al . , 2005 ) . To investigate whether there were any ChAT+ masseter or genioglossus premoter neurons labeled in our experiments , we conducted anti-ChAT immunostaining on samples infected with rabies . We found that only approximately 1–3 cells in the reticular regions in the samples we examined were labeled ( Figure 1—figure supplement 1 ) . Additionally , such rare labeling was inconsistent in terms of rostral–caudal position across samples , suggesting that secondary jumping from such neurons would be very sparse with inconsistent labeling patterns . We therefore focused our quantification and analysis on the premotor regions consistently labeled across all samples . ΔG-RV was prepared and injected as previously described ( Takatoh et al . , 2013 ) . Briefly , male and female mouse pups at postnatal day 1 ( P1 ) were anesthetized by hypothermia , and were injected with 200–400 nL of ΔG-RV into either the masseter muscle or the genioglossus muscle of the tongue ( Figure 1A ) . 1 week post-infection ( Figure 1C , inset ) , mice were deeply anesthetized , transcardially perfused with 4% paraformaldehyde ( PFA ) in 1X phosphate-buffered saline ( PBS ) , post-fixed overnight in 4% PFA , and cryoprotected in 30% sucrose solution in PBS . Brain samples were embedded in Optimal Cutting Temperature compound ( OCT , Tissue-Tek ) and frozen for at least 24 hours at −80°C . Optimal survival times for maximal labeling were assessed prior to collection of results . Periods longer than 1 week resulted in an increase in glial infiltration and cell debris in the motor nuclei suggesting massive cell death , while periods shorter than 5–6 days post-infection resulted in a significant decrease in transsynaptically labeled neurons . Tissue preparation and immunostaining were conducted as previously described ( Takatoh et al . , 2013 ) . 80-µm thick sections were analyzed for tracing results , while ChAT immunostained samples were sectioned at 60 µm . Antibodies used were: goat anti-ChAT ( 1:1000 , Millipore ) , rabbit anti-GFP ( 1:1000 , Abcam ) , rabbit anti-RFP ( 1:1000 , Rockland ) , Alexa488 anti-rabbit ( 1:1000 , Jackson ImmunoResearch ) , Alexa594 anti-goat ( 1:1000 , Jackson ImmunoResearch ) , Alexa488 anti-goat ( 1:1000 , Jackson ImmunoResearch ) . In situ hybridization was performed as previously described with each probe being applied to every sixth 20-µm thick section ( Hasegawa et al . , 2007 ) . GAD1 , GAD2 , GlyT2 , vGluT2 , and Tph2 probes were created as previously described ( Takatoh et al . , 2013 ) . All probes were chosen based on whether they are known to be expressed in the brainstem regions of interest as seen in samples from the Allen Brain Atlas at www . brain-map . org . In addition to immunostaining and in situ hybridization , all sections were stained with DAPI ( 1:2000 ) to visualize the nuclei of all cells . Samples were imaged as previously described ( Takatoh et al . , 2013 ) using a Zeiss 710 inverted confocal microscope at 20X resolution . High resolution bouton images ( Figure 3G–H; Figure 5 ) were obtained as 100X z-stacks . Full-field views of axon collaterals entering central facial , digastric , and contralateral trigeminal nuclei were obtained using a Zeiss inverted epifluorescent microscope at 10X resolution . Infected premotor neurons in each of the brainstem areas were counted manually through serial sections . At least 5 animals were used for quantification . Final results were reported as percent of total premotor neurons labeled , thus normalizing each value to the overall infection level of that sample . Means were calculated from five samples each with primary infection of the masseter or genioglossus . All data are presented as mean percent of total neurons labeled ± standard error of the mean ( SEM ) . Videos were generated using Adobe Photoshop CS6 to align all serial sections and export frames into an MP4 format . The videos were subsequently compressed using proprietary software .	Chewing requires highly coordinated movements of the tongue and jaw . The tongue pushes food around the mouth , keeping it within reach of the teeth , while rhythmic movements of the jaw enable the teeth to grind up food without injuring the tongue itself . However , despite the importance of tongue–jaw coordination , the neuroanatomical circuits that underlie it have not been studied in detail . Stanek et al . have now mapped these neural circuits using a cutting edge tracing technique in mice . Muscles are activated by signals from neurons called motoneurons , which are themselves activated by signals from so-called premotor neurons . To identify the neural circuits responsible for tongue–jaw coordination , Stanek et al . injected a modified version of the rabies virus into the muscles that move the jaw and/or the tongue . This modified virus , which was also labeled with a fluorescent protein , was able to jump ‘backwards’ across the junctions between the muscles and the motoneurons , and then back to the premotor neurons . The fluorescent label allowed the neural circuits to be visualized under a fluorescence microscope . It had been assumed that distinct populations of premotor neurons would control the activity of different muscles . However , when viruses labeled with red fluorescent protein were injected into the muscles on the left side of the jaw , while viruses with green labels were injected into the muscles on the right side , a number of premotor neurons were found to display both red and green fluorescence . This indicates that some premotor neurons control muscles on both sides of the jaw , providing an effective means of coordinating bilateral muscle activity . A similar sharing of premotor neurons was observed between motoneurons that regulate jaw opening and those that trigger tongue protrusion , and between those that regulate jaw closing and tongue retraction . As well as providing new insights into the neuronal circuits that control the movements of the jaw and tongue , the work of Stanek et al . may have identified a general principle—namely the sharing of premotor neurons—that could be common to other circuits that produce coordinated muscle activity .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Monosynaptic premotor circuit tracing reveals neural substrates for oro-motor coordination
We examine the impact of targeted disruption of growth hormone-releasing hormone ( GHRH ) in mice on longevity and the putative mechanisms of delayed aging . GHRH knockout mice are remarkably long-lived , exhibiting major shifts in the expression of genes related to xenobiotic detoxification , stress resistance , and insulin signaling . These mutant mice also have increased adiponectin levels and alterations in glucose homeostasis consistent with the removal of the counter-insulin effects of growth hormone . While these effects overlap with those of caloric restriction , we show that the effects of caloric restriction ( CR ) and the GHRH mutation are additive , with lifespan of GHRH-KO mutants further increased by CR . We conclude that GHRH-KO mice feature perturbations in a network of signaling pathways related to stress resistance , metabolic control and inflammation , and therefore provide a new model that can be used to explore links between GHRH repression , downregulation of the somatotropic axis , and extended longevity . Genetic studies in a variety of organisms have revealed that the endocrine systems play a central role in lifespan determination ( Tatar et al . , 2003 ) . For example , mutations in the insulin/Insulin-like growth factor I ( IGF-I ) signaling can robustly increase longevity in the nematode C . elegans ( Kenyon , 2010 ) , whereas in mouse , disruptions in growth hormone ( GH ) pathway dramatically prolong lifespan ( Bartke , 2011 ) . Ames and Snell dwarf mice are the most studied mutants in which altered GH signals produce dramatic increases in lifespan ( Brown-Borg et al . , 1996; Flurkey et al . , 2001 ) . A number of aging-related phenotypes are also delayed in these mice , including collagen cross-linking , cataract development , kidney diseases , fatal neoplastic diseases and decline in immune function , locomotor activity , learning and memory ( Bartke , 2011 ) . In these models , homozygous mutation of either Prop1 or Pit1 genes cause an abnormal development of the anterior pituitary , which in turn leads to decline in production of GH , thyrotropin ( TSH ) , and prolactin ( PRL ) , with consequent decrease in circulating IGF-I and thyroxine levels ( Bartke , 2011 ) . The specific contribution of GH signaling to lifespan extension in these systems is supported by studies of downstream pathway elements . For instance , mice with disruption of the GH receptor ( Ghr−/− ) have also markedly increased lifespan with concomitant delay of late life diseases and disabilities ( Coschigano et al . , 2003; List et al . , 2011 ) . These findings support the hypothesis that dampening of the GH pathway is the key contributor to lifespan extension in mice . Nevertheless , the associated lack of TSH and prolactin makes these two models less than optimal in conclusively exclude the influence of the lack of these hormones on the delayed aging phenotype . Caloric restriction ( CR ) has been shown to extend lifespan in many species and has been extensively used in experimental gerontology to modulate development of age-related diseases ( Weindruch and Sohal , 1997 ) . In rodents , CR delays the onset of cancer , atherosclerosis , and diabetes , and typically increases lifespan ( by 15% in mice and by 30% in rats ) ( Swindell , 2012 ) . Although this phenomenon was first described over 70 years ago , the molecular basis mediating the effects of CR on the aging process remains incompletely understood . Intriguingly , phenotypic characteristics of the long-lived mutant mice with disrupted GH axis overlap with some effects of CR , suggesting possible mechanistic connections . Shared characteristics include: ( a ) small body size; ( b ) reduced blood glucose and increased insulin sensitivity; and ( c ) reduced or absent levels of various hormones and growth factors , that is , GH , insulin , and IGF-I; ( d ) delaying and/or suppression of the occurrence of several age-related diseases . Nevertheless , longevity phenotypes in different mouse models may rely on CR-sensitive pathways to varying degrees . For instance , 30% CR confers additional life extension in Ames dwarf mice ( Bartke , 2011 ) , but has no additional effect on longevity in male Ghr−/− mice , and only a modest reduction of late-life mortality in Ghr−/− females ( Bonkowski et al . , 2006 ) . In this study , we examined the longevity of mice with isolated GH deficiency due to targeted disruption of the GHRH gene ( GHRH-KO ) . This gene is required for somatotroph cell proliferation and GH secretion ( Alba and Salvatori , 2004 ) . We provide a phenotypic , metabolic and molecular-level characterization of GHRH-KO mice and show that GHRH-KO mutants exhibit lifespan extension comparable to the Ames and Snell dwarf mice . Moreover , we have shown that , in contrast with the Ghr−/− mice , lifespan in GHRH-KO mice is further extended by CR . These findings established the GHRH-KO mice as a novel rodent model for delayed aging and implicate CR-independent mechanisms in longevity assurance . To investigate the effect of isolated GH deficiency on lifespan , we evaluated differences in longevity of GHRH-KO ( KO ) mice and littermate ( wild-type ) control mice on ad libitum ( AL ) standard diet . As shown in Figure 1A , median survival of GHRH-KO mice ( sexes combined ) was increased by 295 days ( or 46% ) relative to that of control mice ( 931 days for KO mice vs 636 days for control mice ) . This difference in survival between KO mice and controls was significant based upon a non-parametric test ( p<0 . 001; log-rank test ) . 10 . 7554/eLife . 01098 . 003Figure 1 . Increased longevity of GHRH-KO mice . Kaplan-Meier survival curves for each genotype: GHRH-KO ( KO ) and Control ( Ct ) mice; each point represents a single mouse . ( A ) Sex pooled survival curves . ( B ) Sex separated survival curves . ( C ) Male survival curves ( N = 56 for controls , N = 39 for KO ) . ( D ) Female survival curves ( N = 52 for controls , N = 58 for KO ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 00310 . 7554/eLife . 01098 . 004Figure 1—figure supplement 1 . Changes in body length of GHRH-KO mice ( Red ) and WT control mice ( Black ) from 1 to 8 weeks of age . Each point represents the mean ± SEM of 14 mice per group . p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 004 Analysis of each sex separately showed that median survival in female GHRH-KO mice was increased by 290 days ( 43%; from 666 to 956 days ) relative to that of control mice ( logrank test , χ2 = 46 . 7 , p<0 . 0001 ) ( Figure 1B , C ) . Male GHRH-KO mice had increased median lifespan by 314 days ( from 614 to 928 days ) or 51% relative to that of control mice ( logrank test , χ2 = 28 . 4 , p<0 . 0001 ) ( Figure 1B , D ) . To evaluate the change of maximal lifespan , we used the Wang/Allison ( Wang et al . , 2004 ) method to compare the proportion of live mice in each group at the age at which only 10% of the population remained alive . GHRH deletion also led to a significant increase in maximum lifespan by 33% for females ( p<0 . 05 ) and by 18% for males ( p<0 . 05 ) relative to controls . The robust increase in both median and maximal lifespan in GHRH-KO mice is consistent with the notion that inhibition of GH signaling ameliorates age-related disease , potentially slowing the aging process . Both male and female GHRH-KO ( KO ) mice were markedly smaller than normal littermate controls ( Figure 2A ) . Although KO mice weighed the same as control animals at birth ( data not shown ) , they gained less weight than control mice ( Figure 2A ) . There was a significant reduction in body length in both male and female KO mice ( Figure 1—figure supplement 1 ) . Food consumption on a per gram body weight basis was not significantly different between KO and control mice . However body composition studies showed increased adiposity in both male and female KO mice ( Figure 2C ) . 10 . 7554/eLife . 01098 . 005Figure 2 . Physiological characteristics . ( A ) Body weight . ( B ) Respiratory exchange ratio ( VCO2/VO2 ) . ( C ) Fat content presented as an absolute values and percentage of body weight . ( D ) Physical activity . ( E ) RQ values plotted as 12-hr averages representing either dark or light periods on both fed and fasted days . RQ = respiratory quotient; ( F ) Glucose tolerance test ( IPGTT ) . 16 hr-fasted mice underwent GTT by intraperitoneal ( i . p . ) injection with 1 g glucose per kg of BW . ( G ) Fasted glucose , plasma insulin , homeostatic model for assessment of insulin resistance ( HOMA-IR ) and IGF-I levels . ( H ) Insulin tolerance test ( IPITT ) . Mice were i . p . injected with 1 IU porcine insulin per kg of BW . ( I ) Representative immunoblot of the indicated proteins from liver lysates . Each lane corresponds to a different mouse . For graph B , D , E , F , H , only data from male mice were shown . N = 8–10/group; each bar represents means ± SEM for 8–10 mice of each group . p<0 . 05 , p<0 . 01 , p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 00510 . 7554/eLife . 01098 . 006Figure 2—figure supplement 1 . Phosphorylation of S6 in the white adipose tissue ( WAT ) and muscle ( MUS ) of GHRH-KO and normal ( little-mate control ) mice . Upper panels: representative images of Western blots for phosphorylated and total forms of S6 protein in GHRH-KO and control mice . Lower panels: quantification of results of Western blots , as means ± SE for eight mice of each genotype . Values represent ratios of phosphoprotein to total protein for each enzyme , relative to the value in the control mice ( with normal mice set at 1 ) . p<0 . 05; *p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 006 To shed light on the physiological mechanisms involved in the energy homeostasis of KO mice , we assessed the effect of isolated GH deficiency on energy expenditure ( EE ) by indirect calorimetry . To estimate fuel utilization , we used the respiratory quotient ( RQ ) , which is a dimensionless ratio comparing the volume of carbon dioxide an organism produces to the volume of oxygen consumed over a given time ( RQ = VCO2/VO2 ) ( Johnston et al . , 2006 ) . The RQ varies inversely with lipid oxidation . A higher fasting RQ , which indicates lowered fat oxidation , is linked to body weight gain , metabolic inflexibility and insulin resistance ( Zurlo et al . , 1990 ) . The KO mice had a significantly decreased RQ compared to their wild-type controls during both the dark and light periods under normal housing conditions ( Figure 2E ) , although their oxygen consumption per gram of body weight ( measured by VO2 ) was not different from controls ( Figure 2B ) . Interestingly , on measures of spontaneous home cage activity ( Figure 2D ) , KO mice were significantly more active than their wild-type counterparts . Heightened insulin sensitivity is a common characteristic of several types of mice with GH-related mutations and extended longevity ( Bartke , 2011 ) . Consistent with this observation , plasma glucose concentrations were significantly reduced in both male and female GHRH-KO mice under fasted conditions ( Figure 2G ) . Furthermore , the lower score of homeostasis model-assessment of insulin resistance ( HOMA-IR ) indicated that GHRH-KO mice had enhanced insulin sensitivity ( Figure 2G ) . As expected from deficiency of pituitary GH , IGF-I levels in the circulation were much lower in GHRH-KO mice than controls . Moreover , insulin tolerance tests ( ITT ) done in mice fasted for 4 hr showed that plasma glucose concentrations decreased more in KO mice than in control mice ( Figure 2H ) despite of normal intraperitoneal glucose tolerance tests ( IPGTT ) results ( Figure 2F ) . Together , these data indicate that isolated GH deficiency results in improved insulin sensitivity and glucose homeostasis . Decreased ribosomal protein S6 kinase 1 ( S6K1 ) activity has been shown to increase insulin sensitivity ( Um et al . , 2006 ) . Since S6K1 signaling is decreased in the tissues of Ames dwarf mice ( Sharp and Bartke , 2005 ) , we speculated that S6K1 signaling may contribute to regulate insulin sensitivity in response to isolated GH deficiency . To investigate this possibility , we examined the levels of phosphorylation of ribosomal protein S6 and phosphorylation of IRS1 ( Figure 2I ) , two downstream targets of S6K1 ( Um et al . , 2006 ) . Hepatic levels of phosphorylation of these proteins were significantly lower in KO mice when compared with controls . Consistent with the evidence in S6k1−/− mice ( Selman et al . , 2009 ) , these data suggest that lower activity of S6K1 in GHRH-KO mice might contribute to the improved insulin sensitivity . Microarray analysis was performed to identify genes altered in liver tissue from GHRH-KO mice as compared to normal littermates ( n = 3 per genotype ) . Overall , we identified 141 genes with elevated expression in the mutants ( FDR < 0 . 05 and FC > 1 . 50; Supplementary file 1A ) , along with 164 genes with decreased expression ( FDR < 0 . 05 and FC < 0 . 67; Supplementary file 1B ) . The most highly increased genes were Sult2a2 , Sult1e1 and Spink3 ( FC > 38; Supplementary file 1A ) , while the most strongly decreased genes were Hsd3b5 , Slco1a1 and Igf1 ( FC <0 . 02; Supplementary file 1B ) . The 141 increased genes were disproportionately associated with GO BP terms related to differentiation , development and proliferation , such as positive regulation of mononuclear cell proliferation ( GO:0032946 ) , regulation of multicellular organismal development ( GO:2000026 ) and regulation of cell differentiation ( GO:0045595 ) ( p<0 . 01 ) ( Figure 3—figure supplement 1 ) . Increased genes were also disproportionately associated with KEGG pathways , including ‘metabolism of xenobiotics by cytochrome P450’ and ‘drug metabolism-cytochrome P450’ ( p<0 . 04 ) . Among the 164 decreased genes , there was overrepresentation of genes associated with oxidation-reduction process ( GO:0055114 ) , xenobiotic metabolic process ( GO:0006805 ) , monocarboxylic acid metabolic process ( GO:0032787 ) , and lipid metabolic process ( GO:0006629 ) ( p<0 . 01 ) ( Figure 3—figure supplement 1 ) . KEGG pathways overrepresented among decreased genes included ‘steroid hormone biosynthesis’ , ‘metabolism of xenobiotics by cytochrome P450’ and ‘drug metabolism—cytochrome P450’ . We compared GHRH-KO-increased and GHRH-KO-decreased genes to genes altered by CR in 13 experiments in which hepatic gene expression was compared between CR and ad lib-fed mice ( Figure 3—figure supplement 1 ) . Surprisingly , GHRH-KO-decreased genes compared more favorably with CR than GHRH-KO-increased genes . For 11 of 13 CR experiments , the 164 GHRH-KO-decreased were biased towards CR-decreased expression , and this trend was significant with respect to 7 of 13 CR experiments ( p<0 . 05 by rank-based GSEA; Figure 3—figure supplement 1 ) . However , we only identified 3 of 13 CR experiments in which the 141 GHRH-KO-increased were biased towards CR-increased expression ( p<0 . 05 by GSEA ) . Hepatic gene expression patterns in GHRH-KO mice were therefore in partial agreement with those of CR , with stronger concordance among GHRH-KO-decreased genes as compared to GHRH-KO-increased genes ( Figure 3—figure supplement 1 ) . We next compared GHRH-KO-increased and -decreased genes to results from a broader set of experiments ( Figure 3A , B ) . This revealed strong correspondence between genes altered in GHRH-KO mice and those altered in other long-lived mouse models . For instance , Cyp2b13 and Igfbp1 were among the genes most strongly elevated in GHRH- KO mice , and expression of both genes was similarly increased in Ames ( mixed background ) ( Amador-Noguez et al . , 2004 ) , Snell ( DW/J Pit1dw × C3H/HeJ Pit1dw-J background ) ( Boylston et al . , 2006 ) , Little ( Ghrhrlit/lit , B6 background ) ( Amador-Noguez et al . , 2004 , 2007 ) , Ghr−/− ( B6 background ) ( Rowland et al . , 2005 ) and Fgf21 Tg mice ( B6 background ) ( Zhang et al . , 2012 ) . Conversely , genes most strongly decreased in GHRH-KO mice were also decreased in Ames , Snell , Little , Ghr−/− and Fgf21 Tg mice ( e . g . , Hsd3b5 , Slco1a1 , Igf1 , Elovl3 , Keg1 and Igfals ) . Consistent with activation of Nrf2 signaling in GHRH-KO mice , genes altered most strongly in GHRH-KO mice were often similarly altered in mice carrying a liver-specific Keap1-KO mutation ( Yates et al . , 2009 ) ( e . g . , see Sult1e1 , Spink3 , Cyp39a1 , Hsd3b5 , Slco1a1 and Igf1; B6 background; Figure 3B ) . We also observed notable trends in which the effects of GHRH-KO on liver gene expression were correspondent with a feminization of gene expression patterns , as well as with the effects of castration in male mice ( A/JCr background; Figure 3B ) ( Rogers et al . , 2007 ) . Finally , we observed an unexpected trend , in which genes increased and decreased in GHRH-KO mice were similarly altered in mice with streptozocin-induced diabetes ( BALB/c background; Figure 3B ) ( Kobori et al . , 2009 ) . 10 . 7554/eLife . 01098 . 007Figure 3 . Hepatic gene expression profiles in GHRH-KO mice . Microarray analysis was used to identify 141 genes with increased expression in GHRH-KO liver tissue ( FDR < 0 . 05 and FC > 0 . 05 ) and 164 genes with decreased expression in GHRH-KO liver ( FDR < 0 . 05 and FC > 0 . 05 ) . ( A ) Venn diagram showing overlap of increased and decreased genes with sets of genes similarly altered in hepatic tissue of CR-fed mice , mice with a liver-specific Keap-1 mutation , and long-lived Ames dwarf mice . The same Affymetrix platform was used in each experiment ( Mouse Genome 430 2 . 0 array ) . ( B ) Genes most strongly increased in GHRH mice ( left ) and genes most strongly decreased in GHRH mice ( right ) . For comparison , heatmap colors show the fold change for each gene in other mutant mouse models and mice provided various treatments ( e . g . , CR , resveratrol , etc ) . Liver tissue was evaluated in all cases . ( C ) Top ranked motifs enriched in 2 kb regions upstream of GHRH-KO-increased genes ( left ) and GHRH-KO-decreased genes ( right ) . Asterisk symbols denote those motifs remaining significant following FDR-adjustment for multiple testing among the 1291 binding sites included within our motif dictionary . ( D ) Sequence logos for the top-ranked motif among GHRH-KO-increased genes ( left , V$HNF1_Q6\|TGGTTAATAATTA ) and the top ranked motif among GHRH-KO-decreased genes ( right , V$NMYC_02\|CATCTG ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 00710 . 7554/eLife . 01098 . 008Figure 3—figure supplement 1 . Characterization of genes altered in GHRH-KO mice . ( A ) Top-ranked gene ontology ( GO ) biological process ( BP ) terms overrepresented among GHRH-KO-increased genes . We identified 141 genes elevated in GHRH-KO mice as compared to normal controls ( FC > 1 . 50 and FDR < 0 . 05 ) . The chart lists GO BP terms most significantly overrepresented among these 141 genes ( p<0 . 05; conditional hypergeometric test ) . The number of GHRH-KO-increased genes associated with each GO BP term is listed in parentheses . The right margin lists example GHRH-KO-increased genes associated with each GO BP term . ( B ) Top-ranked gene ontology ( GO ) biological process ( BP ) terms overrepresented among GHRH-KO-decreased genes . We identified 164 genes decreased in GHRH-KO mice as compared to normal controls ( FC < 0 . 67 and FDR < 0 . 05 ) . The chart lists GO BP terms most significantly overrepresented among these 164 genes ( p<0 . 05; conditional hypergeometric test ) . The number of GHRH-KO-decreased genes associated with each GO BP term is listed in parentheses . The right margin lists example GHRH-KO-decreased genes associated with each GO BP term . ( C ) Comparison between the transcriptional effects of the GHRH-KO mutation and caloric restriction ( CR ) . We identified 141 genes elevated in GHRH-KO mice ( FC > 1 . 50 and FDR < 0 . 05 ) along with 164 genes decreased in GHRH-KO mice ( FC < 0 . 67 and FDR < 0 . 05 ) . We further identified 13 experiments in which microarrays were used to evaluate the effects of CR in mice ( left margin ) . Enrichment of the 141 GHRH-KO-increased genes was evaluated with respect to each of the CR experiments ( left panel ) . Positive enrichment statistics indicate that GHRH-KO-increased genes tend to be CR-increased , while negative statistics indicate that GHRH-KO-increased genes tend to be CR-decreased . Enrichment p values are listed in the right margin ( Wilcoxon rank sum test ) . Similarly , enrichment of the 164 GHRH-KO-decreased genes was evaluated with respect to each of the CR experiments ( right panel ) . Positive enrichment statistics indicate that GHRH-KO-decreased genes tend to be CR-increased , while negative statistics indicate that GHRH-KO-decreased genes tend to be CR-decreased . Enrichment p values are listed in the right margin ( Wilcoxon rank sum test ) . For each CR experiment , the left margin lists the duration of CR , the sex of the mice , and the mouse genotype ( when information is available ) . The Gene Expression Omnibus ( GEO ) or Array Express accession identifier is listed in parentheses , along with the GEO array platform identifier . ( Note: GPL1261 corresponds to the Affymetrix Mouse 430 2 . 0 Genome Array , which is the platform we used to measure gene expression in GHRH-KO mice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 008 We speculated that the shifts in hepatic gene expression in GHRH-KO mice could , at least in part , be due to the activation or repression of key transcription factors . To test this possibility , we scanned 2 kb regions upstream of increased and decreased genes for matches to sites within a dictionary of 1291 motifs associated with known DNA-binding proteins . We then used semi-parametric generalized additive logistic models to identify motifs present at significantly elevated frequency among GHRH-KO-altered genes , as compared to all other hepatic related genes . Among the 141 increased genes , the dominant trend was enrichment of hepatocyte nuclear factor ( HNF ) motifs in upstream regions ( Figure 3C ) . The strongest enrichment was observed for an HNF recognition site with consensus sequence 5-GTTAAT-ATT-3 ( p=7 . 3e-10; Figure 3D ) . Among the 164 decreased genes , enrichment for motifs associated with several DNA-binding proteins was observed ( Figure 3C ) . However , the strongest trend was increased frequency of a MYC binding site in upstream regions , with consensus sequence 5-CA—TG-3 ( p=1 . 8e-5; Figure 3D ) . In agreement with this trend , we also noted enrichment of MYC-MAX dimer binding sites upstream of GHRH-KO-increased genes ( p=2 . 7e-4; Figure 3C ) . Genes altered in GHRH-KO liver were thus characterized by increased frequency of HNF and MYC motifs in upstream regions . Detoxification and elimination of xenobiotics and endobiotics is a major function of the liver and is important for maintaining the metabolic homeostasis of the organism ( Osterreicher and Trauner , 2012 ) . Recent studies in long-lived C . elegans have linked the up-regulation of xenobiotic pathways with increased longevity ( McElwee et al . , 2004 ) . This link has been further supported in mammals , given that several mouse models of delayed aging are characterized by stress resistance and increased xenobiotic gene expression ( Amador-Noguez et al . , 2007; Steinbaugh et al . , 2012 ) . Consistent with this idea , our microarray analysis showed robust elevation in the expression of genes associated with xenobiotic detoxification pathways ( Figure 3; Supplementary file 1A , B ) . To confirm the array results , and to gain further insight into the xenobiotic gene regulation patterns in GHRH-KO mice , we used real-time RT-PCR to compare the expression levels of a set of phase I and phase II xenobiotic detoxification genes in liver and small intestine of GHRH-KO vs control mice . Of 15 such phase I mRNAs evaluated , 10 were found to be elevated in liver of KO mice , and seven of these were dramatically elevated compared to controls ( Figure 4A ) . In contrast , most of the hepatic phase II genes were modestly increased in GHRH-KO mice ( Figure 4B ) . However , Sult2a2 , the gene increased most strongly in microarray screening , was shown by RT-PCR to be elevated by more than 1000-fold in KO mice ( Figure 4B ) . Surprisingly , expression of phase I and II genes was similar in small intestine of KO and control mice , suggesting that the effect of GH on these genes may be liver-specific . 10 . 7554/eLife . 01098 . 009Figure 4 . Alteration in Xenobiotic detoxification genes . ( A ) Expression of phase I xenobiotic metabolism levels in liver and small intestine of KO and control mice . ( B ) Expression of phase II xenobiotic genes . ( C ) GH replacement effects on hepatic xenobiotic gene expression in GH deficient Ames dwarf mice ( df ) and littermates . mRNA was measured using real-time RT-PCR . Data normalized to Gapdh or actin values and were expressed as a ratio ( fold change ) to levels of mRNA in control mice . Bars indicate mean ± SEM for male KO or df and age-matched control male mice . N = 8 mice per group; p<0 . 05 , p<0 . 01 , p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 00910 . 7554/eLife . 01098 . 010Figure 4—figure supplement 1 . Cyp2b10 protein expression is elevated in GHRH-KO mouse liver tissue . ( Upper panel ) Representative immunoblots of Cyp2b10 and β-tubulin total protein expression are shown . ( Lower panel ) The means and SEM of Cyp2b10 expression were normalized to β-tubulin levels ( n = 5; p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 01010 . 7554/eLife . 01098 . 011Figure 4—figure supplement 2 . Increased cytochrome P450 activity in the liver of GHRH-KO mice . Change in resorufin formation per minute was measured in liver microsomes by fluorescence spectrophotometry . Elevations in the rate of methoxyresorufin ( MROD ) , pentoxyresorufin ( PROD ) and ethoxyresorufin ( EROD ) formation were observed in GHRH-KO mice when compared with controls . ( n = 4; p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 011 Our data suggested that the disruption of GH signaling leads to elevated expression of genes involved in hepatic xenobiotic metabolism . Previously , we showed that a short-term GH treatment ( 6 weeks ) in Ames dwarf mice during early stages of postnatal life could both shorten lifespan and decrease the cellular stress resistance ( Panici et al . , 2010 ) . To determine whether GH replacement also influences the expression of hepatic xenobiotic genes , we evaluated each of these mRNAs in liver of Ames dwarf mice that had been exposed to GH replacement for a period of 6 weeks . As shown in Figure 4C , consistent with the pattern in GHRH-KO mice , hepatic expression of these genes was robustly elevated in Ames dwarf mice compared to their littermate controls ( p<0 . 01 ) . GH-treatment of these mutant mice dramatically suppressed the elevation of these genes including Cyp2b9 , Cyp2b10 , Cyp4a14 , Fmo3 and Sult2a2 ( two-tailed t test; p<0 . 01 ) ( Figure 4C ) , but had no effect on housekeeping control genes . These findings suggest that GH is a direct regulator of xenobiotic mRNAs in hepatocytes and that elevation of such genes in GH deficient mice is a direct consequence of attenuated GH signaling . The GH/IGF-1 signaling pathways have been shown to play key role in controlling stress resistance , lifespan , and aging in different organisms ( Kenyon , 2010; Bartke , 2011 ) . We hypothesized that the local action of GH/IGF-1 within specific tissues underlies heightened stress cellular stress responses in GHRH-KO mice . The family of insulin-like growth factor ( IGF ) and related molecules comprises growth factors ( IGF-I , IGF-II ) , their receptors ( IGF-IR , IGF-IIR ) , and six structurally related IGF binding proteins ( IGFBP-1–6 ) ( Hwa et al . , 1999 ) . Using real-time RT-PCR analysis , we found a profound suppression of IGF-I transcription and significant up-regulation of Igf2 , Igfbp1 and Igfbp2 mRNA levels in the liver of GHRH-KO mice relative to controls ( Figure 5A ) . However , such differences were absent in different brain regions , such as cerebral cortex ( Figure 5A ) . This is consistent with previous reports in Ames dwarf mice , indicating that localized regulation of GH/IGF-1-associated mRNAs in the brain is independent of pituitary-derived GH secretion ( Sun et al . , 2005 ) . 10 . 7554/eLife . 01098 . 012Figure 5 . Stress signaling pathways . ( A ) Expression of IGF family related mRNA levels in the liver and cortex of GHRH-KO and control mice using real-time RT-PCR . Data are normalized to Gapdh values and expressed as a ratio to the level seen in control mice . Each bar represents means ± SEM for eight mice of each genotype , p<0 . 05 , p<0 . 01 , p<0 . 001 . ( B ) Representative autoradiographs of Western blots for phosphorylated and total forms of MEK , ERK1/2 and P38 protein in liver lysates of KO and control mice . Each lane corresponds to a different mouse . ( C ) Expression of Egr1 , Jun and Fos mRNA levels in the liver KO and control mice . Data are normalized to 18S values and expressed as a ratio to the level seen in control mice . ( D ) Representative immunoblot of the indicated proteins from cortex lysates . ( E ) Representative autoradiographs of Western blots detecting nuclear accumulation of Nrf2 and Ho1 in the cytoplasm from liver lysates . Gapdh was the cytoplasmic marker and Lamin B was the nuclear marker . Each lane corresponds to a different mouse . ( F ) Expression of Nrf2-dependent mRNA levels in the liver KO and control mice by mRNA Q-PCR . Data are normalized to Gapdh values and expressed as a ratio ( fold change ) to the level seen in control mice . N = 8 male mice per group; p<0 . 05 , p<0 . 01 , p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 012 We further explored the downstream targets/effects of local IGF-I signaling in liver and brain . In comparison to littermate controls , GHRH-KO mice were characterized by decreased hepatic phosphorylation of MAPKs , including the MEK , ERK , and P38 kinases , each of which is known to participate in cellular stress responses ( Figure 5B ) . In contrast , brain tissues from GHRH-KO mice showed no change in phosphorylation of these kinases ( Figure 5D ) . These data further support preservation of local GH/IGF-I signaling in brain regions of KO mice . The transcriptional regulation of immediate early genes ( IEGs ) including Egr1 , Fos , and Jun are dependent on activation of MAPK signaling pathways . As expected , hepatic IEGs mRNA levels are significantly repressed in GHRH-KO mice ( Figure 5C ) whereas no alteration was detected in the brains . Our microarray analysis uncovered correspondence between the hepatic gene expression profiles of GHRH-KO mice and mice carrying a liver-specific Keap1-KO mutation , suggesting that Nrf2 activity might be elevated in GHRH-KO mice ( Figure 3 ) . We thus examined Nrf2 signaling in GHRH-KO tissues . As shown in Figure 5E , Western blot analysis of liver lysates showed that GHRH-KO mice had a higher nuclear level of Nrf2 protein than control animals . To further assess Nrf2 activity in the liver , we next evaluated expression of Nrf2 dependent genes by quantitative RT-PCR ( Figure 5F ) . These genes included those encoding glutamate cysteine ligase modifier subunit 1 ( GCLM ) , quinoneoxidoreductase 1 ( NQO1 ) , thioredoxinreductase ( TXNRD ) , NAD ( P ) H quinoneoxidoreductase 1 ( NQO1 ) and metallothioneins 1 and 2 . Each of these Nrf2-dependent genes was expressed at higher levels in liver from KO mice than in liver from controls . This further supports up-regulation of Nrf2 signaling in GHRH-KO mice . Our microarray analysis of gene expression indicated that disruption of the GHRH gene had effects that were only partly overlapping with those of CR , indicating that isolated GH deficiency and CR may extend lifespan by independent pathways . To evaluate the effects of CR , 199 GHRH-KO mice and 211 of their normal siblings were divided into two groups after ad libitum ( AL ) feeding for the first 12 weeks of life , and then were subjected either to CR or to continued AL feeding . CR produced the expected reduction in body weight in both GHRH-KO and control mice ( Figure 6B , C ) . The relative decrease in body weight in response to CR was comparable between KO and control mice in both sexes . The convergence of average body weighs of N-AL and N-CR males after 80 weeks of age almost certainly represents age-related weight loss in N-AL animals and its delay in the N-CR group . 10 . 7554/eLife . 01098 . 013Figure 6 . Lifespan and growth curve in response to long-term CR . ( A ) Kaplan–Meier survival plot of GHRH-KO ( KO ) and Control mice ( N ) that were fed AL or subjected to long-term CR ( Sex pooled ) ; N-AL ( N = 108 ) , N-CR ( N = 105 ) , KO-AL ( N = 97 ) and KO-CR ( N = 102 ) . Time course of changes in body weight for male ( B ) and female ( C ) GHRH-KO and control mice that were fed AL or subjected to CR . Animals were weighed weekly starting at 4 weeks of age . ( D ) Kaplan–Meier survival plot of male KO and Control mice that were fed AL or subjected to long-term CR; N-AL ( N = 56 ) , N-CR ( N = 53 ) , KO-AL ( N = 39 ) and KO-CR ( N = 42 ) . ( E ) Survival plot of female KO and Control mice; N-AL ( N = 52 ) , N-CR ( N = 52 ) , KO-AL ( N = 58 ) and KO-CR ( N = 60 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 013 Kaplan–Meier survival curves ( Figure 6A ) indicate that in GHRH-KO mice , CR produced significant increase in overall survival , average , median , and maximal longevity when the data from both sexes were combined ( log-rank test , p<0 . 0001 ) . Further analysis of each sex separately showed that CR female KO mice out-live AL KO mice by 21% ( median lifespan: 1156 days vs 956 days; p=0 . 0001; logrank test; p<0 . 0001; Figure 6E ) . In contrast , the median lifespan of male CR KO mice does not significantly exceed that of male AL KO mice ( 945 days vs 928 days Figure 6D ) , but CR still further increased overall survival of male KO mice ( logrank test; p=0 . 0382 ) . As expected , the CR-induced increase in longevity in the control mice was similar to findings in other strains of mice ( p<0 . 0001 ) . Applying the method of Wang/Allison , maximal lifespan of KO mice was increased by CR in both sexes , by 17% for females ( p<0 . 05; Figure 6E ) and by 9% for males ( p<0 . 05; Figure 6D ) . These results , taken together , support the idea that mechanisms underlying extended longevity in GHRH-KO mutants and CR-fed mice are at least partially distinct . Fasting plasma glucose levels were significantly reduced in GHRH-KO AL-fed mice compared with control AL-fed mice ( p<0 . 005; Figure 7A ) ; CR did not significantly affect the glucose level in either genotype . Plasma insulin was greatly decreased in N-CR , KO-AL , and KO-CR relative to N-AL . CR also resulted in significant further reduction of insulin levels in GHRH-KO mice compared with their AL group ( p<0 . 05 ) . As expected , CR resulted in a significant reduction in plasma IGF-I levels in controls , whereas IGF-I levels in GHRH-KO mice remained at similar very low levels regardless of their diet ( Figure 7A ) . The level of adiponectin , a plasma protein secreted specifically from adipocytes , was higher in GHRH-KO than in the control group ( p<0 . 01; Figure 7A ) . Interestingly , CR significantly increased plasma adiponectin levels in both genotypes ( p<0 . 005 and 0 . 01 respectively; Figure 7A ) . Conversely , circulating leptin levels were reduced by CR only in KO mice ( p<0 . 01 ) , while KO AL animals had higher leptin levels than the controls ( p<0 . 05; Figure 7A ) . Moreover , fasting plasma triglyceride and cholesterol levels were significantly decreased in KO mice and further suppressed by CR ( p<0 . 05; Figure 7A ) . 10 . 7554/eLife . 01098 . 014Figure 7 . Blood parameters alteration in response to CR . ( A ) Various plasma parameters from GHRH-KO and Control male mice subjected to caloric restriction . Different superscripts denote significant difference at p<0 . 05 . Data represent the means ± SEM . ( B ) Circulating levels of Fgf21 . ( C ) Hepatic Fgf21 mRNA levels . Data normalized to Gapdh or actin values and were expressed as a ratio to levels of mRNA in control mice . Bars indicate mean ± SEM . N = 8 mice per group; p<0 . 05 , p<0 . 01 , *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 01410 . 7554/eLife . 01098 . 015Figure 4—figure supplement 1 . Hepatic FGF4R and β-Klotho mRNA levels . Data normalized to GAPDH values and were expressed as a ratio ( fold change ) to levels of mRNA in control mice . Bars indicate mean ± SEM N = 8 mice per group . DOI: http://dx . doi . org/10 . 7554/eLife . 01098 . 015 Fibroblast growth factor 21 ( FGF21 ) , a critical metabolic regulator of glucose and lipid metabolism , is secreted primarily from the liver in response to prolonged fasting and plays a major role in coordinating adaptive changes such as mobilizing and burning fatty acids among different tissues ( Potthoff et al . , 2012 ) . A most recent study has shown that overexpression of FGF21 markedly extends lifespan in mice ( Zhang et al . , 2012 ) . Intriguingly , the Fgf21 transgenic mice have been found to be acting to suppress GH action ( Inagaki et al . , 2008 ) . In this context , we further evaluated the effect of GHRH deletion and CR on FGF21 expression . To our surprise , CR profoundly suppressed serum FGF21 levels in both male and female GHRH-KO and control mice ( p<0 . 001; Figure 7B ) whereas isolated GH deficiency had almost no effect on FGF21 levels . Since liver is the major source of circulating FGF21 , hepatic FGF21 mRNA expression was quantified by real-time RT-PCR . Consistent with plasma levels , the transcriptional levels of FGF21 in the liver were markedly inhibited by CR in both GHRH-KO and control mice ( Figure 7C ) . However , no change was detected in the levels of the receptor FGF4R and β-Klotho in liver ( Figure 7—figure supplement 1 ) . Improved insulin sensitivity combined with low insulin and glucose levels are characteristics of a series of long-lived mutant mice and mice with long-term CR ( Bartke , 2011 ) . On the contrary , opposite phenotypic features characteristic of metabolic syndrome , including hyperinsulinemia , and insulin resistance , are associated with increased risk of various age-related diseases and with reduced life expectancy ( Stensvold et al . , 2011; Noale et al . , 2012 ) . Consistent with this notion , insulin sensitivity is also improved in the GHRH-KO mice , further supporting that this is a key contributor of lifespan extension of mice . We suspect that at least some of the gene expression responses observed in hepatic tissue of GHRH-KO mice can be attributed to enhanced insulin signaling . For instance , the GHRH-KO-increased genes we identified included Sox4 and Vcam1 , each of which have been associated with response to glucose stimulus . In addition , among the 164 GHRH-KO-decreased , upstream regions were enriched with motifs recognized by MYC , which is a transcription factor associated with the TOR branch of the insulin signaling pathway . Increased resistance to multiple forms of lethal stress has been consistently associated with lifespan extension in long-lived mutant worms , flies and mice suggested the appealing idea that the longevity seen in these mutants was a consequence of their stress resistance ( Miller , 2009 ) . Our evaluation of changes in hepatic gene expression profiles and stress response pathways also points toward this direction . For instance , several of the genes with increased expression in GHRH-KO liver were associated with the GO biological process term ‘cellular response to stress’ ( e . g . , Blm , Ccnd1 , Dclre1a , Ddx1 , Gpx3 , Il1rn , Krt20 , Morf4l2 , Ppargc1a , Prpf19 , Rtn4 , Slc2a1 , Sox4 and Vldlr ) . Additionally , we noted that several genes with increased expression in GHRH-KO liver had also been induced in experiments where mice had been exposed to injury or toxicity stressors , including partial hepatectomy ( e . g . , Cenpq , Ccnd1 and Cdkn2c ) , the HSF-activating compound quercetin ( e . g . , Gpr146 , Il1rn and Tef ) , and high doses of acetaminophen ( e . g . , Vcam1 , Mafb and Slc2a1 ) . The transcription factor Nrf2 plays a central role in regulating many antioxidant-related genes against multiple types of stress ( Ma , 2013 ) . Under normal conditions , Nrf2 is sequestered in the cytosol by Keap1 , the cytosolic repressor of Nrf2 . In response to various stress , Nrf2/Keap1 binding is disrupted , and Nrf2 translocates into the nucleus to stimulate the activation of various antioxidant-associated genes ( Ma , 2013 ) . Overexpression of SKN-1 ( Nrf2 ortholog ) can prolong C . elegans lifespan ( Tullet et al . , 2008 ) . Disruption of Keap1 , which in turn increase Nrf2 activity , also increases the male Drosophila melanogaster longevity ( Sykiotis and Bohmann , 2008 ) . However , whether Nrf2 can contribute to the mammalian aging and affect lifespan remains undemonstrated . Intriguingly , as revealed by our microarray comparison , liver-specific Keap1 knockout mice have increased resistance to hepatic stressor acetaminophen and exhibit a very similar hepatic gene expression pattern to the one of GHRH-KO mice ( Figure 3A ) ( Okawa et al . , 2006 ) . Genes induced in Keap1 ( −/− ) and GHRH-KO mice include Nqo1 , Adora1 , Cyp39a1 , Trim24 and Gstt3 , while genes decreased in both models include Ttc39c , Tmem19 , Egfr , Mcm10 and Slco1a1 . Moreover , nuclear Nrf2 protein and many ARE genes were upregulated in GHRH-KO mice suggesting that Nrf2 could play an important role in lifespan extension of these long-lived mice . All organisms are continuously exposed to a wide variety of exogenous and endogenous toxic compounds in the environment . Recently , it has been proposed that a generalized up-regulation of xenobiotic detoxification pathways may be a shared feature of long-lived mutants and a key mechanism of longevity assurance ( McElwee et al . , 2004; Shore et al . , 2012; Steinbaugh et al . , 2012 ) . Our findings show both a strong increase and decrease in xenobiotic metabolism genes in GHRH-KO mice . Among GHRH-KO-increased genes , several were associated with P450 drug metabolism ( e . g . , Cyp2b13 , Cyp2c38 and Cyp2c39 ) . Similarly , among GHRH-KO-decreased genes , we noted that several were associated with xenobiotic metabolic processes ( e . g . , Acsl1 , Gstp1 and Ugt2b1 ) . The mechanisms underlying activation of xenobiotic metabolism genes in the absence of environmental toxins remain unknown . Some studies have shown elevation of bile acids ( as endogenous xenobiotics ) level and altered bile acid metabolism in GH deficient mice ( Amador-Noguez et al . , 2004 , 2007 ) . Moreover , several GH-mediated transcriptional factors including Nrf2 , STAT 5b , and hepatic nuclear factor ( HNF ) 4alpha have been shown to play an important role in xenobiotic gene expressions ( Waxman and Holloway , 2009; Wu et al . , 2012 ) . In this study , we postulate that the activation of xenobiotic metabolism pathway , presumably through inhibition of GH signaling and alterations of the corresponding transcriptional factors , protects the tissues from the damaging effects of endogenous and exogenous molecules and thus contributes to maintaining tissue and metabolic homeostasis during aging . We note that an important avenue for future work will be to carry out end-of-life pathology studies of GHRH-KO mice . This will provide information on likely causes of death and how such causes may differ from those in control animals . At this point , however , based upon findings in other GH-related long-lived mutants ( Ikeno et al . , 2003; Vergara et al . , 2004; Ikeno et al . , 2009 ) , we suspect that reduced incidence and/or delayed onset of cancer may have contributed to extended longevity of GHRH-KO mice . Comparison of the impact of CR on longevity in GHRH-KO mice to the results obtained previously in other long-lived GH-related mutants reveals both similarities and differences . Resembling findings in Ames dwarf mice ( Bartke et al . , 2001 ) , CR produced a further significant extension of longevity in both sexes of GHRH-KO animals . In contrast to these findings , in GH resistant Ghr−/− mice , CR does not alter longevity of males and produces only a small , although statistically significant reduction of late mortality in females ( Bonkowski et al . , 2006 ) . Relating these differential responses to CR to the impact of these mutations on insulin levels lends support to our hypothesis that suppression of insulin levels is an important mechanism of life extension by CR . Indeed CR has no or little impact on longevity of animals in which insulin levels are already profoundly suppressed ( Bonkowski et al . , 2006 ) . However , we cannot exclude that differences in the genetic background of animals could also partially explain these differences . The CR-induced extension of longevity in GHRH-KO mice was greater in females than in males , while in GHRH-KO mice subjected to CR , longevity was increased in females only . The mechanism for the sex dimorphism in response to CR is poorly understood . Intriguingly , deletion of several genes related to insulin/IGF-I signaling including S6k1 ( Selman et al . , 2009 ) , IRS1 ( Selman et al . , 2008 ) extends longevity only in females . Perhaps divergent effects of male and female sex hormones on adiposity , adipose tissue distribution , inflammation , cardiac function and/or neuroprotection renders males less responsive to the beneficial effects of reducing insulin , somatotropic and mTOR signaling by dietary or genetic interventions ( Bartke et al . , 2013 ) . It deserves emphasis that the ablation of GHRH in female mice markedly extends longevity even though insulin levels are not significantly reduced . Insulin levels are reduced in Ames dwarf , Snell dwarf , and Ghr−/− , but not in transgenic mice expressing a GH antagonist in which longevity is not increased ( Coschigano et al . , 2003 ) . Suppression of insulin levels is one of the most consistent and robust responses to CR in mice and in other mammalian species . Accordingly , even in humans , insulin sensitivity seems to be a common feature found in centenarians ( Paolisso et al . , 1996 ) . Intriguingly , GHRH-KO mice have a significantly higher percent body fat ( Figure 2 ) throughout their lifespan . In agreement with the increased obesity , KO mice have elevated serum level of leptin . Interestingly , we found that long-term CR reduced circulating leptin level but increased the adiponectin level in GHRH-KO mice . These results suggest that GH signaling interacts with CR affecting gene expressions and secretion patterns of adipose tissues in these animals . FGF21 plays a major role in eliciting and coordinating the adaptive starvation response and promotes similar physiological changes caused by CR , including decreased glucose levels , increased insulin sensitivity , and improved lipid mobilization ( Potthoff et al . , 2012 ) . A most recent study has shown that overexpressing FGF21 in mice prolongs lifespan dramatically ( Zhang et al . , 2012 ) . Furthermore , Fgf21 Tg and CR-fed mice show similarity with respect to their hepatic gene expression profiles ( Zhang et al . , 2012 ) . For these reasons , FGF21 is speculated to increase longevity by partially mimicking the effect of CR in liver and potentially other tissues . In previous studies , however , CR did not induce transcription of hepatic FGF21 , nor did CR alter levels of FGF21 in circulation ( Sharma et al . , 2012; Zhang et al . , 2012 ) . Intriguingly , our data show that long-term ( over 12 months ) 40% CR profoundly suppressed hepatic FGF21 mRNA and plasma FGF21 concentrations in both GHRH-KO and control mice . This result differs from previous studies , potentially due to differences in mouse strains or the use of short-term CR in earlier work . Nevertheless , these results do not support FGF21 as a key mediator of the pro-longevity effects of CR . Previously , FGF21 has been shown to cause GH resistance and inhibit hepatic GH signaling through Stat5-dependent manner ( Inagaki et al . , 2008 ) . Interestingly , we found no alteration in FGF21 expression in GHRH-KO mice compared to controls . Moreover , our hepatic transcriptome analysis has revealed pretty good overlap in genes affected by GHRH deletion and FGF21 overexpression ( Figure 3B ) . Thus , our results further support the idea that FGF21 longevity effect is likely mediated by the GH signaling . Also , the regulatory crosstalk between GH signaling and FGF21 action may play a pivotal role in controlling metabolic homeostasis during starvation and chronic dietary restriction . It will be imperative , in future work , to better understand how CR and GH signaling interact with each other at both the metabolic and cellular levels . Knowledge gained through these experiments will shed light on the mechanism by which aging and lifespan are regulated by the GH pathway . GHRH-KO mice and their littermate controls ( on a mixed C57BL6 and 129SV background ) were bred in a closed colony at the RS’s laboratory ( Alba and Salvatori , 2004 ) , housed under standard conditions ( 12-hr light/12-hr dark cycling and 20–23°C ) , and fed Lab Diet Formula 5001 ( 23% protein , 4 . 5% fat , 6% fiber ) ( Nestle Purina , St . Louis , MO ) . All animals were fed AL for the first ∼12 weeks of life . Thereafter , the mice were either fed AL ( AL groups ) or 40% of AL ( CR groups ) . Mice were weighed approximately 16–20 hr after the CR groups had been fed . Animal protocols were approved by the Animal Care and Use Committee of Southern Illinois University . 16 hr-fasted mice underwent GTT by i . p . injection with 1 g of glucose per kg of body weight ( BW ) . Blood glucose levels were measured at 0 , 15 , 30 , 45 , 60 , and 120 min using a PRESTO glucometer ( AgaMatrix , Salem , NH ) for GTT . Non-fasted mice were injected by i . p . with 1 IU porcine insulin ( Sigma , St . Louis , MO ) per kg of BW Blood glucose levels were measured at 0 , 15 , 30 , and 60 min for ITT . The data for both ITT and GTT are presented as a percentage of baseline glucose . p values were calculated by unpaired , two-tailed Student’s t tests to compare the specific time points . The homeostasis model assessment of insulin resistance ( HOMA-IR ) was calculated using glucose and insulin concentrations obtained after 7 hr of food withdrawal , using the following formula: fasting blood glucose ( mg/dl ) × fasting insulin ( µU/ml ) /405 . Hepatic gene expression in GHRH-KO mice and normal littermates was profiled using the Affymetrix Mouse Genome 430 2 . 0 oligonucleotide array platform ( n = 3 per genotype , males , 10 months of age ) . Standard Affymetrix quality control metrics were calculated for each hybridization , including the percentage of probe sets with signals detected above background ( i . e . , percent present ) , global RNA degradation score , average background , intensity scale factor , and four measures derived from the fitting of probe-level models ( RLE median , RLE IQR , NUSE median and NUSE IQR ) . Hierarchical cluster analysis was also performed to identify potential outliers . All chips were retained for further analyses based upon these quality control evaluations . Expression scores were calculated for each chip using robust multichip average ( RMA ) . The Affymetrix Mouse Genome 430 2 . 0features a total of 45101 probe sets; however , we removed 19956 from consideration since they were not significantly expressed above background in any of the six array hybridizations ( Wilxon signed rank test as implemented in the MAS 5 . 0 algorithm ) . Our analysis was thus based upon 25145 probe sets with expression significantly above background in at least one of the six hybridizations ( 13201 unique mouse genes ) . Probe sets showing increased or decreased expression were identified using empirical Bayes methods as implemented in the limma software package ( R Bioconductor ) . An FDR correction for multiple hypotheses testing among all probe sets was performed using the Benjamini-Hochberg procedure . Since the Mouse Genome 430 2 . 0 includes some ‘sibling’ probe sets that target mRNAs associated with the same gene , we limited redundancy in our results by considering only the most significantly altered probe set associated with a given gene ( i . e . , the probe set with the lowest p value ) . Overall , we identified 365 genes differentially expressed between GHRH-KO mice and normal littermates ( 141 increased and 164 decreased genes; FDR < 0 . 05 with fold-change >1 . 5 or <0 . 67 ) . With respect to these gene sets , we tested for significantly overrepresented Gene Ontology and KEGG terms using a conditional hyper-geometric test as implemented in the GOstats package ( R Bioconductor ) . For these analyses , significant overrepresentation was assessed relative to a background gene set that included only the 12836 liver-expressed genes not identified as differentially expressed . Enrichment of motifs in regions upstream of differentially expressed genes was evaluated by screening a dictionary of 1291 motifs associated with DNA-binding proteins , which had been assembled by pooling motifs available from the Jaspar , UniPROBE and Transfac databases . Motif enrichment in regions upstream of differentially expressed genes was assessed using semi parametric generalized additive logistic regression models ( Swindell , 2012; Swindell et al . , 2012 ) . Protein-coding sequences and repetitive DNA elements were masked in all sequence scans . In all cases , motif enrichment in regions upstream of differentially expressed genes was assessed relative to a background gene set that included only liver-expressed genes not identified as differentially expressed . Quantitative real-time PCR was performed using a Rotor-Gene 3000 system ( Corbett Research ) with a QuantiTect SYBR Green RT-PCR kit ( Bio-rad ) as described ( Sun et al . , 2011 ) . In brief , cells were homogenized with RNA extraction buffer ( TRIZOL reagent; Life Technologies , CA ) to yield total RNA following the manufacturer’s instructions . Total RNA was reverse transcribed with poly-dT oligodeoxynucleotide and SuperScript II . After an initial denaturation step ( 95°C for 90 s ) , amplification was performed over 40–45 cycles of denaturation ( 95°C for 10 s ) , annealing ( 60°C for 5 s ) , and elongation ( 72°C for 13 s ) . Amplification was monitored by measuring the fluorometric intensity of SYBR Green I at the end of each elongation phase . The oligonucleotide-specific primers are shown in Supplementary file 1C . Glyceraldehyde-3-phosphate dehydrogenase ( Gapdh ) or beta-actin expression was quantified to normalize the amount of cDNA in each sample . The change in threshold cycle number ( Ct ) was normalized to the Gapdh reference gene by subtracting CtGapdh from Ctgene . The effect of treatment ( Ct ) was calculated by subtracting Ctnormal from CtTg . Fold induction was determined by calculating 2Ct . Tissues were homogenized in 0 . 5 ml ice-cold T-PER tissue protein extraction buffer ( Thermo Scientific , Rockford , IL ) with protease and phosphatase inhibitors ( Sigma ) . 40 µg of total protein was separated electrophoretically according to size by SDS-polyacrylamide gel electrophoresis using Criterion XT Precast Gel ( Bio-Rad , Hercules , CA ) , and blotted with the antibodies . For visualization of specific bands in the chemiluminescence assays , the membrane was exposed to X-OMAT film; for chemifluorescence the membrane was incubated with ECF ( enhanced chemifluorescence ) substrate and a digital image was generated with the Molecular Dynamics Storm system . Quantification of immunoblot signals was performed using the ImageQuant software package ( Molecular Dynamics , Sunnyvale , CA ) . All groups of discrete , single-measurement data were tested by Student’s t test using GraphPad PRISM 4 . 03 ( GraphPad Software Inc . , La Jolla , CA ) . Insulin tolerance testing results were assessed by analysis of variance ( ANOVA ) with repeated measures using SPSS 17 . 0 ( SPSS Inc . , Chicago , IL ) . Overall survival was tested by logrank test , using GraphPad PRISM 4 . 03 . Maximal survivorship was evaluated as previously described ( Wang et al . , 2004 ) . The following antibodies were obtained for immunoblotting: p38 MAPK , phospho-p38 MAPK ( Thr180/Tyr182 ) , ERK , phospho-ERK ( Thr202/Tyr204 ) , JNK , phospho-JNK ( Thr183/Tyr185 ) , phospho-Akt ( Ser473 ) and Akt , from Cell Signaling Technology ( Beverly , MA ) ; Nrf2 antibody from Novus Biologicals ( Littleton , CO ) ; β-actin , inhibitor from Sigma-Aldrich Corp . ; and goat anti-rabbit and goat anti-mouse antibodies from Santa Cruz Biotechnology , Inc . ( Santa Cruz , CA ) . Methods used to assess enrichment of transcription factor binding sites in 2 kb regions upstream of differentially expressed genes have been described in a recent publication ( Swindell et al . , 2013 ) . In brief , we first assembled a dictionary of 1209 motifs representing the empirically-determined recognition sites of DNA-binding proteins ( e . g . , by chip-chip , chip-RNAseq , or protein binding microarray ) . These motifs were obtained by initially pooling those available in three databases , including Jaspar ( 145 motifs ) , UniPROBE ( 295 motifs ) and TRANSFAC ( 819 motifs ) . This yielded an initial set of 1259 motifs , which were then filtered to exclude repetitive motifs or any motifs fewer than four base pairs in length , yielding the final set of 1209 motifs ( Swindell et al . , 2013 ) . For each mouse gene , we retrieved the 2 kb upstream region based upon coordinates provided in UCSC ref gene files and sequences available from Bioconductor ( package: BSgenome . Mmusculus . UCSC . mm10 ) . Upstream sequences for all mouse genes were scanned for matches to each of the 1209 motifs , respectively ( with masking of protein-coding sequences and repetitive DNA elements ) . Motif matches were identified using the ‘matchPWM’ function ( R package: Biostrings ) , with an 80% match threshold , that is , a motif match was counted if the position weight matrix ( PWM ) similarity score exceeded 80% of the maximum score for that PWM . For each mouse gene , this yielded counts for each PWM indicating the frequency of motif matches in 2 kb upstream regions . The same procedure was applied to determine if motifs were enriched with respect to the 141 GHRH-KO-increased genes and the 164 GHRH-KO-decreased genes , respectively . To assess whether GHRH-KO-increased genes were enriched with respect to the number of occurrences for a given motif , for example , we used semi parametric generalized additive logistic regression models ( Swindell et al . , 2013 ) . For these models , the response variable was an indicator with value 1 if a gene was included among the 141 GHRH-KO-increased , and with value 0 if the gene was not included among the 141 GHRH-KO-increased . Only liver-expressed genes were included in the analysis . Models included two predictor variables , including the number of motif occurrences in the 2 kb upstream region ( x1 ) and the length of sequence scanned ( x2 ) . The variable x1 was estimated using parametric methods , while x2 was included as a non-parametric term with cublic spline smoothing . To assess motif enrichment , we evaluated the significance of the Z statistic associated with the coefficient estimate obtained forx1 . Models of this structure were generated for each of the 1209 motifs . To control the false discovery rate with respect to the 1209 tests , p values associated with Z statistics were adjusted using the Benjamini–Hochberg method . Mice were subjected to indirect calorimetry ( PhysioScan Metabolic System from AccuScan Instruments , Columbus , OH ) as described before ( Westbrook et al . , 2009 ) . This system uses zirconia , infrared sensors and light beams arrays to monitor oxygen ( VO2 ) , carbon dioxide ( VCO2 ) , and spontaneous locomotor activity , respectively inside respiratory chambers in which individual mice were tested . All comparisons are based on animals studied simultaneously in eight different chambers connected to the same O2 , CO2 and light beam sensors in an effort to minimize the effect of environmental variations and calibration on data . After a 24-hr acclimation period , mice were monitored in the metabolic chambers for 24 hr with ad libitum access to standard chow ( Laboratory Diet 5001 ) and water . Gas samples were collected and analyzed every 5 min per animal , and the data were averaged for each hour . Plasma was obtained from blood collected by cardiac puncture at sacrifice and used for measurement of insulin using Mouse Insulin ELISA Kits ( Crystal Chem , Downers Grove , IL ) . Following the manufacturer’s protocol , total ketone bodies and non-esterified free fatty acids ( NEFA ) were measured using colorimetric assays from Wako Chemicals ( Richmond , VA ) ; glycerol was measured using kits from Sigma and triglycerides using kits from Pointe Scientific ( Canton , MI ) , respectively . Adiponectin and resistin levels were assayed using Mouse Adiponectin/Resistin ELISA Kits ( Linco Research , St . Charles , MO ) . Leptin levels were evaluated using Mouse Leptin ELISA Kits ( Crystal Chem Inc . , Downers Grove , IL ) . TNF-α and IL-6 were measured using Mouse TNF-α/IL-6 ELISA Kits ( Biosource , Camarillo , CA ) . Plasma FFAs were assayed using optimized enzymatic colorimetric assays ( Roche , Penzberg , Germany ) . Blood was taken from the tail to measure blood glucose using a glucometer ( AgaMatrix , Salem , NH ) . CYP1A and CYP2B activity was measured by resorufin conversion from methoxyresorufin ( 7 methoxyresorufin O deethylation , MROD ) , ethoxyresorufin ( 7 ethoxyresorufin O deethylation , EROD ) , or pentoxyresorufin ( 7 pentoxy-resorufin O deethylation , PROD ) , as previously described ( Anwar-Mohamed et al . , 2011 ) . Liver samples were collected from mice fed chow containing 1% tBHQ or control chow for 50 days that were exposed to 50 mg/kg diquat or saline for 6 hr prior to dissection . Unfrozen liver samples were briefly homogenized with a Potter-Elvehjem homogenizer and suspended in sucrose buffer . Microsomes were isolated via ultracentrifugation with the commercially available Endoplasmic Reticulum Isolation Kit ( ER0100; Sigma-Aldrich ) . Changes in resorufin fluorescence were measured in a 96-well microtiter plates following the protocol outlined in the commercially available Cytochrome P450 2B Fluorescent Detection Kit ( CYTO2B; Sigma ) .	There is increasing evidence that the hormonal systems involved in growth , the metabolism of glucose , and the processes that balance energy intake and expenditure might also be involved in the aging process . In rodents , mutations in genes involved in these hormone-signaling pathways can substantially increase lifespan , as can a diet that is low in calories but which avoids malnutrition . As well as living longer , such mice also show reductions in age-related conditions such as diabetes , memory loss and cancer . Many of these effects appear to involve the actions of growth hormone . Mice with mutations that disrupt the development of the pituitary gland , which produces growth hormone , show increased longevity , as do mice that lack the receptor for growth hormone . However , these animals also show changes in a number of other hormones , making it difficult to be sure that the reduction in growth hormone signaling is responsible for their increased lifespan . Now , Sun et al . have studied mutant mice that lack a gene called GHRH , which promotes the release of growth hormone . These mice , which have normal levels of all other pituitary hormones , lived for up to 50% longer than their wild-type littermates . They were more active than normal mice and had more body fat , and showed greatly increased sensitivity to insulin . Some of the changes in these mutant mice resembled those seen in animals with a restricted calorie intake , suggesting that the same mechanisms may be implicated in both . However , Sun et al . found that caloric restriction further increased the lifespans of their GHRH knockout mice , indicating that at least some of the effects of caloric restriction are independent of disrupted growth hormone signaling . The results of this study are an important step forward for understanding how growth hormone signaling and caloric restriction regulate aging , both individually and in combination . The GHRH knockout mice are likely to become an important model system for studying these processes and for understanding the complex interactions between diet and hormonal pathways .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2013	Growth hormone-releasing hormone disruption extends lifespan and regulates response to caloric restriction in mice
Animals use the temporal information from previously experienced periodic events to instruct their future behaviors . The retina and cortex are involved in such behavior , but it remains largely unknown how the thalamus , transferring visual information from the retina to the cortex , processes the periodic temporal patterns . Here we report that the luminance cells in the nucleus dorsolateralis anterior thalami ( DLA ) of pigeons exhibited oscillatory activities in a temporal pattern identical to the rhythmic luminance changes of repetitive light/dark ( LD ) stimuli with durations in the seconds-to-minutes range . Particularly , after LD stimulation , the DLA cells retained the entrained oscillatory activities with an interval closely matching the duration of the LD cycle . Furthermore , the post-stimulus oscillatory activities of the DLA cells were sustained without feedback inputs from the pallium ( equivalent to the mammalian cortex ) . Our study suggests that the experience-dependent representation of time interval in the brain might not be confined to the pallial/cortical level , but may occur as early as at the thalamic level . The ability to process and sense the temporal information of external stimuli is fundamental for humans and other animals in enabling them to adapt to the environment . Use of the perceived temporal information from previously experienced stimuli to predict upcoming events has been demonstrated in birds ( Gibbon et al . , 1984; Henderson et al . , 2006; Kalenscher et al . , 2006 ) , rodents ( Crystal and Baramidze , 2007; Agostino et al . , 2011 ) , primates , and humans ( Leon and Shadlen , 2003; Buhusi and Meck , 2005; Janssen and Shadlen , 2005; Penney et al . , 2008 ) . Previous studies on the neural mechanism underlying time perception and representation have mainly focused on the cerebellum ( Malapani et al . , 1998; Harrington et al . , 2004 ) , striatum ( Meck et al . , 2008; Mello et al . , 2015 ) , and cortex of humans and mammals ( Leon and Shadlen , 2003; Xu et al . , 2014 ) . A recent study showed , however , that some retinal ganglion cells in salamander and mouse retinae signal the time at which an omitted stimulus in a sequence of flashes would occur ( Schwartz et al . , 2007 ) . There is also increasing evidence showing that the visual thalamus filters , rather than passively relays , the visual information sensed by the retina as it signals to the cortex ( Cudeiro and Sillito , 2006; Saalmann and Kastner , 2011; Sherman , 2016 ) . Thus , the thalamus may also be involved in coding the temporal information of periodic visual stimuli . However , we have limited knowledge on how the visual thalamus encodes the temporal information from experienced stimuli and represents the time that has elapsed since the previous event . To address this question , we used the pigeon as an animal model and the nucleus dorsolateralis anterior thalami ( DLA ) as the target brain area . The pigeon is a common animal model for behavioral and neurobiological studies of time-dependent cognitive tasks , such as interval timing ( Kalenscher et al . , 2006 ) , sequence learning ( Helduser et al . , 2013; Lissek et al . , 2013 ) , and delayed matching-to-sample ( Browning et al . , 2011 ) . Furthermore , to achieve foraging and safe flight successfully in diverse environments , birds have developed a complex visual system that is superior to that of most vertebrates ( Shimizu and Watanabe , 2012; Wylie et al . , 2015 ) . The avian DLA receives direct retinal inputs , projects onto the pallial Wulst in both hemispheres ( Karten et al . , 1973; Bagnoli and Burkhalter , 1983; Miceli et al . , 1987 , 2008 ) , and receives feedback inputs from the pallial Wulst ( Karten et al . , 1973; Miceli et al . , 1987 ) . The avian retina-DLA-Wulst pathway is comparable to the mammalian retina-lateral geniculate nucleus ( LGN ) -striate visual pathway ( Shimizu and Karten , 1993 ) . The avian Wulst further projects onto the nidopallium caudolaterale ( NCL ) , which is comparable to the mammalian prefrontal cortex ( PFC ) ( Kröner and Güntürkün , 1999 ) ( Figure 1A ) . More importantly , the luminance cells in the pigeon DLA can encode ambient luminance ( Yang et al . , 2005 ) . During visual conditioning , the responses of the DLA luminance cells to the conditioned stimulus ( CS: whole-field light ) are modified by training , and the training-induced changes that occur in response to the CS are in parallel with the acquisition of the behavioral responses of pigeons ( Gibbs et al . , 1986 ) . Using electrophysiological single-unit recordings , we compared the neuronal responses of DLA luminance cells before , during , and after the repetitive presentation of light/dark ( LD ) stimuli with intervals ranging from seconds to minutes ( L/D: 1 s/1 s to 240 s/240 s , 5 to 25 cycles ) . All luminance cells had steady firing rates under constant photic conditions and synchronized their activities with the rhythmic luminance changes of LD stimuli . After LD stimulation , some luminance cells retained the entrained oscillatory activities even when the photic conditions were constant . The post-stimulus replay responses of these cells were dependent on the time interval and number of LD cycles of periodic stimuli applied during LD stimulation . Both Wulst pharmacological inactivation and electrolytic lesions did not affect the post-stimulus oscillatory responses of DLA cells entrained by the periodic LD stimuli , suggesting that the intrinsic circuits in the DLA play the primary role in representing the time interval of periodic events experienced previously . We recorded 190 DLA luminance cells from 19 animals ( Figure 1—figure supplement 1 ) . These cells encoded ambient luminance and sustained steady activities under constant photic conditions . Their firing rates either increased ( light-activated ) or decreased ( light-suppressed ) monotonically when the luminance level of the stimulus was raised in steps ( Figure 1B , C ) . During LD stimulation , all luminance cells synchronized their firing rates with the rhythmic luminance changes of LD stimuli whose temporal frequencies ranged from 0 . 5 Hz ( L/D: 1 s/1 s ) to 0 . 002 Hz ( L/D: 240 s/240 s ) . The oscillation frequencies of the entrained activities of these cells were closely correlated with the temporal frequencies of the luminance changes of LD stimuli ( linear regression , slope = 0 . 99 , R2 = 0 . 99 ) . After LD stimulation , the light-activated cells were tested under light , whereas the light-suppressed cells were tested in darkness . Neuronal activities of each luminance cell after LD stimulation were continuously recorded for 2–3 hr . We found that 54 cells ( 54/190 = 28% ) retained the entrained oscillatory activities under constant photic conditions after LD stimulation . Thus , these cells were referred to as replay cells , which included 25 light-activated and 29 light-suppressed cells . Before LD stimulation , a typical DLA replay cell exhibited steady spontaneous activity in constant darkness ( Figure 1D , top row ) . During LD stimulation , the firing rates of this typical DLA replay cell oscillated at the same frequency ( 0 . 0085 Hz ) as the luminance changes of the LD stimulus ( L/D: 60 s/60 s ) ( middle row ) . As shown in the response histogram , in the first 10 min after 10 LD cycles ( bottom row ) , the neuronal activity of this cell continued to oscillate at almost the same frequency ( 0 . 0073 Hz ) as the LD stimulus . The remaining DLA cells ( 136/190 = 72% ) displayed non-oscillatory response patterns after LD stimulation and were therefore referred to as non-replay cells . For example , a typical DLA non-replay cell ( Figure 1E ) had steady excitatory responses in constant light ( 200 lux ) . This cell synchronized its firing rates with the luminance changes of the periodic photic stimulus ( L/D: 30 s/30 s ) . After 25 LD cycles , this cell first had excitatory responses to the light onset and then returned to a steady firing pattern , rather than continuing the oscillatory activity under constant light . The neuronal responses of replay cells after LD stimulation were regulated by both the temporal frequency of luminance changes and the number of LD cycles during LD stimulation . To evaluate the effects of stimulus parameters on the neuronal activities of replay cells after LD stimulation , the temporal frequency of LD stimuli was varied from 0 . 5 Hz ( L/D: 1 s/1 s ) to 0 . 002 Hz ( L/D: 240 s/240 s ) and the number of LD cycles was varied within a range from 5 to 25 cycles . We found that , first , replay cells did not exhibit oscillatory activities after LD stimulation when the duration of the LD cycle was shorter than 10 s ( L/D: 5 s/5 s ) . The oscillation frequencies of replay cells ( n = 54 cells ) after LD stimulation were highly correlated with the temporal frequencies of LD stimuli ( slope = 0 . 62 , R2 = 0 . 99 ) ( Figure 2A ) . Furthermore , once a replay cell returned to its steady non-oscillatory activity under constant photic conditions , it could be re-entrained by visual stimuli of other frequencies ( 6 cells , Figure 2B ) . The duration of the first replay cycle after LD stimulation was also linearly correlated with the duration of the LD cycle ( slope = 0 . 58 , R2 = 0 . 95 , n = 54 cells ) ( Figure 2C ) . Second , the increase in the number of LD cycles led to an increase in the oscillation time of replay cells after LD stimulation . When the temporal frequency of the LD stimulus was set to 0 . 0165 Hz ( L/D: 30 s/30 s ) , the number of replay cycles increased when the number of LD cycles increased ( slope = 0 . 41 , R2 = 0 . 99 , n = 21 cells ) ( Figure 2D ) . After LD stimulation , the entrained oscillatory activities of replay cells gradually returned to non-oscillatory activities with time . The change was reflected in the decrease of oscillation frequency and response amplitude . Figure 3 shows the whole neuronal response process of a replay cell before , during , and after LD stimulation . This replay cell had steady non-oscillatory activity in darkness before LD stimulation and showed oscillatory activity ( 0 . 0085 Hz ) induced by LD stimuli ( L/D: 60 s/60 s ) . After 10 LD cycles , the entrained oscillatory activity of this cell continued for up to 16 replay cycles ( ~48 min ) in darkness . The oscillation frequency of the post-stimulus responses gradually slowed down over time ( Figure 3A , B ) . In the first 7 min after LD stimulation , this cell displayed 0 . 0073 Hz oscillation , close to the temporal frequency of the LD stimulus ( 0 . 0085 Hz ) . As time elapsed , the oscillation frequency of this cell decreased to 0 . 0061 Hz in the period 19–26 min and to 0 . 0049 Hz in the period 38–45 min after withdrawing the LD stimulus . In addition , the ratio of the duration of the first replay cycle to the duration of the LD cycle ( 120 s ) was 1 . 24 , which increased to 2 . 21 in the last replay cycle ( Figure 3C ) . The peak activity of each replay cycle was divided by the mean response ( 37 . 19 spikes/s ) averaged across all dark periods during LD stimulation . The normalized peak activity of this cell decreased from 1 . 09 in the first replay cycle to 0 . 43 in the last replay cycle ( Figure 3E ) . To further examine the firing pattern changes of replay cells over time , we chose the first three replay cycles after LD stimulation and the last three replay cycles before returning to non-oscillatory activities . We observed that the duration of the replay cycle slowly increased over time ( one-way ANOVA , F5 , 180 = 4 . 38 , p < 0 . 001 , n = 31 cells; Figure 3D ) . For the same group of replay cells , the peak activity of the replay cycle gradually decreased over time ( F5 , 180 = 3 . 59 , p = 0 . 004 ) ( Figure 3F ) . The avian DLA has reciprocal connections with the pallial Wulst . To clarify the possible effect of the inputs from Wulst on the DLA replay cells , the Wulst in both hemispheres was temporarily inactivated by multi-site muscimol injections ( Figure 4—figure supplement 1 ) . We recorded 12 replay cells from 10 animals , and then compared their post-stimulus activities before and after Wulst inactivation . The tested replay cells included five light-activated and seven light-suppressed cells . Wulst inactivation attenuated the spontaneous activities of replay cells , but had no effect on their photic responses . Under constant photic conditions before LD stimulation , the spontaneous firing rates of replay cells were slightly reduced from 3 . 89 ± 0 . 87 spikes/s ( mean ± SEM ) before injection to 2 . 3 ± 0 . 51 spikes/s after injection ( paired t-test , tpre vs . post = 2 . 37 , p = 0 . 03 , n = 12 cells ) . Furthermore , the photic responses of the DLA cells were evaluated by a preference index of their photic responses ( R ) to light ( L = 200 lux ) and dark stimuli ( D = 0 lux ) ( index = ( RL − RD ) / ( RL + RD ) ) . No significant change in the preference index was observed after Wulst inactivation for the light-activated cells ( pre-injection: 0 . 34 ± 0 . 15; post-injection: 0 . 5 ± 0 . 22; Wilcoxon rank sum test , ranksum = 21 , p = 0 . 22 , n = 5 cells ) or for light-suppressed cells ( pre-injection: −0 . 52 ± −0 . 19; post-injection: −0 . 43 ± −0 . 16; ranksum = 46 , p = 0 . 45 , n = 7 cells ) . The entrained post-stimulus oscillatory activities of replay cells were still sustained when the Wulst was temporarily inactivated , as shown for a single replay cell ( Figure 4—figure supplement 2 ) . Neither the linear correlation between the post-stimulus oscillation frequencies of replay cells and the temporal frequencies of LD stimuli ( pre-injection: slope = 0 . 62 , R2 = 0 . 92; post-injection: slope = 0 . 56 , R2 = 0 . 82; Figure 4A ) nor the linear correlation between the duration of the first replay cycle and the duration of the LD cycle were significantly affected by Wulst inactivation ( pre-injection: slope = 0 . 59 , R2 = 0 . 91; post-injection: slope = 0 . 73 , R2 = 0 . 96; Figure 4C ) . After Wulst inactivation , the replay cells could still be re-entrained by periodic stimuli of different frequencies ( n = 4 cells , Figure 4B ) . The number of replay cycles increased as the number of LD cycles ( 30 s/30 s ) increased ( pre-injection: slope = 0 . 62 , R2 = 0 . 68; post-injection: slope = 0 . 54 , R2 = 0 . 64 , n = 10 cells ) ( Figure 4D ) . In addition , for the same periodic stimulus , the post-stimulus activities of replay cells before and after injection showed no significant differences in oscillation frequencies ( paired t-test , tpre vs . post = 0 . 32 , p = 0 . 75 ; Figure 4A ) , duration of the first replay cycle ( tpre vs . post = –1 . 71 , p = 0 . 1; Figure 4C ) , or number of replay cycles ( tpre vs . post = –0 . 94 , p = 0 . 37; Figure 4D ) . Furthermore , the post-stimulus activities of replay cells declined over time before and after injection , which was reflected in the increasing duration of replay cycles ( one-way ANOVA , pre-injection: F5 , 66 = 4 . 05 , p = 0 . 002; post-injection: F5 , 66 = 4 . 55 , p < 0 . 0001 , n = 12 cells; Figure 4E ) and decreasing peak responses of replay cycles over time ( pre-injection: F5 , 66 = 3 . 17 , p = 0 . 01; post-injection: F5 , 66 = 2 . 94 , p = 0 . 01; Figure 4F ) . We compared the spontaneous activities of Wulst cells before and after muscimol injection to prove that muscimol could effectively inhibit these cells . To examine the spatial range and temporal course of the muscimol inhibitory effect in the Wulst , Wulst cell activities were determined by the mean firing rates of eight recording positions . Each recording site was 0 . 5 mm from the injection site ( Figure 4—figure supplement 3 ) . We measured the spontaneous activity of Wulst cells averaged for eight recording sites surrounding the injection site before and after injection . Before the injection , the spontaneous firing rates of the Wulst cells surrounding the injection sites were 6 . 55 ± 0 . 85 spikes/s ( n = 3 injection sites ) . Their firing rates between 0 . 25 hr and 1 . 5 hr after injection were only 15 . 9% ± 3 . 7% ( 1 . 08 ± 0 . 37 spikes/s ) of those before injection , and slowly increased to 26 . 3% ± 0 . 3% ( 1 . 72 ± 0 . 22 spikes/s ) between 1 . 5 hr and 3 hr after injection . Taken together , these data illustrate that muscimol effectively inhibited the Wulst cells , but that successful Wulst inactivation did not affect the entrained post-stimulus oscillatory activities of the DLA replay cells . The pharmacological inactivation of the pallial Wulst had no significant effect on the post-stimulus oscillatory activities of the DLA replay cells . To exclude the possible impact of incomplete Wulst inactivation , electrolytic lesions were applied in the Wulst of both hemispheres ( Figure 5—figure supplement 1 ) . We recorded 29 replay cells from nine Wulst-lesioned animals , which included 14 light-activated and 15 light-suppressed cells . By comparing the neuronal responses of the DLA replay cells in normal and Wulst-lesioned animals , we found that Wulst lesions did not affect the spontaneous activities or photic responses of the DLA replay cells . Under constant photic conditions before LD stimulation , no significant firing rate changes were observed between the replay cells recorded in the normal animals ( 3 . 39 ± 0 . 43 spikes/s ) and those in Wulst-lesioned animals ( 3 . 54 ± 0 . 36 spikes/s ) ( t-test , tnormal vs . lesion = –0 . 26 , p = 0 . 39 , nnormal = 54 cells , nlesion = 29 cells ) . Furthermore , the photic responses of the DLA replay cells were not affected by Wulst lesions , which was reflected in the comparable preference indices of light-activated cells ( normal animals: 0 . 46 ± 0 . 06 , n = 25 cells; lesion animals: 0 . 5 ± 0 . 05 , n = 14 cells , t-test , tnormal vs . lesion = –0 . 46 , p = 0 . 64 ) and those of light-suppressed cells ( normal animals: −0 . 52 ± −0 . 06 , n = 29 cells; lesion animals: −0 . 42 ± −0 . 03 , n = 15 cells , tnormal vs . lesion = -1 . 38 , p = 0 . 17 ) . The post-stimulus oscillatory activities of replay cells were still retained after the application of Wulst lesions in both hemispheres . After LD stimulation , the replay cells still showed oscillatory activities , and the oscillation frequencies were linearly correlated with the temporal frequencies of LD stimuli ( slope = 0 . 62 , R2 = 0 . 99 , n = 29 cells ) ( Figure 5A ) . Furthermore , replay cells ( six cells ) could be re-entrained by periodic stimuli of different frequencies ( Figure 5B ) . The duration of the first replay cycle of these cells was linearly correlated with that of the LD cycle ( slope = 0 . 58 , R2 = 0 . 96 , Figure 5C ) . The number of replay cycles increased when the number of LD cycles ( L/D: 30 s/30 s ) increased ( slope = 0 . 86 , R2 = 0 . 94 , n = 20 cells; Figure 5D ) . Furthermore , the post-stimulus oscillatory activities of the replay cells declined with time . The duration of the replay cycles increased ( one-way ANOVA , F5 , 168 = 4 . 03 , p = 0 . 002 , n = 29 cells; Figure 5E ) and the peak activities of the replay cycles decreased over time ( F5 , 168 = 9 . 96 , p < 0 . 0001; Figure 5F ) . Therefore , these data from the Wulst-lesioned animals further confirmed that the pallial Wulst did not participate in the modulation of the post-stimulus oscillatory response of the DLA replay cells . Previous study has shown that retinal ganglion cells in salamander and mouse retinae respond to an omitted stimulus in a sequence of flashes ( Schwartz et al . , 2007 ) . However , the retinal response to the omitted stimulus is not observed more than once and it is irrelevant to whether the recorded cells respond to the flashes throughout the flash sequence . Compared to the retinal cells , the pigeon thalamic cells in our experiment not only accurately signaled the time of each luminance change of repetitive stimuli during LD stimulation , but also showed reliable replay responses after the periodic stimulation . The omitted stimulus response in the retina occurs for repetitive stimuli with short intervals ( 20 ms–100 ms ) ( Schwartz et al . , 2007 ) , but thalamic cells can be entrained by temporal patterns with long intervals ( 10 s–8 min ) . Furthermore , the replay responses of thalamic cells can be maintained for more than one cycle . For example , the replay responses of the DLA cell in Figure 3 continued for up to 48 min and 16 replay cycles in constant darkness after 20 min of periodic LD stimulation ( L/D: 60 s/60 s , 10 cycles ) . In addition , the number of replay cycles increased with the increase in the number of LD cycles . The post-stimulus oscillatory responses of thalamic replay cells entrained by periodic stimuli with the interval in the seconds-to-minutes range are unlikely to originate from the retina . The omitted stimulus potential ( OSP ) is traditionally regarded as a sign of expectation of a stimulus at the due-time . In human EEG recordings , the OSP has been observed under low ( < 2 Hz ) and fast ( > 5 Hz ) stimulus rates ( Bullock et al . , 1994 ) . The fast OSP arises in the retina ( Bullock et al . , 1990; Schwartz et al . , 2007 ) , but it is not clear whether the slow OSP also arises in the retina . When a periodic stimulus with an interval in the order of seconds was presented to zebrafish larvae in vivo , the retinal ganglion cells did not show post-stimulus rhythmic activity ( Sumbre et al . , 2008 ) . In addition , during visual conditioning , the responses of retinal ganglion cells evoked by CS ( a few seconds of whole-field light ) did not change with the training of pigeons ( Wild and Cohen , 1985 ) . These studies imply that retinal ganglion cells might not retain the entrained activity after the presentation of repetitive stimuli with the interval in the seconds-to-minutes range . In addition to the thalamic cells reported in the present study , the post-stimulus replay response has also been observed in the optic tectum of zebrafish larvae ( Sumbre et al . , 2008 ) . There are two major visual pathways linking the eyes to the brain: one projects to the visual thalamus and the other to the optic tectum of vertebrates or the superior colliculus of mammals ( Grüsser et al . , 1975; Jessell et al . , 2000 ) . In visual conditioning , after the repetitive CS presentation of seconds in duration , neuronal ensembles in the zebrafish tectum show rhythmic activities with an interval matching the duration of the CS . Correspondingly , the visuomotor behavior of zebrafish larvae is highly correlated with the post-CS rhythmic neuronal activities in the tectum ( Sumbre et al . , 2008 ) . Therefore , the experience-dependent representation of time interval might not be confined to the pallial/cortical level , but may occur as early as the subcortical levels in the brain . Although repetitive visual stimuli induce similar replay responses in the tectum and thalamus , the temporal information encoded in these two nuclei may contribute to different time-dependent tasks . Through descending outputs to the hindbrain , the vertebrate optic tectum/mammalian superior colliculus can use detected temporal information to accurately control the fast and immediate movements of animals , such as the eye-head coordination of monkeys ( Klier et al . , 2003 ) , the prey capture and visuomotor behavior of zebrafish ( Gahtan et al . , 2005; Sumbre et al . , 2008 ) , and the looming-object detection and avoidance of pigeons , cats , and mice ( Wu et al . , 2005; Liu et al . , 2011; Shang et al . , 2015 ) . However , the avian DLA/mammalian LGN mainly project to the pallium/cortex , thus it is more likely that the visual thalamus participates in the perceptual and cognitive tasks performed by the pallium/cortex . In the traditional view , the thalamus is thought to passively transfer ongoing visual information from the retina to the cortex ( Derrington , 2001; Liu et al . , 2008; Naito et al . , 2013 ) . By contrast , the thalamic replay cells in the present study not only followed the changes of current stimuli , but also retained a copy of the periodic events exposed previously . Here , these cells acted like a time-adjustable alarm clock . Although the external periodic events were vanished , the replay cells continuously signaled the time of upcoming events that would occur . Given the recent findings that the visual thalamus participates in many dynamic processes in the visual pathway ( Cudeiro and Sillito , 2006; Guillery and Sherman , 2011; Saalmann and Kastner , 2011; Sherman , 2016 ) , the timing signal that we observed in the visual thalamus might be recruited by the pallium/cortex in the time-dependent task . Given that neither pharmacological inactivation nor electrolytic lesions of the Wulst in both hemispheres affected the post-stimulus oscillatory activities of the DLA cells entrained by external , slow frequency LD periodic stimulation , the time-interval representation of avian thalamic cells in the order of minutes is unlikely to be modulated by the pallium Wulst . Like the mammalian LGN , the avian DLA is involved in far more than the simple transmission of visual information from the retina to the visual pallium . In addition to the retinal and pallial Wulst projections , the DLA also receives afferent supplies from the suprachiasmatic nucleus ( SCN ) and the optic tectum ( Miceli et al . , 2008; Cantwell and Cassone , 2006 ) . However , we do not know which cognitive functions of visual thalamus are modulated by the inputs from SCN and tectum in the pigeon . Therefore , further neurophysiological evidence is required to reveal the possible modulating inputs to DLA cells that are needed for time-interval representation . The intrinsic electrical properties of thalamic cells might determine their oscillatory responses to periodic stimuli . In mammals , thalamocortical cells ( TC cells ) are excitatory and project to the cortex , whereas the local interneurons in the LGN are GABAergic and exert inhibitory influence on TC cells ( Uhlrich and Cucchiaro , 1992; Sherman , 2016 ) . Previous studies on thalamic slices suggest that TC cells and interneurons exhibit voltage-dependent intrinsic oscillation ( Zhu et al . , 1999; Llinás and Steriade , 2006 ) . In guinea pig thalamic slices , TC cells show voltage-sensitive ionic conductance and can generate two distinct functional states: repetitive spiking and bursting modes ( Llinás and Jahnsen , 1982 ) . By adjusting the membrane potential , the firings of TC cells can be switched from one state to the other . The interplay between low-threshold Ca2+ ( ICa ) and Na+-K+ current ( INa+K ) is crucial for the low-frequency oscillation ( < 4 Hz ) of TC cells ( McCormick and Pape , 1990; Soltesz et al . , 1991 ) . In comparison with TC cells , the interaction between ICa and the calcium-activated non-selective cation current ( ICAN ) is essential for the oscillatory burst firing of interneurons ( Bal and McCormick , 1993; Zhu et al . , 1999 ) . Our experiment provides further evidence that the intrinsic circuit in the DLA , rather than the feedback inputs from the pallial Wulst , plays the primary role in the post-stimulus replay responses of DLA cells . The avian DLA receives direct retinal inputs and has reciprocal connections with the pallial Wulst . The retinal inputs are unlikely to contribute to the entrained post-stimulus activity in the seconds-to-minutes range , as discussed above . Neither pharmacological inactivation nor electrolytic lesions of the Wulst in both hemispheres had any significant effect on the post-stimulus oscillatory activities of DLA replay cells . To understand the potential mechanism underlying the post-stimulus replay responses of the DLA cells , we introduced a simplified computational model ( see Materials and methods ) based on the electrophysiological properties of thalamic cells reported in previous studies ( Zhu et al . , 1999; Llinás and Steriade , 2006 ) and our current results . The simple two-cell system included two model neurons ( R-neuron and NR-neuron ) that respectively simulated a replay cell and a non-replay cell . The DLA in pigeons has an abundance of GABAA , GABAB , and benzodiazepine-binding sites ( Veenman et al . , 1994 ) . In the model , we proposed that the R-neuron received inhibitory synaptic inputs from the NR-neuron ( Isyn ) , but the NR-neuron did not receive synaptic inputs from the R-neuron ( Figure 6A ) . By adjusting the model parameters , the model neurons captured most of the response features of the thalamic cells observed in the present experiment ( Figure 6B ) . During LD stimulation , both the R-neuron and NR-neuron exhibited oscillatory activities synchronous with the periodic LD stimuli . After LD stimulation , the R-neuron continued the entrained oscillatory activity , whereas the NR-neuron returned to the non-oscillatory firing pattern . Further analyses of the post-stimulus activities of the R-neuron ( R-neuron [u , Isyn] ) showed that the model neuron had a response pattern consistent with that of the real cell illustrated in Figure 3 , as reflected in the increasing duration and decreasing mean activity of each replay cycle over time ( Figure 6C , D ) . Without inhibitory inputs ( Isyn ) , the model neuron ( R-neuron [u] ) still showed oscillatory activity after LD stimulation , but the duration and mean activity of each replay cycle did not change over time ( Figure 6C , D ) . The computational model is not fully conclusive and only provides a possible explanation for the neural mechanism underlying the declining oscillatory responses of thalamic cells after periodic stimulation . In addition to the intrinsic membrane currents applied in the present model ( Marder et al . , 1996 ) , there are several factors that might affect the persistent responses of thalamic cells in the absence of periodic inputs , such as N-methyl-D-aspartate ( NMDA ) currents ( Wang , 2001; Bottjer , 2005 ) and short-term synaptic plasticity in the neural network ( Mongillo et al . , 2008; Szatmáry and Izhikevich , 2010 ) . In addition , interneurons can exert inhibitory influences on neurons projecting to the Wulst in the avian DLA ( Miceli et al . , 2008 ) . It is plausible that the inhibitory inputs from the NR-neuron to the R-neuron in the model were mediated by inhibitory interneurons . In the present study , the computational model was simplified to include only two cells ( R-neuron and NR-neuron ) . In the real brain , however , the post-stimulus oscillatory responses of replay cells are very likely to be generated by neural networks composed of a population of neurons in the DLA rather than a single R-neuron ( Gutnisky and Dragoi , 2008; Wang et al . , 2011; Benucci et al . , 2013 ) . As slow wave activity ( < 1 Hz ) emerges in the thalamic cells of both anesthetized ( Steriade et al . , 1993 ) and awake animals ( Albrecht et al . , 1998; Filippov and Frolov , 2005 ) , one could assume that the anesthetic state induced the oscillatory activities of thalamic cells observed in the present experiment . However , our results suggest that the post-stimulus replay response of the thalamic cells was evoked by the external stimuli rather than by the anesthetic state for six reasons: ( 1 ) both replay and non-replay cells exhibited steady spontaneous activities under constant photic conditions before periodic stimulation; ( 2 ) both replay and non-replay cells showed synchronous activities with luminance changes during LD stimulation; and ( 3 ) under constant photic conditions after periodic stimulation , replay cells continued the entrained oscillatory activities in contrast to non-replay cells . The post-stimulus replay responses of replay cells were maintained for a long period of time , and then gradually returned to the non-oscillatory responses over time . ( 4 ) The replay responses of thalamic cells after LD stimulation were regulated by the temporal frequency and number of LD cycles applied during LD stimulation . ( 5 ) The time interval of LD stimuli shorter than 10 s ( L/D: 5 s/5 s ) could not induce the post-stimulus oscillatory responses of replay cells . ( 6 ) During recording , the depth of anesthesia was monitored and additional top-up doses of anesthetic were applied as required . Moreover , animals were also isolated from any other sensory stimuli , such as auditory , olfactory , and taste stimuli in the environment . Taken together , the periodic stimulus was the only factor that induced the photic and replay responses of the thalamic cells . In addition , previous studies have reported that the abilities of thalamic cells to discriminate visual features are less affected by the brain state ( alert/non-alert ) of animals . Although thalamic cells have higher firing rates when animals are awake than during anesthetic state , their sensitivities to stimuli with different spatial and temporal contrasts are comparable under the two conditions ( Cano et al . , 2006; Alitto et al . , 2011 ) . The present study sheds light on the neural mechanism of translating temporal information from external and variable photic events into internal timing in the brain . Our study provides the first evidence that the visual thalamus is involved in time perception and in memorizing the timing of periodic events that have occurred previously . The retina-DLA-Wulst pathway in birds participates in spatial and sun compass orientation as well as in light-dependent navigation ( Budzynski et al . , 2002; Heyers et al . , 2007; Watanabe et al . , 2011; Keary and Bischof , 2012; Bischof et al . , 2016 ) . The entrained replay responses of thalamic cells may contribute to the navigation behaviors of birds by signaling the time of expected events previously experienced in the environment . All experiments performed on the 38 adult homing pigeons ( Columba livia ) were in accordance with the guidelines for the care and use of animals established by the Society for Neuroscience and approved by the Institutional Animal Care and Usage Committee ( IACUC ) of the Institutes of Biophysics , Chinese Academy of Sciences ( SYDK2016-07 ) . Each pigeon was initially anesthetized by injecting ketamine hydrochloride ( 40 mg/kg ) and xylazine hydrochloride ( 5 mg/kg ) into the pectoral muscles , and was supplemented with ketamine hydrochloride ( 20 mg/kg ) and xylazine hydrochloride ( 2 mg/kg ) per hour . The animal was gently wrapped in a bag and placed on a foam-lined holder in a stereotaxic apparatus . The depth of anesthesia was monitored by breathing patterns and reflex from pinching the toe . Body temperature was maintained at 41°C by a warming pad . The wound edges and muscles were infiltrated periodically with lidocaine . The telencephalon overlying the DLA ( anterior [A]: 6 . 25 to 7 . 25 , lateral [L]: 2 . 80 to 3 . 80 , height [H]: 7 . 00 to 8 . 00 ) and the Wulst ( A 8 . 00 to 14 . 50 , 0 to L3 . 00 , 0 to H3 . 00 ) in both hemispheres was exposed with a dental drill and surgical forceps ( Karten and Hodos , 1967 ) . To investigate the possible contributions of the inputs from the pallial Wulst to the post-stimulus oscillatory activities of the DLA cells , the Wulst in both hemispheres was pharmacologically inactivated or electrolytically lesioned . ( 1 ) Pharmacological inactivation was achieved via the administration of a 1 μl Hamilton syringe filled with 2% muscimol ( Abcam , UK ) . Multi-unit recordings show that 1 µl of muscimol ( 2% ) can completely inactivate neuronal activities in a 2 mm diameter area around the cannula tip for ~3 hr , and can significantly attenuate the neuronal activities in a 4 mm diameter area ( Partsalis et al . , 1995; Arikan et al . , 2002 ) . To inactivate the Wulst in both hemispheres , we injected 1 µl of muscimol ( 2% ) at four different coordinates: A10 , L1 . 5 , H1 . 5 and A12 , L1 . 5 , H1 . 5 in the left hemisphere , and A10 , L1 . 5 , H1 . 5 and A12 , L1 . 5 , H1 . 5 in the right hemisphere . The locations of injection sites were confirmed by injecting 0 . 2 µl of direct-blue 15 ( 2% direct-blue 15 in 0 . 5 M sodium acetate solution , Sigma , USA ) . ( 2 ) Electrolytic lesions ( CH-HI Cautery , Advanced Meditech International , USA ) were applied to the Wulst in both hemispheres . During surgery , the physical condition of the anesthetized animals was strictly monitored , with indicators including breathing and heart rate . After surgery , the lesion area was covered with sterilized medical hemostatic sponge . The electrophysiological recordings started 1 hr after surgery . One eye of each animal was occluded , and the other eye was stimulated by a light-emitting diode ( LED ) . The center of the LED was in line with the optical axis of the viewing eye and 1 . 5 cm from the eyeball . The light from the LED was diffused over the whole visual field of the viewing eye . A rubber eye cap enclosing the LED was fitted closely to the eye orbit rim . The eye-cap and LED light constituted the probe of a custom-designed multifunctional visual photostimulator ( Institute of Biophysics , Chinese Academy of Sciences , Beijing , China ) . For the photic stimulus ( 423 nm–688 nm ) , the luminance level was adjusted in six steps ( 0 , 20 , 50 , 100 , 200 , and 400 lux ) . The duration of each LD cycle ( 2 , 4 , 10 , 20 , 30 , 60 , 120 , 240 , and 480 s ) and number of LD cycles ( 5–25 cycles ) were also adjusted . In the routine experiments , the dark was 0 lux . The luminance of the photic stimulus was measured by a digital light meter ( LX-1330B , Shenzhen TONDAJ Instrument Co . , China ) . Single-unit recordings were made in the pigeon DLA/Wulst using tungsten-in-glass microelectrodes made in the laboratory ( 2–3 MΩ ) . Luminance cells in the DLA were first isolated with a flashlight . Two rigorous criteria were used to identify luminance cells: ( 1 ) in constant light or darkness , the steady firing of the recorded cell was sustained over a long period of time ( usually 20–60 min ) before the repetitive LD stimulus was presented; and ( 2 ) the firing rates of the recorded cell either increased or decreased monotonically when the luminance level was increased from 0 to 400 lux in 3–6 steps . The firing rates for each luminance level were averaged for 5–10 repeats . The activities of each luminance cell were examined before , during , and after 5–25 LD cycles of photic LD stimulation . After LD stimulation , the responses of each cell were continuously recorded for 2–3 hr . Light-activated cells were tested under light , and light-suppressed cells were tested in darkness until the cells completely returned to spontaneous activity . When the photostimulator sent a switch signal to the LED , it also sent the signal simultaneously to the computer to mark the time of stimulus onset/offset . Neuronal spikes were amplified and fed into an oscilloscope ( 54622A , Agilent Technologies Inc . , USA ) for observation and a computer for data collection and off-line analyses . Recording sites were identified via electrolytic lesions that were generated by applying positive currents of 30–100 µA for 20–30 s . To verify the electrolytic lesion sites and direct-blue marks in the brain , the animals were euthanized by an overdose of urethane ( 4 g/kg ) via intraperitoneal injection after the experiment . The brains removed from the skulls were fixed in 4% paraformaldehyde for 6–12 hr , and soaked in 30% sucrose solution in a refrigerator ( 4°C ) overnight . Frontal sections were cut on a freezing microtome ( Leica CM1850 , Germany ) at 40 μm thickness and counterstained with cresyl violet ( Sigma , USA ) . They were dehydrated and covered for subsequent microscopic observations . Neuronal spikes and photostimulator switch signals were sampled at 8000 Hz with Cool Edit software ( Version 2 . 0 , Syntrillium Software Co . , USA ) . The data were quantitatively analyzed off-line by Spike2 software ( CED , Cambridge Electronic Design Ltd . , UK ) and custom-made MATLAB routines ( R2009a , MathWorks , USA ) . Single units were classified on the basis of full wave templates and clustered by principle component analysis and direct waveform feature measures . Only well-isolated units were included in this study . To determine whether there was a significant oscillation in each response histogram ( bin = 0 . 2 s ) , we constructed a control group composed of 1000 resampled histograms computed by bootstrap re-sampling of the original response histogram . The original response and reconstructed histograms in the control group were transformed into the frequency domain by Fast Fourier Transformation ( FFT ) using Hanning windowing . The oscillation frequency of the original histogram detected by FFT analysis was considered statistically significant only when its FFT amplitude was larger than the mean +4 SDs of the control . To separate each replay cycle of the recorded cell precisely , the response histogram was filtered by a zero-phase , low-pass Butterworth filter . The cutoff frequency of the filter was the maximal oscillation frequency detected by shifting an analysis window ( 7 min ) at 1 min steps from the end of the periodic LD stimulation to the end of the recording . The start and end time points of each replay cycle was the intersection of the filtered histogram with the mean value of the filtered histogram itself . To measure the correlation between the responses of each cell and the luminance level of the photic stimulus , the mean response of the recorded cell to each luminance level ( 5–10 repeats ) was normalized by the cell's maximal response to all tested luminance levels . To quantify the changes in oscillatory activities of the real or model cells after LD stimulation , the duration of each replay cycle of the real and model cells was normalized by the duration of the LD cycle . Correspondingly , the mean or peak response of the replay cycle of the light-activated/light-suppressed cells was normalized by the mean response averaged across all light/dark periods during LD stimulation , respectively . The model neuron ( Morris and Lecar , 1981; Sherman and Rinzel , 1992; Izhikevich , 2007 ) was described by the following equations:CV˙=I+I1+IK+INa+ICaI1=g1 ( V1−V ) Ik=gks ( Vk−V ) INa=gNam∞ ( V ) ( VNa−V ) ICa=gCau ( VCa−V ) s˙=λ ( V ) ( w∞ ( V ) −s ) w∞ ( V ) =12 ( 1+tanhV−V1V2 ) m∞ ( V ) =12 ( 1+tanhV−V3V4 ) λ ( V ) =13coshV−V12V2 where V ( mV ) and C ( μF ) are the membrane potential and capacitance of the model neuron , respectively; I , Il , Ik , INa , and ICa are the applied current and the currents of leak , K+ , Na+ , and Ca2+ respectively; gl , gk , gNa , and gCa are the leak , K+ , Na+ , and Ca2+ conductances through membrane channels , respectively; Vl , Vk , VNa , and VCa are the equilibrium potentials of the relevant leak , K+ , Na+ , and Ca2+ channels , respectively; V2 and V4 are the reciprocal of the slope of voltage dependence of w∞ ( V ) and m∞ ( V ) , and V1 and V3 are the potentials whose values are dependent on V2 and V4 , respectively . By adjusting the dynamic variable ‘u’ , the model neuron shows continuous spiking or bursting . A constant u value induces the repetitive spiking of the model neuron . When the changes in the u value satisfy the following equation:u˙=μR/NR ( w∞ ( V ) −u ) the model neuron shows bursting , where μR/NR determines the inter-burst interval of the R-neuron or NR-neuron , respectively . The u value slowly increases when the neuron is depolarized and slowly decreases when the neuron is hyperpolarized , which periodically switches the model neuron between the active and silent phases ( Sherman et al . , 1988 ) . Using the model neuron described above , we set up a simple two-cell system composed of two model neurons ( R-neuron and NR-neuron ) that simulated a replay cell and a non-replay cell , respectively ( Figure 6A ) . We proposed that the model R-neuron received inhibitory synaptic inputs from the model NR-neuron , with synaptic weight ( wR , NR ) and coupling strength ( σR ) , as described by the following equations:CV˙R=I+I1+Ik+INa+ICa+IsynIsyn=σRwR , NR ( Esyn−VR ) w˙R , NR=arctan[VR ( VNR−VRwR , NR ) ] The model NR-neuron did not receive synaptic inputs from the model R-neuron , and its neuronal activity can be described by the following equation:CV˙NR=I+I1+IK+INa+ICa where VR or VNR denotes the membrane potential of the model R-neuron or NR-neuron , respectively . The Esyn value depends on whether the synapse is excitatory or inhibitory . To simulate the real thalamic cells observed in the present experiment , the parameters in the equations were adjusted: C = 3 . 33 μF , V = -0 . 041 mV , s = 0 . 5048 , Ilight/dark = 0 . 06 μA/0 . 1 μA , gCa = 1 mS , VCa = −0 . 7 mV , V = −0 . 041 mV , V1 = 0 . 1 mV , V2 = 0 . 05 mV , V3 = −0 . 01 mV , V4 = 0 . 15 mV , gl = 0 . 5 mS , Vl= −0 . 5 mV; gk = 2 mS , Vk = −0 . 7 mV , gNa = 1 . 2 mS , VNa = 1 mV , Esyn = −0 . 7 mV , wR , NR = 0 . 05 . For the R-neuron , σR = 0 . 036 , u = 0 . 075 before LD stimulation , σR = 0 . 086 , μR = 0 . 01 during LD stimulation . The σR and μR values changed with time ( t ) : σR ( t ) = 0 . 0091–0 . 0005e0 . 00008t , μR ( t ) = 0 . 008+0 . 001log ( 0 . 005 t ) /log0 . 05 after LD stimulation . For the NR-neuron , u = 0 . 075 before and after LD stimulation , and μNR = 0 . 01 during LD stimulation .	Being able to track the passage of time enables animals to predict when events will occur in the future . This in turn helps them optimize their behavior . Hummingbirds , for example , schedule their foraging visits so that they return to each flower after it has had time to replenish its supply of nectar . In the laboratory , rats and monkeys can learn to delay their responses to a cue until a specific period of time has elapsed in order to earn a reward . But how does the brain keep track of time ? To discover what is happening in the world around us , we rely on our eyes and other sense organs to collect sensory information . These organs send this information to the thalamus , a structure deep below the surface of the brain . The thalamus processes and filters the sensory information , and then forwards it to the cortex located in the brain’s outer layer . Experiments have shown that , after training with events that occur at regular intervals , cells in the cortex and the eye can signal the time at which the next event would occur . But it was not known if cells in the thalamus could do this too . To answer this question , Yang , Wang et al . recorded from the thalamus of pigeons while exposing the birds to alternating periods of light and darkness . Pigeons were chosen because they have good eyesight and perform well on time-tracking tasks . A set of cells in the pigeon thalamus changed their activity levels to follow each light/dark switch . The cells tracked switches that occurred every few minutes as accurately as those that occurred every second . Next , Yang , Wang et al . stopped switching the light on and off , and instead left the light either on or off for 2-3 hr . Even in constant light or darkness during the 2-3 hr , some of the cells maintained their previous pattern of firing . In other words , the cells continued to signal the time when the light/dark switches should occur long after the switching had been stopped . Inactivating the pigeon’s equivalent area of the mammalian cortex in the brain had no effect on this response . The findings of Yang , Wang et al . suggest that the thalamus – like the cortex and the eye – can track events that occur at regular intervals , at least in pigeons . The next step is to determine whether the thalamus encodes time intervals in other species as well , and how this might help the animals to optimize time-dependent behaviors , such as foraging and navigation .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2017	Representation of time interval entrained by periodic stimuli in the visual thalamus of pigeons
Endoplasmic Reticulum ( ER ) -derived COPII coated vesicles constitutively transport secretory cargo to the Golgi . However , during starvation-induced stress , COPII vesicles have been implicated as a membrane source for autophagosomes , distinct organelles that engulf cellular components for degradation by macroautophagy ( hereafter called autophagy ) . How cells regulate core trafficking machinery to fulfill dramatically different cellular roles in response to environmental cues is unknown . Here we show that phosphorylation of conserved amino acids on the membrane-distal surface of the Saccharomyces cerevisiae COPII cargo adaptor , Sec24 , reprograms COPII vesicles for autophagy . We also show casein kinase 1 ( Hrr25 ) is a key kinase that phosphorylates this regulatory surface . During autophagy , Sec24 phosphorylation regulates autophagosome number and its interaction with the C-terminus of Atg9 , a component of the autophagy machinery required for autophagosome initiation . We propose that the acute need to produce autophagosomes during starvation drives the interaction of Sec24 with Atg9 to increase autophagosome abundance . Autophagy is a highly conserved catabolic process that uses membrane traffic to target proteins and organelles for degradation . Basal levels of autophagy continuously replenish the cellular pool of amino acids and other metabolites to maintain homeostasis . However , when cells are starved for nutrients , autophagy is quickly upregulated . This upregulation leads to a dramatic intracellular reorganization to meet the high demand for membrane required to form autophagosomes , distinct organelles that target cellular components for degradation ( Nakatogawa et al . , 2009 ) . Induction of autophagy leads to the formation of a double-membrane structure , called the isolation membrane , that forms adjacent to a pre-autophagosome structure ( PAS ) where autophagy related proteins ( Atg ) are recruited in a hierarchical manner ( Nakatogawa et al . , 2009 ) . As the isolation membrane expands , it engulfs cytoplasmic proteins and organelles targeted for degradation before it seals to form an autophagosome . The autophagosome then fuses with the vacuole/lysosome , releasing its contents for degradation ( Lamb et al . , 2013; Nakatogawa et al . , 2009 ) . Although the assembly pathway of the Atg proteins is known , the mechanism by which membranes are directed to the autophagy pathway remains a central unanswered question in the field . Autophagosome biogenesis has been linked to COPII vesicles and an ER subdomain called the ER exit sites ( ERES ) where COPII vesicles are formed ( Graef et al . , 2013; Suzuki et al . , 2013; Tan et al . , 2013; Ge et al . , 2014; Wang et al . , 2015; Lemus et al . , 2016 ) . How COPII vesicles , which are faithfully targeted to the Golgi , can be reprogrammed to function on an alternate trafficking pathway during nutrient deprivation remains enigmatic . COPII coated vesicle formation is initiated at the ER with the recruitment of an inner coat layer comprising the Sec23/Sec24 complex . Coat polymerization and vesicle budding occur when Sec23/Sec24 recruits a second complex ( Sec13/Sec31 ) that forms the outer shell of the coat . Sec24 , the major cargo adaptor of the COPII coat , recruits biosynthetic cargo and SNAREs ( which mediate vesicle fusion ) into vesicles that are delivered to the Golgi ( Lord et al . , 2013 ) . After vesicle fission , the coat lingers on the vesicle to facilitate vesicle targeting to the Golgi ( Cai et al . , 2007; Lord et al . , 2011 ) . COPII vesicle budding mutants , as well as other mutants that disrupt ER-Golgi traffic , also disrupt autophagy ( Hamasaki et al . , 2003 ) . However , a direct functional link between COPII vesicles , the COPII coat and autophagy has been difficult to demonstrate . Multiple COPII coat subunits , including Sec23 and Sec24 , are phosphorylated by Hrr25 ( Bhandari et al . , 2013; Lord et al . , 2011 ) , a kinase required for ER-Golgi traffic and autophagy ( Lord et al . , 2011; Murakami et al . , 1999; Wang et al . , 2015; Yu and Roth , 2002 ) . Previously , we showed coat phosphorylation is required for COPII vesicle fusion ( Lord et al . , 2011; Wang et al . , 2015 ) . Given the recently identified role of COPII vesicles in autophagosome formation , and the observation that Hrr25 is required for autophagy , we asked if coat phosphorylation also functions to regulate vesicle traffic during autophagy . Here we find that phosphorylation of the membrane distal surface of Sec24 promotes the interaction of Sec24 with the C-terminus of Atg9 , which is needed for autophagy . Failure to phosphorylate this Sec24 site leads to a decrease in autophagosome number , but not autophagosome expansion . This phosphorylation event is independent of the assembly of the Atg machinery at the PAS . Together these studies reveal a surprising role for coat phosphorylation in reprogramming core trafficking machinery to fulfill a separate function during starvation induced stress . To address whether coat phosphorylation allows COPII vesicles to function in autophagy versus ER-Golgi traffic , we purified the COPII inner coat from yeast cells induced for autophagy . Our analysis initially focused on Sec24 as it is the major COPII cargo adaptor . Using mass spectrometry , we identified 27 high confidence Sec24 phosphorylation sites in vivo ( Supplementary file 1 ) and subsequently tested if they specifically affect autophagy but not ER-Golgi traffic . Many of the identified Sec24 phosphosites were conserved in the closely related paralog Iss1 , which is also a cargo adaptor ( Kurihara et al . , 2000 ) ( Supplementary file 1 ) . Two of the conserved Sec24 phosphosites ( S730 and S735 ) map to a region of the protein that comprises one of the four well-characterized yeast cargo-binding sites ( Miller et al . , 2003 , 2005; Pagant et al . , 2015 ) . The so-called A-site ( Figure 1—figure supplement 1A ) packages a SNARE , Sed5 , needed for ER-Golgi traffic and autophagy ( Miller et al . , 2005; Mossessova et al . , 2003; Tan et al . , 2013 ) , marking these residues as candidate regulatory sites . COPII vesicles formed in vitro with Sec24-S730A/S735A showed normal capture of Sed5 and other cargo , whereas those formed with Sec24-S730D/S735D contained reduced amounts of Sed5 and were modestly impaired in their fusion efficiency with the Golgi ( Figure 1—figure supplement 1B , C ) . Additionally , a strain harboring the Sec24-S730D/S735D mutations had reduced autophagic activity , whereas the Sec24-S730A/S735A mutation had no effect ( Figure 1—figure supplement 1D , E ) . Because the Sec24-S730D/S735D mutations had effects on both ER-Golgi traffic and autophagy , these sites are unlikely to specify the pathway the vesicle takes and instead highlight the importance of Sed5 packaging into COPII vesicles in both pathways . A second set of conserved phosphosites , S645/S678 , located near a known cargo binding site in mammalian Sec24 ( mSec24 ) ( Figure 1A ) , did not disrupt the packaging of any cargo tested ( Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 21167 . 003Figure 1 . Identification of Sec24 phosphorylation sites required for autophagy . ( A ) Ribbon diagram of Sec23 ( lime ) and Sec24 ( lavender ) with groups of key Sec24 phosphorylation sites ( green , pink , and teal ) . The mSec24 membrin-binding site and conserved cargo binding B-site are colored light brown . ( B ) Sec24 phosphorylated residues conserved in Iss1 were mutated to alanine and introduced into sec24∆ and sec24∆iss1∆ deletion strains as described in the Materials and Methods . Sec24 and Iss1 were aligned using MAFFT alignment program and residues were considered conserved if either serine or threonine . Mutants in the sec24∆iss1∆ background were screened for autophagy defects 2 to 4 hr after nitrogen starvation using vacuolar alkaline phosphatase activity as a marker . Assays were performed in triplicate biological replicates and mutants were considered defective if p<0 . 05 using Student’s paired t-test . ( C ) Plasmids encoding SEC24 ( WT ) or mutant sec24 were expressed in SFNY2201 ( left ) or SFNY2202 ( right ) and grown on 5-FOA at 25°C to select against the WT balancing plasmid as described in the Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 00310 . 7554/eLife . 21167 . 004Figure 1—figure supplement 1 . Phosphorylation of Sec24 S730/S735 disrupts Sed5 packaging . ( A ) Two different views of a ribbon diagram of Sec24 ( lavender ) bound to a Sed5 peptide ( green ) with S730 and S735 highlighted ( pink ) . ( B ) Total membranes ( T ) and budded vesicles were collected from in vitro budding reactions performed with wild-type ( WT ) or mutant Sec24 proteins as indicated . Incorporation of Sed5 and other cargo was determined by immunoblotting ( left ) and quantitated using fluorescent secondary antibodies ( right ) . For each sample , cargo abundance in the vesicle pool was normalized to the reaction containing WT Sec24 . Note , vesicles do not form in the presence of GDP . Averages and s . d . are shown for five biological replicates; p=0 . 012 for Sed5 ( S730A/S735A and S730D/S735D ) , Student’s unpaired t-test . ( C ) Vesicles were generated in vitro with WT or mutant Sec24 harboring mutations in S730 and S735 and tested for their ability to fuse with acceptor membranes . Averages and s . e . m for six biological replicates are shown . p-values=0 . 011 ( WT and S730D/S735D ) , 0 . 006 ( S730A/S735A and S730D/S735D ) , Student’s unpaired t-test . ( D ) Vacuolar phosphatase activity was assayed in lysates prepared from WT and mutant Sec24 . The activity of WT 2 hr after starvation was set as 100% and time 0 was subtracted , Averages and s . e . m . are shown for three biological replicates; p-value = 0 . 003 , Student’s unpaired t-test . ( E ) The translocation of GFP-Atg8 to the vacuole was examined in WT Sec24 , Sec24-730D/735D and Sec24-S730A/S735A cells 30 min after nitrogen starvation at 37°C . Upon autophagy induction , GFP-Atg8 is incorporated into the autophagosomal membrane and cleaved after it is delivered to the vacuole . Scale bar 2 µm ( left ) . 300 cells were quantitated in three biological replicates ( right ) . WT was set to 100% for each experiment and had an average of 75% vacuolar localization . Averages and s . e . m . are shown; p-values=0 . 032 ( WT and S730D/S735D ) and 0 . 006 ( S730D/735D and S730A/S735A ) , Student’s paired t-test . p<0 . 05 , p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 00410 . 7554/eLife . 21167 . 005Figure 1—figure supplement 2 . Sec24 S645/S678 does not affect cargo packaging . Vesicles were collected from in vitro budding reactions performed with WT ( lane 1 ) , no Sec24 ( lane 2 ) or mutant Sec24 proteins ( lanes 3–4 ) . The incorporation of a panel of cargo was determined by immunoblotting . Cropped ( left ) and uncropped ( right ) western blots are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 005 Having ruled out known cargo-packaging sites on Sec24 as a functional switch for COPII vesicles from ER-Golgi traffic to autophagy , we next screened a panel of alanine mutations in the remaining phosphorylation sites conserved in Iss1 . The effects of mutations in these sites on autophagy were monitored in a sec24Δiss1Δ double mutant background by assaying Pho8Δ60 activity . Pho8Δ60 is a cytosolic form of vacuolar alkaline phosphatase that is delivered to and activated in the vacuole when autophagy is induced ( Klionsky , 2007 ) . Only combinations of T324A , T325A and T328A were defective in autophagy ( Figure 1B , red box ) . These three sites form a patch located on the membrane distal surface of Sec24 ( Figure 1A ) , making them unlikely to regulate cargo packaging . None of the additional 13 phosphosites tested showed autophagy defects ( Figure 1B ) . When we dissected the role of individual residues on this membrane distal surface of Sec24 , single and double mutant combinations of T324A/T325A/T328A showed a range of defects in Pho8Δ60 activity , with the triple alanine mutant showing the most dramatic defect ( Figure 2A ) . We were unable to test the full range of phosphomimetic combinations , as some caused lethality ( Figure 1C and Supplementary file 1 ) . However , none of the viable phosphomimetic mutants showed a defect in autophagy ( Figure 2B , C ) . Sec24-T324A/T325A/T328A ( hereafter referred to as Sec24-3A ) was further characterized because it had the most dramatic defect in Pho8Δ60 activity . To confirm that phosphorylation of the Sec24 membrane distal surface is required for autophagy , we examined a second autophagy marker , Atg8 . Cytosolic GFP-Atg8 is lipidated and incorporated into membrane at the PAS before being delivered to the vacuole ( Klionsky et al . , 2007 ) . Translocation of GFP-Atg8 to the vacuole , following nitrogen starvation , was significantly reduced in a strain expressing Sec24-3A ( Figure 2D ) . This defect was confirmed using western blot analysis by monitoring the cleavage of GFP-Atg8 to GFP ( Figure 2E ) . The GFP-Atg8 translocation defect was less severe in the presence of Iss1 ( Figure 2D , right ) , suggesting functional complementation by the paralogous protein . 10 . 7554/eLife . 21167 . 006Figure 2 . Phosphorylation of T324/T325/T328 in Sec24 is required for autophagy , but not ER-Golgi transport . ( A ) Vacuolar alkaline phosphatase activity was assayed in lysates prepared from a sec24∆iss1∆ deletion strain harboring sec24 alanine mutations . The activity of wild-type ( WT ) 2 hr after starvation was set as 100% and 0 time-point values were subtracted . Averages and s . e . m . are shown for 3 ( or four for T324A/T325A ) biological replicates . p-values=0 . 006 ( T324A ) , 0 . 012 ( T325A ) , 0 . 009 ( T328 ) , 0 . 008 ( T324A/T328A ) , 0 . 02 ( T324A/T325A ) , 0 . 01 ( T325A/T328A ) , 0 . 006 ( T324A/T325A/T328A ) , Student’s paired t-test . ( B , C ) As in ( A ) except activity was assayed in extracts from phosphomimetic mutations in sec24∆iss1∆ ( B ) or sec24∆ ( C ) deletion strains . Averages and s . e . m . are shown for three biological replicates . p-values=0 . 88 ( T325E ) , 0 . 78 ( T328E ) , 0 . 32 ( T324E/T328E ) , 0 . 26 ( T324E/T325E ) , Student’s paired t-test . ( D ) The translocation of GFP-Atg8 to the vacuole was examined 1 hr after nitrogen starvation at 25°C in sec24∆iss1∆ and sec24∆ deletion strains in either the presence of WT Sec24 or Sec24-3A . Representative images ( left ) and quantification from 300 cells ( right ) are shown . Scale bar 2 µm . WT was set to 100% for each experiment and had an average vacuolar localization of 76% ( sec24∆iss1∆ ) and 86% ( sec24∆ ) . Averages and s . e . m . are shown for three biological replicates . p-values=0 . 006 ( sec24∆iss1∆ ) , 0 . 02 ( sec24∆ ) , Student’s paired t-test . ( E ) Cleavage of GFP-Atg8 was examined in sec24∆iss1∆ cells expressing Sec24 or Sec24-3A 1 hr after starvation at 25°C ( left ) . The ratio of free GFP to GFP-Atg8 was quantitated . The cleavage in WT was set to 1 ( right ) . Averages and s . e . m . are shown for three biological replicates . p-value = 0 . 015 , Student’s unpaired t-test . ( F ) sec24∆iss1∆ cells expressing Sec24 ( lanes 1–4 ) or Sec24-3A ( lanes 5–8 ) were pulse-labeled for 4 min and chased for the indicated times ( left ) . The p1 ( ER ) , p2 ( Golgi ) and m ( vacuolar ) forms of CPY are labeled . Quantitation of the ratio of p2/p1 CPY for the 5 and 10 min time points are shown ( right ) . Averages and s . e . m . are shown for three biological replicates . p-values=0 . 08 ( 5 min ) , 0 . 66 ( 10 min ) , Student’s unpaired t-test . p<0 . 05; *p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 00610 . 7554/eLife . 21167 . 007Figure 2—figure supplement 1 . ER-Golgi transport is delayed during autophagy . ( A ) WT cells were grown in nutrient rich media ( SMD ) ( lanes 1–4 ) or starved for nitrogen ( SD-N ) for 1 hr at 25°C ( lanes 5–8 ) . Cells were pulse-labeled for 4 min and chased at the indicated times as described in the Materials and Methods before CPY was immunoprecipitated . ER ( p1 ) , Golgi ( p2 ) and vacuolar ( m ) forms of CPY are labeled ( left ) . Ratio of p2 to p1 CPY was determined ( right ) . Averages and s . e . m . are shown for three biological replicates , p-values=0 . 0005 ( 5 min ) and 0 . 009 ( 10 min ) , Student’s unpaired t-test . ( B ) As in ( A ) except sec24∆iss1∆ cells expressing Sec24 or Sec24-3A were starved for nitrogen for 1 hr at 25°C ( left ) . Ratio of p2 to p1 CPY was determined ( right ) . Averages and s . e . m . are shown for three biological replicates . p-values=0 . 46 ( 5 min ) and 0 . 96 ( 10 min ) , Student’s unpaired t-test . ( C ) As in ( A ) except Sec24-3A cells were grown in nutrient rich media ( SMD ) ( lanes 1–4 ) or starved for nitrogen ( SD-N ) for 1 hr at 25°C ( lanes 5–8 ) . Averages and s . e . m . are shown for three biological replicates , p-values=0 . 0015 ( 5 min ) and 0 . 01 ( 10 min ) , Student’s unpaired t-test . p<0 . 05; p<0 . 01; p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 00710 . 7554/eLife . 21167 . 008Figure 2—figure supplement 2 . Phosphorylation of the Sec23 membrane distal sites is not required for autophagy . ( A ) Ribbon diagram of Sec23 ( lime ) and Sec24 ( lavender ) with membrane distal phosphorylation sites highlighted ( red ) . ( B ) Plasmids encoding SEC23 ( WT ) or mutant sec23 were expressed in SFNY1948 and grown on 5-FOA at 25°C for 2–3 days to select against the WT plasmid . ( C ) Vacuolar phosphatase activity was measured in protein extracts of cells expressing WT Sec23 and mutant Sec23-T146A/T147A/S149A or Sec23-T146E/S147D/S149D . The activity of WT 2 hr after starvation was set as 100% and time 0 was subtracted . Averages and s . e . m are shown for three biological replicates , p-values=0 . 48 ( T146A/T147A/S149A ) , 0 . 6 ( T146E/S147D/S149D ) . ( D ) Translocation of GFP-Atg8 to the vacuole was examined in WT and the sec23 mutants 1 hr after nitrogen starvation at 25°C ( left ) . Scale bar 2 µm . 300 cells were quantitated ( right ) . WT was set to 100% for each experiment and had an average of 75% vacuolar localization . Averages and s . e . m are shown for three biological replicates , p-values=0 . 34 ( T146A/T147A/S149A ) , 0 . 74 ( T146E/S147D/S149D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 008 We next demonstrated that phosphorylation of the membrane distal sites on Sec24 is specifically required for autophagy by examining ER-Golgi traffic in cells containing Sec24-3A . The processing of carboxypeptidase Y ( CPY ) , as it traffics from the ER ( p1 ) , Golgi ( p2 ) , and vacuole ( m ) is kinetically indistinguishable from wild-type ( Figure 2F ) , demonstrating that phosphorylation of the Sec24 membrane distal patch regulates a novel function of Sec24 that is specific to autophagy . Supporting the model that a subpopulation of COPII vesicles is diverted from the secretory pathway during autophagy , ER to Golgi traffic was delayed , but not blocked , during starvation ( Figure 2—figure supplement 1A ) . Expression of Sec24-3A did not rescue this delay even in an iss1∆ strain background ( Figure 2—figure supplement 1B , C ) suggesting that additional sites and/or factors also participate in reprogramming COPII vesicles during starvation . Since Sec24 robustly co-purifies with Sec23 , our mass spectrometry data also included information on Sec23 , which contained three phosphosites on an alpha helix structurally equivalent to the Sec24 membrane distal patch ( Figure 2—figure supplement 2A ) . Mutation of these Sec23 sites ( T146/S147/S149 ) caused no defects in growth ( Figure 2—figure supplement 2B ) , Pho8∆60 activity ( Figure 2—figure supplement 2C ) or translocation of GFP-Atg8 to the vacuole ( Figure 2—figure supplement 2D ) , suggesting the effects we observed on autophagy are specific to Sec24 and its paralog , Iss1 . We next wanted to address whether phosphorylation of the Sec24 membrane distal sites is specifically required for autophagosome formation during starvation-induced upregulation of autophagy . To begin to address this question , structured illumination microscopy ( SIM ) was used to examine GFP-Atg8 puncta formation in the absence ( nutrient rich ) or presence of rapamycin . GFP-Atg8 puncta mark autophagosomes , as well as the PAS and isolation membrane ( Kirisako et al . , 1999 ) . In nutrient-rich media , autophagosome-like structures , called Cvt vesicles , form . These vesicles traffic a precursor form of aminopeptidase I ( prApe1 ) to the vacuole in an Atg11-dependent manner ( He et al . , 2006 ) , where it is proteolytically activated ( mApe1 ) . GFP-Atg8 puncta were smaller ( Figure 3A ) , less numerous , and dependent on Atg11 ( Figure 3B ) in nutrient rich medium . Consistent with the possibility that the GFP-Atg8 puncta may represent Cvt vesicles , Atg11 was only required for their formation in rich medium and not in rapamycin-treated cells ( compare Figure 3B , C ) . 10 . 7554/eLife . 21167 . 009Figure 3 . Phosphorylation of Sec24 regulates autophagosome frequency during starvation . ( A ) Representative images from WT cells expressing GFP-Atg8 treated with 400 ng/ml rapamycin for 1 hr at 25°C ( left top ) or untreated ( left bottom ) . Deconvolved images are shown . Scale bar , 1 µm . 100 GFP-Atg8 puncta were measured in WT cells treated with or without rapamycin ( right ) . Averages and s . e . m . are shown for three biological replicates; p=0 . 0002 , Student’s unpaired t-test . ( B ) WT Sec24 and Sec24-3A expressed in the sec24Δiss1Δ deletion strain , and WT ( +Atg11 ) and atg11Δ cells ( -Atg11 ) expressing GFP-Atg8 were imaged and the number of puncta per cell was quantitated from 300 cells . Averages and s . e . m . are shown for three biological replicates; p-values=0 . 81 ( Sec24-3A ) , 0 . 009 ( -Atg11 ) , Student’s unpaired t-test . ( C ) As in ( B ) except cells were treated with 400 ng/ml rapamycin for 1 hr at 25°C . Averages and s . e . m . are shown for three biological replicates; p-values=0 . 02 ( Sec24-3A ) , 0 . 68 ( -Atg11 ) , Student’s unpaired t-test . ( D ) As in ( C ) only the diameter of 100 GFP-Atg8 puncta was measured from cells expressing WT Sec24 and Sec24-3A in sec24Δiss1Δ deletion strains treated with 400 ng/ml rapamycin for 1 hr at 25°C . Averages and s . e . m . are shown for four biological replicates; p-value = 0 . 057 , Student’s unpaired t-test . ( E ) Representative images of autophagic bodies in cells expressing Sec24 ( left ) and Sec24-3A ( right ) in sec24Δiss1Δpep4Δ deletion strains after 1 . 5 hr of nitrogen starvation at 30°C . Scale bar represents 500 nm . ( F ) Histogram showing the distribution of the number of autophagic bodies per vacuole section in Sec24 and Sec24-3A . The number of autophagic bodies was quantitated for 78 vacuole sections for each strain ( left ) . p-value = 0 . 00012; Mann-Whitney Test . Box plot of the number of autophagic bodies per vacuole section . Bars show data between the lower and upper quartiles , the median is a horizontal line within the box . Whiskers indicate the smallest and largest observations ( right ) . ( G ) The diameter of autophagic bodies was determined . For Sec24 N = 398 , for Sec24-3A N = 342 . Averages with error bars as s . e . m . are shown . p<0 . 05; p<0 . 01; p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 00910 . 7554/eLife . 21167 . 010Figure 3—figure supplement 1 . GFP-Atg8 puncta formation is affected in the sec12-4 mutant during autophagy induction . ( A ) Cells with ( left ) and without ( right ) Iss1 were lysed as described in the Materials and methods and Ape1 processing was assayed by western blot analysis . Ape1 is processed in the sec24 alanine mutants to the same extent as wild-type . ( B ) WT and sec12-4 mutant cells expressing GFP-Atg8 were treated with 400 ng/ml rapamycin ( left ) or untreated ( right ) for 1 hr at 37°C . The number of GFP-Atg8 puncta per cell was determined for 300 cells . Averages and s . e . m . are shown for three biological replicates; p-value = 0 . 002 ( rapamycin ) 0 . 51 ( nutrient rich ) , Student’s unpaired t-test . p<0 . 01 . ( C ) The size of GFP-Atg8 puncta from WT and sec12-4 cells treated with rapamycin was determined for 100 puncta . Averages and s . e . m . are shown for four biological replicates; p-value = 0 . 13 , Student’s unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 010 Cells expressing Sec24-3A were not defective in GFP-Atg8 puncta formation in nutrient rich medium ( Figure 3B ) , however , fewer puncta formed in the rapamycin-treated Sec24-3A cells ( Figure 3C ) . While Sec24-3A significantly reduced the number of puncta formed , it did not significantly affect their size ( Figure 3D ) . Consistent with Sec24-3A not affecting Cvt vesicle formation , processing of Ape1 was unaffected in sec24 alanine mutants ( Figure 3—figure supplement 1A ) . Similar results were obtained when we blocked COPII vesicle budding at 37°C in the temperature-sensitive mutant sec12-4 mutant ( Figure 3—figure supplement 1B , C ) . The observation that COPII vesicles are not needed on the Cvt pathway is consistent with our earlier studies ( Wang et al . , 2015 ) and those of Ishihara et al . ( 2001 ) We used a second assay to address the effect of Sec24-3A on the frequency of autophagosome formation during starvation . Autophagosome number and size can be assessed using transmission electron microscopy by analyzing autophagic bodies , fully formed autophagosomes that have fused with the vacuole . Upon deletion of the PEP4 gene , which encodes a protease that is required for the activation of multiple vacuolar hydrolases , autophagic bodies accumulate in the vacuole ( Backues et al . , 2014 ) . After starvation , fewer autophagic bodies accumulated in cells expressing Sec24-3A compared to WT Sec24 ( Figure 3E , F ) . Although autophagic body number was significantly reduced , autophagic body size was not affected ( Figure 3E , F , G ) . Together these findings indicate that phosphorylation of the Sec24 membrane distal sites regulates autophagosome number , while autophagosome size or expansion is largely unaffected . The role of Sec24 in autophagy is likely to be conserved in mammalian cells as T324 and T328 are conserved in mSec24A ( Figure 4A ) . 10 . 7554/eLife . 21167 . 011Figure 4 . The Sec23/Sec24 complex binds the C-terminus of Atg9 . ( A ) Structure of yeast Sec24 and mSec24a with conserved residues in membrane distal sites ( red ) . ( B ) ypt7∆ cells expressing Atg9-13myc were grown in nutrient rich ( SMD ) or starvation ( SD-N ) media for 4 hr and Sec24 was immunoprecipitated and blotted for Atg9-13myc ( left ) . Precipitated Atg9-13myc was quantitated and normalized to the amount of Sec24 in the precipitate . SMD was set as one for each experiment . ypt7∆ cells were used as autophagosomes fail to fuse with the vacuole in the absence of Ypt7 and accumulate in the depleted cells ( Kirisako et al . , 1999 ) . Averages and s . e . m . are shown for four biological replicates , p-value = 0 . 037 , Student’s unpaired t-test . ( C ) Schematic showing cytosolic domains of Atg9 . ( D ) Equimolar amounts ( 200 nM ) of purified GST , GST-Sec31 ( aa878-1114 ) or GST-Atg9 fragments were incubated with 50 or 100 nM of Sec23/Sec24-His6 . ( E ) Equimolar amounts ( 100 nM ) of purified GST or GST-Atg9C was incubated with increasing amounts of Sec23/Sec24-His6 . ( F ) Same as ( D ) except His6-Sec23 was used . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01110 . 7554/eLife . 21167 . 012Figure 4—figure supplement 1 . Ponceau staining of in vitro bindings in Figure 4D , E and F . ( A ) Ponceau staining of Figure 4D . Equimolar amounts ( 200 nM ) of purified GST , GST-Sec31 ( aa878-1114 ) or GST-Atg9 fragments were incubated with 50 or 100 nM of Sec23/Sec24-His6 . Asterisks denote GST fusion protein and arrowheads denote bound Sec24-His6 . ( B ) Ponceau staining of Figure 4E . Equimolar amounts ( 100 nM ) of purified GST or GST-Atg9C was incubated with increasing amounts of Sec23/Sec24-His6 . Asterisks denote GST fusion protein . ( C ) Ponceau staining of Figure 4F . Equimolar amounts ( 200 nM ) of purified GST , GST-Sec24 or GST-Atg9 fragments were incubated with 50 or 100 nM of His6-Sec23 . Asterisks denote GST fusion protein and arrowhead denotes bound His6-Sec23 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 012 We previously proposed that COPII vesicles fuse with Atg9 vesicles at the PAS ( Tan et al . , 2013 ) . When autophagy is induced , Atg9 vesicles traffic from the late Golgi to the PAS and localize adjacent to the ER exit sites that produce COPII vesicles ( Graef et al . , 2013; Suzuki et al . , 2013; Yamamoto et al . , 2012 ) . Interestingly , similar to our observations with Sec24 , Atg9 was recently shown to regulate autophagosome number , but not size ( Jin et al . , 2014 ) . Proteomics also revealed an interaction between multiple COPII coat subunits and Atg9 in detergent lysates ( Graef et al . , 2013 ) . However , it remained unclear from these studies which coat subunit mediated this interaction and whether it was required for autophagy . Therefore , we first confirmed that Sec24 co-precipitates with Atg9 and tested whether this interaction is regulated by starvation . Sec24 was immunoprecipitated from cells expressing Atg9-13myc that were grown in nutrient rich conditions ( SMD ) or starved for nitrogen ( SD-N ) , and the precipitate was then blotted with anti-myc antibody . Atg9-13myc co-immunoprecipitated with Sec24 , but not the pre-immune control ( Figure 4B ) . Moreover , approximately 2 . 5-fold more Atg9-13myc co-immunoprecipitated with Sec24 from lysates prepared from nitrogen starved cells ( Figure 4B ) , demonstrating that this interaction is upregulated during starvation . Atg9 is a six transmembrane protein with N-terminal , C-terminal and core or middle ( M ) cytoplasmic domains ( Figure 4C ) . While the middle and C-terminal domains are present in mammalian cells , the N-terminus is largely absent ( Young et al . , 2006 ) . In order to determine which domain of Atg9 interacts with Sec24 , the cytoplasmic domains of Atg9 were fused to GST and the purified fusion proteins were incubated in vitro with purified Sec23/Sec24 complex . The Sec23/Sec24 complex was used for these studies , since Sec24-His6 is unstable in the absence of Sec23 . Sec24 was predominantly unphosphorylated , as it was purified without phosphatase inhibitors and stored in the freezer following purification . A GST fusion to a Sec31 fragment ( aa878-1114 ) , which was previously shown to interact with the Sec23/Sec24 complex ( Bi et al . , 2007 ) , served as a positive control . The C-terminus of Atg9 bound to Sec23/Sec24 , while the N-terminal and middle hydrophilic Atg9 domains did not ( Figure 4D , Figure 4—figure supplement 1A ) . Furthermore , binding of Sec23/Sec24 to the C-terminus increased with increasing concentrations of the Sec23/Sec24 complex and appeared to be saturable ( Figure 4E , Figure 4—figure supplement 1B ) . This interaction was also dependent on Sec24 , as Sec23 alone did not interact with GST-Atg9C ( Figure 4F , Figure 4—figure supplement 1C ) . As the crystal structure of Sec23 and Sec24 are identical whether they are in a complex or not , allosteric effects are unlikely ( Bi et al . , 2002 ) . Next we used the phosphomimetic mutations to ask if the interaction between Sec24 and Atg9 is enhanced by phosphorylation of the membrane distal sites . While the most dramatic effects on autophagy were seen with the sec24 triple alanine mutant ( Figure 2 ) , the sec24 triple phosphomimetic mutant is inviable . Therefore , to cover all three phosphosites of interest in our binding studies , we used Sec23/Sec24-T325E and Sec23/Sec24-T324E/T328E . Consistent with the notion that phosphorylation enhances the Sec24-Atg9 interaction , both T325E and T324E/T328E increased the interaction of Sec24 with GST-Atg9C ( Figure 5A , Figure 5—figure supplement 1A ) . To determine whether the Sec24-Atg9 interaction is regulated by phosphorylation in vivo , wild-type or mutant sec24 cells co-expressing Atg9-13myc were starved for nitrogen and Sec24 was immunoprecipitated in the presence of phosphatase inhibitors . Sec24-T324A/T325A ( Figure 5B ) , but not Sec24-T324E/T325E ( Figure 5—figure supplement 1B ) disrupted the interaction of Sec24 with Atg9-13myc . This defect was more pronounced in Sec24-3A , where Sec24-T328 is also mutated ( Figure 5C ) . Sec24-3A did not impair the trafficking of Atg9 to the PAS ( Figure 5—figure supplement 2A , C ) , indicating that Sec24 does not regulate Atg9 traffic . Together these findings imply that the C-terminus of Atg9 interacts with phosphorylated Sec24 via its membrane distal surface after Atg9 has been recruited to the PAS . 10 . 7554/eLife . 21167 . 013Figure 5 . Phosphorylation of the Sec24 membrane distal sites regulates the Sec24-Atg9 interaction . ( A ) GST-Atg9C ( 200 nM ) was incubated with 37 . 5 nM of WT Sec23/Sec24-His6 , Sec23/Sec24-T325E-His6 or Sec23/Sec24-T324E/T328E-His6 ( left ) . Ratio of Sec24 bound to GST-Atg9C was quantified from three biological replicates . Averages and s . e . m . are shown ( right ) . WT Sec23/Sec24 was set as one for each experiment; p-value = 0 . 018 ( T325E ) , 0 . 0008 ( T324E/T328E ) , Student’s unpaired t-test . ( B ) Sec24 ( WT ) and Sec24-T324A/T325A or ( C ) Sec24-T324A/T325A/T328A ( Sec24-3A ) were immunoprecipitated from lysates expressing Atg9-13myc as described in the Materials and Methods . Precipitated Atg9-13myc was quantitated and normalized to the amount of Sec24 in the precipitate . WT Sec24 was set as one for each experiment . Averages and s . e . m . are shown for 4 ( B ) or 5 ( C ) biological replicates; p-value = 0 . 006 ( B ) , 0 . 0008 ( C ) Student’s unpaired t-test . ( D ) Alignment of the region surrounding T324/T325/T328 ( shown in red ) with Sec24 orthologues . ( E ) Same as ( B ) except Sec24 was immunoprecipitated from WT or hrr25-5 lysates . Averages and s . e . m . are shown for five biological replicates . p-value = 0 . 0002 , Student’s unpaired t-test . p<0 . 01; p<0 . 01; p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01310 . 7554/eLife . 21167 . 014Figure 5—figure supplement 1 . GST negative control for in vitro binding in Figure 5A . ( A ) Equimolar amounts of GST or GST-Atg9C ( 200 nM ) were incubated with 37 . 5 nM of WT Sec23/Sec24-His6 , Sec23/Sec24-T325E-His6 or Sec23/Sec24-T324E/T328E-His6 . ( B ) Sec24 T324E/T325E does not disrupt the Sec24-Atg9 interaction in vivo . Sec24 T324E/T325E was immunoprecipitated from lysates expressing Atg9-13myc as described in the Materials and Methods ( left ) . Precipitated Atg9-13myc was quantitated and normalized to the amount of Sec24 in the precipitate ( right ) . Precipitated WT Sec24 was set as one for each experiment . Averages and s . e . m . are shown for five biological replicates; p-value = 0 . 39 , Student’s unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01410 . 7554/eLife . 21167 . 015Figure 5—figure supplement 2 . Sec24-3A does not affect Atg assembly at the PAS . ( A ) Sec24 and Sec24-3A cells expressing Ape1-RFP and GFP tagged Atgs were treated with 400 ng/mL rapamycin for 1 hr at 25°C and the percent of Ape1-RFP colocalized with Atgs was determined in 300 cells . Arrowheads point to Ape1 puncta that colocalize with the Atg . Scale bar 2 µm . ( B ) The six Atg hierarchy groups are the Atg1 complex; the phosphatidylinositol 3-phosphate kinase complex ( PI3K ) ; the Atg2/Atg18 complex; the transmembrane protein Atg9; and two different ubiquitin-like conjugating systems , Atg12/Atg5/Atg16 and Atg8-PE . Asterisks denote Atgs that were examined . ( C ) Quantitation of data in ( A ) . Averages and s . e . m . are shown for three biological replicates; p-value = 0 . 78 ( Atg2 ) , 0 . 38 ( Atg5 ) , 0 . 75 ( Atg9 ) , 0 . 7 , ( Atg13 ) 0 . 69 ( Atg14 ) , Student’s unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01510 . 7554/eLife . 21167 . 016Figure 5—figure supplement 3 . Sec24-3A does not affect ERES formation . ( A ) Cells with Sec13-GFP and expressing either Sec24 or Sec24-3A were grown in nutrient rich medium ( SMD ) or starved for nitrogen ( SD-N ) for 2 hr at 25°C . Scale bar 2 µm . ( B ) The number of Sec13-GFP puncta per cell was quantitated . Over 300 cells were quantitated from three biological replicates . Averages and s . e . m . are shown; p-value = 0 . 8 ( SMD ) , 0 . 35 ( SD-N ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01610 . 7554/eLife . 21167 . 017Figure 5—figure supplement 4 . Representative blots for quantitation shown in Figure 5E . Sec24 was immunoprecipitated from lysates prepared from WT and hrr25-5 cells expressing Atg9-13myc . Cropped western blot ( top ) . Uncropped western blot ( bottom ) from top showing pre-immune controls ( left boxed area ) for samples in lanes 5 and 6 ( right boxed area ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01710 . 7554/eLife . 21167 . 018Figure 5—figure supplement 5 . Autophagic body number is reduced in the hrr25-5 mutant . ( A ) Representative images of autophagic bodies in pep4Δ ( left ) or hrr25-5pep4Δ ( right ) cells 1 . 5 hr after nitrogen starvation at 37°C . Scale bar represents 500 nm . ( B ) Histogram showing the distribution of the number of autophagic bodies per cell section in WT and the hrr25-5 mutant . The number of autophagic bodies was quantitated for 100 cell sections for each strain ( left ) . p-value = 0 . 0016; Mann-Whitney Test . Box plot of the number of autophagic bodies per cell section . Bars show data between the lower and upper quartiles , the median is a horizontal line within the box . Whiskers indicate the smallest and largest observations ( right ) . ( C ) The diameter of autophagic bodies was determined . For WT N = 396 , for hrr25-5 N = 254 . Averages with error bars as s . e . m . are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 018 Microscopy-based studies have linked ERES to autophagosome formation ( Graef et al . , 2013; Suzuki et al . , 2013 ) , however , the ERES are also a subdomain of the ER where COPII vesicles bud ( Budnik and Stephens , 2009 ) . To ask if Sec24-3A disrupts the formation of ERES , the localization of Sec13-GFP was examined in cells expressing wild-type Sec24 or Sec24-3A . Sec13-GFP , a component of the outer COPII coat , predominantly localizes to ERES ( Shindiapina and Barlowe , 2010 ) . ERES localization of Sec13-GFP was not affected by Sec24-3A in nutrient rich or starvation conditions , suggesting Sec24-3A does not disrupt ERES formation ( Figure 5—figure supplement 3 ) . Thus , our findings imply that the major function of the Sec24 membrane distal surface is to regulate the interaction of COPII vesicles with the Atg machinery . Hrr25 , the only kinase known to phosphorylate Sec24 in yeast , is required for COPII vesicle fusion in ER-Golgi traffic and autophagy ( Lord et al . , 2011; Wang et al . , 2015 ) . Recent studies , however , have suggested Hrr25 has an additional role in autophagy , upstream of COPII vesicle fusion at the PAS ( Wang et al . , 2015 ) . Specifically , we found that while COPII vesicles accumulate at the PAS in some mutants that disrupt autophagy , they failed to accumulate in the hrr25-5 mutant . Additionally , epistasis studies revealed Hrr25 acts upstream of the key autophagy kinase , Atg1 ( Wang et al . , 2015 ) . Interestingly , Sec24-T328 fits the CK1 consensus motif ( pS/pT-X-X-S/T , Figure 5D ) ( Knippschild et al . , 2005 ) and was found with low confidence to be phosphorylated by Hrr25 in vitro . Consistent with a role for Hrr25 in phosphorylating the Sec24 membrane distal patch , less Atg9 co-immunoprecipitated with Sec24 from an hrr25-5 mutant lysate ( Figure 5E , Figure 5—figure supplement 4 ) . Additionally , like the sec24 triple mutant ( Figure 3E , F , G ) , we observed a reduction in autophagic body number , but not size in the hrr25-5 mutant using transmission electron microscopy ( Figure 5—figure supplement 5 ) . To determine if Hrr25 regulates the Sec24-Atg9 interaction via Sec24 phosphorylation , we asked whether the Sec24 phosphomimetic mutations could rescue the Sec24-Atg9 interaction defect in hrr25-5 . Sec24 T325/T328 was chosen for this analysis as it contains T328 , which is phosphorylated by Hrr25 . WT Sec24 , Sec24 T325A/T328A or Sec24 T325E/T328E was ectopically expressed in hrr25-5 cells containing Atg9-13myc and Sec24 was immunoprecipitated . Sec24 T325E/T328E significantly rescued the Sec24-Atg9 interaction in hrr25-5 , whereas Sec24 T325A/T328A did not ( Figure 6A ) . Additionally , Sec24 T325E/T328E alleviated the autophagy defect in hrr25-5 as the vacuolar localization of GFP-Atg8 induced by starvation was partially rescued by Sec24 T325E/T328E , but not Sec24 T325A/T328A ( Figure 6B ) . To confirm the fluorescence results , cleavage of GFP-Atg8 was also examined . Sec24 T325E/T328E almost fully rescued the GFP cleavage defect in hrr25-5 , while Sec24 T325A/T328A had no effect ( Figure 6C , Figure 6—figure supplement 1 ) . Therefore , although Hrr25 is required for COPII vesicle fusion ( Lord et al . , 2011; Wang et al . , 2015 ) , phosphorylation of Sec24 is a primary function of this kinase during starvation induced autophagy . These findings also indicate the Sec24-Atg9 interaction is needed for autophagy . 10 . 7554/eLife . 21167 . 019Figure 6 . Hrr25 regulates autophagy via phosphorylation of the Sec24 membrane distal sites . ( A ) Sec24 was immunoprecipitated from WT or hrr25-5 cells expressing Atg9-13myc and either WT Sec24 or Sec24 T325A/T328A or Sec24 T325E/T328E . Precipitated Atg9-13myc was quantitated and normalized to the amount of Sec24 in the precipitate . WT was set as one for each experiment . Averages and s . e . m . are shown for three biological replicates . p-value = 0 . 01 ( hrr25-5 + T325E/T328E ) , 0 . 81 ( hrr25-5 + T325A/T328A ) , Student’s unpaired t-test . ( B ) Vacuolar localization of GFP-Atg8 was examined 2 hr after nitrogen starvation at 37°C in WT or hrr25-5 cells expressing either WT Sec24 or Sec24 T325A/T328A or Sec24 T325E/T328E . Scale bar , 2 µm ( left ) . Over 300 cells were quantitated from three biological replicates . WT was set as 100% for each experiment . Averages and s . e . m . are shown . p-value = 0 . 006 ( hrr25-5 + T325E/T328E ) , 0 . 17 ( hrr25-5 + T325A/T328A ) , Student’s unpaired t-test . ( C ) Cleavage of GFP-Atg8 in hrr25-5 cells expressing WT Sec24 or Sec24 T325A/T328A or Sec24 T325E/T328E were examined 2 hr after nitrogen starvation at 37°C ( left ) . The ratio of GFP to GFP-Atg8 was quantitated from three biological replicates . The cleavage in WT was set to 1 . Averages and s . e . m . are shown . p-value = 0 . 03 ( hrr25-5 + T325E/T328E ) , 0 . 99 ( hrr25-5 + T325A/T328A ) , Student’s unpaired t-test . p<0 . 05; *p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 01910 . 7554/eLife . 21167 . 020Figure 6—figure supplement 1 . Nutrient rich controls for GFP-Atg8 cleavage in Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 02010 . 7554/eLife . 21167 . 021Figure 6—figure supplement 2 . Hrr25 is not regulated during starvation . ( A ) Lysates were prepared from WT cells starved for nitrogen for the indicated time periods by incubating 2 . 5 OD600 units of cells with 200 µl 0 . 1 M NaOH for 5 min at room temperature . The precipitate was pelleted and heated in sample buffer for 5 min at 95°C . Lysates were immunoblotted with anti-Hrr25 ( top ) and anti-Bos1 ( bottom ) antibodies . The SNARE Bos1 was used as a loading control . ( B ) Hrr25-HA was immunoprecipitated from cells grown in nutrient rich media ( SMD ) or starved for nitrogen ( SD-N ) for 1 hr at 25°C . The kinase activity of Hrr25 was assayed in vitro using myelin basic protein ( MBP ) as a substrate as described in the Materials and Methods . Asterisk denotes contaminate band from HA resin . Hrr25 activity was quantitated and normalized to amount of Hrr25-HA in the precipitate ( right ) . SMD was set as one for each experiment . Averages and s . e . m . are shown for four biological replicates; p-value = 0 . 76 , Student’s paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 02110 . 7554/eLife . 21167 . 022Figure 6—figure supplement 3 . The hrr25-5 mutant does not affect Atg assembly at the PAS . ( A ) WT and hrr25-5 cells expressing Ape1-RFP and GFP tagged Atgs were starved for nitrogen for 2 hr at 37°C and the percent of Ape1-RFP colocalized with Atgs was determined in 300 cells . Arrowheads point to Ape1 puncta that colocalize with the Atg . Scale bar 2 µm . ( B ) Quantitation of data in ( A ) . Averages and s . e . m . are shown for three biological replicates; p-value = 0 . 4 ( Atg2 ) , 0 . 66 ( Atg5 ) , 0 . 2 ( Atg9 ) , 0 . 6 ( Atg13 ) , 0 . 79 ( Atg14 ) , Student’s unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 02210 . 7554/eLife . 21167 . 023Figure 6—figure supplement 4 . The sec12-4 mutant does not affect Atg2 or Atg14 PAS localization . ( A ) WT and sec12-4 cells expressing Ape1-RFP and Atg2-GFP or Atg14-GFP were starved for nitrogen for 2 hr at 37°C and the percent of Ape1-RFP colocalized with Atg2 or Atg14 was determined in 300 cells . Arrowheads point to Ape1 puncta that colocalize with the Atg protein . Scale bar 2 µm . ( B ) Quantitation of data in ( A ) . Averages and s . e . m . are shown for three biological replicates; p-value = 0 . 5 ( Atg2 ) , 0 . 19 ( Atg14 ) , Student’s unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 023 Despite the role of Hrr25 in regulating the Atg9-Sec24 interaction , neither Hrr25 expression , nor activity was altered upon nitrogen starvation ( Figure 6—figure supplement 2 ) . However , as CK1 kinases act preferentially on substrates that are already phosphorylated at the N-3 position ( Figure 5D ) ( Knippschild et al . , 2005 ) , during autophagy Hrr25 may work with other kinases that create additional Sec24 phosphosites . To determine if Hrr25 mediated phosphorylation of Sec24 is required for Atg complex assembly ( Figure 5—figure supplement 2B ) , we visualized the PAS recruitment of a member of each of the major groups of Atg proteins required for autophagosome formation in the hrr25-5 and sec24 triple alanine mutants . We found no effect of hrr25-5 ( Figure 6—figure supplement 3 ) or Sec24-3A ( Figure 5—figure supplement 2A , C ) on the PAS recruitment of Atg2 , Atg5 , Atg9 , Atg13 or Atg14 . Consistent with our findings , blocking COPII vesicle formation in the sec12-4 mutant also had no effect on the PAS localization of Atg2 and Atg14 ( Figure 6—figure supplement 4 ) . Thus , while COPII vesicles are needed for autophagy , they are dispensable for the assembly of the Atg hierarchy . Together these findings demonstrate that Sec24 phosphorylation does not indirectly affect autophagy by disrupting the trafficking of Atg proteins to the PAS . Although significant progress has been made in defining the upstream events leading to the assembly of Atg proteins at the PAS ( Nakatogawa et al . , 2009 ) , the membrane rearrangements that occur during autophagy remain poorly understood . Here we show that phosphorylation of a conserved regulatory domain of the major COPII cargo adaptor Sec24 reprograms the function of COPII vesicles by modulating its interaction with the C-terminus of Atg9 , a key component of the autophagy machinery . The C-terminus of Atg9 is present in vertebrates and is essential for autophagy ( He et al . , 2008; Young et al . , 2006 ) . Atg9 functions early in autophagosome initiation ( Suzuki et al . , 2007; Yamamoto et al . , 2012 ) and affects autophagosome number , but not size ( Jin et al . , 2014 ) . Similarly , failure to phosphorylate the Sec24 regulatory sites , which disrupts the Sec24-Atg9 interaction , specifically affects autophagosome number . Consistent with our proposal that autophagosome formation requires the fusion of COPII vesicles with Atg9 vesicles ( Tan et al . , 2013; Figure 7 ) , we see an accumulation of COPII coated structures at the PAS in the atg9Δ mutant ( Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 21167 . 024Figure 7 . Phosphorylation of a conserved regulatory surface of Sec24 enhances the ability of the COPII coat to recognize the autophagy machinery . Phosphorylation of the Sec24 membrane distal sites regulates the interaction of Sec24 with the C-terminus of Atg9 . During starvation , the Sec24-Atg9 interaction is needed to increase autophagosome number . If Sec24 is not phosphorylated , it is unable to efficiently interact with Atg9 and COPII vesicles traffic to the Golgi . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 02410 . 7554/eLife . 21167 . 025Figure 7—figure supplement 1 . Sec13-GFP accumulates at the PAS in the atg9∆ mutant . ( A ) WT and atg9∆ cells expressing Sec13-GFP and Ape1-RFP were starved for nitrogen for 2 hr at 30°C and the percent of Ape1-RFP that colocalized with ( left ) or was adjacent to ( right ) Sec13-GFP was determined for over 300 cells from three biological replicates . When the COPII vesicles accumulate in the atg9∆ mutant , fewer are found adjacent to the PAS . Averages and s . e . m . are shown for three biological replicates . p-value = 0 . 018 ( colocalized ) , 0 . 01 ( adjacent to ) , Student’s unpaired t-test . p<0 . 05 . ( B ) Representative images quantitated in ( A ) . Scale bar 2 µm . Arrowheads mark the PAS . DOI: http://dx . doi . org/10 . 7554/eLife . 21167 . 025 For technical reasons , we have been unable to directly address if the Sec24 membrane distal sites are phosphorylated as a consequence of inducing autophagy . However , given that these sites are only required for autophagosome formation during starvation-induced autophagy , and both phosphorylation and autophagy enhance the Sec24-Atg9 interaction , it seems likely the membrane distal Sec24 surface is specifically phosphorylated when autophagy is induced . As autophagosomes form in proximity to each other in nutrient rich and starved cells , the regulated phosphorylation of Sec24 would prevent the inappropriate fusion of COPII vesicles with Atg9 membranes during constitutive growth . Phosphorylation of the Sec24 membrane distal surface is regulated at least in part by Hrr25 . We have shown Hrr25 is required for autophagy and functions upstream of COPII vesicle delivery to the PAS ( Wang et al . , 2015 ) , but how Hrr25 regulates COPII vesicles on this pathway has been unclear . The findings we present here show that , during autophagy , Hrr25 acts through Sec24 by regulating its interaction with Atg9 . As Hrr25 has been linked to the TOR network ( Breitkreutz et al . , 2010 ) , a key negative regulator of autophagy ( Loewith and Hall , 2011 ) , it is tempting to speculate that Hrr25 works with one or more kinases that act downstream of Tor during nutrient deprivation . We propose that the heightened need to rapidly produce autophagosomes during starvation leads to phosphorylation of the Sec24 membrane distal surface , which enhances the Sec24-Atg9 interaction , resulting in an increase in autophagy to maintain homeostasis ( Figure 7 ) . These events are independent of the assembly of the Atg hierarchy that occurs when autophagy is induced . Previous work has established that ERES and COPII vesicles are required for autophagy in both yeast and mammalian cells ( Ge et al . , 2014; Graef et al . , 2013; Lemus et al . , 2016; Suzuki et al . , 2013; Tan et al . , 2013; Wang et al . , 2015 ) . However , because it has been difficult to fully tease apart the function of COPII vesicles on the secretory pathway from their role in autophagy , their contribution to autophagy has been problematic to address . Our finding that phosphorylation of a novel conserved regulatory surface of Sec24 is specifically required for autophagy and not ER-Golgi transport conclusively demonstrates that COPII vesicles play a direct role in autophagy , rather than an indirect role in maintaining the trafficking of autophagy machinery through the secretory pathway . Additionally , we can now ascribe an autophagy-specific role to COPII vesicles in enhancing autophagosome number during nutrient deprivation , which is likely to be conserved . In mammalian cells , nucleation of autophagosomes occurs in the vicinity of the ER and the Atg9 compartment ( Karanasios et al . , 2016 ) . Interestingly , membrane fractionation identified the ER-Golgi intermediate compartment ( ERGIC ) , which resides adjacent to the ER , as the site of lipidation of the Atg8 homologue LC3 ( Ge et al . , 2014 ) . During starvation , the ERES were found to translocate to the ERGIC fraction , and although the ERGIC is typically a location of COPI vesicle budding ( Lorente-Rodríguez and Barlowe , 2011 ) , COPII vesicle budding from this ERGIC fraction was shown to be required for LC3 lipidation ( Ge et al . , 2014 ) . These observations suggested that , in higher eukaryotes , the COPII vesicles used in autophagy are spatially separated from those that traffic to the Golgi . The findings we report here could explain how these spatially segregated COPII vesicles engage the Atg machinery . Other components of the secretory pathway have also been implicated in autophagy ( Amaya et al . , 2015; Geng et al . , 2010; Nair et al . , 2011 ) . It remains to be seen whether these components directly contribute membrane for autophagosome formation or if they only affect the trafficking of autophagy machinery , such as Atg9 , to sites of autophagosome formation . It is known that another component of the autophagy machinery , Atg8 , controls autophagosome size , but not number ( Xie et al . , 2008 ) . However , it is unclear if Atg8 is sufficient to drive autophagosome expansion or if additional membranes are required for this event . In conclusion , our findings highlight an unexpected role for phosphorylation in regulating the reorganization of membrane trafficking pathways during starvation and demonstrate that the COPII coat is a key target of this regulation . The findings we present here ascribe a new role for Sec24 and the COPII coat , and provide a possible explanation for why the coat is retained on the vesicle subsequent to vesicle scission ( Cai et al . , 2007; Lord et al . , 2011 ) . Identification of a cellular mechanism that redirects the flow of membrane during autophagy makes it possible to now study how these complex membrane rearrangement events culminate in the formation of a distinct organelle , the autophagosome . Future work will be needed to determine if the diverted COPII vesicles contain autophagy-specific cargo , or if the role of these vesicles in autophagy is solely to bring certain lipids and SNAREs to this pathway . Strains used in this study are listed in Supplementary file 2 . Yeast cells were grown at 25°C in rich media ( YPD: 1% yeast extract , 2% peptone , and 2% dextrose ) or synthetic minimal media ( SMD: 0 . 67% yeast nitrogen base , 2% dextrose and auxotrophic amino acids as needed ) . Nitrogen starvation was induced in synthetic minimal medium lacking nitrogen ( SD-N: 0 . 17% yeast nitrogen base without amino acids , 2% dextrose ) . For galactose induction of Sec24 , the growth conditions are described in the next section . For solid media , agar was added to a final concentration of 2% . Sec24-His6 was purified from SFNY2181 ( Supplementary file 2 ) as described previously ( Kurihara et al . , 2000 ) with the following modifications . Cells were grown overnight to log phase in SC-Ura-Leu medium with 10% glycerol as the carbon source , and induced at a starting OD600 of 0 . 6–0 . 8 with 0 . 2% galactose for 5 hr at 30°C . These growth conditions were found to induce autophagy in the absence of rapamycin . Cells were lysed in approximately 20 ml HSLB ( 0 . 75 M potassium acetate , 50 mM HEPES pH 7 . 0 , 0 . 1 mM EGTA , 20% glycerol ) with protease and phosphatase inhibitor cocktails ( Sigma , St . Louis , MO ) . The cleared lysate was incubated with Ni-NTA beads for 1 hr at 4°C and washed with 20 ml B-II and 20 ml B-III ( Kurihara et al . , 2000 ) . Sample preparation for mass-spectrometry was carried out as described before ( Guttman et al . , 2009 ) , then liquid chromatography coupled tandem mass spectrometry analysis ( LC-MS/MS ) was performed as described previously ( Meyer et al . , 2014 ) . The high confidence Sec24 mass spectrometry data from three runs are compiled in Supplementary file 3 . Note , potential sites buried in the Sec24 structure were not analyzed further . Mutations in pSFN1915 ( SEC24 , HIS3 , CEN ) were made using the QuikChange Site-directed mutagenesis kit ( Agilent technologies , Santa Clara , CA ) and all mutations were confirmed by DNA sequencing . Plasmids were then introduced into SFNY2201 and SFNY2202 ( Supplementary file 2 ) and grown for two rounds on 5-fluoroorotic acid ( 5-FOA ) plates at 25°C to select against pLM22 ( SEC24 , URA3 , CEN ) . To observe growth defects , sec24 mutants were compared to WT SEC24 at 25°C after two rounds of 5-FOA . For purification of mutant coat proteins , the mutations were made on pSFNB1895 ( GAL1-SEC24-His , LEU2 , CEN ) and co-transformed with pSFNB1894 ( GAL1-SEC23 , URA3 , CEN ) into SFNY2367 ( Supplementary file 2 ) . All structures were accessed through the protein data bank ( PDB ) and analyzed with PyMol software . The Sec23/Sec24 structure , PDB ID 1M2V , was reported in a previous study ( Bi et al . , 2002 ) . The Sec24 structure with the Sed5 peptide , PDB ID 1PD0 , was reported in a previous study ( Mossessova et al . , 2003 ) . COPII proteins ( Sar1 , Sec23/Sec24 and Sec13/Sec31 ) were purified ( Miller et al . , 2003 ) and used to generate vesicles from microsomal membranes prepared as described ( Barlowe et al . , 1994 ) . Vesicle budding assays were performed as described previously ( Miller et al . , 2002 ) . Vesicle fusion was monitored by measuring α−1 , 6-mannose modification of 35S-labeled pro-α-factor as described previously ( Barlowe et al . , 1994 ) . Alkaline phosphatase assays were performed as previously described ( Klionsky , 2007 ) . Cells were grown overnight at 25°C to log phase , OD600 between 0 . 7 and 1 . 0 , washed with 10 ml SD-N medium and incubated in SD-N medium for 2 to 4 hr at 25°C or 37°C to induce autophagy . 2 . 5 OD600 units of cells were collected and washed , and lysed in 250 µl of lysis buffer ( 20 mM PIPES pH 7 . 2 , 0 . 5% TritonX-100 , 50 mM KCl , 100 mM potassium acetate , 10 mM MgS04 , 10 µM ZnSO4 , and 1 mM PMSF ) using glass beads . Lysates , with a protein concentration around 0 . 5 mg/ml , were spun for 5 min at 13 , 000 rpm , and 100 µl of lysate was assayed at 37°C in 400 µl reaction buffer ( 1 . 25 mM p-nitrophenyl phosphate , 250 mM Tris-HCl pH 8 . 5 , 0 . 4% Triton X-100 , 10 mM MgSO4 , and 10 µM ZnSO4 ) . The reaction was stopped with 500 µl of stop buffer ( 1 M glycine/KOH pH 11 . 0 ) , and the OD400 value was determined . The data were normalized to protein concentration using the Bradford method and IgG as a standard . For GFP-Atg8 vacuolar localization , cells were grown overnight at 25°C in SC-Ura to early log phase , OD600 between 0 . 6 and 1 . 0 . Cells were washed and resuspended in SD-N and incubated for 1 hr at 25°C ( Sec24-3A ) , 30 min at 37°C ( Sec24-S730/S735 ) , or 2 hr at 37°C ( hrr25-5 mutant ) . For co-localization of Ape1-RFP with Atg proteins , cells were grown overnight to early log phase at OD6000 . 6–1 . 0 and treated with 400 ng/mL rapamycin for 1 hr at 25°C ( Sec24-3A ) , or starved for nitrogen for 2 hr at 37°C ( hrr25-5 and sec12-4 mutants ) . Cells were then visualized at 25°C with a Zeiss Axio Imager Z1 fluorescence microscope using a 100×1 . 3 NA oil-immersion objective . Images were captured with a Zeiss AxioCam MRm digital camera and analyzed with AxioVision software . To examine GFP-Atg8 puncta by structured illumination ( SIM ) microscopy , cells were grown overnight at 25°C in SC-Ura to early log phase , OD600 between 0 . 6 and 1 . 0 . For autophagy induction , cells were treated with 400 ng/ml rapamycin for 1 hr at 25°C or for 1 hr at 37°C for temperature-sensitive mutants . Cells were pelleted and incubated in 3 . 7% formaldehyde for 30 min at 25°C and visualized on an Applied Precision DeltaVision OMX Super Resolution System using an Evolve 512 EMCCD camera . The data was acquired and processed using Delta Vision OMX Master Control software and SoftWoRx reconstruction and analysis software . To determine autophagosome size , deconvolved images were analyzed with Image J software . To observe cleavage of GFP-Atg8 during autophagy , cells were grown overnight to early log phase , washed , resuspended in SD-N and incubated for 1 hr at 25°C ( Sec24-3A ) or 2 hr at 37°C ( hrr25-5 mutant ) . The cells ( 2 . 5 OD600 units ) were then pelleted , resuspended in 0 . 1 M NaOH and incubated for 5 min at room temperature . The samples were spun and heated in sample buffer for 5 min at 95°C before SDS PAGE . To monitor Ape1 processing , cells were grown in nutrient rich media to early log phase and lysed as described above . Cells were grown overnight in minimal media at 25°C to early log phase and 16 OD600 units of cells were pelleted and resuspended in 3 . 6 ml of fresh minimal media . For starved samples , cells were washed and shifted to SD-N for 1 hr at 25°C . 16 OD600 units of cells were then pelleted and resuspended in 3 . 6 ml of fresh SD-N . Cells were pulse labeled with 400 µCi of S35-methionine for 4 min at 25°C , and 700 µl of the cell suspension was removed and added to 700 µl of ice-cold 20 mM sodium fluoride/sodium azide ( 0 min time-point ) . 250 µl of chase mix ( 250 mM methionine , 250 mM cysteine ) was added to the remaining sample , and then 700 µl of cells were removed at 5 , 10 and 15 min . Cells were pelleted and washed with 1 ml of cold 10 mM sodium fluoride/sodium azide , resuspended in 150 µl spheroplasting buffer ( 1 . 4 M sorbitol , 100 mM sodium phosphate pH 7 . 5 , 0 . 35% β-mercaptoethanol and 0 . 2 mg/ml zymolyase ) and incubated at 37°C for 45 min . Spheroplasts were spun for 3 min at 6500 rpm and heated for 5 min at 95°C in 100 µl 1% SDS . 900 µl of PBS plus 2% Triton X-100 was added to the lysates before they were spun for 15 min at 14 , 000 rpm . CPY antibody ( 3 µl of anti-Rabbit serum prepared against CPY ) was added to 920 µl of cleared lysate and incubated for 1 hr at 4°C with rotation . 50 µl of 50% Protein-A sepharose was added and incubated for 1 hr . The protein-A beads were washed twice with 1 ml of PBS , followed by two washes with 1 ml of 1% β-mercaptoethanol and heated in 70 µl of 1x sample buffer for 5 min at 95°C . Samples were normalized to cpm in the cell lysate , then loaded onto an 8% SDS-PAGE gel and processed for autoradiography . Protein bands were quantified using Image J software . Cells were grown overnight in YPD to an OD600 of 1 . 0 and shifted to SD-N for 1 . 5 hr at 30°C ( Sec24-3A ) or 1 . 5 hr at 37°C ( hrr25-5 mutant ) . 30 OD600 units of cells were pelleted , resuspended in 1 mL of 1 . 5% KMnO4 and incubated for 30 min at 4°C with nutation . Cells were then pelleted and resuspended in 1 mL of 1 . 5% KMnO4 and incubated overnight at 4°C with nutation . Samples were dehydrated in ethanol , embedded in Durcupan epoxy resin ( Sigma-Aldrich ) and sectioned at 60 nm on a Leica UCT ultramicrotome . Sections were picked up on Formvar and carbon-coated copper grids and stained with 2% uranyl acetate for 5 min and Sato's lead stain for 1 min . Grids were viewed using a Tecnai G2 Spirit BioTWIN transmission electron microscope equipped with an Eagle 4k HS digital camera ( FEI , Hilsboro , OR ) . Autophagic body number and size were determined with Adobe Photoshop and Image J software as described previously ( Backues et al . , 2014 ) . Cells were grown overnight to early log phase . For starvation , cells were shifted to SD-N for 4 hr at 30°C , or for the hrr25-5 mutant 2 hr at 37°C . 100 OD600 units of cells were pelleted , resuspended in 2 ml of spheroplasting buffer ( 1 . 4 M sorbitol , 100 mM sodium phosphate pH 7 . 5 , 0 . 35% β-mercaptoethanol and 0 . 5 mg/ml zymolyase ) and incubated for 30 min at 37°C . Spheroplasts were loaded on top of a 5 ml sorbitol cushion ( 1 . 7 M sorbitol , 100 mM HEPES pH 7 . 2 ) and spun for 5 min at 3000 rpm . Cells were lysed in 1 ml of lysis buffer II ( 20 mM Hepes pH 7 . 4 , 150 mM NaCl , excess protease inhibitors ( Roche , Switzerland and Sigma mix ) , phosphatase inhibitors ( Sigma ) ) with a dounce homogenizer on ice . Cell debris was cleared by a 10 min spin at 500 xg . When cross-linking was performed , lysates were incubated on ice with 100 mM dithiobis ( succinimidyl propionate ) for 30 min . To quench excess crosslinker , 100 mM Tris pH 7 . 6 was added and incubated for 15 min on ice . Triton X-100 was added to a final concentration of 1% and incubated on ice for 30 min followed by a 15 min spin at 15 , 000xg . To immunoprecipitate Sec24 , 2 mg of lysate was incubated with 10 µl of Sec24 antibody ( rabbit polyclonal prepared against GST-Sec24 ) or 10 µl of pre-immune serum for 2 hr at 4°C with rotation . 50 µl of 50% protein A-sepharose was added and incubated for 45 min at 4°C with rotation . The protein-A beads were pre-incubated with 1 mg/ml BSA for 30 min at 4°C before they were added to the sample to reduce background . The beads were then washed five times with 1 ml of lysis buffer with 1% Triton X-100 and heated in 40 µl of 3x sample buffer for 5 min at 95°C . Note that similar amounts of Atg9-13myc co-immunoprecipitated with Sec24 without the use of crosslinker . To induce expression of GST fusion proteins , bacterial cells were incubated overnight at 18°C with 0 . 5 mM isopropyl β-D-1-thiogalactopyranoside . Cells were collected , resuspended in 1x phosphate-buffered saline ( PBS ) with 1 mM DTT and protease inhibitors , then sonicated for a total of 2 min with 15 s on/off bursts on ice . Triton X-100 was added to a final concentration of 1% and lysates were incubated on ice for 15 min . The lysates were cleared by a 15 min centrifugation at 15 , 000 rpm . The supernatant was incubated with 1 mL of 50% glutathione sepharose beads ( GE Healthcare , United Kingdom ) that was prewashed with PBS for 1 hr at 4°C with rotation before the beads were washed extensively with PBS and stored at 4°C . His6-Sec23 was purified as described above for the GST fusion proteins , except cells were lysed in 50 mM Hepes pH 7 . 2 , 150 mM NaCl , 15 mM Imidazole , 1 mM DTT with protease inhibitors and incubated with Ni2+-NTA resin ( Qiagen , Germany ) . Protein was eluted from the resin with 50 mM Hepes pH 7 . 2 , 150 mM NaCl , 250 mM Imidazole . Equimolar amounts ( 0 . 2 µM ) of GST fusion proteins were incubated with rotation in binding buffer ( 50 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X-100 , 1 mM MgCl2 , 1 mM EDTA , 1 mM DTT , protease inhibitors ) for 4 hr at 4°C with increasing amounts of His6-Sec23 that was purified from bacteria or Sec23/Sec24-His6 purified from yeast as described before ( Miller et al . , 2002 ) . The beads were washed 3–4 times with binding buffer and eluted in 50 µL of sample buffer by heating for 5 min at 95°C . Cells expressing Hrr25-HA were grown overnight in SC-Ura to early log phase , OD6000 . 6–0 . 8 , then washed and shifted to SD-N medium for 1 hr at 25°C to induce autophagy . The cells were harvested by centrifugation , washed with 20 mM Tris pH 7 . 4 , resupsended in 5 ml of spheroplasting buffer and incubated at 37°C for 30 min . Spheroplasts were pelleted through a 10 ml sorbitol cushion , resuspended in 5 ml of lysis buffer III ( 50 mM Tris-HCl pH 7 . 4 , 100 mM NaCl , 5 mM EDTA , 1 mM PMSF , 1% Triton X-100 , 1X protease inhibitor mixture ( Roche ) ) and lysed with a dounce homogenizer on ice . Lysates were then cleared by a 15 min centrifugation at 14 , 000 rpm . To immunoprecipitate Hrr25-HA , lysates were incubated with 20 µl anti-HA resin ( Sigma ) for 2 hr at 4°C with rotation . The beads were washed three times with lysis buffer and two times with kinase buffer ( 50 mM HEPES pH 7 . 4 , 5 mM MgCl2 , 0 . 2% NP-40 and 1 mM DTT ) . The kinase activity of immunopurified Hrr25-HA was assayed in a 50 µl reaction volume using 1 µg of myelin basic protein ( MBP ) as substrate as described before ( Wang et al . , 2013 ) .	When cells experience stressful conditions , such as a shortage of nutrients , they can digest their own material via a ‘self-eating’ process called autophagy and then recycle the products for further use . When autophagy is triggered , a new membrane structure called the autophagosome forms within the cell as it engulfs the material that is to be digested . The autophagosome delivers these materials to a compartment where they are broken down into smaller parts and the resulting raw materials are reused as needed . The membranes that make up the autophagosome are derived from other membranes within the cell . These include small membrane-bound compartments called vesicles , which carry proteins from one part of the cell to another , or to the outside of the cell . COPII vesicles , for example , carry out the first transport step in the pathway that leads out of the cell – the so-called secretory pathway . Recently it was found that , when cells are starving , COPII vesicles can be diverted to the autophagy pathway and provide a source of membrane to build the autophagosome . However , it was not understood how the membrane of a COPII vesicle is reprogrammed so that it can interact with the cellular machinery that builds autophagosomes . Using genetic and biochemical methods , Davis et al . have now teased apart the distinct roles of COPII vesicles in autophagy and the secretory pathway in budding yeast . The results show that a protein called Sec24 , a component of the coat on the vesicles , interacts with another protein called Atg9 , which is needed for the first steps of autophagosome formation . Davis et al . observed that Sec24 could be modified by the attachment of phosphate groups at a distinct site on the surface of Sec24 . This modification promotes Sec24 to interact with Atg9 and increases the number of autophagosomes that form when cells are starving . Davis et al . also found that the enzyme casein kinase 1 is one of the enzymes responsible for attaching phosphate groups to Sec24 . Following on from this work , it will be important to test whether modification of vesicle coat proteins is a widespread mechanism for reprogramming membranes for different uses in other situations as well .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2016	Sec24 phosphorylation regulates autophagosome abundance during nutrient deprivation
Although thrombosis has been extensively studied using various animal models , our understanding of the underlying mechanism remains elusive . Here , using zebrafish model , we demonstrated that smarca5-deficient red blood cells ( RBCs ) formed blood clots in the caudal vein plexus . We further used the anti-thrombosis drugs to treat smarca5zko1049a embryos and found that a thrombin inhibitor , argatroban , partially prevented blood clot formation in smarca5zko1049a . To explore the regulatory mechanism of smarca5 in RBC homeostasis , we profiled the chromatin accessibility landscape and transcriptome features in RBCs from smarca5zko1049a and their siblings and found that both the chromatin accessibility at the keap1a promoter and expression of keap1a were decreased . Keap1 is a suppressor protein of Nrf2 , which is a major regulator of oxidative responses . We further identified that the expression of hmox1a , a downstream target of Keap1-Nrf2 signaling pathway , was markedly increased upon smarca5 deletion . Importantly , overexpression of keap1a or knockdown of hmox1a partially rescued the blood clot formation , suggesting that the disrupted Keap1-Nrf2 signaling is responsible for the RBC aggregation in smarca5 mutants . Together , our study using zebrafish smarca5 mutants characterizes a novel role for smarca5 in RBC aggregation , which may provide a new venous thrombosis animal model to support drug screening and pre-clinical therapeutic assessments to treat thrombosis . The erythrocytes , or red blood cells ( RBCs ) , are highly differentiated cells produced during erythropoiesis . Mature RBCs are characterized for their abundance of hemoglobin , which can deliver oxygen to surrounding tissues . Importantly , the flexible structure of RBCs makes it capable of traveling through all blood vessels including capillaries by deformation ( Rodríguez-García et al . , 2016 ) . On the benefit of accumulated hemoglobin and the deformation ability , RBCs are essential for organism development by facilitating tissue oxygen delivery and transporting carbon dioxide into the respiration tissues . Moreover , RBCs participate in the maintenance of thrombosis and hemostasis ( Weisel and Litvinov , 2019 ) . Epigenetic regulation of RBC-related genes is fundamental for normal development and maintenance of RBCs ( Hewitt et al . , 2014 ) . In this process , the regulation of chromatin accessibility is a prerequisite for gene transcription and is regulated by chromatin remodelers . For instance , Brg1 could regulate α- and β-globin gene transcription in primitive erythrocytes in mice ( Bultman et al . , 2005; Griffin et al . , 2008 ) . The nucleosome remodeling and histone deacetylase ( NuRD ) is identified to activate human adult-type globin gene expression ( Miccio and Blobel , 2010 ) . Disorder of the gene regulation in RBCs will lead to cellular defects , thereby causing multiple diseases , such as hemoglobinopathy-induced anemia , RBC lysis-induced hemolytic anemia , and thrombosis ( Kato et al . , 2018; Roumenina et al . , 2016; Weisel and Litvinov , 2019 ) . Among them , thrombosis is a leading cause of death worldwide ( Wendelboe and Raskob , 2016 ) . In contrast to arterial thrombi , which are rich in platelets , the venous thrombi are enriched in fibrin and RBCs ( Mackman , 2008; Mackman et al . , 2020 ) . Moreover , venous thrombi can break off , travel , and lodge in the lung , thereby causing pulmonary embolism ( Wolberg et al . , 2015 ) . Currently , the ligature-based inferior vena cava models , free radical thrombosis models and genetic knockout models are widely used in mice to study deep vein thrombosis ( Diaz et al . , 2019; Grover and Mackman , 2019 ) . These disease models are generated mainly through disrupting blood flow , endothelium and blood coagulability . Taking advantage of the conserved hemostatic system and the transparency of embryos , zebrafish has also been used to generate thrombosis models . For instance , phenylhydrazine-treated zebrafish develop severe thrombosis in the caudal vein ( Zhu et al . , 2016 ) . Mechanistically , phenylhydrazine causes externalization of phosphatidylserine on plasma of RBC membrane and generates oxidative radicals , thereafter , resulting in the thrombosis formation . These studies in animal models shed light on the understanding and treatment of vaso-occlusion phenotype in patients with RBC defects . However , the detailed kinetics and underlying mechanism of thrombosis formation in these models are not fully explored . In our previous study , genetic deletion of an epigenetic regulator-smarca5 ( smarca5zko1049a ) resulted in abnormal chromatin accessibility , and we observed disruption of hematopoietic transcription factor binding in the genome , finally leading to defects in fetal hematopoietic stem and progenitor cells ( HSPCs ) ( Ding et al . , 2021 ) . However , whether the other hematopoietic cell types are regulated by smarca5 is unknown . Here , we develop a new zebrafish RBC aggregation model with a deletion of smarca5 , loss of which leads to the formation of blood clots in the caudal vein plexus ( CVP ) . We further present how exactly the change in the subcellular structure of smarca5-deficient RBCs occurred using transmission electron microscopy ( TEM ) , and uncovered the disintegration of cristae in mitochondria in RBCs . To explore the regulatory mechanism of smarca5 in RBC homeostasis , we profiled the chromatin accessibility landscape and transcriptome features by performing Assay for Transposase-Accessible Chromatin with high-throughput sequencing ( ATAC-seq ) and RNA sequencing ( RNA-seq ) analyses in RBCs from smarca5zko1049a and their siblings . Mechanistically , loss of smarca5 led to the decreased chromatin accessibility at keap1a promoter and thus decreased transcriptional expression of keap1a . Keap1 is a suppressor protein of Nrf2 , which regulates the expression of oxidative response genes . A downstream target of Keap1-Nrf2 , hmox1a , showed a markedly increased expression upon smarca5 deletion . Moreover , overexpression of keap1a or knockdown of hmox1a partially rescued the blood clot formation , supporting that the disrupted Keap1-Nrf2 signaling in smarca5 mutants led to the RBC aggregation . Collectively , our smarca5-deficient zebrafish model may serve as a new venous thrombosis model for drug screening in clinical therapy . In our previously generated smarca5zko1049a mutants ( Ding et al . , 2021 ) , we observed that the blood clots were formed in CVP at 2 days post fertilization ( dpf ) , which was not present in their sibling embryos ( Figure 1A ) . Our whole mount in situ hybridization ( WISH ) data showed that scl was expressed in blood clots , indicating that cells in the observed blood clots were primitive RBCs in smarca5zko1049a ( Figure 1B ) . To directly observe the blood clot formation in the CVP , we used the transgenic line ( Tg ) ( gata1:dsRed;kdrl:GFP ) to label RBCs and endothelial cells , in smarca5zko1049a and in siblings . Confocal imaging analysis showed that the blood clots were formed inside the blood vessels ( Figure 1C ) . Notably , there was no difference in the distribution of myeloid cells labeled by Tg ( coro1a:GFP ) or Tg ( mpo:GFP ) in caudal hematopoietic tissue ( CHT ) between smarca5zko1049a and their siblings , and we did not observe accumulation of myeloid cells in the blood clots of smarca5zko1049a ( Figure 1—figure supplement 1A ) . To further determine whether smarca5 is involved in the development of primitive hematopoiesis , we examined the expression level of gata1 and pu . 1 , which are the erythrocyte and myeloid marker genes , respectively , in smarca5zko1049a and their siblings . WISH and quantitative PCR ( qPCR ) analyses showed that the expression level of gata1 and pu . 1 was comparable between smarca5zko1049a and their siblings at 33 hours post fertilization ( hpf ) ( Figure 1D–E and Figure 1—figure supplement 1B-C ) . Moreover , the expression level of ikaros and scl , which are two primitive erythrocyte markers , was normal ( Figure 1D–E ) , as well as the expression of globin genes in smarca5zko1049a ( Figure 1E ) . In addition , the myeloid markers pu . 1 , lyz and mfap4 were normally expressed in smarca5zko1049a at 33 hpf and 2 dpf ( Figure 1—figure supplement 1B-C ) . Thus , the early development of primitive erythrocytes and myeloid cells , is not affected upon the loss of smarca5 in zebrafish embryos . Taken together , these results show that smarca5 is functionally required for normal behaviors of primitive erythrocytes and the blood clotting is formed by erythrocytes in smarca5zko1049a . To visualize how smarca5-deficient RBCs formed blood clots in the CVP of smarca5zko1049a , we performed time lapse imaging using Tg ( gata1:dsRed ) . We tracked the behavior of circulating RBCs in siblings ( , Video 1 ) and smarca5zko1049a ( Video 2 ) from 36 hpf to 2 dpf . The results showed that smarca5-deficient RBCs tended to clump in the CVP at around 40 hpf , after which these clots will migrate or break off under blood flow at the early stage . As the blood clots formed with larger size , these clots will finally lodge in the vein ( Video 2 ) . The snapshot of Tg ( gata1:dsRed ) showed the process of blood clots formation from 36 hpf to 2 dpf in smarca5zko1049a and their siblings ( Figure 2A ) . These results show that the clumping of RBCs precedes their sequestration in CVP , suggesting that the formation of blood clots might be independent of vascular niche . To further explore whether the blood clots formed in smarca5zko1049a were not resulted from the abnormal niche environment , we performed parabiosis experiment using smarca5zko1049a and their siblings . The parabiotic embryo pairs with smarca5zko1049a and siblings share a common bloodstream so that the behavior of circulating cells could reflect the influence of niche environment on these cells . We found that the blood clots , which occurred in smarca5 mutants , were observed in both smarca5zko1049a and their siblings in parabiosis pairs ( Figure 2B ) , indicating the smarca5-deficient RBCs form blood clots largely independent of niche environment . To specifically label the RBCs in smarca5zko1049a and their siblings , the Tg ( gata1:dsRed ) or Tg ( gata1:GFP ) transgenic line was used , respectively . The results showed that smarca5-deficient RBCs labeled by gata1:GFP aggregated both in smarca5zko1049a and in their siblings in parabiosis pairs ( Figure 2C ) . Although several sibling RBCs labeled by gata1:dsRed were found trapped in blood clots , the vast majority of gata1:dsRed+ cells were normally circulating in blood stream both in smarca5zko1049a and their siblings ( Figure 2C ) . Overall , these results indicate that the blood clots in smarca5zko1049a are formed largely in RBC-autonomous manner . To further explore whether thrombocytes participate in the formation of blood clots , we detected the blood clots using Tg ( CD41:GFP ) . The imaging data showed that no CD41:GFPhigh-labeled thrombocytes were present in the blood clots ( Figure 2D ) . The CHT is a hematopoietic tissue critical for HSPC development . We thus wanted to know whether the blood clots formed in smarca5zko1049a could influence the structure of CHT , further leading to HSPC defects . As observed previously , the structure of CHT was normal in smarca5zko1049a and the number of cmyb:GFP+ HSPCs in CHT at 2 dpf was comparable between smarca5zko1049a and their siblings ( Figure 2—figure supplement 1 ) , indicating that the formation of blood clots in smarca5zko1049a is dispensable for HSPC development in CHT . Zebrafish is a useful model to screen drugs for preclinical applications . In our smarca5-deficient zebrafish model , we observed blood clots in veins , raising questions regarding whether there was a thrombus-like phenotype . To this end , we tried to test the clinically used anti-thrombosis drugs to treat smarca5zko1049a embryos . We tested reagents including heparin , aspirin , and argatroban that have been reported to target thrombosis to examine whether the blood clots in smarca5zko1049a can be alleviated after chemical treatment . The embryos were incubated in aspirin or injected with heparin or argatroban at 36 hpf and the phenotype was examined at 2 dpf . As a result , we found that a direct thrombin inhibitor , argatroban , but not an antithrombin-dependent drug , heparin , or a platelet aggregation inhibitor , aspirin , partially prevented blood clot formation in smarca5zko1049a at 2 dpf ( Figure 2E-G ) . These results suggest that the RBC aggregation in smarca5zko1049a is more relevant to venous thrombosis and the smarca5-deficient zebrafish model may serve as a venous thrombosis model to screen drugs in preclinical setting . Both quantitative and qualitative changes in RBCs have been linked to thrombosis ( Weisel and Litvinov , 2019 ) . To identify whether smarca5 deletion will lead to the quantitative changes of RBCs , we performed fluorescence activating cell sorter ( FACS ) analysis of the percentage of gata1:dsRed+ cells in smarca5zko1049a and their siblings . Deletion of smarca5 did not lead to the significant changes in RBC counts at 2 dpf ( Figure 3—figure supplement 1A-B ) . These data suggest that the blood clots in smarca5zko1049a are formed by RBC aggregation with no overt cell number change . To explore whether there exist qualitative changes in smarca5-deficient RBCs , we performed blood-smear and Giemsa-staining analysis . The results showed that the morphology of RBCs had no obvious changes in smarca5zko1049a ( Figure 3—figure supplement 1C ) . And the statistical analysis showed that the nucleocytoplasmic ratio was normal in smarca5-deficient RBCs ( Figure 3—figure supplement 1D ) , indicating that the differentiation of RBCs at 2 dpf was not evidently impaired upon smarca5 loss . To further investigate the changes in subcellular structure of erythrocytes in smarca5zko1049a , we performed TEM analysis . Compared with smarca5 sibling embryos in which the circulating RBCs had normal organization in mitochondria ( Figure 3A–B ) , we found that the smarca5-deficient erythrocytes displayed disintegration of cristae in mitochondria while nuclear integrity was preserved in smarca5zko1049a ( Figure 3C–E ) . The area of mitochondria was not significantly changed and the number of mitochondria was slightly increased but not significantly changed in smarca5-deficent RBCs ( Figure 3F–G ) . We propose that the erythrocytes in smarca5zko1049a may have undergone cellular damages , such as oxidative stress , which could lead to the disintegration of mitochondria ( Lewerenz et al . , 2018 ) . It is also possible that the mitochondrial defects may further exacerbate oxidative stresses ( Dan Dunn et al . , 2015; Yang et al . , 2016 ) , thereafter leading to the erythroid defects caused by smarca5 deletion . Thus , the morphological disruption in mitochondria suggests the disorder of cellular homeostasis in erythrocytes after smarca5 deletion . Smarca5 typically regulates nucleosome spacing , further affecting gene transcription ( Clapier et al . , 2017 ) . To decipher how loss of Smarca5 affects the transcriptome , RNA-seq was used to profile sorted erythrocytes labeled by gata1:dsRed from smarca5zko1049a and their siblings at 2 dpf , respectively ( Figure 4A ) . Principal components analysis ( PCA ) indicated clear separation of the smarca5zko1049a and sibling samples ( Figure 4—figure supplement 1A ) . A total of 1506 genes were upregulated and 633 genes were downregulated significantly ( Log2 ( fold change ) > 1 , adjusted p-value < 0 . 05 ) in smarca5-deficient erythrocytes compared to erythrocytes from siblings ( Figure 4B ) . Gene set variation analysis ( GSVA ) revealed a strong enrichment of terms related to ‘Gata1 targets’ , ‘autophagy’ , ‘erythrocytes take up carbon dioxide and release oxygen’ and ‘erythrocytes take up oxygen and release carbon dioxide’ in sibling erythrocytes; for smarca5zko1049a , while the ‘apoptosis’ , ‘environmental stress response’ , ‘senescence’ , and ‘cell oxidation’ were markedly increased ( Figure 4C ) . The enrichment plots showed the decreased expression of genes related to ‘erythrocyte homeostasis’ in smarca5zko1049a , whereas the expression of genes related to ‘inflammatory response’ was increased ( Figure 4D ) . These results suggest that the disrupted pathways in smarca5-deficient RBCs were highly related to erythrocyte function and cellular homeostasis . RBCs have specialized proteome , which is enriched in hemoglobin . We then focused on the expression of hemoglobin complex related genes . The expression level of embryonic globin genes , including hbae1 , hbae3 , hbbe1 , hbbe2 , and hbbe3 , was not obviously affected in smarca5zko1049a and the expression level of the adult globin genes , including hbaa1 , hbba1 , and hbba2 , was comparable between smarca5zko1049a and their siblings ( Figure 4—figure supplement 1B ) . The WISH results also showed the comparable expression of embryonic and adult globin genes in smarca5zko1049a and their siblings ( Figure 4—figure supplement 1C ) . Moreover , the level of hemoglobin detected by O-dianisidine staining was comparable between smarca5zko1049a and their siblings ( Figure 4—figure supplement 1D ) . Therefore , smarca5 deletion does not lead to obvious hemoglobinopathy in smarca5zko1049a at 2 dpf and the smarca5-deficient RBCs does not have the obvious developmental delay . In addition , we observed the persistent expression of spi1a , spi1b , mfap4 , and lyz markers characteristic of myeloid cells in smarca5-defecient erythrocytes ( Figure 4—figure supplement 1E ) . Perturbation of the exquisite control by smarca5 likely causes ‘hybrid’ primitive erythrocytes that resemble partial transcriptional properties of myeloid cells . One possible mechanism for this phenotype is the regulation of SMARCA5 and CTCF at the enhancer of PU . 1 ( Dluhosova et al . , 2014 ) , thereby blocking of smarca5 leads to the upregulation of pu . 1 gene expression . However , despite the inappropriate expression of myeloid genes in smarca5-deficient RBCs , the development of myeloid lineage was not obviously impaired in smarca5zko1049a manifested with normal expression pattern of pu . 1 and lyz at 33 hpf and 2 dpf ( Figure 1—figure supplement 1B-C ) , suggesting the unaltered lineage choices at the primitive stage . To further explore whether the inappropriate expression of myeloid genes in smarca5-deficient RBCs caused RBC aggregation , we tried to knockdown of pu . 1 in smarca5zko1049a . The results showed that knockdown of pu . 1 cannot rescue the RBC aggregation phenotype in smarca5zko1049a ( Figure 4—figure supplement 1F-G ) . Taken together , smarca5 deletion leads to the disrupted pathways related to erythrocyte function and cellular homeostasis . To explore the mechanism through which Smarca5 in regulating the chromatin accessibility in RBCs , we performed the ATAC-seq in FACS-purified RBCs from smarca5zko1049a and their siblings at 2 dpf . Density heatmaps of mapped ATAC-seq reads showed that fragments less than 100 bp in length clustered immediately upstream of transcriptional start sites ( TSSs ) throughout the zebrafish genome in both mutant and sibling RBC nuclei ( Figure 5—figure supplement 1A-B ) . The PCA analysis was performed for ATAC-seq samples and the results showed that the mutant samples or sibling samples can be grouped together , respectively ( Figure 5—figure supplement 1C ) . The feature distributions of mutant-ATAC-seq peaks and sibling-ATAC-seq peaks across the genome were identified by ChIPseeker ( Figure 5—figure supplement 1D ) . We then calculated the number of genes with changes in chromatin accessibility after smarca5 deletion ( Figure 5A ) . The chromatin accessibility at promoters of 256 genes was decreased in smarca5zko1049a , while there were 439 genes with increased chromatin accessibility at promoters after smarca5 deletion . Next , we screened the motifs enriched in sibling RBC-specific accessible chromatin regions . We found that the erythrocyte master regulator-Gata1 motif was on the top list ( Figure 5B ) . Thus , deletion of smarca5 might affect the binding of hematopoietic transcription factors in erythrocytes , such as Gata1 . It has been reported that Smarca5 could interact with Gata1 in erythrocytes ( Rodriguez et al . , 2005 ) . We propose that Smarca5 might be recruited by Gata1 and mediate the chromatin accessibility of Gata1-binding sites in target genes . We further detected the genes in which the chromatin accessibility at promoters or distal regions and their transcription were both increased or decreased after smarca5 deletion ( Figure 5C and Figure 5—figure supplement 1E ) . The results showed that the chromatin accessibility at promoters and transcription of 84 genes , such as il34 , cox4i2 , skap2 , vclb , and acbd7 , were increased , while the chromatin accessibility at promoters and transcription of 36 genes , such as trim2a , keap1a , acox3 , igfbp1a , and ada , were decreased in smarca5-deficient RBCs ( Figure 5D ) . The lack of overlap between changes in gene expression and ATAC-seq signals may partially due to the complex interactions between cis-regulatory elements and trans-regulatory elements in the regulation of gene expression ( Gibson and Weir , 2005; Hill et al . , 2021; Wittkopp , 2005; Wittkopp et al . , 2004 ) . Moreover , cells exhibit signiﬁcant variations in gene expression and the underlying regulation of chromatin because of intrinsic and extrinsic factors ( Ma et al . , 2020 ) . The accessibility of peaks and the expression of genes are not exactly matched , which may contribute to explaining the lack of overlap between changes in gene expression and ATAC-seq signals . Taken together , smarca5 deletion leads to the disrupted chromatin accessibility and transcriptome in RBCs . Based on the screening results , the chromatin accessibility at keap1a promoters , which contains Gata1 motif , was decreased in smarca5zko1049a ( Figure 6A ) . The transcription level of keap1a detected by qPCR was also decreased in smarca5-deficient RBCs ( Figure 6B ) . Given that keap1 was previously identified to correlate with human venous thrombosis ( Akin-Bali et al . , 2020 ) , we propose that keap1a may act as a downstream target of Smarca5 in RBCs . Keap1-Nrf2 system is an evolutionarily conserved defense mechanism in oxidative stress ( Itoh et al . , 1997; Itoh et al . , 1999 ) . In cytoplasm , Keap1 could anchor to Nrf2 to facilitate the Nrf2 degradation , while oxidative stress leads to the proteasomal degradation of Keap1 and release of Nrf2 to the nucleus , thereafter activate the expression of oxidation defense factors . Both our RNA-seq and qPCR analysis showed the downregulation of keap1a and as a downstream target of Nrf2 , hmox1a showed a markedly increase in gene expression upon smarca5 deletion ( Figure 6C ) , suggesting the disruption of Keap1-Nrf2 signaling pathway . It is worthy of note that , although the upregulated expression of Keap1-Nrf2 downstream targets can protect cells from oxidative damage , the excessive activation of hmox1a , which catalyzes the degradation of heme to biliverdin , carbon monoxide , and Fe2+ , could even lead to the oxidative stress ( Hassannia et al . , 2019 ) . Thus , we propose that the unbalanced Keap1-Nrf2 signaling , especially the upregulation of hmox1a , could increase oxidative damage in smarca5-deficient RBCs . We next performed functional validation of keap1a in smarca5zko1049a by overexpression of hsp70:keap1a-EGFP . Heat shock was performed at 24 hpf and 36 hpf , and the phenotype was examined at 2 dpf . The results showed that overexpression of keap1a in smarca5zko1049a could partially rescue the blood clots phenotype ( Figure 6D–E ) . In addition , knockdown of hmox1a , the downstream target of Keap1-Nrf2 , can also partially rescue the blood clots phenotype in smarca5zko1049a ( Figure 6F–G ) , further supporting that the Keap1-Nrf2 signaling pathway downstream of Smarca5 is essential for blood clot formation ( Figure 6H ) . To further identify the conserved role of SMARCA5 in mammalian erythrocyte homeostasis , we used K562 cells ( human erythroleukemic cells ) to perform further analysis . Treatment of hemin induced the hemoglobinization of most K562 cells , suggesting the efficient erythroid differentiation ( Figure 6—figure supplement 1 ) . We then knocked down SMARCA5 in hemin-induced K562 cells using SMARCA5 short interfering RNA ( siRNA ) and the qPCR and western blot analyses showed that both the RNA and protein levels of SMARCA5 were decreased significantly after siSMARCA5s ( siSMARCA5-1 , siSMARCA5-2 and siSMARCA5-3 ) transfection ( Figure 6—figure supplement 1 ) . In addition , the expression of KEAP1 was decreased while HMOX1 was obviously upregulated after SMARCA5 knockdown ( Figure 6—figure supplement 1 ) , indicating the conserved role of SMARCA5 in human erythrocyte homeostasis . Considering the role of Keap1-Nrf2 signaling pathway in oxidative stress regulation , we further asked whether the oxidative stress could be a trigger for blood clot formation in smarca5 mutants . Then , we used a free radical scavenger glutathione to determine the mechanisms of smarca5-deficiency induced blood clots . We found that glutathione obviously prevented RBC aggregation in smarca5zko1049a ( Figure 6—figure supplement 2 ) , implying that free radical generation may play an important role in RBC aggregation in smarca5zko1049a . Taken together , loss of smarca5 leads to the disruption of keap1a expression and excessive activation of hmox1a in smarca5zko1049a , which together contribute to the formation of blood clots . In this work , we develop a zebrafish RBC aggregation model with a deletion of an epigenetic regulator-smarca5 . The blood clots are formed in the CVP of smarca5zko1049a and the erythrocytes manifest disintegration of cristae in mitochondria . Further transcriptome and chromatin accessibility analysis show that keap1a acts as a downstream target of Smarca5 . Moreover , the elevated expression of the downstream target of Keap1-Nrf2 , hmox1a , leads to the aggregation of smarca5-deficient RBCs . Together , these results demonstrate the protective role of Smarca5 in regulating erythrocyte homeostasis and that the smarca5 loss-of-function zebrafish mutant may serve as a new thrombosis model to screen molecular drugs for clinical therapy . Considering the conserved coagulation and anticoagulation signaling pathway , the zebrafish model has been used to study the physiology of thrombosis ( Hanumanthaiah et al . , 2002; Jagadeeswaran et al . , 1999; Sheehan et al . , 2001 ) . The ferric chloride and laser injury methods are widely used in zebrafish to generate thrombus in the circulation ( Gregory et al . , 2002 ) . Phenylhydrazine-treated zebrafish also develop thrombosis in the caudal vein ( Zhu et al . , 2016 ) . Moreover , zebrafish is an ideal model to explore novel players in thrombosis based on genetic manipulation . For example , mutation of anti-thrombin III gene in zebrafish can mimic disseminate intravascular coagulation ( Liu et al . , 2014 ) . miR-126 was identified as a regulator of thrombi generation in zebrafish ( Zapilko et al . , 2020 ) . Importantly , the transparency of zebrafish embryo makes it feasible to image the kinetics of thrombus formation . In our study , the gata1:dsRed-labeled RBCs were imaged during blood clot formation . Thus , the zebrafish thrombosis model is a great asset for exploring the underlying mechanisms in thrombosis formation . Unlike Brg1 , which is essential for mouse erythrocyte development by regulating globin gene expression ( Bultman et al . , 2005; Griffin et al . , 2008 ) , Smarca5 is required for primitive erythrocyte homeostasis at the erythrocyte differentiation stage . Deletion of smarca5 does not lead to the gross changes in RBC morphology and viability , but specifically results in the RBC aggregation phenotype . The mechanistic details for different chromatin remodelers functioning in the different processes during erythropoiesis warrant further investigation . Previous evidence suggests that chromatin remodeler NuRD is required to maintain lineage fidelity during erythroid-megakaryocyte ontogeny ( Gao et al . , 2010; Gregory et al . , 2010 ) . Our results show that , despite the normal lineage choice for primitive erythrocytes in smarca5zko1049a , the aberrant activation of myeloid genes occurred in RBCs after smarca5 deletion . The exquisite cell lineage control by smarca5 may be due to the regulation of SMARCA5 at the enhancer of PU . 1 ( Dluhosova et al . , 2014 ) . The RBCs are sensitive to mitochondrial biogenesis and function . A previous study shows that during human erythrocyte specification , the pathways related to mitochondrial biogenesis are enhanced through post-transcriptional regulation ( Liu et al . , 2017 ) . Deletion of mitochondria factors resulted in metabolic changes and histone hyperacetylation , further leading to the impaired erythrocyte differentiation . Moreover , the transcription elongation factor TIF1γ directly regulates mitochondrial genes and histone methylation during erythrocyte differentiation ( Rossmann et al . , 2021 ) . These studies suggest that the mitochondria biogenesis and function are highly regulated during normal erythropoiesis through transcriptional , epigenetic and post-transcriptional mechanisms , which may explain the specific defects observed in smarca5-deficient erythroid cells . Besides the conserved role of Keap1-Nrf2 system in oxidative stress , Keap1-Nrf2 is also demonstrated to act as a regulator in cell development and differentiation across multiple tissues and cell types . For instance , Keap1-Nrf2 signaling pathway is indispensable for hematopoietic stem cell ( HSC ) lineage commitment in mice ( Murakami et al . , 2014 ) . Knockout of Keap1 in HSCs showed enhanced granulocyte-monocyte differentiation ability at the expense of lymphoid and erythrocyte differentiation . And the expression level of erythrocyte and lymphoid genes was decreased in Keap1-deficient HSCs . Importantly , the abundance of Hmox1 is upregulated during erythrocyte differentiation , and Hmox1 expression must be tightly regulated at appropriate level for efficient erythropoiesis ( Garcia-Santos et al . , 2014 ) . Overexpression of Hmox1 impairs hemoglobin synthesis , while lack of Hmox1 leads to the enhancement of hemoglobinization . Here , we show that the disruption of keap1a expression and excessive activation of hmox1a in smarca5zko1049a contribute to RBC aggregation . Besides the free radical generation , which may play an important role in RBC aggregation in smarca5zko1049a , we cannot rule out other possibilities that may also be involved in the observed phenotype , such as the regulation of Keap1-Nrf2 signaling pathway in erythrocyte gene expression . In summary , we have demonstrated , for the first time , that deletion of smarca5 in zebrafish leads to the RBC aggregation by regulating the Keap1-Nrf2 signaling pathway in RBCs . These findings raise the possibility that zebrafish smarca5 mutant may serve as a new venous thrombosis model for drug screening and pre-clinical therapeutic assessment . Zebrafish strains including Tubingen , Tg ( CD41:GFP ) ( Lin et al . , 2005 ) , Tg ( gata1:dsRed ) ( Traver et al . , 2003 ) , Tg ( kdrl:mCherry ) ( Bertrand et al . , 2010 ) , Tg ( gata1:dsRed;kdrl:GFP ) ( kindly provided by Stefan Schulte-Merker ( Hubrecht Institute , Utrecht , The Netherlands ) ) , Tg ( mpo:GFP ) ( Renshaw et al . , 2006 ) , Tg ( coro1a:GFP ) ( Li et al . , 2012 ) , smarca5zko1049a heterozygous mutants ( Ding et al . , 2021 ) were raised under standard conditions ( 28 . 5 °C in system water ) . The zebrafish embryos were raised in incubator at 28 . 5 °C . The present study was approved by the Ethical Review Committee of the Institute of Zoology , Chinese Academy of Sciences , China . The human K562 cells ( ATCC:CCL-243 ) were cultured in RPMI-1640 medium supplemented with 10 % FBS at 37 °C in 5 % CO2 . WISH was performed as previously described ( Wang et al . , 2011 ) . The Digoxigenin-labeled RNA probe genes including gata1 , ikaros , scl , pu . 1 , lyz , hbae1 , hbae3 , hbbe1 , hbbe2 , hbbe3 , hbaa1 , hbba1 , and hbba2 were cloned from zebrafish cDNA and ligated to the T-vector , then in vitro transcribed using T7 or SP6 polymerase . Total RNAs were extracted from smarca5zko1049a and their sibling embryos using TRIzol reagent ( Life technologies , 15596018 ) or from sorted RBCs using QIAGEN RNeasy Mini Kit ( Cat . No . 74104 ) . The cDNA was reverse transcribed using M-MLV Reverse Transcriptase ( Promega , M1701 ) . The detailed primers used for qPCR are listed in Supplementary file 1A . The antisense MOs were purchased from GeneTools . The sequences of MOs were used as previous described , these gene-specific MOs include hmox1a MO and pu . 1 MO . The detailed sequence and dosage used in this work are listed in Supplementary file 1B . Parabiosis experiment was performed by following the previous published procedures ( Demy et al . , 2013; Hagedorn et al . , 2016 ) . Briefly , smarca5zko1049a and their sibling embryos between the 128 cell blastula and 30 % epiboly stages were removed out of chorions and gently transferred into methylcellulose drop under fish water . Then , detach a few cells from each embryo at the contact points using the pulled glass micropipette and move these two embryos contact each other properly until they fusion together . O-dianisidine staining , Giemsa-staining , and Benzidine staining smarca5zko1049a and their sibling embryos at 2 dpf were stained with o-dianisidine staining solution for 15 min in the dark as previously described ( Detrich et al . , 1995 ) . The blood cells from smarca5zko1049a and their sibling embryos at 2 dpf were collected from heart and caudal vein and attached to slides . The dried slides will be stained by Fast Giemsa Stain ( Yeasen Biotech Co . , Ltd , CAT: 40 , 751ES02 ) following the standard manufacturer’s instructions . The K562 cells were collected and washed once using PBS . Then the cells were suspended using 500 μl PBS . Subsequently , add 10 μl 0 . 4 % benzidine , 1 μl 30 % H2O2 , and 1 μl 5 % sodium nitroferricyanide dihydrate and incubate for 3 min , 5 min and 3 min , respectively . Then the cells were attached to slides for further imaging . Argatroban ( Sigma , A0487 ) , dissolved in DMSO ( 2 mg/ml ) , was injected into smarca5zko1049a and their sibling embryos at 36 hpf at the dosage of 4 nl/embryo . The control embryos were injected with DMSO alone at the same dosage . Heparin ( Sigma , H3393 ) , dissolved in H2O ( 2 . 5 mg/ml ) , was injected into smarca5zko1049a and their sibling embryos at 36 hpf at the dosage of 4 nl/embryo . For aspirin treatment , the smarca5zko1049a and sibling embryos at 36 hpf were incubated with aspirin ( Sigma , A2093 ) at the concentration of 5 μg/ml . The smarca5zko1049a and sibling embryos at 36 hpf were incubated with Glutathione ( Sigma , PHR1359 ) at the concentration of 0 . 5 mg/ml . Confocal microscopy was performed using Nikon confocal A1 laser microscope ( Nikon ) and Andor high speed confocal ( dragonfly , Belfast , UK ) . The embryos were embedded in 1 . 2 % low melting agarose . For overexpression experiment , the full length CDS of keap1a was cloned into pDestTol2pA2 with a hsp70 promoter and an EGFP reporter by DNA assembly ( NEBuilder HiFi DNA Assembly Master Mix , E2621S ) . The plasmids together with tol2 mRNA were injected into zebrafish embryos at one-cell stage to generate Tg ( hsp70:flag-keap1a-EGFP ) . Control and SMARCA5 siRNAs were synthesized by GenePharma Corporation . The K562 cells were maintained in RPMI-1640 medium supplemented with 10 % FBS and stimulated with hemin ( Sigma , 51280 , 30 μM ) for 3 days to induce erythroid differentiation . Then , the hemin-induced K562 cells were transfected with siRNAs using Lipofectamine RNAiMAX Reagent ( Invitrogen , 13778–030 ) following the manufacturer’s instructions . The detailed sequences are listed in Supplementary file 1C . The western blotting was performed to detect the protein level of SMARCA5 in K562 cells after siRNA transfection . The antibodies used were as followings: anti-Smarca5 antibody ( Santa Cruz , H-300: sc-13054 ) , anti-β-Actin antibody ( Cell Signaling Technology , 4967 ) . The smarca5zko1049a and their sibling embryos with Tg ( gata1:dsRed ) background at 2 dpf were collected and washed by Ringers buffer . After digesting into single-cell suspension using 0 . 5 % trypsin , the reaction was stopped by adding CaCl2 up to 1 M and fetal calf serum up to 10 % . Then the cells were filtered through 300 Mesh nylon cell-strainer to make single-cell suspension . The RBCs ( gata1:dsRed+ ) were sorted using MoFlo XDP ( Beckman Coulter ) and collected into PBS containing 1 % FBS . RNA-seq was performed in FACS-purified RBCs from smarca5zko1049a and their siblings at 2 dpf . A total of 50 , 000 RBCs were used per sample for RNA-seq experiments . The RNAs of sorted HSPCs were isolated using QIAGEN RNeasy Mini Kit ( Cat . No . 74104 ) following the standard manufacturer’s instructions . The mRNA libraries were constructed using NEBNext Ultra RNA Library Prep Kit for Illumina and sequenced under Illumina HiSeq X Ten with pair end 150 bp ( PE150 ) . Raw RNA-seq reads data were trimmed using the fastp ( Chen et al . , 2018 ) ( v2 . 4 ) ( parameter: with default parameters ) , and aligned to ‘Danio_rerio GRCz10’ cDNA reference sequence using the STAR ( Dobin et al . , 2013 ) ( v 2 . 7 . 7 a ) with the default parameters . Read counts for each gene were quantified as the total number of reads mapping to exons using featureCounts ( Liao et al . , 2014 ) ( subread v1 . 5 . 3 ) . DESeq2 ( Love et al . , 2014 ) was applied to perform differential expression analysis with raw counts quantified by featureCounts . We used Benjamini-Hochberg adjusted p-value < 0 . 05 and log2 fold change >1 as the threshold for significant difference . Gene set enrichment analysis was performed using GSEA function in the clusterProfiler ( Yu et al . , 2012 ) package ( v 3 . 18 . 0 ) . Gene set variation analysis was performed by the GSVA ( Hänzelmann et al . , 2013 ) package ( v 1 . 38 . 0 ) . The gene sets we used were exported by the msigdbr package ( v 7 . 2 . 1 ) . The differences in pathway activities scored between smarca5zko1049a and their sibling RBCs were calculated with limma ( Ritchie et al . , 2015 ) package ( v 3 . 46 . 0 ) . ATAC-seq was performed in FACS-purified RBCs from smarca5zko1049a and their siblings at 2 dpf . A total of 50 , 000 RBCs were used per sample for ATAC-seq library preparation using TruePrep DNA Library Prep Kit V2 for Illumina ( Vazyme , TD501 ) as previously described ( Ding et al . , 2021 ) . Firstly , wash the sorted RBCs using 1xPBST . Then , the cell pellet was lysed using 50 μl cold lysis buffer ( 10 mM Tris-HCl ( pH 7 . 4 ) , 10 mM NaCl , 3 mM MgCl2 and 0 . 15% NP-40 ) for 5 min on ice . Centrifuge and discard the supernatant to get the cell pellet ( about 2 μl ) . Then , the transposition reaction system combining 5xTTBL ( 10 μl ) , TTE Mix ( 5 μl ) , and H2O ( 33 μl ) was added immediately to the cell pellet and pipetted up and down gently for several times . After the incubation at 37 °C for 30 min , the DNA was extracted with chloroform-phenol . After the purification , the DNA was amplified using TruePrep DNA Index Kit V2 for Illumina ( Vazyme , TD202 ) . After the fragments length purification using VAHTS DNA Clean Beads ( Vazyme , N411 ) , The DNA libraries are under sequencing under Illumina NovaSeq with pair end 150 bp ( PE150 ) . Raw ATAC-seq reads were trimmed using cutadapt ( v 2 . 4 ) ( parameter: -q 20 m 20 ) and mapped to the danRer10 reference genome using Bowtie2 ( Langmead and Salzberg , 2012 ) ( v 2 . 3 . 4 . 2 ) ( default parameters ) . Sorting , removal of PCR duplicates and conversion from SAM to BAM files were performed using SAMtools ( Li et al . , 2009 ) ( v 1 . 3 . 1 ) . For quality assessment of ATAC-seq libraries , we applied an R package ATACseqQC ( Ou et al . , 2018 ) ( v 1 . 6 . 4 ) to check the fragment size distributions , Transcription Start Site ( TSS ) enrichment scores , and plot heatmaps for nucleosome positions . We employed deepTools2 ( Ramírez et al . , 2016 ) ( v 2 . 5 . 7 ) to check the reproducibility of the biological replicates and generated bigwig files from BAM output to visualize mapped reads . Peaks were called using MACS2 ( Zhang et al . , 2008 ) ( v2 . 1 . 2 ) ( parameter: --nomodel --nolambda --gsize 1 . 4e9 --keep-dup all --slocal 10000 ) . Differentially accessible regions were identified using an R package DiffBind ( Ross-Innes et al . , 2012 ) ( v 2 . 10 . 0 ) with a log2 fold change threshold of 0 . 5 , and Benjamini-Hochberg adjusted p-value < 0 . 1 . Peak annotation was performed by an R package ChIPseeker ( Yu et al . , 2015 ) ( v 1 . 18 . 0 ) . We identified the enriched de novo motifs across the whole genomic regions using the findMotifsGenome . pl function of HOMER ( Heinz et al . , 2010 ) ( parameter: -size 500 -len 8 , 10 , 12 -mask -dumpFasta ) . The tail region of smarca5zko1049a and their siblings at 2 dpf were fixed with 2 . 5 % ( vol/vol ) glutaraldehyde and 2 % paraformaldehyde in phosphate buffer ( PB ) ( 0 . 1 M , pH 7 . 4 ) . After washing with PB for four times , the tissues were immersed in 1 % ( wt/vol ) OsO4 and 1 . 5 % ( wt/vol ) potassium ferricyanide aqueous solution at 4 °C for 1 hr . After washing , the tissues were incubated in ﬁltered 1 % thiocarbohydrazide ( TCH ) aqueous solution ( Sigma-Aldrich ) at room temperature for 30 min , followed by 1 % unbuffered OsO4 aqueous solution at 4 °C for 1 hr and 1% UA aqueous solution at room temperature for 2 hr . The tissues were dehydrated through graded alcohol ( 30% , 50% , 70% , 80% , 90% , 100% , 100% , 10 min each , at 4 °C ) . Then , transfer the tissues into pure acetone for 10 min ( twice ) . Tissues were infiltrated in graded mixtures of acetone and SPI-PON812 resin ( 21 ml SPI-PON812 , 13 ml DDSA and 11 ml NMA ) ( 3:1 , 1:1 , 1:3 ) , then transfer the tissues into pure resin . Finally , the tissues were embedded in pure resin with 1 . 5 % BDMA and polymerized at 45 °C for 12 hr , followed by at 60 °C for 48 hr . The ultrathin sections ( 70 nm thick ) were sectioned with microtome ( Leica EM UC6 ) , and examined by a transmission electron microscope ( FEI Tecnai Spirit120kV ) . Raw image data were processed using ImageJ , photoshop CC 2018 and Adobe Illustrator CC 2018 . All the statistical analysis was performed for at least three independent biological repeats . GraphPad Prism six was used to analyze the data . Data are mean ± s . d . p Values calculated by two-tailed unpaired Student’s t-test were used to indicate the significance if not clarified in figure legends .	After an injury , cells in our blood ( called red blood cells ) often stick together to form clots to stop us from bleeding and prevent infection . These clots , however , can sometimes develop in veins and arteries , resulting in a condition known as thrombosis . If left untreated , these blockages can be life-threatening and lead to a heart attack or stroke . To study the physical effects of venous thrombosis and test different treatments , researchers often use animal models . In particular , the transparent embryos of zebrafish , as it easy to see how blood flows through their circulatory system . However , it is difficult to explore the underlying mechanisms that cause red blood cells to aggregate together using these models . To overcome this , Ding et al . developed a new model for venous thrombosis by deleting the gene for a protein called Smarca5 . They found that red blood cells lacking this gene were more likely to clump together in the veins of zebrafish . Further experiments showed that this mutation reduced the activity of the gene for a protein called Keap1a , which suppresses the activity of Nrf2 . Nrf2 switches on a number of genes involved in blood clotting , including the gene for the protein Hmox1a . Ding et al . discovered that increasing the activity of the gene that encodes the Keap1a protein , or decreasing the activity of the gene for Hmox1a , partially stopped red blood cells from sticking together in the zebrafish model . These findings suggest that the blood clots formed in the zebrafish model are due to the disrupted connection between Keap1a and Nrf2 . This model could be used to screen new drugs for treating venous thrombosis . However , further experiments are still needed to see how similar the blood clots in the zebrafish are to the ones found in patients with this disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2021	The chromatin-remodeling enzyme Smarca5 regulates erythrocyte aggregation via Keap1-Nrf2 signaling
The actin cytoskeleton mediates mechanical coupling between cells and their tissue microenvironments . The architecture and composition of actin networks are modulated by force; however , it is unclear how interactions between actin filaments ( F-actin ) and associated proteins are mechanically regulated . Here we employ both optical trapping and biochemical reconstitution with myosin motor proteins to show single piconewton forces applied solely to F-actin enhance binding by the human version of the essential cell-cell adhesion protein αE-catenin but not its homolog vinculin . Cryo-electron microscopy structures of both proteins bound to F-actin reveal unique rearrangements that facilitate their flexible C-termini refolding to engage distinct interfaces . Truncating α-catenin’s C-terminus eliminates force-activated F-actin binding , and addition of this motif to vinculin confers force-activated binding , demonstrating that α-catenin’s C-terminus is a modular detector of F-actin tension . Our studies establish that piconewton force on F-actin can enhance partner binding , which we propose mechanically regulates cellular adhesion through α-catenin . Cells probe and respond to the mechanical properties of their surroundings through cytoskeletal networks composed of actin filaments ( F-actin ) , myosin motor proteins , and dozens of actin-binding proteins ( ABPs ) . These networks transmit forces through cell-matrix focal adhesions ( Humphrey et al . , 2014 ) and cell-cell adherens junctions ( Charras and Yap , 2018 ) , plasma membrane-associated many-protein assemblies which serve as hubs for the conversion of mechanical cues and stimuli into biochemical signaling cascades ( mechanotransduction ) . Defects in mechanotransduction are associated with numerous diseases ( Jaalouk and Lammerding , 2009 ) , including muscular dystrophies , cardiomyopathies , and metastatic cancer , yet therapeutics which specifically target these pathways are largely absent due to our ignorance of the mechanisms that transduce mechanical signals through the cytoskeleton . Diverse force regimes modulate the polymerization dynamics , micron-scale architecture , and protein composition of actin assemblies by molecular mechanisms that remain unclear ( Harris et al . , 2018; Romet-Lemonne and Jégou , 2013 ) . Cellular-scale pressures in the range of hundreds of pascal regulate the assembly and mechanical power of branched-actin networks generated by the ARP2/3 complex in vitro ( Bieling et al . , 2016 ) and in vivo ( Mueller et al . , 2017 ) , tuning network geometric properties including filament length , density , and distribution of filament orientations . Live-cell imaging studies identified a subset of ABPs that preferentially localize to the cytoskeleton in vivo in response to this magnitude of mechanical stimulation ( Schiffhauer et al . , 2016 ) , which were also postulated to recognize properties of network geometry . Molecular-scale forces in the piconewton range have also been shown to modulate the activity of F-actin polymerization ( Courtemanche et al . , 2013; Jégou et al . , 2013; Risca et al . , 2012; Zimmermann et al . , 2017 ) and severing ( Hayakawa et al . , 2011 ) factors in vitro , suggesting that molecular components of actin networks could be mechanically regulated by physiological forces . While F-actin binding by the actin-depolymerization factor cofilin was reported to be directly regulated by tension across the filament to modulate its activity ( Hayakawa et al . , 2011 ) , a recent study has suggested that cofilin is tension-insensitive , instead having its severing activity regulated by filament bending ( Wioland et al . , 2019 ) . F-actin has been reported to adopt a structural landscape of co-existing conformations in cryo-electron microscopy ( cryo-EM ) studies ( Galkin et al . , 2010b ) , leading to speculation that actin filaments could themselves serve as tension sensors by presenting distinct binding interfaces to ABPs in the presence of load ( Galkin et al . , 2012 ) . It nevertheless remains unclear to what extent mechanical modulation of functional interactions between ABPs and F-actin occurs through direct regulation of F-actin’s binding interactions by force . Furthermore , structural mechanisms enabling ABPs to detect force on F-actin , to our knowledge , are unknown . Inspired by the report that the enhanced cytoskeletal localization response of ABP isoforms differed substantially when cells were mechanically stimulated ( Schiffhauer et al . , 2016 ) , we reasoned that biophysical and structural analysis of closely-related ABPs would be a powerful approach for mechanistically dissecting mechanically regulated F-actin binding . Here we specifically investigate enhanced binding to F-actin when the load is applied solely to the filament , which we refer to as ‘force-activated binding’ . Our studies focus on the homologous adhesion proteins α-catenin ( Kobielak and Fuchs , 2004 ) and vinculin ( Ziegler et al . , 2006 ) , which are major ABP components found in adherens junctions ( α-catenin and vinculin ) and focal adhesions ( vinculin ) that are critical for the force-dependent strengthening of cellular adhesion and mechanotransduction ( Dumbauld et al . , 2013; Yonemura et al . , 2010 ) . Vinculin is a strictly auto-inhibited globular protein that must engage with multiple adhesion partners to be activated and bind F-actin ( Johnson and Craig , 1995 ) , stabilizing adhesion through incompletely defined mechanisms . On the other hand , α-catenin exists in two distinct populations maintained in dynamic equilibrium in the cell ( Drees et al . , 2005 ) . It serves as a central component of the membrane-anchored heterotrimeric α-catenin–β-catenin–cadherin complex ( the ‘cadherin-catenin complex’ ) at adherens junctions , which lacks F-actin-binding activity in traditional assays when isolated ( Yamada et al . , 2005 ) . It also forms a soluble homodimer with constitutive modest F-actin-binding activity , thought to play a role in the generation and maintenance of actin bundles by cross-linking filaments and inhibiting ARP2/3 binding , thereby suppressing branched-actin formation ( Drees et al . , 2005 ) . The structural mechanisms by which forces transmitted through the actin cytoskeleton modulate the complex networks of binding interactions formed by α-catenin and vinculin during adhesion maturation remain unknown . Both proteins are entirely α-helical ( Bakolitsa et al . , 2004; Rangarajan and Izard , 2013 ) and are composed of a large N-terminal ‘head’ domain , which engages in protein-protein interactions with other adhesion molecules ( Kobielak and Fuchs , 2004; Ziegler et al . , 2006 ) and a smaller C-terminal 5-helix bundle F-actin binding ‘tail’ domain ( Bakolitsa et al . , 1999; Ishiyama et al . , 2013; Figure 1—figure supplement 1 , Helices H1–H5 ) , connected by a flexible linker . The isolated actin-binding domains ( ABDs ) , which we utilize in our study , retain their structures and actin-binding activities ( Bakolitsa et al . , 1999; Ishiyama et al . , 2013 ) , engaging a similar site on the filament surface ( Hansen et al . , 2013; Janssen et al . , 2006; Kim et al . , 2016; Thompson et al . , 2014 ) . Recent single-molecule force-spectroscopy studies reported that both α-catenin in the cadherin-catenin complex ( Buckley et al . , 2014 ) and vinculin ( Huang et al . , 2017 ) form catch bonds with F-actin , characterized by increased bond lifetime when moderate forces around 10 pN are applied across the ABP-actin interface . These studies provided one potential resolution to the apparent contradiction between biochemical studies demonstrating the cadherin-associated α-catenin population lacks F-actin binding activity ( Drees et al . , 2005 ) and cellular studies suggesting α-catenin plays a key role in linking adherens junctions to the cytoskeleton ( Kobielak and Fuchs , 2004 ) . However , in vivo both vinculin and α-catenin primarily engage contractile actomyosin bundles , whose component actin filaments are constitutively exposed to myosin-generated forces . This led us to hypothesize that force-activated binding to tensed F-actin could also be a key regulatory mechanism at adhesions , suitable for promoting the formation of initial attachments to actomyosin through membrane-anchored ABPs ( e . g . the cadherin complex ) before the ABP-F-actin interface itself coming under load . Furthermore , it could also serve as a mechanism to enrich soluble ABPs ( e . g . the α-catenin dimer ) whose binding geometry is fundamentally incompatible with catch-bond formation . Here using simultaneous optical trapping and confocal fluorescence microscopy , we show that tension on the order of 1 pN across actin filaments directly enhances F-actin binding by human αE-catenin , but not vinculin . Utilizing a novel Total Internal Reflection Fluorescence ( TIRF ) microscopy in vitro reconstitution assay , we further show that physiological forces generated by myosin motor proteins activate α-catenin F-actin binding . Approximately 3 Å resolution cryo-EM structures of both proteins bound to F-actin ( to our knowledge , the highest-resolution structures of ABP-F-actin complexes reported to date ) reveal they share an overlapping major actin-binding site . However , they undergo markedly different rearrangements at their flexible N- and C-termini upon binding , resulting in distinct contacts mediated by their C-terminal extensions ( CTEs ) which re-fold on the actin surface . Truncating α-catenin’s CTE results in constitutive strong F-actin binding regardless of force , and a chimeric construct of vinculin featuring α-catenin’s CTE gains force-activated binding , demonstrating the α-catenin CTE to be a modular ‘force-detector’ which negatively regulates low-force binding . Together , our studies indicate piconewton force on F-actin can be discriminated by flexible elements in ABPs to mediate direct force-activated binding . To determine whether the actin-binding activity of vinculin or α-catenin could be regulated solely by force across F-actin , we performed correlative force and fluorescence measurements with a commercial instrument ( Hashemi Shabestari et al . , 2017 ) that combines dual-trap optical tweezers and confocal fluorescence microscopy ( Figure 1A ) . In these experiments , Alexa-555 phalloidin-stabilized biotinylated actin filaments were captured from a laminar stream across a microfluidic flow-cell between two optically trapped streptavidin-coated beads . Tethered filaments were then transferred to a reservoir containing Halo-tagged ABD labeled with Alexa-488 and subjected to a constant velocity pulling protocol while simultaneously recording confocal movies of both the actin and ABD fluorescence signals . We frequently observed a monotonic increase in force once a threshold extension was reached , followed by an instantaneous return to baseline ( Figure 1B , C; Figure 1—figure supplement 2A , B ) , consistent with resistance from a tether composed of one or more actin filaments extended beyond their resting length ( visible as straightening in the actin fluorescence channel ) followed by force-induced breakage . To correlate force with ABD binding , we calculated the background-subtracted ratio of the ABD fluorescence intensity to the actin intensity ( which we refer to as ‘IABD/Iactin’ ) in each frame of the confocal movies ( Materials and methods; Figure 1B , C , bottom ) , as well as the corresponding average force during frame acquisition ( Figure 1B , C , top , black connected points ) . Apparent binding by the vinculin ABD fluctuated and did not change in response to load during individual pulling trajectories ( Figure 1B ) , consistent with a previous report that mechanical stimulation does not promote cytoskeletal localization of vinculin in cells ( Schiffhauer et al . , 2016 ) . However , we observed an apparent steady increase in binding along individual tethers by the α-catenin ABD ( Figure 1C ) , consistent with force-activated binding . To quantify this phenomenon , due to the inherent fluctuations of individual IABD/Iactin traces in this assay ( potentially due to dynamic changes in ABD binding density along individual tethers and instability in the focus during confocal imaging ) , we first focused our analysis on a subset of long-lived tethers which featured at least 10 frames . We examined the difference between a ‘high-force’ average , defined as the mean IABD/Iactin value from the five frames before the final tether rupture , and a ‘low-force’ average , defined as the mean IABD/Iactin value from the five preceding frames during each recording ( Figure 1B , C ) . Paired analysis of low-force/high force averages ( Materials and methods ) showed a significant increase consistent with force-activated binding only by the α-catenin ABD ( Figure 1D and E; Figure 1—figure supplement 2A ) . Next , we sought to determine the impacts of force on ABD binding at the level of individual actin filaments . As our tether assembly procedure captures filaments from solution , in principle each tether could be composed of a single filament or multiple filaments . We thus elaborated the possible multi-filament tether configurations and designed a series of criteria to identify and exclude them from the analysis . The first plausible multi-filament tether configuration ( 1 ) is composed of multiple filaments of different lengths , all attached to the trapped beads at both ends ( Figure 1—figure supplement 2B ) , which we identified by the presence of multiple breaking peaks in the force curve , generated as the shortest remaining filament in the tether reaches full extension then ruptures ( Figure 1—figure supplement 2B ) . The next plausible configuration ( 2 ) is multiple filaments of different lengths , not all of which are attached at both ends ( Figure 1—figure supplement 2C ) , which we identified by fluorescence line scans of the actin intensity , where we observe step-like reductions which we interpret to correspond to the ends of non-bridging filaments ( Figure 1—figure supplement 2C ) . The third and final configuration ( 3 ) , composed of multiple filaments , all of essentially identical length and attached at both ends ( Figure 1—figure supplement 2D ) , is the most difficult to rigorously identify . We observed significant heterogeneity in the raw actin fluorescence intensity values between tethers ( potentially due to slight differences in focus ) , with no significant difference between configuration one tethers and those featuring a single rupture force peak ( data not shown ) , suggesting raw actin intensity cannot be used as a proxy for filament number . We thus analyzed the distribution of final breaking forces for all trials ( pooling data collected in the presence of both the α-catenin and vinculin ABD ) , reasoning that configuration three tethers composed of multiple fully-extended actin filaments in parallel at the final break would generally rupture at higher forces . Consistent with this prediction , we observed a single major peak in the breaking force distribution centered at ~6 pN , which is likely to primarily consist of single filaments , as well as a long tail of ≥16 . 5 pN breaking forces , which we interpret to primarily encompass configuration three tethers . We thus additionally excluded all tethers with ≥16 . 5 pN breaking force from further analysis . Based on these criteria , only 21 tethers ( 11 vinculin ABD and 10 α-catenin ABD ) out of 183 total trials were included . This subset very likely predominantly contains tethers composed of single filaments . Consistent with the expected relative fragility of single-filament tethers with low breaking forces , most of these recordings only contained a limited number of frames , precluding quantification of force-dependent binding changes in individual tethers . We thus instead examined the relationship between force magnitude and ABD binding by pooling our recordings for each protein and plotting normalized IABD/Iactin versus force ( Figure 1F , G ) . This analysis revealed no apparent correlation between vinculin ABD binding and force ( Figure 1F ) . However , the α-catenin ABD plot showed an apparent step-like transition from force-uncorrelated binding to consistent strong binding above a threshold force of approximately 1 pN ( Figure 1G ) . To quantify this phenomenon , we initially examined the distribution of intensity ratios above and below the subjectively identified 1 pN threshold , finding a significant difference only for the α-catenin ABD ( Figure 1F , G ) . To objectively define the threshold force , we performed K-means clustering analysis ( Materials and methods ) which revealed that the α-catenin ABD force-fluorescence distribution could be optimally divided into two clusters with a threshold force of 1 . 8 pN . We found that the intensity ratios were also significantly different between these clusters ( Figure 1G ) . Although there was no obvious correlation between force and binding by the vinculin ABD , as a control we nevertheless used K-means to divide the data into two clusters , which separated at a threshold force of 2 . 2 pN . Consistent with the lack of a force-binding correlation for the vinculin ABD , we find no significant difference between the distribution of intensity ratios in these two clusters ( Figure 1F ) . This analysis , which is insensitive to the exact force threshold employed , supports a force-dependent increase in F-actin binding only for the α-catenin ABD . Collectively , these data suggest that piconewton-level tensile force along individual actin filaments is sufficient to activate α-catenin’s F-actin binding . As ~1 pN is the magnitude of force generated by individual myosin motor domains ( Finer et al . , 1994 ) , we hypothesized α-catenin’s F-actin binding would also be enhanced by physiological forces generated by myosins . To test this hypothesis , we developed a novel adaptation of the gliding filament assay ( Kron and Spudich , 1986 ) to apply force to filaments mimicking actomyosin contractility in vivo ( Figure 2A ) . In our preparation , plus-end directed myosin V motor proteins and minus-end directed myosin VI motor proteins are randomly surface-immobilized inside a flow chamber assembled on a cover glass for TIRF , resulting in a configuration where the motors are poised to engage in tug-of-war along non-stabilized , rhodamine-labeled actin filaments . A Halo-tagged , JF-646 ( Grimm et al . , 2015 ) labeled ABD is then flowed into the chamber in the absence of ATP , and a 2-color TIRF movie is recorded to visualize the basal level of actin binding when filaments are anchored by the rigor-state motors . The ABD is then re-introduced into the same chamber in the presence of ATP to activate the motors , and a second movie is recorded to visualize binding in the presence of force generation . Visual inspection of -ATP/+ATP TIRF movie pairs for the vinculin and α-catenin ABDs suggest they respond to motor-generated forces on F-actin distinctly . Actin localization of the vinculin ABD did not change in response to motor activity ( Figure 2B; Figure 2—video 1 ) ; however , α-catenin ABD’s actin localization was enhanced upon motor activation , indicative of force-activated actin binding ( Figure 2C; Figure 2—video 2 ) . As we expected the random deposition and inherent stochasticity of molecular motors in our assay to give rise to a distribution of forces and ABD binding states , we implemented an image analysis procedure to quantify ABD association by automatically identifying and tracking tens to hundreds of ‘filament regions’ through time ( Materials and methods , Figure 2—figure supplement 1A ) . For each region , IABD/Iactin was calculated in each frame , then averaged over all frames in which the region was detected . Consistent with our qualitative interpretation , histograms of IABD/Iactin distributions before and after ATP addition to individual flow chambers showed no shift for the vinculin ABD ( Figure 2D , Figure 2—figure supplement 1B ) . However , this analysis demonstrated a clear shift toward higher values upon ATP addition for the α-catenin ABD , supporting force-activated binding ( Figure 1E , Figure 2—figure supplement 1B ) . While the reported trends were consistent across experiments for both ABDs , we nevertheless observed variability between trials ( Figure 2—figure supplement 1B ) , potentially due to differences in the background intensities in both channels resulting from inconsistencies in cover-glass surface preparation ( Materials and methods ) . We thus performed paired analysis of the mean IABD/Iactin between the -ATP/+ATP conditions for each chamber ( Figure 2F and G ) , which demonstrated a significant increase only for the α-catenin ABD ( Figure 2G ) . Our optical-trapping experiments suggest that force along individual filaments is sufficient to activate α-catenin binding . However , in the cellular context , both α-catenin and vinculin primarily engage actin-myosin bundles . In our TIRF assay , visual inspection supports increased F-actin bundling by both vinculin and α-catenin upon motor activation ( Figure 2B and C ) , presumably due to motility facilitating encounters between filaments . Although IABD/Iactin measurements are internally normalized for the local density of F-actin in each region , we are aware that inter-filament ABD bundling contacts could in principle enhance apparent binding . Additionally , while the ATP-dependence of this enhanced F-actin binding strongly suggests it is activated by force , allosteric remodeling of actin filament structure due to local deformations imposed by motor binding could also potentially contribute ( Gurel et al . , 2017 ) . To decouple these effects , we performed assays in the presence of individual motors . In the presence of both myosin V alone ( Figure 2—figure supplement 2 , Figure 2—video 3 and Figure 2—video 4 ) and myosin VI alone ( Figure 2—figure supplement 2 , Figure 2—video 5 and Figure 2—video 6 ) we observe ATP-dependent formation of bundles . However , we observe no significant increase in apparent vinculin ABD or α-catenin ABD binding in either condition . This strongly suggests that force-activated α-catenin ABD binding is dependent upon the tug-of-war between motors of opposed directionality , mimicking the forces generated by bi-polar myosin II filaments in vivo . We note that forces generated by the randomly distributed force generators in the dual-motor assay are complex , and can in principle include tension , compression , and torsional forces ( Beausang et al . , 2008; Sun et al . , 2007 ) . While our optical trapping studies suggest tensile forces are sufficient to activate α-catenin ABD binding ( Figure 1 ) , future studies will be required to explicitly dissect the contribution of compression and torsion in the presence of myosin motors . Additionally , we note that there are caveats associated with our TIRF experiments . First , the relatively low apparent affinity of our α-catenin ABD construct in this assay necessitates utilizing a high working concentration . The commensurate high fluorescence background is refractory to definitively measuring the intensity level constituting F-actin binding saturation in the absence or presence of force generation . This limits our ability to establish the binding stoichiometry range in which force-activated actin binding occurs , which has the potential to constrain or discriminate between plausible molecular mechanisms ( see Discussion ) . In contemporary studies , Xu and colleagues have reported a series of α-catenin ABD N-terminal deletion constructs with substantially increased F-actin affinity in the absence of force , notably a construct consisting of residues 698–906 , which boosts affinity ~18-fold ( Xu et al . , 2020 ) . Presuming this construct maintains force-activated binding activity , it may be a useful tool to dissect the impact of α-catenin’s binding stoichiometry . Second , due to the strong propensity of the α-catenin ABD to bundle actin filaments in our TIRF experiments , it remains to be conclusively determined if myosin motor activity can also activate α-catenin binding along single filaments . It may be possible to achieve a sufficiently low surface density of actin filaments to avoid bundling in the dual-motor assay in presence of the actin stabilizing drug phalloidin , which would enable probing α-catenin’s association with individual actin filaments in the presence of active force generation . We will pursue these studies , and their outcome will be presented in a follow-up report . Regardless , the data we present here collectively suggest physiological forces generated by myosin motor proteins in an appropriate configuration can directly activate α-catenin binding to F-actin , and that force-activated α-catenin binding also occurs in the context of actin bundles , the primary cytoskeletal architecture engaged by the protein in vivo . Hypothesizing that differences in the F-actin-binding interfaces of vinculin and α-catenin could underlie their differential force-activated actin binding , we pursued structural studies of both ABDs bound to F-actin with cryo-EM ( Figure 3; Figure 3—figure supplement 1; Table 1 ) . As optimizing the density of fully-decorated , well-separated individual filaments in cryo-EM images is a major bottleneck for single-particle analysis of F-actin-ABP complexes , we chose to use the ABD of the vinculin splice variant metavinculin for these studies , where a 68 amino-acid insert displaces the H1 helix and replaces it with helix H1’ , producing a protein which retains actin binding but completely loses actin bundling activity ( Janssen et al . , 2012; Kim et al . , 2016; Oztug Durer et al . , 2015; Figure 1—figure supplement 1B ) . Previous studies have suggested these isoforms engage an essentially identical site on the F-actin surface with equivalent affinity ( Janssen et al . , 2012; Kim et al . , 2016 ) , and we further found the metavinculin ABD lacks force-activated actin-binding activity in our TIRF assay , validating its use in these studies ( Figure 2—figure supplement 1C , D ) . We were able to obtain fields of individual decorated filaments using this construct ( Figure 3A , top ) . After careful optimization ( Materials and methods ) , we were also able to acquire cryo-EM images of filaments decorated with the α-catenin ABD ( Figure 3A , bottom ) , although persistent bundling by this construct necessitated the collection of substantially more images to obtain a sufficient dataset of individual segments to obtain a high-resolution reconstruction ( Table 1; Figure 3—figure supplement 1 ) . Using the Iterative Helical Real Space Reconstruction ( IHRSR ) approach ( Egelman , 2007 ) as implemented in Relion 3 . 0 ( He and Scheres , 2017; Zivanov et al . , 2018; Materials and methods , Figure 3—figure supplement 1 ) , we obtained reconstructions of the metavinculin ABD ( residues 879–1134 ) –F-actin complex ( Figure 3B , left ) at 2 . 9 Å overall resolution ( Figure 3—figure supplement 1 ) and the α-catenin ABD ( residues 664–906 ) –F-actin complex ( Figure 3B , right ) at 3 . 2 Å overall resolution ( Figure 3—figure supplement 1 ) . As local resolutions ranged from 2 . 7 Å to 3 . 6 Å , radially decaying outward from the helical axis ( Figure 3—figure supplement 1C , D ) for both reconstructions , atomic models for the complete sequence of Mg-ADP α-actin and continuous segments of ABD residues 981–1131 for metavinculin ( Figure 3C ) and 699–871 for α-catenin ( Figure 3C ) were built and refined into the maps ( Figure 3—figure supplement 1 ) . In their contemporary work , Xu and colleagues also reported a 3 . 6 Å resolution cryo-EM reconstruction of the α-catenin ABD bound to F-actin ( Xu et al . , 2020 ) , which shows a highly similar conformation of the complex to that presented here ( actin , 0 . 5 Å RMSD; α-catenin residues 711–842 , corresponding to helices H2–H5 , 1 . 0 Å RMSD ) . Superposition of the actin-bound metavinculin ABD with the full-length vinculin crystal structure ( Bakolitsa et al . , 2004 ) confirms that actin binding by both vinculin isoforms is auto-inhibited by intramolecular interactions between the N-terminal head and C-terminal ABD tail domains ( Johnson and Craig , 1995; Figure 4—figure supplement 1A , B ) , as the head domain clearly clashes with F-actin in the crystallized conformation . Full-length α-catenin was crystallized as an asymmetric ‘left-handshake dimer’ , characterized by differential relative orientations between the head and tail domains of each protomer ( Rangarajan and Izard , 2013 ) . Comparing the actin-bound α-catenin ABD with both conformers in the asymmetric dimer crystal structure also reveals severe clashes between the α-catenin head and actin ( Figure 4—figure supplement 1C–F ) , suggesting that full-length α-catenin must also undergo substantial conformational rearrangements to bind F-actin . Next , we compared the metavinculin ABD-F-actin and α-catenin ABD-F-actin structures ( Figure 3C ) , confirming previous low- and moderate-resolution studies ( Janssen et al . , 2006; Janssen et al . , 2012; Kim et al . , 2016; Thompson et al . , 2014 ) that both ABDs engage a major site spanning the longitudinal interface of 2 actin protomers , which we term Actin I and Actin II ( numbered from the plus end of the filament ) . In turn , each actin protomer also contacts 2 ABDs , leading to a 1:1 binding stoichiometry at saturation . Our structures establish this region is almost identical between the two ABDs , comprising 2040 Å2 of buried surface area for the metavinculin ABD and 1920 Å2 for the α-catenin ABD . However , our high-resolution models also reveal a previously unobserved minor interface for each ABD ( confirming a recent computational prediction in the case of metavinculin [Krokhotin et al . , 2019] ) , mediated by residues in their flexible C-terminal extensions ( CTEs ) , which are entirely distinct between the two proteins ( Figure 1—figure supplement 1A ) . Consistent with our previous medium-resolution structural studies ( Kim et al . , 2016 ) , we find the metavinculin ABD undergoes substantial conformational remodeling upon F-actin engagement , characterized by displacement of helix H1’ from the 5-helix bundle to license a rearrangement of helices H2–H5 to relieve clashes with F-actin ( Figure 4A , B; Figure 4—video 1 and Figure 4—video 2 ) . N-terminal residues 879–980 are not visible in the map ( Figure 4A , transparent brown ) , and are presumably disordered in the actin-bound state , while residues 981–985 ( Figure 4A , brown ) undergo a slight rearrangement , extending helix H2 by 1 . 5 turns ( five residues ) . Our high-resolution map reveals this contact to be mediated by the ( meta ) vinculin CTE ( Figure 4 ) . The CTE is released from its pre-bound position , extending helix H5 by two turns ( six residues ) , then undergoing an approximately 60° swing to engage a site along actin subdomain one proximal to H5 ( Figure 4A and B; Figure 5—video 1 ) . Coupled to this transition , helices H2–H5 slightly rearrange to accommodate actin binding and avoid steric clashes ( Figure 4B and C; Figure 4—video 2 ) . We find that the α-catenin ABD also undergoes an order-to-disorder transition at its N-terminus upon actin binding ( Figure 4A , B; Figure 4—video 1 ) , as no density for residues 664–698 , the majority of H0-H1 , is present in our map ( Figure 3D; Figure 4A , B; Figure 4—video 1 ) , confirming recent reports that this region is important for activating α-catenin’s actin engagement ( Ishiyama et al . , 2018; Xu et al . , 2020 ) . This is accompanied by a twisting rearrangement of helices H2–H4 reminiscent of that found in ( meta ) vinculin ( Figure 4B; Figure 4—video 2 ) , as well as the extension of H4 by three turns ( nine residues ) through folding of the H3-H4 loop , to sculpt a major actin-binding interface sterically compatible with the filament ( Figure 4B; Figure 4—video 1 ) . This suggests that N-terminal helix release allosterically coupled to ABD helical-bundle rearrangement is a fundamentally conserved mechanistic feature of actin binding by members of the vinculin/α-catenin family . However , we observe distinct rearrangements in the α-catenin CTE , which undergoes a slight lateral shift and helical unfurling , rather than a swing , to engage a site spanning a different surface of actin subdomain 1 ( Figure 4A and B; Figure 5—video 1 ) . Metavinculin H5 extension is facilitated by binding interactions with Actin I ( Figure 5A , B; Figure 5—video 1 ) , notably a hydrophobic interaction between metavinculin I1114 and Actin Y91 , and a salt bridge between metavinculin R1117 and actin E100 . This positions the CTE to form an extended interface with actin subdomain 1 , contiguous with that mediated by H5 , with metavinculin W1126 buried in a proximal hydrophobic pocket formed by actin residues A7 , P102 , P130 , A131 , and W356 , bolstered by a distal salt bridge between metavinculin R1128 and actin E361 ( Figure 5B ) . By contrast , the α-catenin CTE retains an overall conformation similar to its pre-bound state ( Figure 5A and C; Figure 5—video 1 ) . An extensive hydrophobic network we term a ‘tryptophan latch’ embraces CTE residue W859 in both pre-bound and post-bound states , preventing α-catenin CTE unfurling ( Figure 5C , right; Figure 5—video 1 ) . A single turn of helix H1 on the N-terminal side of the ABD remains folded , with H1 residue W705 packing against CTE residue M861 , encircling W859 along with CTE residues L852 and L854 , as well as residues W705 , I712 , I763 , L776 , P768 , V833 , and Y837 from neighboring regions of the helical bundle , facilitating coordinated conformational transitions between the N- and C-terminal flexible regions of the α-catenin ABD upon actin binding . A putative hydrogen bond is also maintained between S840 and the single nitrogen atom in W859’s indole ring , maximizing the binding potential of this residue . The latch positions the neighboring region of the CTE to bind a distinct site on Actin I’s subdomain 1 ( Figure 5C , left; Figure 5—video 1 ) mediated by proximal salt bridges ( α-catenin E865 – actin R28 , α-catenin K866 – actin D24/D25 , α-catenin K867 – actin E93 ) and distal hydrophobic interactions ( α-catenin L869/V870 – actin P333/Y337 ) . Consistent with a key role for the latch in coordinating conformational transitions that enable F-actin binding , Xu et al . report that mutating W859 to alanine reduces α-catenin’s F-actin-binding affinity 10-fold ( Xu et al . , 2020 ) . Superposition of the actin-bound conformation of the α-catenin ABD with the pre-bound conformation of the ( meta ) vinculin ABD ( Figure 4—figure supplement 1G ) reveals a striking positional overlap between α-catenin W859 and metavinculin W1126 ( vinculin W1058 is identically positioned , not shown ) , which is also engaged by a sparser latch in the pre-bound conformation ( Figure 5—video 1 ) . We thus speculate the extensive latch of α-catenin prevents W859 release and the extension of its CTE to engage the same site as metavinculin W1126 . The complete non-overlap of the ( meta ) vinculin and α-catenin minor actin-binding interfaces mediated by their CTEs lead us to hypothesize that the CTEs could be involved in differential force-activated actin binding . To identify other potential contributing structural elements , we undertook a detailed comparison of their major actin-binding interfaces mediated by helices H4–H5 in both proteins ( Figure 5D ) . The Actin II binding interface is almost identical between the two ABDs ( Figure 5E ) , mediated by an extensive series of conserved hydrophobic contacts: α-catenin I792/metavinculin I1065 – actin I345 , α-catenin V796/metavinculin V1069 – actin I341 , and α-catenin V800/metavinculin M1073 – actin P333/E334 . The Actin I interface , on the other hand , is more variable and characterized by few clear residue-level binding interactions ( Figure 5F ) , notably likely weak long-distance salt bridges ( metavinculin R1044 – actin E83 and α-catenin K845 – actin E99 ) specific to each protein , despite the overall shape complementarity across the interface . Each ABD also features a unique hydrophobic interaction with actin Y91 ( metavinculin I1114/α-catenin V838 ) . The Actin I interface extends into contacts with the actin D-loop ( Figure 5G ) , a flexible region of actin which mediates structurally polymorphic longitudinal interactions between protomers ( Galkin et al . , 2010b ) reported to be modulated by actin nucleotide state ( Chou and Pollard , 2019; Merino et al . , 2018 ) and ABPs ( Dominguez and Holmes , 2011; Oda et al . , 2019 ) . Both ABDs form a potential weak long-distance interaction with actin D-loop residue K50: in metavinculin , a hydrogen bond through N1048 , and in α-catenin , a salt bridge through D775 . The D-loop then adopts subtly different conformations between the two interfaces centered at M47 . Although clear residue-level binding interactions are not readily apparent , the conformation of M47 at the metavinculin interface would clash with α-catenin Y786 , a position occupied by the smaller residue I1059 in metavinculin , suggesting local sterics unique to each ABD determine compatibility with a distinct D-loop conformation . Comparison of the actin conformation observed in a similar-resolution structure of ADP F-actin in isolation ( ‘F-actin alone’ , M . S . in preparation ) versus when bound to metavinculin or α-catenin , as well as comparison of the metavinculin-bound and α-catenin-bound conformations reveals minimal rearrangements throughout the majority of the structure ( Figure 5H ) . This contrasts with a previous report of α-catenin-induced structural changes in F-actin inferred from low-resolution cryo-EM analysis ( Hansen et al . , 2013 ) , but it is consistent with the high-resolution studies of Xu et al . , 2020 . The sole region featuring rearrangements greater than 1 Å RMSD is a 3–4 residue stretch of the D-loop centered on M47 . As force across the filament could feasibly modulate D-loop structure to regulate ABP binding , we hypothesized ABD residues mediating D-loop interactions could also mediate differential force-activated actin binding . To investigate whether D-loop interactions contribute to α-catenin force-activated binding , we designed a triple point-mutant α-catenin ABD construct where three residues in close proximity to the D-loop were replaced by those in vinculin: α-cat ABDA778Q Y779V Y786I . In force reconstitution assays , this construct did not visibly associate with actin in either the –ATP or +ATP condition in the concentration regime accessible by TIRF ( Figure 5—figure supplement 1A ) . While these data suggest that the α-catenin D-loop interactions contribute to overall affinity for F-actin , the complete lack of binding is refractory to determining whether this interface has a separable role in force-activated actin binding . We thus returned to our initial hypothesis that differential force-activated binding could be mediated by the CTEs . Although we were unable to accurately model the final three residues of the metavinculin CTE , weak density is clearly present ( Figure 5—figure supplement 1B , red ) , suggesting the entire CTE engages F-actin . By contrast , density for the α-catenin CTE is only present until K871 ( Figure 5C; Figure 5—figure supplement 1C ) . Notably , all three human α-catenin isoforms have highly homologous CTEs that extend an additional 35 amino acids ( Figure 1—figure supplement 1A ) , diverging in sequence and length from the vinculin CTE . Consistent with previous primary-structure-function analysis ( Pokutta et al . , 2002 ) showing that residues after P864 , which bear no homology to vinculin , are necessary for actin binding , our structure shows that residues 865–871 are in direct contact with actin ( Figure 5C; Figure 5—video 1 ) , forming an extensive interface . We therefore hypothesized that distal residues 872–906 , a 35-residue element unique to α-catenin ( Figure 1—figure supplement 1A ) that was not resolved in our cryo-EM analysis and is thus presumably conformationally flexible , could uniquely contribute to force-activated actin binding . To test whether distal residues 872–906 have a separable role in force-activated binding , that is as a ‘force detector’ , we first truncated them from the α-catenin ABD ( α-cat ABDΔC ) . Consistent with a regulatory role for this segment , α-cat ABDΔC constitutively associated with F-actin in TIRF assays ( Figure 6A; Figure 6—figure supplement 1A; Figure 6—video 1 ) , with no significant increase in binding upon ATP addition , suggesting that this region is necessary for force-activated binding by negatively-regulating low-force binding . This contrasts with the observations of Xu et al . , who report a modest ( ~2-fold ) reduction in F-actin-binding affinity when this region is truncated in solution co-sedimentation assays , notably in the background of a construct where H0 has also been truncated in order to boost affinity overall ( Xu et al . , 2020 ) . Possible sources of this discrepancy include differences in α-catenin binding behavior between solution assays and our TIRF assays , where filaments are immobilized . Additionally , coordination between rearrangements in the CTE and H0–H1 ( potentially through allosteric mechanisms ) may be necessary to mediate the force-detector’s negative regulatory effects . To investigate the sufficiency of the force-detector , we generated chimeric ABDs featuring the H2–H5 bundle region of vinculin and the flexible termini of α-catenin . A vinculin ABD construct where only the CTE was substituted was non-functional ( data not shown ) . However , consistent with structural coordination between the α-catenin N-terminal segment and the CTE through the tryptophan latch ( Figure 5C; Figure 5—video 1 ) , a construct featuring both the α-catenin N-terminal segment and the CTE ( vinc ABD-NCSwap , Materials and methods ) gained force-activated binding activity , with diminished low-force binding observed ( Figure 6B; Figure 6—figure supplement 1B; Figure 6—video 2 ) in contrast to the wild-type vinculin ABD ( Figure 1B ) . A C-terminal truncation of this construct ( vinc ABD-NCSwapΔC ) equivalent to α-cat ABDΔC reverted to constitutive binding regardless of force ( Figure 6C; Figure 6—figure supplement 1C; Figure 6—video 3 ) , supporting the α-catenin CTE as the key determinant of force-activated binding , in which 872–906 serves as the force-detector . We thus conclude that the distal C-terminus ( residues 872–906 ) of α-catenin is necessary and sufficient for force-activated actin binding through negative regulation of low-force binding . While our studies pinpoint the final 35 amino acids of α-catenin as a force-detector , the exact molecular mechanism by which this segment negatively regulates low-force binding to F-actin remains to be elucidated . Here we propose two potential , non-exclusive conceptual models for this modulation . As we observe the distal tip of the ordered region of the CTE to be in close apposition to the next ABD binding site along the actin filament , with potential contacts between CTE residues V870 and K871 with the H4–H5 loop and the N-terminal tip of H5 in longitudinally adjacent ABD ( Figure 5—figure supplement 1C ) , the first model invokes steric exclusion ( Figure 7A ) . In the absence of force , the force-detector ( Figure 6; Figure 7 ) physically blocks the adjacent binding site through steric hindrance , which can be relieved by an increase in protomer axial spacing in the presence of tension , consistent with prior truncation studies suggesting residues 884–906 may inhibit α-catenin’s actin binding ( Pappas and Rimm , 2006 ) . As the force-detector likely represents a flexibly tethered conformational ensemble , in the presence of thermal fluctuations we envision this would manifest as a tension-dependent increase of the binding on-rate at the site due to its increased fractional availability . Although saturating the filament for structural studies could lead to non-physiological inter-ABD interactions , cooperative F-actin binding by the α-catenin ABD has previously been reported under non-saturating conditions ( Hansen et al . , 2013 ) , and supplemental soluble ABD enhanced catch-bonding by the cadherin complex ( Buckley et al . , 2014 ) , suggesting communication between actin-bound ABDs is likely to be physiologically relevant . We note that cooperative and inhibitory inter-ABD communication are not a priori mutually exclusive , and the interplay of these opposing effects could produce differential outcomes as a function of ABD concentration and filament load , an important subject for future studies . The second model invokes a conformational change in the actin protomer that specifically occurs in the presence of mechanical load , which is recognized and preferentially bound by the force-detector , relieving inhibition ( Figure 7B ) . Although our studies suggest only minor actin conformational changes when the binding is driven by mass action ( Figure 5H ) , they do not rule out as yet unobserved actin conformations specifically evoked by force . Furthermore , while for simplicity we have framed both models in terms of discrete transitions between structural states , low piconewton forces could also modulate the intrinsic structural fluctuations of F-actin to control α-catenin engagement through either mechanism , as has previously been speculated for cofilin ( Galkin et al . , 2012; Hayakawa et al . , 2011; Wioland et al . , 2019 ) . Although currently technically prohibitive , structural studies of the α-catenin ABD–F-actin interface in the presence of active force generation , as well as supporting functional experiments , will be necessary to dissect the interplay of these models . Our finding that approximately 1 pN of tension along individual filaments is sufficient for force-activated binding suggests that the actin-binding interface of α-catenin has been evolutionarily optimized to sense contractile forces generated by myosin motors ( Finer et al . , 1994 ) . We speculate force-activated binding enables the cadherin-catenin complex to recognize and preferentially engage pre-stressed actomyosin cables adjacent to the plasma membrane at adherens junctions ( Figure 7C ) , providing a mechanism for initial engagement between actin and the cadherin-associated population of α-catenin , as well as rapidly strengthening adhesion after preliminary attachments are formed . This could facilitate the transition from nascent cell-cell contacts to mature adherens junctions , as punctate nascent adhesions associated with radial actin cables coalesce and spread along a developing circumferential band of tensed actomyosin bundles ( Vaezi et al . , 2002 ) , as well as support dynamic adherens junction remodeling during epithelial morphodynamics ( Lecuit and Yap , 2015 ) . It also provides a mechanism for concentrating the soluble , cytoplasmic α-catenin population at sites of cytoskeletal tension ( Figure 7C ) . Previous studies have demonstrated this population is enriched on actomyosin bundles linked to adherens junctions , where it suppresses lamellipodium activity and promotes adhesion maturation ( Drees et al . , 2005 ) . While enrichment has been speculated to occur via local release from the cadherin complex , this model is difficult to reconcile with its low cellular concentration and the high concentration of soluble α-catenin required to inhibit Arp2/3-mediated actin branching in vitro ( Drees et al . , 2005 ) . Force-activated binding provides a feasible cellular mechanism for this enrichment of the homodimeric cytoplasmic α-catenin population , driven by enhanced affinity for tensed F-actin rather than mass action effects due to local concentration . Our identification of a functionally separable force-detector in α-catenin with amino-acid level precision will facilitate a detailed examination of these models in cell lines and in vivo . We note that force-activated binding is an additional , rather than alternative , mechanism to catch-bond formation ( Buckley et al . , 2014 ) for mechanical regulation of F-actin binding . Our finding that vinculin lacks this activity despite forming catch-bonds with F-actin ( Huang et al . , 2017 ) strongly suggests that these two modes of mechanical regulation operate by unique structural mechanisms , likely to fulfill distinct biological functions . Vinculin’s lack of force-activated actin-binding activity is consistent with the ordered sequence of mechanically-regulated binding events underlying its coordination with α-catenin at adherens junctions , where only after preliminary attachments form through the cadherin complex and come under load is vinculin recruited and activated to bind F-actin ( Yonemura et al . , 2010 ) . The lower force threshold for force-activated binding than catch-bond formation ( ~1 pN vs . ~10 pN ) supports a model in which the formation of initial attachments to the cytoskeleton through the cadherin complex is stimulated through force-activated binding , which is subsequently strengthened by catch-bonding through both α-catenin and vinculin during adhesion maturation . In their parallel work , Xu et al . speculate that the displacement of H1 from the α-catenin ABD upon actin binding could be stabilized or induced by force ( Xu et al . , 2020 ) , providing a possible structural mechanism for catch-bond formation which we have previously proposed could also be employed by vinculin ( Kim et al . , 2016; Swaminathan et al . , 2017 ) . This model predicts that the force required to dissociate H1 from the α-catenin ABD should be lower than the force required to displace the ABD in the mechanically-reinforced , strongly bound state from F-actin . It furthermore predicts that the constitutive high F-actin affinity α-catenin ABD construct ( residues 711–842 ) lacking H0-H1 reported by Xu et al . , which they propose mimics the strongly-bound state ( Xu et al . , 2020 ) , should no longer form catch bonds with F-actin due to its anticipated inability to switch between weakly and strongly bound states . Thus , this proposed structural mechanistic framework is now well-positioned to be subjected to explicit experimental scrutiny . We believe dissecting the interplay between force-activated binding and catch-bond formation in vitro and in vivo , as well as the structural basis for coordinated actin catch-bonding by α-catenin and vinculin , are important subjects for future studies . Our direct observation of force-activated binding to F-actin mediated by a short , flexible sequence element suggests this mechanism could feasibly be employed by other ABPs . Proteins from the large Calponin Homology ( CH ) domain ABP superfamily have diverse functions as cytoskeletal cross-linkers and plasma membrane/organelle tethers ( Liem , 2016; Razinia et al . , 2012 ) . Their ABDs have also been reported to have sequence elements that undergo folding transitions associated with filament engagement ( Avery et al . , 2017; Galkin et al . , 2010a; Iwamoto et al . , 2018 ) which can sterically regulate their actin binding ( Harris et al . , 2019 ) . This suggests members of this family could plausibly employ force-activated binding mechanisms similar to α-catenin to coordinate diverse mechanotransduction pathways throughout the cell , as could other ABPs with similar properties . The experimental strategy established here should be broadly useful for identifying force-sensitive ABPs and defining their force-detector-F-actin interfaces , such as the α-catenin CTE-F-actin interaction , in atomistic detail . This will facilitate elucidating the molecular and cellular mechanisms of cytoskeletal mechanotransduction with sufficient precision for guiding inhibitor development . Further information and requests for resources and reagents should be directed to the corresponding author , Gregory M . Alushin ( galushin@rockefeller . edu ) . All reagents generated in this study are available from the corresponding author without restriction . Globular actin ( G-actin ) monomers were purified from chicken skeletal muscle as described previously ( Pardee and Spudich , 1982 ) and maintained in G-Ca buffer: G buffer ( 2 mM Tris-Cl pH 8 , 0 . 5 mM DTT , 0 . 2M ATP , 0 . 01% NaN3 ) supplemented with 0 . 1 mM CaCl2 , at 4°C before use . C-terminally GFP-tagged mouse myosin V HMM and myosin VI S1 were purified from SF9 insect cells using published protocols ( Wang et al . , 2000 ) . All other proteins were heterologously expressed in Rosetta2 ( DE3 ) E . coli cells ( Novagen ) grown in LB media as described in Method details . Fluorescent dye JF-646 ( Grimm et al . , 2015 ) NHS-ester building block ( TOCRIS ) was conjugated with Halo-tag ligand amine O4 ( Promega ) by synthetic chemistry according to published protocols ( Grimm et al . , 2017 ) . Briefly , 1 . 5 equivalents of amine O4 ligand were added to one equivalent of the JF-646 NHS-ester in DMF followed by adding 5% triethylamine . The reaction was vigorously stirred for 16 hr at room temperature and the product was purified by silica gel chromatography , dried by SpeedVac ( ThermoFisher ) , and reconstituted in DMSO . For optical trapping/confocal microscopy assays , HaloTag Alexa Fluor 488 Ligand ( Promega ) was utilized as described above , followed by desalting through a PD SpinTrap G-25 column ( GE Healthcare ) according to the manufacturer’s protocol to remove unreacted dye before use . To label the Halo-tagged actin-binding proteins with Halo-JF-646 for TIRF microscopy assays , two equivalents of synthesized Halo-JF-646 dye was added to the protein solution , followed by incubation for at least 2 hr in the dark at 4°C before use . Subsequent removal of excess dye was not required , as JF-646 is a fluorogenic dye ( Grimm et al . , 2015 ) . Experiments were performed at room temperature ( approximately 25°C ) on a LUMICKS C-Trap instrument combining confocal fluorescence with dual-trap optical tweezers ( Hashemi Shabestari et al . , 2017; Wasserman et al . , 2019 ) . The optical traps were cycled through pre-set positions in the five channels of a microfluidic flow cell by an automated stage ( Figure 1A ) . Channels 1–3 were separated from each other by laminar flow , which we utilized to form actin filament tethers between two 4 μm-diameter streptavidin-coated polystyrene beads ( Spherotech ) held in optical traps with a stiffness of 0 . 3 pN/nm . We first captured a single bead in each trap in channel 1 . The traps were then transferred to channel 2 , containing 5–20 nM Alexa 555 phalloidin-stabilized , 10% biotinylated F-actin in motility buffer ( ‘MB’: 20 mM MOPS pH7 . 4 , 5 mM MgCl2 , 0 . 1 mM EGTA , 1 mM DTT ) supplemented with 1 µM dark phalloidin , where tethers were formed by briefly moving 1 of the two traps toward the other trap against the direction of flow , followed by rapidly moving the traps to channel 3 , which contained only buffer ( MB + 1 μM dark phalloidin ) . The presence of a tether was verified by carefully separating the traps and observing an associated increase in force when monitoring the force-extension curve , applying the minimum extension feasible to avoid prematurely rupturing the tether . The traps were then moved to orthogonal channel 4 or 5 , which contained 2 μM fluorescently labeled vinculin ABD or α-catenin ABD ( diluted in MB supplemented with 1 μM dark phalloidin ) , respectively , and flow was ceased during data acquisition . Force data were acquired at 200 Hz during constant velocity ( 0 . 1 μm/s ) pulling experiments while simultaneously acquiring 2-color confocal fluorescence scans at 33 ms line scan time , exciting Alexa Fluor 488 HaloTag ligand and Alexa Fluor 555 phalloidin fluorophores with laser lines at 488 nm and 532 nm , respectively . Data analysis was performed using ImageJ and custom software provided by LUMICKS . Force data from the two traps were averaged and binned to the confocal frame interval . The intensity values are measured by drawing a box in ImageJ to measure the fluorescence intensities of actin and ABP in both channels with background subtraction , calculating the background from an equal-sized box from that frame in an area devoid of filaments or beads . For paired analysis , the ‘high-force’ and ‘low-force’ averages were calculated only from the final tether to rupture , as long as the entire force trace has at least 10 quantifiable confocal image frames . For fluorescence-force correlation plots , only single-filament tethers selected based on Figure 1—figure supplement 2A–C were used . IABP/IActin values were normalized by dividing the values in each recording by the largest value observed during that recording . K-means clustering analysis was performed to identify the cutoff force in the correlation plots for each ABP . Briefly , the force-fluorescence data were grouped into two clusters using scikit-learn ( Pedregosa et al . , 2011 ) . For both α-catenin and vinculin , the data separated along the force axis , providing a threshold force for each ABP . Furthermore , a silhouette analysis for both ABPs confirmed that the data should not be clustered into more than two clusters ( Kaufman and Rousseeuw , 2009 ) . Glass coverslips ( Rectangular: Corning 22 × 50 mm #1½ Cover Glass; Square: Fisherbrand 22 × 22 mm #1½ Microscope Cover Glass ) were cleaned by 30 min 100% acetone wash , 10 min 100% ethanol wash , and 2 hr 2% Hellmanex III liquid cleaning concentrate ( HellmaAnalytics ) wash in a bath sonicator followed by rinsing with water . The cleaned glass coverslips were coated with 1 mg/mL mPEG5K-Silane ( Sigma ) in a 96% ethanol , 10 mM HCl solution for at least 16 hr . After coating , the coverslips were rinsed with 96% ethanol and water , then air-dried and stored at 4°C until use . Flow cells were prepared with one square and one rectangular coverslip , both coated with mPEG-Silane . Double-sided adhesive tape ( 3M ) was used to make ~4 mm-wide flow chambers between the coverslips , which were open on both sides to facilitate buffer exchange when adding components for imaging . For each assay , 6 mg/mL anti-GFP antibody ( Sigma #G1546 ) solution reconstituted in water was first introduced into the flow chamber and incubated for 2 min . Subsequently , MB containing 0 . 075 μM GFP-myosin V S1 and 0 . 15 μM GFP-myosin VI S1 were flowed into the chamber and incubated for another 2 min . A solution of 1 mg/mL bovine serum albumin ( BSA ) in MB was then flowed into the flow chamber and incubated for at least 2 min , after which 1 μM rhodamine-labeled F-actin in MB was flowed into the chamber and incubated for 20–30 s . The flow chamber was rinsed with MB buffer to remove F-actin not bound to the rigor-state motors , then imaging buffer ( MB without ATP , supplemented with 0 . 01% Nonidet P-40 substitute [Roche] , 1 μM calmodulin , 15 mM glucose [Sigma] , 1 μg/mL glucose oxidase [Sigma] , and 0 . 05 μg/mL catalase [Sigma] ) containing 2 μM fluorescently labeled ABP was flowed into the chamber . The first movie ( -ATP , no force ) was then recorded . A second imaging buffer , identical to the first but now including 100 μM ATP , was then introduced into the same chamber , and a second movie ( +ATP , with force ) was recorded . For each solution that was introduced , complete buffer exchange was facilitated by applying a filter paper at the other end of the flow chamber while pipetting . Dual-color TIRF image sequences ( movies ) were recorded at room temperature ( approximately 25°C ) using a Nikon TiE inverted microscope equipped with an H-TIRF module and an Agilent laser launch , driven by Nikon Elements software . Images were taken every 2 s with an Apo TIRF 60 × 1 . 49 NA objective ( Nikon ) on an Andor iXon EMCCD camera; Rhodamine and JF646 fluorophores were excited by laser lines at 561 nm and 640 nm , respectively . To quantify ABP association with individual ‘filament regions’ of TIRF movies , we developed a custom ImageJ ( Schneider et al . , 2012 ) plugin ( Figure 2—figure supplement 1A ) which features a graphical user interface ( GUI ) . The plugin takes as input two movie files , the actin channel and the ABP channel from a dual-color TIRF experiment , as well as an adjustable set of parameters ( set by default in the GUI to the optimized values used in this study ) . To identify regions of interest ( ROIs ) in each frame , the actin channel image series was preprocessed ( unsharp mask , median filter , rolling ball subtraction ) , binarized , then segmented into contiguous regions of pixels using the built-in ImageJ plugin ‘Analyze Particles’ . ROIs whose centroids were fewer than 30 pixels from the edge of the field-of-view were excluded from further analysis due to incompatibility with downstream background subtraction procedures . The ROIs were then tracked through the image series and sorted into filament regions ( representing individual filaments or small groups of filaments ) by shortest Euclidean distance between ROI centroids in neighboring frames , with a maximum distance cutoff of 24 pixels . ROIs that were not matched with a pre-existing filament region by this criterion ( i . e . whose centroid was greater than 24 pixels away from any ROI in the previous frame ) were considered to represent a newly appeared filament region . Although this may result in overcounting the absolute number of filaments ( should this be of interest; we do not believe this caveat impacts the conclusions of the present study ) , we find this procedure reliably handles events such as filament breakages and desultory motion in a completely automated fashion . To account for poorly tethered filaments fluctuating in and out of the evanescent field , only filament regions detected in at least 10 consecutive frames were included in the analysis . Intensity in both the actin channel and the ABP channel were then quantified for each region . For each channel , the local background for each filament region in each frame was calculated as the mean gray value of the pixels from a 60 by 60 pixel box centered on the region’s centroid , excluding pixels belonging to the region itself or any other filament region detected in the frame . Background-subtracted mean gray values were then calculated , followed by the ratio of these values across all frames in which the filament region was detected and their average , which we here report as the overall IABP/IActin for that filament region . The program then outputs all frame and average IABP/IActin values of the tracked filaments sorted by filament number for analysis , as well as a file containing all tracked regions . F-actin was polymerized in G-Mg and KMEI from 5 µM unlabeled actin monomers at room temperature for 1 hr and then diluted to 0 . 6 μM in KMEI before use . Purified metavinculin ABD ( 879–1134 ) was diluted in KMEI to 10 μM before use . After screening grids prepared with finely sampled ABP concentrations , we found diluting purified α-catenin ABD ( 664-906 ) to 20 μM in KMEI before use gave an optimal balance between filament decoration and bundling . Immediately before sample preparation , CF-1 . 2/1 . 3-3Au 300-mesh gold C-flat holey carbon cryo-TEM grids ( Protochips ) were plasma cleaned with a Hydrogen/Oxygen mixture for 5 s in a Gatan Solarus . Actin ( 3 μL ) was first applied to the grid in the humidified chamber of a Leica EM GP plunge freezer and incubated for 60 s at 25°C . Actin-binding protein ( 3 μL ) was then applied and incubated for 30 s . Solution ( 3 μL ) was then removed and an additional 3 μL of the same actin-binding protein solution was applied . After an additional 30 s , 3 μL of solution was removed , then the grid was back-blotted for 5 s , plunge-frozen in ethane slush , and stored in liquid nitrogen until imaging . Cryo-EM data were recorded on a Titan Krios ( ThermoFisher/FEI ) operated at 300 kV equipped with a Gatan K2 Summit camera . SerialEM ( Mastronarde , 2005 ) was used for automated data collection . Movies were collected at a nominal magnification of 29 , 000X in super-resolution mode resulting in a calibrated pixel size of 1 . 03 Å/pixel ( superresolution pixel size of 0 . 515 Å/pixel ) , over a defocus range of −1 . 5 to −3 . 5 μm; 40 frames were recorded over 10 s of exposure at a dose rate of 6 electrons per pixel per second ( 1 . 5 electrons per Å2 per second ) for a cumulative dose of 60 electrons per Å2 . Unless otherwise noted , all image processing was performed within the RELION-3 . 0 package ( Zivanov et al . , 2018 ) . Movie frames were aligned and summed with 2 × 2 binning using the MotionCor2 algorithm ( Zheng et al . , 2017 ) as implemented in RELION ( Zivanov et al . , 2019 ) , utilizing subframe motion correction with 5 × 5 patches . Contrast transfer function ( CTF ) parameters were estimated from non-doseweighted summed images with CTFFIND4 ( Rohou and Grigorieff , 2015 ) . Bimodal angular searches around psi angle priors were utilized in all subsequent 2D and 3D alignment/classification procedures . Around 2000 segments were initially manually picked , extracted , and subjected to 2D classification to generate templates for auto-picking . Helical auto-picking was then performed utilizing a step-size of 3 asymmetric units with a 27 Å helical rise . Segments were extracted from dose-weighted ( Grant and Grigorieff , 2015 ) sum images in 512 × 512 pixel boxes which were not further down-sampled , then a second round of 2D classification followed by auto-picking with featureful class averages was performed . A total of 237 , 503 particles from 1708 images ( for the metavinculin ABD-actin dataset ) and 540 , 553 particles from 7317 images ( for the α-catenin ABD-actin dataset ) were then extracted and subjected to whole-dataset 2D classification ( Figure 3—figure supplement 1 ) using a 200 Å tube diameter and 300 Å mask diameter . 234 , 703 segments from the metavinculin ABD-actin dataset and 428 , 335 particles from the α-catenin ABD-actin dataset contributed to featureful class averages and were selected for 3D analysis . All subsequent 3D classification and 3D auto-refine steps were primed with estimates of helical rise and twist of 27 . 0 Å and −167 . 0° , respectively , utilizing an initial reference low-pass filtered to 35 Å resolution , with the outer tube diameter set to 200 Å , inner tube diameter set to −1 , and the mask diameter set to 300 Å . The first round of 3D classification into three classes was performed utilizing a reconstruction of a bare actin filament ( EMBD-7115 ) as the initial reference . A second iteration of 3D classification was then performed as above , utilizing a featureful class with clear ABP density produced by the first round as the initial reference . For both datasets , this second round of 3D classification yielded two classes with helical parameters similar to the initial estimates and well-resolved 3D features , and one junk class with aberrant helical parameters and distorted features ( Figure 3—figure supplement 1 ) . Segments contributing to the two good classes were then pooled ( 215 , 369 particles for the metavinculin ABD-actin dataset and 414 , 486 particles for the α-catenin ABD-actin dataset ) for 3D auto-refinement . The first round of auto-refinement was then performed using one of the two good 3D averages as an initial reference . All masks for subsequent post-processing steps were calculated with 0 pixel extension and a six pixel soft edge from the converged reconstruction produced by that round of refinement , low-pass filtered to 15 Å and thresholded to fully encompass the density map . The first-round post-processing was performed with a 50% z length mask , followed by CTF refinement without beam-tilt estimation and Bayesian polishing ( Zivanov et al . , 2019 ) . A second round of auto-refinement was then performed using the converged reconstruction from the first round as the initial reference . The second-round post-processing was performed with a 30% z length mask , followed by a second round of CTF refinement with beam-tilt estimation and Bayesian polishing . A final round of auto-refinement was then performed using the converged reconstruction from the second round as the initial reference . We found that this iterative procedure of tightening the mask for polishing resulted in substantial resolution improvements , potentially by mitigating the effects of medium-range disorder in F-actin previously speculated to limit the resolution of reconstructions of this filament ( Galkin et al . , 2012; Merino et al . , 2018 ) . The final reconstructions converged with helical rise of 27 . 1 Å and twist of −167 . 1° for the metavinculin ABD–F-actin complex , and a helical rise of 27 . 0 Å and twist of −166 . 9° for the α-catenin ABD-F-actin complex , consistent with our finding that actin rearrangements evoked by these ABPs are minimal ( Figure 5H ) . Final post-processing was performed with a 30% z length mask , leading to global resolution estimates of 2 . 9 Å for the metavinculin ABD–F-actin complex and 3 . 2 Å for the α-catenin ABD–F-actin complex by the gold-standard Fourier shell correlation ( FSC ) 0 . 143 criterion ( Figure 3—figure supplement 1 ) . B-factors of both datasets estimated during post-processing were then used to generate sharpened , local-resolution filtered maps with RELION . The key statistics summarizing cryo-EM image processing are reported in Table 1 . Asymmetric focused classification ( without alignment ) utilizing masks isolating the ABD region showed no evidence of segments with unoccupied binding sites ( data not shown ) , suggesting that decoration of actin filaments by both 10 μM metavinculin ABD and 20 μM α-catenin ABD was essentially complete , with 100% occupancy at the limit of detection of current methods . Sharpened , local-resolution filtered maps as described above were used for model building . The 2 . 9 Å and 3 . 2 Å density maps were of sufficient quality for de novo atomic model building . As structures of components were available , initial models of actin ( PDB 3j8a ) , metavinculin ABD ( PDB 3jbk ) truncated to residues 981–1131 and α-catenin ABD ( PDB 4igg chain B ) truncated to residues 699–871 were fit into the density map using Rosetta ( Wang et al . , 2016 ) . Models were subsequently inspected and adjusted with Coot ( Brown et al . , 2015; Emsley et al . , 2010 ) , and regions that underwent significant conformational rearrangements were manually rebuilt . The models were then subjected to several rounds of simulated annealing followed by real-space refinement in Phenix ( Adams et al . , 2010; Afonine et al . , 2018 ) alternating with manual adjustment in Coot . A final round of real-space refinement was performed without simulated annealing . The key statistics summarizing model building , refinement , and validation are reported in Table 1 . Structural figures and movies were prepared with ChimeraX ( Goddard et al . , 2018 ) . Per-residue RMSD analysis was performed with UCSF Chimera ( Pettersen et al . , 2004 ) as previously described ( Zhang et al . , 2015 ) . The surface area of actin-binding interfaces was calculated with PDBePISA ( Krissinel and Henrick , 2007; ( EMBL-EBI ) . Model quality was assessed with EMRinger ( Barad et al . , 2015 ) and MolProbity ( Chen et al . , 2010 ) as implemented in Phenix . Protein sequences of human vinculin ( UniProt Accession Code P18206-2 ) , human metavinculin ( P18206-1 ) , human αE-catenin ( P35221 ) , human αN-catenin ( P26232 ) , and human αT-catenin ( Q9UI47 ) were aligned with ClustalOmega ( Sievers and Higgins , 2014; EMBL-EBI ) . Plotting and statistical analysis of data from TIRF force reconstitution assays and force-spectroscopy/confocal microscopy assays was performed with GraphPad Prism 8 . All the details can be found in the figure legends of these figures and in the Method details . The data collection and refinement statistics of the cryo-EM structures can be found in Table 1 . Resolution estimations of cryo-EM density maps and statistical validation performed on the deposited models are described in the Method details . The atomic coordinates for the metavinculin ABD–F-actin complex and α-catenin ABD–F-actin complex have been deposited in the Protein Data Bank ( PDB ) with accession codes 6UPW and 6UPV , and the corresponding cryo-EM density maps in the Electron Microscopy Data Bank ( EMDB ) with accession codes EMD-20844 and EMD-20843 . The code for analyzing TIRF movies is freely available as an ImageJ plugin with a graphical user interface at https://github . com/alushinlab/ActinEnrichment . ( copy archived at https://github . com/elifesciences-publications/ActinEnrichment; Alushinlab , 2020 ) . All other data are available in the manuscript or supplementary materials .	All of the cells in our bodies rely on cues from their surrounding environment to alter their behavior . As well sending each other chemical signals , such as hormones , cells can also detect pressure and physical forces applied by the cells around them . These physical interactions are coordinated by a network of proteins called the cytoskeleton , which provide the internal scaffold that maintains a cell’s shape . However , it is not well understood how forces transmitted through the cytoskeleton are converted into mechanical signals that control cell behavior . The cytoskeleton is primarily made up protein filaments called actin , which are frequently under tension from external and internal forces that push and pull on the cell . Many proteins bind directly to actin , including adhesion proteins that allow the cell to ‘stick’ to its surroundings . One possibility is that when actin filaments feel tension , they convert this into a mechanical signal by altering how they bind to other proteins . To test this theory , Mei et al . isolated and studied an adhesion protein called α-catenin which is known to interact with actin . This revealed that when tiny forces – similar to the amount cells experience in the body – were applied to actin filaments , this caused α-catenin and actin to bind together more strongly . However , applying the same level of physical force did not alter how well actin bound to a similar adhesion protein called vinculin . Further experiments showed that this was due to differences in a small , flexible region found on both proteins . Manipulating this region revealed that it helps α-catenin attach to actin when a force is present , and was thus named a ‘force detector’ . Proteins that bind to actin are essential in all animals , making it likely that force detectors are a common mechanism . Scientists can now use this discovery to identify and manipulate force detectors in other proteins across different cells and animals . This may help to develop drugs that target the mechanical signaling process , although this will require further understanding of how force detectors work at the molecular level .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2020	Molecular mechanism for direct actin force-sensing by α-catenin

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.